Geomorphology 55 (2003) 1 – 4

www.elsevier.com/locate/geomorph

Editorial
Introduction to the special issue: mountain
geomorphology—integrating earth systems

It were as well to be educated in the shadow of a lithosphere, and spatial techniques for characteriza-
mountain as in more classical shades. Some will tion, modeling, and scientific visualization. A total of
remember, no doubt, not only that they went to the 19 invited papers are presented here, representing all
college, but that they went to the mountain. Every but two of the oral papers, as well as three invited
visit would, as it were, generalize the particular papers based on outstanding poster presentations.
information gained below, and subject it to more Also presented here is the paper based on the sympo-
catholic tests. H.D. Thoreau, 1844. sium’s keynote address, presented by Professor Peter
W. Birkeland, University of Colorado.

1. Introduction
2. A festschrift for John D. (‘‘Jack’’) Vitek
The Binghamton Geomorphology Symposium has
been described as the most influential gathering of The 32nd Annual Binghamton Symposium was
geomorphologists in North America. The symposium also a festschrift in honor of Professor John D.
has a pronounced reputation for excellence earned (‘‘Jack’’) Vitek of Oklahoma State University (Fig.
over a quarter century of meetings and outstanding 1). Jack Vitek has served as an inspiration and mentor
scholarship. The theme of the 32nd Binghamton Geo- to each of the co-organizers of the 32nd Symposium
morphology Symposium, ‘‘Mountain Geomorphol- (Professors Butler, Walsh, and Malanson), as well as
ogy—Integrating Earth Systems,’’ and its timing in to many prominent geomorphologists. His collabora-
late 2001 were to commemorate the United Nations’ tive partners have included such names as Don
declared International Year of the Mountain in 2002. Coates, ‘‘Dusty’’ Ritter, Jack Shroder, Rick Giardino,
While all places integrate the Earth’s subsystems and Dick Marston. As a student of Neil Salisbury at
(atmosphere, hydrosphere, biosphere, and litho- the University of Iowa, Jack began his career in
sphere), mountain landscapes provide important per- mountain geomorphology with his first publications
spectives and opportunities for research. At any given on earth mounds of south – central Colorado in the
extent, mountains are more heterogeneous than other early 1970s. By the late 1970s, he had found his
landscapes; thus many, not only topographic, but also ‘‘special place’’, the Sangre de Cristo Range of
other gradients are steep and spatial interactions are south –central Colorado, and specifically the Blanca
concentrated. This concentration intensifies the inter- Massif (visible as the isolated circular mountain to his
actions among the subsystems and also raises scale left in the photograph). The bulk of his professional
questions in and of itself. Mountain landscapes are publications on mountain geomorphology have dealt
thus valuable for advancing our understanding of the with aspects of the periglacial geomorphology of this
Earth system in general. The symposium featured alpine region, with special attention given to perigla-
invited papers and contributed posters on the topics cial patterned ground, and rock glaciers. He brought
of mountain atmosphere, hydrosphere, biosphere, rigorous statistical analyses to his work, and has

0169-555X/03/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0169-555X(03)00128-4
2 Editorial

Fig. 1. Jack Vitek in his office. The Sangre de Cristo Mountains and Blanca Massif of south – central Colorado (USA) are shown on the satellite
image on the wall behind him.

always been willing to employ new field approaches Binghamton meetings are the stuff of legend, as are
and new methods in the burgeoning fields of remote his marathon 16 –20 h drives from Stillwater, Okla-
sensing and Geographic Information Sciences in homa, to attend the meetings held in Buffalo or
order to better understand this mountain environment. Binghamton, New York! Jack has been one of the
His publications include numerous papers in out- Binghamton Symposia series’ strongest, most visible,
standing outlets such as Geomorphology, Arctic and and most enthusiastic supporters. It is in this spirit that
Alpine Research, Zeitschrift für Geomorphologie, we dedicate the proceedings of the 32nd Annual
Geografiska Annaler, Environmental Geology, and Binghamton Geomorphology Symposium, on the
Journal of Quaternary Science. He co-edited the appropriate topic ‘‘Mountain Geomorphology—Inte-
landmark book Rock Glaciers with Jack Shroder grating Earth Systems’’, as a festschrift in his honor.
and Rick Giardino, and co-edited two books compris- Thank you, Jack, from all of us!
ing the papers of previous Binghamton Symposia,
Thresholds in Geomorphology, with Don Coates
(1980), and Geomorphology—The Research Frontier 3. Organization of this special issue
and Beyond, with Rick Giardino (1993). Jack has also
maintained active research interests, with numerous This issue of Geomorphology is subdivided into
excellent attendant publications, in the history of the groups of papers representing each of the Earth’s
discipline of geomorphology, natural hazards, carto- systems and their interactions with other Earth sys-
graphic and GIS display of geomorphological phe- tems; hydrosphere, biosphere, lithosphere, and atmos-
nomena, and earth science education. phere, as well as papers that utilize spatial digital
Jack has served as an inspiration through his technologies within the context of GIScience tech-
dedication and service to our discipline. He served niques. The first set of papers, representing research
as Book Review Editor of Geomorphology from 1992 on the hydrosphere in mountain environments, can be
through 1995, and as Editor-in-Chief for the journal subdivided into two groups. The first group, repre-
from 1994 to 1998. Of particular significance to the sented by papers by Harden and Scruggs, Cornwell et
Binghamton Symposia in Geomorphology, Jack has al., Schrott et al., Marston et al., and Fonstad, exam-
served as a member of the Steering Committee for 2 ines sediment transport and hydrology in steep moun-
decades. His keen interest in, and support for, the tain environments including the Ecuadorian Andes,
 Editorial 3

the southern Appalachian Mountains (USA), the tus to many key geomorphological processes in
Luquillo Mountains of Puerto Rico, the Nanga Parbat mountain environments. James examines the role of
Himalaya (Pakistan), the Bavarian Alps (Germany), Pleistocene glacial processes and the inferred linkages
and the Sangre de Cristo Mountains (New Mexico, to erosional forms in the Sierra Nevada of California
USA). The paper by Lenzi and Comiti, examining the (USA). Slaymaker et al. utilize hydroclimatological
role of check dams in steep channels of the Italian and sediment budget approaches based on sediments
Alps, offers a logical segue from this first group to a from alpine lakes in British Columbia (Canada) to
group of papers that focuses on step-pool and pool- discern patterns in Holocene sedimentation, soil
riffle sequences in mountain streams. These papers, development, and glaciation. On a much briefer time
by Rathburn and Wohl, Chin, and Marion and Weirich scale, Wilkerson and Schmid describe the meteoro-
examine mountain stream pool systems in a variety of logical conditions associated with the production of
climatic settings, including northern Colorado (USA), debris flows and their attendant hazards in the steep
the Mediterranean mountains of southern California terrain of Glacier National Park, Montana (USA).
(USA), and the humid-climate Ouachita Mountains of The paper by Birkeland et al., resulting from the
central Arkansas (USA). keynote address by Peter Birkeland, itself represents
The biosphere in mountain environments is repre- an integration of all the Earth spheres operating and
sented by four papers that examine the interactions interacting in mountain environments, through its
among living organisms and geomorphic processes in examination of soils and geomorphology. The paper
mountain environments. Dorn describes the role of is based on many years of intensive and rigorous
wildfire as a weathering and erosion agent in the fieldwork by Birkeland and his co-authors in the Front
Sierra Ancha Mountains (Arizona, USA). Two papers Range of Colorado (USA), and stands as a testament
examine the role of geomorphology and substrate in to Birkeland’s outstanding contribution to the field of
influencing plant distributions; Perez describes the pedology, weathering, and geomorphological research
role of unstable volcanic substrate on the distribution in mountain environments.
of the Hawaiian Silversword on the island of Maui, The final set of papers illustrates emerging spatial
Hawaii (USA), and Butler et al. illustrate the interact- digital technologies associated with the examination
ing roles of lithology, structure, and geomorphology of geomorphic processes and landforms in mountain
in constraining the development of ribbon forest environments. Bishop et al. use remote sensing and
patterns in the Rocky Mountains of northern Montana geomorphometric measurements for studying relief
(USA). Hall and Lamont remind us that animals, too, production at Nanga Parbat (Pakistan). Marcus et al.
are a primary component of the biosphere, and illustrate the utility of high spatial resolution, hyper-
describe the role of animals as agents of erosion in spectral imagery for the mapping of in-stream hab-
the Canadian Rockies of British Columbia. itats and associated characteristics in riffle/pool
The lithosphere, of course, is the very core of streams (again returning to this significant theme in
mountain environments, the medium with which mountain geomorphology) in the mountains of Yel-
mountains are created. Montgomery and Lopez- lowstone National Park (USA). The final paper, by
Blanco examine lithospheric controls on drainage Walsh et al., also returns to prior themes of the
capture in the southern Sierra Madre Occidental importance of interactions between atmosphere and
(Mexico). Cruden describes the importance of the lithosphere, through the use of GIScience techniques
sedimentary lithology and structure in determining in examining the scale and pattern of relict solifluc-
the shapes of mountains in the Canadian Rockies. In tion treads and risers in the Rocky Mountains of
a milder climatic region with far older mountains, Montana (USA).
Mills also stresses the importance of bedrock control These papers illustrate interactions among subsys-
on the topography of the southern and central Appa- tems and substantiate our view that mountain land-
lachian Mountains (USA). scapes are important for Earth systems science. As we
The atmosphere, through medium and long-term move beyond 2002, the United Nations International
climate conditions and changes as well as through Year of the Mountain, we can use these works as a
short-term meteorological events, provides the impe- guide to enriching all of geomorphology.
4 Editorial

Acknowledgements Sonoma State University; Daniel Muhs, U.S. Geo-

logical Survey; Alan Nelson, U.S. Geological Survey;
The 2001 Binghamton International Geomorphol- Joseph W. Nicholas, Mary Washington College;
ogy Symposium was held on October 20 and 21, Francisco L. Perez, University of Texas; P. Daniel
2001, on the campus of the University of North Royall, University of Alabama; and John D. Vitek,
Carolina – Chapel Hill. Financial sponsorship for the Oklahoma State University.
meeting came from the James and Marilyn Lovell
Center and the Department of Geography at South-
west Texas State University, the Department of
Geography and the College of Arts and Sciences at
the University of North Carolina, and the Department
of Geography at the University of Iowa (the academic
homes of the three co-organizers). Support was also
provided by the Mountain GeoDynamics Research
Group and the International Year of the Mountains.
We would also like to extend our appreciation to
Christine Erlien, University of North Carolina, for her
David R. Butler*
assistance with local arrangements, and we offer a
Department of Geography,
very special thanks to the many individuals who
The James and Marilyn Lovell Center,
willingly gave their time and efforts in reviewing the
Southwest Texas State University,
papers for this special issue. In addition to all papers
601 University Drive, ELA 120-B,
being reviewed by the guest editors and North
San Marcos, TX 78666-4616, USA
American Editor Dick Marston, each paper was also
E-mail address: [email protected]
reviewed by one or more of the following individuals
(some of whom reviewed more than one paper):
Stephen J. Walsh
Daniel A. Cenderelli, University of Alabama; Anne
Department of Geography,
Chin, Texas A&M University; Dennis Dahms, Uni-
University of North Carolina,
versity of Northern Iowa; Lisa M. DeChano, Western
Chapel Hill, NC 27599-3220, USA
Michigan University; Richard W. Dixon, Southwest
Texas State University; James Doerner, University of George P. Malanson
Northern Colorado; Ronald Dorn, Arizona State Department of Geography, University of Iowa,
University; Andrew J. Elmore, Brown University; Iowa City, IA 52242, USA
Mark Fonstad, Southwest Texas State University; L.
Allan James, University of South Carolina; David
Leigh, University of Georgia; Ross K. Meentemeyer, * Corresponding author.
 Geomorphology 55 (2003) 5 – 24
www.elsevier.com/locate/geomorph

Infiltration on mountain slopes: a comparison

of three environments
Carol P. Harden *, P. Delmas Scruggs
Department of Geography, University of Tennessee, 304 Burchfiel Geography Building, Knoxville, TN 37996-0925, USA
Received 27 November 2001; received in revised form 21 June 2002; accepted 10 March 2003

Abstract

Water is well established as a major driver of the geomorphic change that eventually reduces mountains to lower relief
landscapes. Nonetheless, within the altitudinal limits of continuous vegetation in humid climates, water is also an essential
factor in slope stability. In this paper, we present results from field experiments to determine infiltration rates at forested sites in
the Andes Mountains (Ecuador), the southern Appalachian Mountains (USA), and the Luquillo Mountains (Puerto Rico). Using
a portable rainfall simulator – infiltrometer (all three areas), and a single ring infiltrometer (Andes), we determined infiltration
rates, even on steep slopes. Based on these results, we examine the spatial variability of infiltration, the relationship of rainfall
runoff and infiltration to landscape position, the influence of vegetation on infiltration rates on slopes, and the implications of
this research for better understanding erosional processes and landscape change.
Infiltration rates ranged from 6 to 206 mm/h on lower slopes of the Andes, 16 to 117 mm/h in the southern Appalachians,
and 0 to 106 mm/h in the Luquillo Mountains. These rates exceed those of most natural rain events, confirming that surface
runoff is rare in montane forests with deep soil/regolith mantles. On well-drained forested slopes and ridges, apparent steady-
state infiltration may be controlled by the near-surface downslope movement of infiltrated water rather than by characteristics of
the full vertical soil profile. With only two exceptions, the local variability of infiltration rates at the scale of 10j m overpowered
other expected spatial relationships between infiltration, vegetation type, slope position, and soil factors. One exception was the
significant difference between infiltration rates on alluvial versus upland soils in the Andean study area. The other exception
was the significant difference between infiltration rates in topographic coves compared to other slope positions in the tabonuco
forest of one watershed in the Luquillo Mountains. Our research provides additional evidence of the ability of forests and forest
soils to preserve geomorphic features from denudation by surface erosion, documents the importance of subsurface flow in
mountain forests, and supports the need for caution in extrapolating infiltration rates.
D 2003 Elsevier Science B.V. All rights reserved.

Keywords: Geomorphology; Infiltration; Denudation; Andes; Southern Appalachians

1. Introduction mountain terrain. As three-dimensional features

extending thousands of meters in height and thou-
Although the high peaks tend to capture our sands of kilometers in length, mountains encompass a
attention, they comprise only a small portion of variety of microenvironments. A major challenge to
geomorphologists is to identify the ranges of values
* Corresponding author. Fax: +1-865-974-6025. and scales of spatial variability of geomorphic pro-
E-mail address: [email protected] (C.P. Harden). cesses and their controls on mountain slopes. In this

0169-555X/03/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0169-555X(03)00129-6
6 C.P. Harden, P.D. Scruggs / Geomorphology 55 (2003) 5–24

paper, we focus on forested mountain slopes in humid (ii) Vegetation: plentiful moisture supports the growth
temperate to tropical regions. Soils on these slopes of continuous and luxuriant vegetation, which, in
comprise the ‘‘skin’’ of the mountains, and their turn, alters the moisture regime of slopes through
biophysical characteristics exert important control on interception and evapotranspiration and alters soil
rates of degradational processes. We seek to better biophysical characteristics to better retain and
understand the role of water as an agent of erosion, drain moisture.
and the characteristics of slope surfaces that partition
rainwater into surface and subsurface flow paths. The role of water as a driver of erosion has received
Precipitation and gravity are well established as considerable scientific study. The more complex and
major drivers of the geomorphic change that even- less direct roles of water in promoting slope stability
tually reduces mountains to lower relief landscapes. through its effects on vegetation have received less
Rainfall and water from snowmelt promote denuda- attention than they merit in geomorphology. Under-
tion in three ways: standing the balance that determines whether water has
a stabilizing or destabilizing effect on mountain slopes
(i) Weathering: moisture serves as a reactant and a at micro- to subregional scales requires better under-
transport agent in weathering processes in which standing the role of the solum in integrating climatic,
slopes lose strength and rocks become frag- biologic, and geological components of the hillslope
mented. environmental system. Recent attention to mountain
(ii) Erosion: the erosive energy of water striking and regions recognizes that the world’s largest rivers orig-
flowing across the land surface entrains and inate in mountains and that at least half of the world’s
transports particles downslope. population depends on water flowing in or from moun-
(iii) Mass wasting: water entering pore spaces in slope tains (Price, 1999). The importance of mountains as
surface materials contributes to the potential for sources of fresh water further underscores the need for
mass movement by adding mass, increasing pore better understanding the water cycle, including infiltra-
water pressure, and reducing strength. tion processes, on mountain slopes.
In this paper, we present results from field experi-
At the same time, however, especially in non-arid ments of rainfall runoff and infiltration in the Andes
mountain regions and within the altitudinal limits of Mountains (Ecuador), the southern Appalachian
continuous vegetation, water promotes slope stability: Mountains (USA), and the Luquillo Mountains (Puerto
Rico). We hypothesized that mid-scale (hillslope to
(i) Soil: water contributes to the development of soil, km) differences in geologic, edaphic, topographic, and
which stores moisture and promotes low energy, biotic conditions control the spatial variability of infil-
non-erosive, subsurface water movement. tration rates in forested mountain regions; and we ex-

Table 1
Study area locations and characteristics
Study area Jatun Sacha, Southern Appalachians Luquillo Mountains
Andes Mountains
Country Ecuador USA Puerto Rico
Latitude Longitude 01jS 78jW 36jN 84jW 18jN 66jW
Soil parent material Sedimentary formations, Limestone, dolostone, Volcanoclastic sediments,
alluvium (Tertiary) and shale formations tuff units, and dioritic
(Cambrian to Ordovician) intrusions (Cretaceous, Tertiary)
Annual rainfall (mm) 4100 1360 2600 – 3600
Elevation of study 350 – 450 250 – 350 250 – 1050
sites (m)
Vegetation type Tropical rainforest Mixed hardwood forest, except Tropical forest vegetation;
two sites in pines, one in grass, four main associations
and one in disturbed site
 C.P. Harden, P.D. Scruggs / Geomorphology 55 (2003) 5–24 7

pected relationships between infiltration rates and site position, the spatial heterogeneity of infiltration, the
factors to support a scientific basis for extrapolating influence of vegetation on infiltration, and the implica-
infiltration rates, determined at points, to broader ex- tions of this research for erosion and landscape change.
tents of montane forest. We also hypothesized that In the hydrologic cycle, rainwater returns to the
infiltration capacities (maximum rates) would be high atmosphere through evaporation and transpiration,
enough to absorb rainfall and prohibit surface runoff remains on land (detention storage) and vegetation
during most rainfall events. Based on the results of our surfaces (as interception), or percolates into the soil.
fieldwork in the three study areas, we examine the rela- The movement of water into the soil, called infiltration
tionship of rainfall runoff and infiltration to landscape (I), is generally measured indirectly. If evapotranspira-

Fig. 1. Jatun Sacha Biological Reserve showing locations of groups of sites. The area of the reserve is unshaded.
8 C.P. Harden, P.D. Scruggs / Geomorphology 55 (2003) 5–24

tion, detention, and interception are minor or absent, steady rate of infiltration under gravitational potential,
infiltration (I) can be calculated as I = R
RO, where R and B is a time-dependent term representing the
is rain and RO is a measure of runoff. Infiltration is not hydraulic potential gradient at the advancing wetting
a single process but an assemblage of processes involv- front. Rates of infiltration are usually compared by
ing gravity and forces of molecular attraction between comparing the A (steady-state) term (Whipkey and
soil and water molecules. It integrates three independ- Kirkby, 1978).
ent processes: (i) entry through the soil surface, (ii) Environmental factors that control infiltration rates
storage within the soil, and (iii) transmission through are rainfall rates, soil properties (including texture, pore
the soil (Dunne and Leopold, 1978). Infiltration rates characteristics, organic matter content, and structure),
are known to decline to a steady or quasi-steady state as vegetation, land use, depth of soil, and initial moisture
a soil becomes increasingly moist over the period of a (Betson, 1964; Dunne and Leopold, 1978). Most
storm or experimental wetting. The widely used Philip mountain slopes in humid regions are covered by
equation (Philip, 1957) gives the infiltration rate (I) as a forest, which contributes organic matter to the soil
function of time t in the form and increases soil drainage by promoting soil particle
aggregation and supporting soil fauna. Organic litter
I ¼ A þ Bt
1=2 ð1Þ protects the soil surface from compression and sealing
by raindrop impact. Other environmental factors
where A and B are constants that depend on the soil and increasing infiltration rates in forests are those that
its initial moisture distribution. A mainly represents the create cracks and voids, such as earthworms and tree

Fig. 2. The Oak Ridge Reservation showing locations of groups of sites.

 C.P. Harden, P.D. Scruggs / Geomorphology 55 (2003) 5–24 9

roots (Knapp, 1978). The effects of these factors all as a significant control on spatial variability of
vary spatially. subsurface runoff. Infiltration capacity has been
As interest in using geographic information sys- shown to change with topographic position, but the
tems and modeling environmental processes across trend of the change is not always the same. Grah et
broader spaces grows, so does recognition of the al. (1983) found infiltration capacity to decrease
spatial variability of infiltration rates and the differ- downslope. In contrast, Dunne et al. (1991) found it
ence between the net hydrologic performance of a to increase downslope. Because infiltration is defined
slope compared to that of discrete points within it as the vertical movement of water into soil (Hillel,
(Hawkins and Cundy, 1987). Jetten et al. (1993) 1971) and most commonly measured on horizontal
found the sample variance of infiltration rates for surfaces, studying the relationship of infiltration to
tropical rainforest soils to be so large that it was not slope position in mountain regions poses significant
possible to predict infiltration rate as a simple func- research challenges.
tion of soil properties. Loague and Gander (1990) Knowledge of hillslope hydrology has been ham-
analyzed 247 infiltration rate measurements at 25-, pered by the lack of measurements of soil hydraulic
5-, and 2-m spacings from a grassland catchment in properties, especially in the humid tropics (Bonell and
Oklahoma. They found that variations in infiltration Balek, 1993). Our research was undertaken to better
rates were not explained by soil texture and sug- document and understand apparent infiltration rates in
gested that animal activity, vegetation, and climate soils in tropical and temperate mountainous regions.
strongly affected the distribution of infiltration Specific research objectives were to explore the rela-
rates. tionships of infiltration rates with vegetation type,
In mountain environments, slope position may soils, and slope position as a step towards better
contribute to the spatial variability of infiltration understanding the spatial variability of infiltration rates
rates. Woods et al. (1997) identified slope position in forested mountain environments.

Fig. 3. The Luquillo Experimental Forest showing locations of groups of sites.

10 C.P. Harden, P.D. Scruggs / Geomorphology 55 (2003) 5–24

2. Study areas December – February period (Jatun Sacha Foundation,

1992). Beneath the canopy of the tropical forest at Jatun
The experiments and results come from three Sacha, we compared infiltration rates of soils formed
separate research projects: one on the lower eastern on recent alluvium to upland soils, principally Ultisols,
flank of the Andes Mountains in Ecuador, one in the to test the hypothesis that they would differ signifi-
southern Appalachian Mountains, and one in the cantly. Surface soils of both groups were loamy in
Luquillo National Forest in Puerto Rico. General texture.
characteristics of the three research areas are given Fieldwork in the southern Appalachians was con-
in Table 1. All are humid and forested; none are ducted in the summer of 1997 for the purpose of
known to have been glaciated. determining infiltration rates and investigating the
Field research at the Jatun Sacha Biological Reserve spatial variability of infiltration on the Oak Ridge
was conducted in June 1993 as part of a larger project Reservation, located 28 km NW of Knoxville, TN
investigating infiltration and hydrologic variability of (Fig. 2). We chose research sites on the f 15,000 ha
soils in several different humid tropical environments Oak Ridge Reservation to represent major differences
(Buhi, 1997) and comparing runoff on- and off-trails in lithology, vegetation, topographic position, and land
(Wallin and Harden, 1996). The Biological Reserve use history to begin to determine whether these factors
covers a discontinuous area of about 750 ha in the could be used to extrapolate infiltration rates across the
steeply rolling foothills of the Andes and extends onto Reservation. The Oak Ridge Reservation covers a
alluvial terraces and the flood plain of the Napo River series of NE – SW-trending ridges and valleys, with
(Fig. 1). Soil parent materials are Tertiary sedimentary bedrock geology of Early Cambrian to Mississippian
rocks (shales and marls of Chalcana and Tena Forma- sedimentary rocks. Structurally, the area is dominated
tions) deformed by the Andes Mountains (Ecuador, by thrust faults that formed when the Blue Ridge thrust
1982), and Andean-origin alluvium on terraces of the sheet pushed westward and deformed the Valley and
Napo River. Mean annual rainfall of f 4100 mm is Ridge sedimentary formations in front of it (Hatcher et
fairly distributed across the year, with a slightly drier al., 1992). Bedrock is predominantly carbonate (dolo-

Table 2
Comparison of research goals and methods
Study area Jatun Sacha Reserve, Andes, Ecuador Southern Appalachian Mountains, Luquillo National Forest,
Oak Ridge, TN Puerto Rico
Research goals (1.) Compare runoff and infiltration (1.) Determine spatial variability (1.) Determine infiltration
rates on-and off-trails of infiltration and runoff on rates for Luquillo Forest
Oak Ridge Reservation
(2.) Compare results of parallel (2.) Examine relation of infiltration (2.) Compare infiltration
research in Ecuadorian and rates to other variables rates of different forest
Costa Rican sites (soil, slope, geology, litter, types and slope positions
topographic position)
Research method Constant head single-ring Rainfall simulation Rainfall simulation
infiltrometer and
rainfall simulation
Number of trials 27 with ring infiltrometer, 21 73 54
with simulator
Antecedent moisture Wet Wet Wet
condition
Median intensity of 36.4 87 91.8
simulated
rainfall (mm/h)
Number of bulk density 15 73 54
sample sites
Bulk density depth (cm) 0 – 10 0 – 10 0 – 10
 C.P. Harden, P.D. Scruggs / Geomorphology 55 (2003) 5–24 11

We conducted field research in the Luquillo Exper-

imental Forest in Puerto Rico (Fig. 3) in January 1998
to develop a database of infiltration rates in support of
water balance modelling for the forest and to begin to
test the hypothesis that relationships between infiltra-
tion rates and vegetation association or slope position
could provide a sound basis for extrapolating infiltra-
tion rates across the forest. The Luquillo Mountains in
eastern Puerto Rico have steep slopes, with the high-
est point in the Luquillo Mountains, El Toro Peak
(1975 m), only 9 km from the coast. Vegetation in the
Luquillo Experimental Forest falls into four catego-
ries: tabonuco (Dacryodes excelsa) forest in the sub-
tropical wet and subtropical rainforest life zones;
Colorado (Cyrilla racemiflora) forest in the lower
montane wet and rainforest life zones; palm (Pretoea
montana) forest, which is associated with subtropical
and lower montane wet and rainforest life zones,
especially on steep and on poorly drained sites; and
dwarf (cloud) forest, which occurs above 900 m
(Wadsworth, 1951). Larsen et al. (1999) reported that
much of the Luquillo area remained forested during
the expansion of plantation and subsistence agricul-
ture in Puerto Rico. Two catchments in the NE
section of the experimental forest, the Bisley water-
sheds, have been the focus of long-term ecological
and silvicultural research programs (Scatena, 1989).
Fig. 4. Photograph of the McQueen rainfall simulator – infiltrometer
in the Luquillo Mountains, Puerto Rico.

Table 3
Characteristics and results of rainfall simulation experiments
mite, limestone), much of which is weathered to Study area Jatun Sacha Oak Ridge Luquillo
saprolite (isovolumetrically chemically weathered bed- Biological Reservation, Experimental
rock), and shale (Hatcher et al., 1992). Soils are formed Reserve, USA Forest,
in colluvial residuum (weathering products lacking the Ecuador Puerto Rico
isovolumetric characteristic of saprolite), with some of Number of trials 21 72 54
the carbonate ridge soils formed above 30 m or more of Number with NO 15 34 15
saprolite. Soils are predominantly Ultisols, Alfisols, runoff
Median rainfall 36.4 87 91.8
and Inceptisols, and typically present a cherty, silt loam rate (mm/h)
A horizon over a clayey to loamy B horizon (Lee et al., Median infiltration 31 69 38
1988). Present-day vegetation at most sites is second- rate (sites with infill
growth, mixed hardwood forest. Settlers of European < 100% rain)
descent moved into the area in the mid-1880s and had (mm/h)
Median runoff (% rain) 0 1 25
extensively cleared and farmed the land by 1880 (SCS, for all trials
1936). In 1942, when the US government took over the Range of infiltration 17 – 34 16 – 113
2 – 99
land for the development of atomic weapons and rates (only from
energy, much of the land had been abandoned and runoff-generating trials)
had returned to forest (Lafon, 1995). (mm/h)
12 C.P. Harden, P.D. Scruggs / Geomorphology 55 (2003) 5–24

Table 4
Runoff and infiltration results from Jatun Sacha, Ecuador
Site Soil Bulk Runoff: rainfall Infiltration rateb: Infiltration ratec: Infiltration range
densitya simulator (mm/h) simulator (mm/h) ring (mm/h) in site group
(g/cc) min.
max. (mm/h)
1.3 Upland 0.72 0 >38 ...
1B Upland 0.91 ... ... 20
1B Upland 0.91 ... ... 23
1B Upland 0.91 ... ... 36
1B Upland 0.91 ... ... 38
1B Upland 0.91 ... ... 41
1B Upland 0.91 ... ... 55
1B Upland 0.91 ... ... 79
1B Upland 0.91 ... ... 120 20 – 120
2.52 Upland 0.71 0 >47 ...
2.53 Upland 0.71 0 >36 ...
2.54 Upland 0.71 0 >32 ...
2.55 Upland 0.71 0 >48 ...
2.72 Upland 1.03 0 >18 ...
2.73 Upland 1.03 4 43 ...
2.92 Upland 0.89 0 >23 ...
2.93 Upland 0.89 0 >47 ...
2B Upland 0.96 ... ... 27
2B Upland 0.96 ... ... 35
2B Upland 0.96 ... ... 47
2B Upland 0.96 ... ... 70
2B Upland 0.96 ... ... 127 27 – 127
5.192 Upland 0.78 0 >45 ...
5.193 Upland 0.78 0 >11 ...
5.194 Upland 0.78 0 >30 ...
5B1 Upland 0.78 ... ... 8
5B2 Upland 0.78 ... ... 88
5B3 Upland 0.78 ... ... 98
5B4 Upland 0.78 ... ... 59 8 – 98
3.112 Alluvial 0.77 10 17 ...
3.113 Alluvial 0.77 63 31 ...
3.131 Alluvial 0.63 0 >19 ...
3.132 Alluvial 0.63 0 >43 ...
3.142 Alluvial 0.63 Trace >19 ...
3.143 Alluvial 0.63 7 34 ...
3.152 Alluvial 0.71 Trace >22 ...
3.153 Alluvial 0.71 178 31 ...
3.154 Alluvial 0.71 119 25
3B1 Alluvial 0.65 ... ... 82
3B2 Alluvial 0.65 ... ... 125
3B3 Alluvial 0.65 ... ... 136
3B4 Alluvial 0.87 ... ... 156
3B5 Alluvial 0.87 ... ... 156 0 – 156
4B1 Alluvial 0.87 ... ... 77
4B2 Alluvial 0.87 ... ... 86
4B3 Alluvial 0.87 ... ... 175
4B4 Alluvial 0.87 ... ... 206 77 – 206
a
Mean bulk densities of top 5 – 10 cm of soil determined from constant volume auger samples.
b
35-min trials on wet soil, off-trail on forest floor.
c
Steady rate from constant head in single ring on prewet soil.
 C.P. Harden, P.D. Scruggs / Geomorphology 55 (2003) 5–24 13

Rainfall increases with elevation from about 2600 in situ measurements of runoff in remote and steep
mm year
1 at the base to nearly 5000 mm year
1 at locations. The rainfall simulator – infiltrometer’s plex-
the summit (Brown et al., 1983). The clay-rich soils iglass reservoir, rainulator, and windscreen are
are principally Ultisols (50% of area) and Inceptisols mounted on a standard surveyor’s tripod (Fig. 4).
(45%), with Entisols and Oxisols comprising the From the reservoir, the water flows through a bub-
other 5% (Brown et al., 1983). We chose sampling bler system that maintains a constant head in the
sites to include as many combinations of the forest rainulator to ensure uniform application of ‘‘rain’’
type and topographic position as possible; however, throughout the duration of the experiment. The rain-
the dwarf (cloud) forest is limited to high ridges. ulator forms 5-mm diameter drops on 91 pins
inserted in precisely drilled holes, and the drops fall
through the cylindrical windscreen to the soil sur-
3. Research methods face. We placed the infiltrometer ring directly under
the windscreen, penetrating the soil surface by 2 –3
Each of the three studies discussed in this paper cm to define the 182.4-cm2 sample plot and prevent
had a slightly different set of research questions and lateral movement of the simulated rain water on the
research design; yet all three involved measuring soil surface.
rainfall runoff and infiltration rates and used identical Most plots were wet to at least field capacity from
or similar research methods. Research methods for recent and ongoing rainfall, or, in a few cases, we
the three sites are summarized in Table 2. In Puerto prewet them. We further prepared plots by removing
Rico, in Tennessee, and at 21 of the Ecuadorian twigs and undecomposed leaves so that simulated rain
sites, we used a McQueen (1963)-type rainfall sim- would not be blocked from reaching the soil surface.
ulator with a ring infiltrometer to determine the Rainfall intensity, determined by the head of water in
infiltration rate of wet soils on the forest floor. The the rainulator, was constant throughout each experi-
McQueen rainfall simulator – infiltrometer is light- ment. During the experiments, we noted incremental
weight and can be hand carried to the field, allowing rainfall rates and measured runoff by extracting and

Fig. 5. Box plot showing infiltration rates at Jatun Sacha for sites on two different soils, as determined by single-ring infiltrometer. Horizontal
line inside the box is the median value.
14 C.P. Harden, P.D. Scruggs / Geomorphology 55 (2003) 5–24

Table 5
Runoff and infiltration results from Oak Ridge Reservation, USA
Site Slope Bulk density Simulated Runoff Runoff ratio Duration of Infiltration rate Infiltration Vegetation
position (g/cc) rain (mm) (ml) (ml/ml) trial (min) (mm/h) range (mm/h) and geology
C12.11 Middle 2.18 21 0 0.00 30 >41 >36 – 98a Under canopy,
C12.12 Middle 1.20 51 38 0.04 30 98 deciduous forest,
C12.13 Middle 1.08 44 208 0.25 30 66 on Eidson
C12.21 Middle – 18 0 0.00 30 >36 member (shale)
C12.22 Middle 1.28 43 114 0.14 30 73
C12.23 Middle 1.16 44 132 0.16 30 74
C12.14 Bottom 0.96 51 244 0.25 30 76 59 – 109 Under canopy,
C12.15 Bottom 1.17 55 6 0.01 30 109 deciduous forest,
C12.24 Bottom 1.13 50 64 0.07 30 93 on Eidson
C12.25 Bottom 0.96 43 250 0.31 30 59 member (shale)
C17.11 Top – 23 0 0.00 30 >45 no runoff Under canopy,
C17.12 Top – 40 0 0.00 30 >80 >32 – >82 deciduous forest,
C17.21 Top – 16 0 0.00 30 >32 on Copper Ridge
C17.22 Top – 41 0 0.00 30 >82 formation (carbonate)
7.1.11 Bottom 0.92 44 0 0.00 30 >88 26 – >91 Under canopy,
7.1.12 Bottom 1.18 54 606 0.59 30 44 deciduous forest,
7.1.13 Bottom 1.29 43 30 0.04 30 84 on Nolichucky
7.1.14 Bottom 1.03 46 0 0.00 30 >91 formation (shale)
7.1.21 Bottom 1.12 49 526 0.56 35 26
7.1.22 Bottom 1.06 41 0 0.00 30 >81
7.1.23 Bottom 1.15 36 7 0.01 35 61
7.1.24 Bottom 1.27 28 3 0.01 30 56
7.2.11 Top 1.26 43 13 0.02 30 84 >51 – >91 Under canopy,
7.2.12 Top 1.05 46 0 0.00 30 >91 deciduous forest,
7.2.13 Top 1.04 40 0 0.00 30 >79 on Rogersville
7.2.21 Top 1.18 26 0 0.00 30 >52 formation (shale)
7.2.22 Top 0.85 26 0 0.00 30 >51
7.2.23 Top 1.19 76 11 0.01 60 75
7.3.11 Middle 1.10 97 77 0.04 60 89 27 – 89 Open, formerly
7.3.12 Middle 1.23 52 618 0.62 30 40 managed as grassland,
7.3.13 Middle 1.13 61 738 0.63 35 27 on Dismal Gap
7.3.21 Middle 1.22 52 0 0.00 50 >62 formation (shale)
7.3.22 Middle 1.13 44 380 0.46 35 33
7.4.11 Top 1.17 55 35 0.03 30 107 68 – 107 Under canopy,
7.4.12 Top 1.01 50 48 0.05 30 95 deciduous forest,
7.4.21 Top 1.12 36 18 0.03 30 71 on Dismal Gap
7.4.22 Top 1.13 36 41 0.06 30 68 formation (shale)
wb1.11 Top 1.03 40 0 0.00 30 >79 >48 – 94a Under canopy,
wb1.21 Top 1.27 35 0 0.00 30 >70 deciduous forest,
wb2.11 Top 1.10 48 14 0.02 30 94 on Copper Ridge
wb2.21 Top 1.11 24 0 0.00 30 >48 formation (carbonate)
wb3.11 Middle 1.21 99 82 0.04 55 97 >52 – 97a Under canopy,
wb3.12 Middle 1.32 43 97 0.12 30 75 deciduous forest,
wb3.21 Middle 1.13 35 0 0.00 30 >69 on Kingsport
wb3.22 Middle 0.47 26 0 0.00 30 >52 formation (carbonate)
wb4.11 Top 0.87 45 0 0.00 30 >89 no runoff Under canopy,
wb4.12 Top 1.17 46 0 0.00 30 >92 >92 deciduous forest,
wb4.21 Top 1.04 29 0 0.00 30 >58 on Kingsport
wb4.22 Top 0.70 33 0 0.00 30 >66 formation (carbonate)
 C.P. Harden, P.D. Scruggs / Geomorphology 55 (2003) 5–24 15

Table 5 (continued )
Site Slope Bulk density Simulated Runoff Runoff ratio Duration of Infiltration rate Infiltration Vegetation
position (g/cc) rain (mm) (ml) (ml/ml) trial (min) (mm/h) range (mm/h) and geology
bc1.11 Top 0.90 58 0 0.00 40 >87 16 – >87 Under young canopy,
bc1.12 Top 1.07 41 473 0.80 30 16 deciduous forest,
bc1.13 Top 1.12 41 190 0.25 30 61 on Dismal Gap
bc1.21 Top 1.05 46 0 0.00 40 >68 formation (shale)
bc1.22 Top 1.18 50 24 0.02 40 72
bc1.23 Top 1.01 59 30 0.03 40 84
bc2.11 Middle 1.71 50 723 0.76 30 24 23 – 56 In pine forest,
bc2.12 Middle 1.20 48 695 0.76 30 23 over shale
bc2.21 Middle 1.10 38 428 0.59 30 31 (Nolichucky)
bc2.22 Middle 1.21 45 134 0.15 40 56
bc3.11 Middle 1.09 47 0 0.00 30 >94 48 – >94 In pine forest, over
bc3.12 Middle 1.13 47 0 0.00 30 >93 shale (Dismal Gap)
bc3.21 Middle 1.24 36 12 0.02 30 72
bc3.22 Middle 1.12 39 293 0.39 30 48
bc4.21 Bottom 1.00 43 0 0.00 30 >86 >117 Under canopy,
bc4.11 Bottom 1.28 58 0 0.00 30 >117 no runoff deciduous forest,
on Rogersville
formation (shale)
bc5.11 Middle 1.20 50 191 0.20 30 80 50 – 80 Open, disturbed by
bc5.21 Middle 1.72 32 141 0.23 30 50 vehicles Dismal
Gap formation (shale)
pc1.11 Middle 0.91 43 0 0.00 30 >86 43 – 113a Under canopy,
pc1.12 Middle 1.07 77 36 0.02 40 113 deciduous forest,
pc1.13 Middle 0.81 54 0 0.00 30 >109 on Copper Ridge
pc1.21 Middle 1.01 30 0 0.00 30 >60 formation (carbonate)
pc1.22 Middle 1.21 32 18 0.03 40 43
pc1.23 Middle 0.90 18 0 0.00 30 >36
a
Maximum may be underestimated because runoff did not occur in all trials.

measuring all water ponded in the infiltrometer ring at etation in the southern Appalachians, and slope posi-
5-min intervals. For sites in which runoff was gener- tion and forest type in the Luquillo Experimental
ated, we calculated the infiltration rate over each 5-min Forest. We conducted multiple trials at groups of sites
interval by subtracting runoff volume from rain vol- in each study area so that we could compare the
ume. In Puerto Rico, we continued the experiments for variability of infiltration rates within and between site
30 min or until the runoff rate became steady. The groups.
experiments in Ecuador and most in Tennessee were In Tennessee and Puerto Rico, we used a root auger
discontinued after 30 – 35 min. For rainfall simulation (8 cm diameter) to extract soil cores at two depths (0 –
experiments in Tennessee and Puerto Rico, we used 10 and 10 –25 cm) before and after rainfall simula-
high rain intensities to deliberately exceed infiltration tions. We oven dried and weighed the cores to
rates; but in the Ecuadorian study, we used lower quantify antecedent moisture and changes in moisture
intensities in order to compare runoff thresholds on- resulting from infiltration and to calculate bulk den-
and off-trails. Only the off-trail results are reported in sity. In the Ecuadorian study, we used a hammer-
this paper. driven constant volume (8.5 cm long by 4.85 cm
The portability of the rainfall simulator– infiltrom- diameter) auger for bulk density samples.
eter allowed us to conduct infiltration experiments at An additional set of in situ measurements was
sites throughout the study areas. Site selection was made in the Ecuadorian study using a single-ring
designed to achieve reconnaissance-level comparisons infiltrometer pressed 10 cm into the soil with a
between site factors: upland versus alluvial soils in constant head of 4 cm. Although the research was
Ecuador, different slope positions, lithology, and veg- completed during the wet season, we prewet the
16 C.P. Harden, P.D. Scruggs / Geomorphology 55 (2003) 5–24

surrounding soil to minimize lateral seepage. The 4.2. Oak Ridge Reservation, southern Appalachian
amount of water added to maintain a constant head Mountains
was monitored, and the infiltration rate recorded
when it reached a steady state (Buhi, 1997). At the southern Appalachian study area, apparent
infiltration rates of runoff-producing trials varied from
16 to 117 mm/h (Table 5, Fig. 6A). A salient feature of
4. Results these data is the within-site variability. Using only data
from runoff-producing trials, a Kruskal – Wallis test
At each study area, we found infiltration rates to indicated that median infiltration rates did not differ
vary considerably within site groups. Often, the full significantly (a = 0.05) between trial locations. The
range of infiltration rates for a study area occurred high rate of rainfall application proved to be less than
within a single site. Table 3 summarizes the results of that required to generate runoff, even under moist soil
rainfall simulation experiments at all three study conditions, at nearly half (34 of 72) of the trials. We
areas. At 34 of 72 sites in Tennessee, 15 of 54 in obtained only one steady-state infiltration rate (94 mm/
Puerto Rico, and 15 of 21 in Ecuador, the rate of rain h) in 12 trials on carbonate ridge tops compared to 10
applied was not enough to generate surface runoff. rates, ranging from 16 to 107 mm/h, from the 16 trials
For those sites, we report that infiltration capacities on shale ridges.
exceed the rate of rainfall applied. Because the Infiltration rates of runoff-producing trials did not
research goals, designs, and findings differed between differ significantly (Kruskal –Wallis test, a = 0.05) by
study areas, we present the additional results from topographic position (Fig. 6B), but more of the trials
each area separately. with no runoff were on ridgetops (55%) than mid-
(29%) or low-slope (21%) positions. For the Oak Ridge
4.1. Jatun Sacha, Ecuadorian Andes Reservation, we found no significant linear relation-
ships between infiltration rates and bulk density. Ante-
On the tropical forest floor at Jatun Sacha, the litter cedent moisture contents of surface (0 –10 cm) cores
layer was typically just one leaf deep. Under natural showed no relationship to topographic position or
rainfall, we observed that the larger leaves (e.g., infiltration rate, but subsurface (10 – 20 cm) cores from
Cecropia) block infiltration, causing rainfall runoff ridgetop sites typically had slightly lower gravimetric
over short distances. With intact leaves removed from moisture contents than those from slope bottom sites
the soil surface for our experiments, apparent infiltra- (median 23 g/g on ridge compared to 26 g/g on bottom,
tion rates at Jatun Sacha ranged from 17 to 43 mm/h with one outlier removed in each case).
in the six (of 21) rainfall simulation experiments that
produced runoff and from 6 to 206 mm/h in the 4.3. Luquillo Experimental Forest, Puerto Rico
constant head experiments (Table 4). Rainfall records
for 1987 – 1992 at the research station indicated that Apparent infiltration rates in the Luquillo Exper-
the relatively light rainfall intensities used in these imental Forest ranged from 0 to >106 mm/h and varied
experiments occurred multiple times (eight estimated) considerably within vegetation zones (Table 6, Fig.
per year (Wallin, 1995). 7A). We found no significant difference in infiltration
Infiltration rates determined with the constant head rates between different forest types (Kruskal Wallis,
infiltration ring differed significantly (a = 0.05, Mann a = 0.05), although cloud forest soil conditions visibly
Whitney U-test) between sites with alluvial and differed from those at other montane sites. Antecedent
upland soils (Fig. 5). Alluvial surface soils had higher moisture was higher in the cloud forest sites, and soil
infiltration rates, with a mean of 134 F 38 mm/h, bulk density seemed lower, although it was also very
compared to the mean of 57 F 35 mm/h for the upland low at occasional sites in the tabonuco and Colorado
soils. Within-site trials demonstrated a high degree of forests. Although we did not measure soil organic
very local variation of surface infiltration rates at the matter, we observed that cloud forest soils contained
scale of f 1 m, on both alluvial and upland soils much more visible organic matter than soils in other
(Buhi, 1997). types of forest. Across the experimental forest, we did
 C.P. Harden, P.D. Scruggs / Geomorphology 55 (2003) 5–24 17

Fig. 6. Box plots of infiltration rates on the Oak Ridge Reservation, as determined from rainfall simulation experiments that produced surface
runoff: (A) shows infiltration rates by site group; (B) shows infiltration rates for all runoff-producing trials plotted by topographic position.

not find a statistically significant relationship between higher (a = 0.05) infiltration rates than the coves they
antecedent moisture and infiltration rates or a signifi- surrounded (Fig. 7B). Likewise, none of the 15 (of
cant relationship between apparent infiltration rate and 54) sites with no runoff in the experimental forest
soil bulk density. were in cove positions. In the Bisley watersheds
An intriguing topographic pattern that emerged in (tabonuco and palm forests), we observed that slopes
the Bisley 1 watershed of the Luquillo Experimental drained efficiently without producing surface runoff
Forest is that ridges and side slopes had significantly in steady rain; but coves became much wetter and
18 C.P. Harden, P.D. Scruggs / Geomorphology 55 (2003) 5–24

Table 6
Infiltration data from Luquillo Experimental Forest, Puerto Rico
Forest type Site Bulk Topographic Slope Simulated rain Infiltration rate Infiltration rate
density positiona (j) (mm/h) (mm/h) range (mm/h)
(g/cc)
Colorado C.1.01 0.58 R 9 107.2 83 1 – >105
Colorado C.2.01 1.17 S 15 69.4 1
Colorado C.1.02 0.50 R 15 105.0 >105
Colorado C.2.02 0.90 R 9 74.8 >75
Colorado C.1.03 0.58 S 27 101.2 82
Colorado C.2.03 0.55 S 18 77.4 29
Colorado C.1.04 0.41 S 36 113.2 8
Colorado C.2.04 0.45 S 34 68.6 51
Colorado C.2.05 0.80 C 7 55.6
2
Cloud forest EP.1.01 0.76 R 2 100.0 6 6 – 42
Cloud forest EP.1.02 nd R 22 105.0 21
Cloud forest EP.1.03 nd R 21 124.6 7
Cloud forest EP.1.04 0.42 R 16 135.4 42
Cloud forest EP.1.05 nd R 10 102.0 6
Cloud forest EP.1.06 0.30 R 20 109.2 32
Cloud forest EP.1.07 0.37 R 21 94.6 39
Palm P.1.02 0.67 S 25 96.6 83 18 – 83
Palm P.1.03 0.71 S 21 115.6 18
Palm P.2.01 0.65 S 9 87.4 81
Palm P.2.02 0.64 S 25 74.6 60
Palm P.2.03 0.57 S 20 79.8 71
Tabonuco B.1.01 0.65 C 5 79.2 4 0 – 69
Tabonuco B.2.01 0.78 C 10 71.4 0
Tabonuco B.1.05 0.56 C 5 106.0 69
Tabonuco B.1.11a 0.48 C 14 114.0 11
Tabonuco B.1.11b 0.60 C 12 107.8 41
Tabonuco B.1.02 0.65 W 3 68.0 5 3–5
Tabonuco B.2.02 0.68 W 2 60.4 3
Tabonuco B.1.03 nd R 40 77.6 >78 no runoff
Tabonuco B.2.03 0.71 R 40 62.0 >62 >106
Tabonuco B.1.04 0.62 R 23 106.0 >106
Tabonuco B.2.04 nd R 32 81.0 >81
Tabonuco B.2.06 0.71 S 30 78.0 >78 76 – >99
Tabonuco B.1.07 0.56 S 3 98.6 >99
Tabonuco B.1.07b 0.60 S 36 82.2 76
Tabonuco B.2.07 0.64 S 20 69.2 >69
Tabonuco B.2.07b 0.66 S 46 101.8 77
Tabonuco B.1.08 0.59 S 20 81.6 21 10 – 94b
Tabonuco B.2.08 0.64 S 13 88.0 >88
Tabonuco B.1.09 0.57 S 5 100.6 94
Tabonuco B.2.09 0.61 S 32 106.6 10
Tabonuco B.2.10A 0.28 S 30 98.0 83
Tabonuco B.2.10B 0.55 S 36 59.8 >60
Tabonuco B.1.12 0.59 R 4 83.2 >83 74 – 99b
Tabonuco B.2.12 0.81 R 7 85.0 >85
Tabonuco B.1.13 0.24 R 53 102.2 99
Tabonuco B.2.13 0.73 R 30 81.4 74
Tabonuco B.1.14 0.43 R 4 96.2 >96
Tabonuco B.2.14 0.63 R 0 92.0 >92
a
C = cove, R = ridge, S = side slope, W = riparian.
b
Maximum may be underestimated because runoff did not occur in all trials.
 C.P. Harden, P.D. Scruggs / Geomorphology 55 (2003) 5–24 19

Fig. 7. Box plots of infiltration rates in the Luquillo Experimental Forest: (A) shows infiltration rates by forest type, using rainfall intensity as a
minimum estimate for infiltration rate for 15 sites without runoff; (B) shows infiltration rates for all runoff-producing sites plotted by
topographic position.

generated surface flow. During one sustained rain- the surface but did observe a consistently wet zone
storm, we climbed around in the Bisley 1 watershed in the top 3– 5 cm of soil (sometimes as deep as 10
with a soil auger to study the response to natural cm), where water was visibly draining downslope
rain. We observed little to no runoff flowing across through a near-surface fine root zone. Soil below this
20 C.P. Harden, P.D. Scruggs / Geomorphology 55 (2003) 5–24

depth was not saturated, in spite of the steady rain. produced no runoff in our experiments seems ex-
Throughout our experiments in the Luquillo Exper- tremely low. At the US Forest Service’s Coweeta
imental Forest, we found numerous earthworms in Hydrological Laboratory in the southern Appala-
our samples. According to National Forest personnel, chians in NC, 13% of annual rainfall is lost through
earthworms are considered to be the faunal species interception (Coweeta, 1984). Very high infiltration
with greatest biomass in the Luquillo Experimental rates observed in many of our experiments at Oak
Forest. Ridge are consistent with previous research on the
reservation that documented the importance of subsur-
face flow during rain events. Wilson et al. (1990)
5. Discussion found a perched saturated zone above the soil/sapro-
lite interface ( f 2 m deep) during storms, and Mul-
5.1. Jatun Sacha, Andes holland et al. (1990) reported evidence that vadose
zone and saturated soil zone flowpaths as well as
Infiltration rates in the tropical rainforests at Jatun bedrock zone flowpaths contributed to stream flow in
Sacha and Puerto Rico are comparable in magnitude the Walker Branch watershed at Oak Ridge.
and local variability to those of forested sites in For the 34 Oak Ridge sites with infiltration capaci-
Tennessee. The most notable trend we observed at ties exceeding the rate of rainfall application (87 mm/
Jatun Sacha was that the alluvial soils had higher h median), we did not detect other site factors that
apparent infiltration rates than the older upland soils. could provide a sound basis for spatially extrapolating
Even in the upland, the most intense rainstorms of a infiltration rates. Twelve of the sixteen site groups at
typical year would only generate runoff at a small Oak Ridge had at least one trial yielding no runoff,
proportion of the off-trail, forested sites tested if all of and 15 of 16 had at least one apparent infiltration rate
the rain reached the land surface. In this closed in excess of 80 mm/h. Such local variability is
canopy forest, however, interception by trees makes consistent with findings by Loague and Gander
surface runoff even less likely. Although only one of (1990) and Robertson et al. (1997) for other locations
nine rainfall simulation trials at upland sites yielded in the US. On the Oak Ridge Reservation, infiltration
runoff, rainfall simulation experiments using higher processes at different slope positions may be affected
rainfall intensities were not undertaken because the by the geologic situation in which most of the ridges
Jatun Sacha research had different objectives. (deep saprolite from limestone and dolostone forma-
As expected in the tropical forest, soil surface tions, and fractured carbonate bedrock) are very
horizons at Jatun Sacha contained little visible organic permeable, while the valleys are underlain by much
matter. Although the research station calls much of the less permeable shale formations. Further research
forest ‘‘bosque primario’’ (primary forest), the pres- seeking statistically significant differences in infiltra-
ence of potsherds at our study sites served as a tion rates between soil types, topographic positions,
reminder that this site, too, has a legacy of human or land use history at Oak Ridge or our other study
use, which may have affected soil compaction and soil areas would require a much larger set of samples with
infiltration capacities. much more water applied. Even then, microscale
differences may continue to dominate over broader-
5.2. Southern Appalachian Mountains scale patterns, as occurred in the work by Jetten et al.
(1993).
Infiltration rates exceeded the normal range of All of the reservation sites have a land use history
rainfall intensities across much of the Oak Ridge involving forest clearing at some time prior to 1941, so
Reservation. The median intensity of applied rainfall some microscale differences in infiltration rates may
(87 mm/h) approximated that of a 30-min rain with a reflect site-specific historic anthropogenic effects. The
10-year recurrence interval in the region (Dunne and first Roane County soil survey (SCS, 1936) reported
Leopold, 1978). Given that trees would intercept some that unsustainable farming practices caused slopes in
of the natural rainfall, the likelihood of surface runoff the area to be highly eroded by around 1900. Slopes
occurring during the full leaf season at sites that across the reservation contain deep gullies, some of
 C.P. Harden, P.D. Scruggs / Geomorphology 55 (2003) 5–24 21

which are still active. Gullies are visible on an air throughfall to be 59% and stemflow to be 2.3% of
photo taken in the late 1930s (Lafon, 1995); but the annual rainfall, and showed the forest to have a high
high infiltration rates determined in our study indicate frequency, low intensity rainfall regime (Scatena,
that surface runoff is rare in the current landscape, and 1990). The observed effectiveness of near-surface
new gully formation would not be expected under drainage during a heavy rain suggests that apparent
present conditions. steady-state infiltration rates of well-drained slopes
The apparent erodibility, evident in gullying, of are controlled by downslope subsurface drainage
deeply weathered ridge sites in the southern Appala- rather than soil profile characteristics alone. Thus,
chian Mountains poses the interesting geomorphic lateral subsurface flow, which is not considered to
problem of why such soft ridges have continued to be a component of infiltration for horizontal sites
exist in this humid environment. We suggest that a (Hillel, 1971), is integral to infiltration processes on
continuous presence of forest cover has protected mountainsides, and strict definitions of horizontal
these ridges from erosion over millennia. Settlers infiltration (by matric suction) and vertical infiltration
arriving in the area in the mid-1800s found an (suction and gravity) do not adequately describe
extensively forested region, which had been inhabited infiltration phenomena on mountain slopes.
since about 11,000 years BP by small nomadic bands Lower infiltration rates in cove sites appear to be
and later, by about 4500 years BP, by small, scattered related to the relatively recent colluvial origin of their
Native American communities practicing subsistence soil materials, probably from thin slope failures, and
agriculture (SAMAB, 1996). Pollen records indicate the reduced presence of roots and visible organic
that this region was continuously forested, at times by matter. As low points in the surrounding landscape,
more boreal species, even throughout the last glacial the coves are sites of topographic convergence of
period (Delcourt and Delcourt, 1987). Thus, forest subsurface flow lines. We suggest that the fine collu-
clearance in the mid-1800s may have been the first vial deposits in the coves serve as reservoirs of
major forest-removing disturbance in >10 ky. Slopes moisture whose capacities are exceeded in wet times
in the Oak Ridge Reservation returned to forest and from which moisture is supplied to streams via
when fields and pastures were abandoned, many in subsurface drainage in drier times. Further study of
the first decades of the 1900s (SCS, 1936) and the these cove-floor colluvial deposits would extend
rest in 1942 when the reservation was created. The understanding of the hydrology of this forest and
fact that these erodible ridges have withstood mil- other mountain headwater regions as sources and
lennia of rains in what is now a humid temperate regulators of fresh water.
environment highlights the important role of forests
in preserving an otherwise more transient topo- 5.4. Broader-scale patterns and implications
graphic form and leads us to a greater appreciation
of the geomorphic role of forest ecosystems in Beyond the noise of fine-scale spatial variability,
slowing denudation on deep, well-drained soils in we had hypothesized broader-scale controls of infil-
humid mountain environments. tration rates at the three study areas. Quantitative
analyses were limited by the difficulty of quantifying
5.3. Luquillo Experimental Forest infiltration rates that exceeded our rainfall application
rates and by the resulting small number of trials that
Our experiments and observations in the Luquillo achieved steady or quasi-steady state rates. Only two
Experimental Forest reinforced patterns observed in broader trends were significant in our limited data. In
the Andes and southern Appalachians, confirming the Ecuador, infiltration rates differed significantly be-
near absence of surface runoff on forested slopes and tween alluvial and upland soils; and in Puerto Rico,
the great spatial heterogeneity of infiltration rates. The within the tabonuco forest of the Bisley 1 watershed,
likelihood of surface runoff is even lower than indi- apparent infiltration rates were lower in topographic
cated by our infiltration results, which do not account coves than on slopes. The strong signal from these
for rainfall interception by the forest. Previous data and from our field observations is of microsite
research in the Luquillo Experimental Forest revealed control of infiltration rates, presumably dominated by
22 C.P. Harden, P.D. Scruggs / Geomorphology 55 (2003) 5–24

root and faunal macropores that exist in association more evident. Nonetheless, the end results of high and
with the forest ecosystem. spatially heterogeneous infiltration rates were similar.
Although soil scientists have characterized infil- Except at sites of subsurface flow path convergence
tration rates of US soils for decades, such tests have and at which soils are very shallow or overlie an
not occurred in forests or on steep slopes because the impermeable formation, surface runoff would be rare
commonly used double-ring infiltrometers are neither in a temperate or tropical mountainside forest. High
very portable nor suited for use on sloping surfaces local variability of infiltration rates on mountain
or in soils containing roots. Thus, most knowledge slopes suggests that rainfall that reaches the ground
about infiltration rates is based on flat land studies but does not infiltrate into the soil at first would most
outside of forests. The portable sprinkling infiltrom- likely be absorbed by higher infiltration capacity soil
eters and single infiltration ring enabled us to extend within a short distance as it flows downhill.
our knowledge of infiltration to forested mountain If litter, faunal activity, and the distribution of
slopes. On flat surfaces, the gravitational component soil macropores are dominant controls of rainfall
of water movement in soil is limited to an area infiltration and subsurface flow in these mountain
directly beneath the infiltration ring, even though forests, land use conversions would be likely to
water in tension also moves laterally. On mountain- reduce subsurface flow and decrease infiltration
sides, however, the gravitational component of sub- capacity. This was demonstrated by Spaans et al.
surface flow continued to drain water from the site (1990), who found saturated hydraulic conductivity
throughout the experiments, as it would throughout of soil to drop from 1000 to 50 cm/day following
natural storms. In a flat field, soil pores fill with conversion from a tropical rainforest to a 3-year-old
water and the steady-state infiltration capacity pasture in Costa Rica. The potential for a similarly
becomes a function of the positive pressure of the dramatic decrease in infiltration rates in our study
hydraulic head and the hydraulic conductivity of the areas was demonstrated by three rainfall simulation
soil medium. On a long mountainside, the gravity- trials we conducted in a pasture outside the Luquillo
driven flow of water in the near subsurface can National Forest in Puerto Rico. The pasture trials
continue to freely drain water from the soil below yielded infiltration rates of 19.5 – 44 mm/h, which
the infiltration ring; and saturation may never be are lower than those (92 mm/h median) in the
achieved, except in topographic settings where flow montane forest.
lines converge or at microsites where flow encoun- Classic denudation studies described the relation-
ters a barrier. Downslope drainage was clearly dem- ship between sediment yield and mean annual pre-
onstrated to us in the field where, in one case in the cipitation (or runoff), noted the importance of
Luquillo Forest, extracting a surface soil core pro- vegetation in protecting the soil from erosion in
duced an audible suction 1 m downslope from the humid regions (Langbein and Schumm, 1958; Doug-
auger. las, 1967), and recognized the key role of soil in
decreasing the conversion of rainfall to runoff, espe-
5.5. Infiltration rates and vegetation cially in arid environments (Yair and Anzel, 1987).
Our studies broaden the geographic scope of local-
Evidence from these studies suggests that forest scale data on infiltration rates and extend understand-
vegetation is critically important to hillslope hydrol- ing of the geomorphic significance of infiltration and
ogy in humid mountain regions. The forest environ- subsurface flow on mountain slopes. Well-drained,
ment creates root macropores, including those of the porous forest soils protect mountain slopes from
fine root zone near the soil surface and those created denudation by surface water erosion, while forest
by faunal activity. Low bulk densities of surface forest disturbances, including fire, blow down, impairment
soils facilitate downslope drainage. Compared to the by pests or contaminants, climate fluctuations, and
deep litter layer and distinct A horizon in the temper- clearing by humans, contribute to denudation and, if
ate mixed deciduous forest of the southern Appala- widespread, can initiate a cycle of positive feedback
chians, leaf litter in the tropical forests was very thin, leading to lower infiltration and higher denudation
but faunal (ant, termite, earthworm) activity was much rates.
 C.P. Harden, P.D. Scruggs / Geomorphology 55 (2003) 5–24 23

Acknowledgements dor. 1:1,000,000. Ministerio de Recursos Naturales y Energéti-

cos, Quito.
Grah, O., Hawkins, R., Cundy, T., 1983. Distribution of infiltration
The research was funded by the US Forest Service, on a small watershed. Advances in Irrigation and Drainage:
the National Geographic Research and Exploration Surviving External Pressures. ASCE Irrigation and Drainage
Fund, and a US EPA grant to Michael Huston at the Specialty Conference, Jackson, WY.
University of Tennessee. We thank Fred Scatena, US Hatcher Jr., R.D., Lemiszki, P., Dreier, R., Ketelle, R., Lee, R.,
Leitzke, D., McMaster, W., Foreman, J., Lee, S. 1992. Status
Forest Service, for logistical support in the Luquillo
Report on the Geology of the Oak Ridge Reservation. Environ-
Experimental Forest; and the Jatun Sacha Biological mental Sciences Division Publication, vol. 3860. Oak Ridge
Reserve and the Oak Ridge Reservation for permis- National Laboratory, US Department of Energy, Oak Ridge,
sion to work at those sites. We appreciate the TN. 247 pp.
assistance of Roger Clapp, who reviewed an earlier Hawkins, R., Cundy, T.W., 1987. Steady-state analysis of infiltra-
version of the manuscript, the UT Cartographic tion and overland flow for spatially-varied hillslopes. Water
Resources Bulletin 23 (2), 251 – 256.
Services Laboratory for drafting maps, and the Hillel, D., 1971. Soil and Water: Physical Principles and Processes.
following people for their help in the field and lab: Academic Press, New York, pp. 131 – 153.
Christopher Buhi and Tom Wallin in Ecuador; Judy Jatun Sacha Foundation, 1992. Jatun Sacha Biological Station.
Grable, Katie Hall, Evan Hart, Charles Lafon, and Ecuador: Information for Visitors. Jatun Sacha Foundation, Qui-
Kim Raia in Puerto Rico; and Judy Grable, Charles to, Ecuador.
Jetten, V.G., Riezebos, H., Hoefsloot, F., van Rossum, J., 1993.
Lafon, Daryl Neergaard, Ken Orvis, Kim Raia, and Spatial variability of infiltration and related properties of trop-
Dan Royall on the Oak Ridge Reservation. ical soils. Earth Surface Processes and Landforms 18, 477 – 488.
Knapp, B.J., 1978. Infiltration and storage of soil water. In:
Kirkby, M.J. (Ed.), Hillslope Hydrology. Wiley, Chichester,
UK, pp. 43 – 72.
References Lafon, C., 1995. Fifty years of succession on abandoned pastures at
the Oak Ridge Reservation, Tennessee: influences of pre-aban-
Betson, R., 1964. What is watershed runoff? Journal of Geophysical donment land use on vegetation development. MS Thesis, De-
Research 69 (8), 1541 – 1552. partment of Geography, University of Tennessee, Knoxville.
Bonell, M., Balek, J., 1993. Recent scientific developments and Langbein, W.B., Schumm, S., 1958. Yield of sediment in relation to
research needs in hydrological processes of the humid tropics. mean annual precipitation. Transactions, American Geophysical
In: Bonell, M., Balek, J. (Eds.), Hydrology and Water Manage- Union 39 (6), 1076 – 1984.
ment in the Humid Tropics. Cambridge Univ. Press, Cambridge, Larsen, M., Torres-Sánchez, A., Concepción, I., 1999. Slopewash,
UK, pp. 168 – 260. surface runoff and fine-litter transport in forest and landslide
Brown, S., Lugo, A., Silander, S. and Liegel, L., 1983. Research scars in humid – tropical steeplands, Luquillo Experimental For-
History and Opportunities in the Luquillo Experimental Forest. est, Puerto Rico. Earth Surface Processes and Landforms 24,
USDA Forest Service, Southern Forest Experiment Station, 481 – 502.
General Technical Report SO-44, New Orleans, LA. Lee, S., Lietzke, D., Ketelle, R., Ammons, J., 1988. Soil and
Buhi, C., 1997. Soil hydraulic properties in two tropical lowland Surficial Geology Guidebook to the Oak Ridge Reservation,
forests: Jatun Sacha, Ecuador and La Selva, Costa Rica. MS Oak Ridge, TN. Environmental Sciences Division Publication,
Thesis, Department of Geography, University of Tennessee, vol. 3112. Oak Ridge National Laboratory, US Department of
Knoxville. Energy, Oak Ridge, TN. 89 pp.
Coweeta, 1984. Coweeta Hydrologic Laboratory: A Guide to the Loague, K., Gander, G., 1990. R-5 revisited: 1. Spatial variability of
Research Program. US Forest Service, Southeastern Forest Ex- infiltration on a small rangeland catchment. Water Resources
periment Station, Otto, NC. 11 pp. Research 26 (5), 957 – 971.
Delcourt, P., Delcourt, H., 1987. Long-term Forest Dynamics of the McQueen, I., 1963. Development of a hand-portable rainfall simu-
Temperate Zone: A Case Study of Late Quaternary Forests in lator infiltrometer. US Geological Survey Circular, vol. 482. US
Eastern North America. Springer, New York. Geological Survey, Washington, DC.
Douglas, I., 1967. Man, vegetation and the sediment yields of riv- Mulholland, P., Wilson, G., Jardine, P., 1990. Hydrogeochemical
ers. Nature 215, 925 – 928. response of a forested watershed to storms: effects of preferen-
Dunne, T., Leopold, L., 1978. Water in Environmental Planning. tial flow along shallow and deep pathways. Water Resources
Freeman, San Francisco. Research 26 (12), 3021 – 3036.
Dunne, T., Zhang, W., Aubry, B., 1991. Effects of rainfall, vegeta- Philip, J.R., 1957. The theory of infiltration: 5. Sorptivity and al-
tion, and microtopography on infiltration and runoff. Water Re- gebraic infiltration equations. Soil Science 84, 257 – 267.
sources Research 27 (9), 2271 – 2285. Price, M., 1999. Global Change in the Mountains. Parthenon Pub-
Ecuador, 1982. National Geological Map of the Republic of Ecua- lishing Group, New York.
24 C.P. Harden, P.D. Scruggs / Geomorphology 55 (2003) 5–24

Robertson, G.P., Klingensmith, K., Klug, M., Paul, E., Crum, J., Wadsworth, F.H., 1951. Forest management in the Luquillo Moun-
Ellis, B., 1997. Soil resources, microbial activity, and primary tains: I. The setting. Caribbean Forester 12, 93 – 114.
production across an agricultural ecosystem. Ecological Appli- Wallin, T., 1995. Quantifying trail-related soil erosion at two sites in
cations 7 (1), 158 – 170. the humid tropics: Jatun Sacha, Ecuador, and La Selva, Costa
Scatena, F., 1989. An Introduction to the Physiography and History Rica. MS Thesis, Department of Geography, University of Ten-
of the Bisley Experimental Watersheds in the Luquillo Moun- nessee, Knoxville.
tains of Puerto Rico. Forest Service General Technical Report Wallin, T., Harden, C., 1996. Quantifying trail-related soil erosion
SO-72, US Department of Agriculture Forest Service, Southern at two sites in the humid tropics: Jatun Sacha, Ecuador, and La
Forest Experiment Station, New Orleans, LA. Selva, Costa Rica. Ambio 25 (7), 517 – 522.
Scatena, F., 1990. Watershed scale rainfall interception on two for- Whipkey, R.Z., Kirkby, M.J., 1978. Flow within soil. In:
ested watersheds in the Luquillo Mountains of Puerto Rico. Kirkby, M.J. (Ed.), Hillslope Hydrology. Wiley, Chichester,
Journal of Hydrology 113, 89 – 102. UK, pp. 121 – 144.
Soil Conservation Service (SCS), 1936. Soil Survey, Roane County, Wilson, G., Jardine, P., Luxmoore, R., Jones, J., 1990. Hydrology
Tennessee. US Department of Agriculture Soil Conservation of a forested watershed during storm events. Geoderma 46,
Service, Washington, DC. 119 – 138.
Southern Appalachian Man and the Biosphere (SAMAB), 1996. Woods, E., Sivapalan, M., Robinson, J., 1997. Modeling the spatial
The Social, Economic, and Cultural Technical Report, Report variability of subsurface runoff using a topographic index. Water
4, Southern Appalachian Assessment. US Department of Agri- Resources Research 33 (5), 1061 – 1073.
culture, Forest Service, Atlanta, GA, p. 220. http://sunsite. Yair, A., Anzel, Y., 1987. The relationship between annual rainfall
utk.edu/samab/saa/saa_reports.html. and sediment yield in arid and semiarid areas. The case of the
Spaans, E., Bouma, J., Lansu, A., Wielemaker, W., 1990. Measur- northern Negev. In: Ahnert, F. (Ed.), Geomorphological Models:
ing soil hydraulic properties after clearing of tropical rain forest Theoretical and Empirical Aspects. Catena Supplement, vol. 10.
in a Costa Rican soil. Tropical Agriculture 67 (1), 61 – 65. Catena Verlag, Cremlingen, Germany, pp. 121 – 135.
 Geomorphology 55 (2003) 25 – 43
www.elsevier.com/locate/geomorph

Drainage, sediment transport, and denudation rates on the

Nanga Parbat Himalaya, Pakistan
Kevin Cornwell a,*, Doug Norsby b, Richard Marston c
a
Department of Geology, California State University-Sacramento, Sacramento, CA 95819-6043, USA
b
Department of Public Works, Watershed Division, Topeka, KS 66215, USA
c
School of Geology, Oklahoma State University, Stillwater, OK 74078, USA
Received 10 February 2002; received in revised form 30 August 2002; accepted 10 March 2003

Abstract

The Nanga Parbat Himalaya presents some of the greatest relief on Earth, yet sediment production and denudation rates have
only been sporadically addressed. We utilized field measurements and computer models to estimate bank full discharge,
sediment transport, and denudation rates for the Raikot and Buldar drainage basins (north slope of Nanga Parbat) and the upper
reach of the Rupal drainage basin (south slope).
The overall tasks of determining stream flow conditions in such a dynamic geomorphic setting is challenging. No gage data
exist for these drainage basins, and the overall character of the drainage basins (high relief, steep flow gradients, and turbulent
flow conditions) does not lend itself to either ready access or complete profiling.
Cross-sectional profiles were surveyed through selected reaches of these drainage basins. These data were then incorporated
into software (WinXSPRO) that aids in the characterization (stage, discharge, velocity, and shear stress) of high altitude, steep
mountain stream conditions.
Complete field measurements of channel depths were rarely possible (except at several bridges where the middle of the
channel could actually be straddled and probed) and, when coupled with velocity measurements, provided discrete points of
field-measured discharge calculations. These points were then used to calibrate WinXSPRO results for the same reach and
provided a confidence level for computer-generated results.
Flow calculations suggest that under near bank full conditions, the upper Raikot drainage basin produces discharges of
f 61 cm and moves about 11,000 tons day
1 (9980 tons day
1) of sediment through its channel. Bank full conditions on
the upper portion of the Rupal drainage basin generate discharges of f 84 cm and moves only about 3800 tons day
1 (3450
tons day
1) of sediment. Although the upper Rupal drainage basin moves more water, the lower slope of the drainage basin
(0.03) generates a much smaller shear stress (461 Pa) than does the higher slope (0.12) of the upper Raikot drainage basin
(1925 Pa).
Dissolved and suspended sediment loads were measured from water/sediment samples collected throughout the day and
night over a period of 10 days at the height of the summer melt season but proved to be a minor variable in transport flux.
Channel bed loads were measured using a pebble count method of bank material and then used to generate ratings curves of bed
loads relative to discharge volumes. When coupled with discharge data and basin area, mean annual sediment yield and
denudation rates for Nanga Parbat are produced. Denudation rates calculated in this fashion range from 0.2 mm year
1 in the

* Corresponding author. Tel.: +1-916-278-6667; fax: +1-916-278-4650.

E-mail address: [email protected] (K. Cornwell).

0169-555X/03/$ - see front matter D 2003 Elsevier B.V. All rights reserved.
doi:10.1016/S0169-555X(03)00130-2
26 K. Cornwell et al. / Geomorphology 55 (2003) 25–43

slower, more sluggish Rupal drainage basin to almost 6 mm year
1 in the steeper, faster flowing Raikot and Buldar drainage
basins.
D 2003 Elsevier B.V. All rights reserved.

Keywords: Sediment transport; Pakistan; Mountain denudation

1. Introduction basin area. Burbank et al. (1996), using exposure ages

of incised bedrock surfaces along the Indus River,
The Nanga Parbat massif is one of the largest derived incision rates of 2– 12 mm year
1 across the
topographic features in the western Himalaya of Nanga Parbat uplift. Shroder et al. (1999) measured
northern Pakistan with its main summit rising to the volume of small debris fans on the south slopes of
8126 m and several lesser peaks (Raikot, Chongra, Nanga Parbat and constrained the timing of their
Ganalo, and Mazeno) rising to 7000 m. The massif deposition using cosmogenic dating of glacially
supports 69 separate glacier systems and produces exposed rocks and sediment to produce denudation
vertical drops of 7000 m over a 21-km horizontal rates of about 2 mm year
1. Shroder and Bishop
distance, representing the steepest relief above sea (2000), measuring rates of local incision around the
level in the world (Shroder, 1989). The 1250-km2 Nanga Parbat massif (15 different localities), deter-
massif is drained by nine major rivers that drain the mined maximum incision rates for the massif of
north slope (Patro, Raikot, and Buldar), the east slope 22 F 11 mm year
1.
(Astore, Goricot, Lotang, and Sachen), the south slope The significance of denudation rates and processes
(Rupal), and the west slope (Diamir) (Fig. 1). These in high elevation environments is becoming more
rivers (with the exception of the Astore, which gen- apparent as science better understands the nature of
erates some of its discharge from terrains beyond the tectonic uplift, long-term isostatic adjustment, and
massif) are all fed by high elevation glacial melt erosion rates that increase mechanical and chemical
waters that produce heavily sediment-laden dis- weathering processes and likely influence the climatic
charges, and yet sediment production and denudation trends that have fluctuated so dramatically during late
rates have only been sporadically addressed. Cenozoic time (Raymo et al., 1988; Raymo and
Geologic mapping, petrographic analysis, and Ruddiman, 1992). Substantial chemical weathering
cooling-age pattern studies (Zeitler et al., 1993; (a potentially large sink for atmospheric CO2) gen-
Schneider et al., 1999a,b; Poage et al., 2000) suggest erated from uplift and erosion of vast mountainous
that the massif has experienced rapid exhumation in areas such as the Himalayas and Tibet may have
the recent geologic past. On the basis of geochrono- contributed pronouncedly to the onset of colder cli-
logic evidence, Zeitler et al. (1989) and Winslow et al. matic conditions.
(1994) suggested that the entire massif had been This paper outlines our efforts to generate conser-
exhumed 10 km in the past 10 Ma with 6 km of that vative estimates of modern-day discharge conditions,
exhumation occurring during the last 1.3 Ma. This measures fluvially stored sediment, calculates sedi-
evidence coupled with the relatively young, high- ment transport fluxes, and determines denudation
grade metamorphic and igneous rocks exposed at high rates from three basins that drain Nanga Parbat. The
elevation (Zeitler et al., 1993) reflects a most rapid Raikot (299 km2) and Buldar (212 km2) river basins
rate of uplift and consequent denudation. (which drain a portion of the north side of the massif)
Denudation studies in such extreme topography are and the Rupal (315 km2) River basin (which drains a
few. Sediment transport and yield conditions were significant portion of the south side) were chosen for
evaluated for the Raikot glacier on the north slope of this analysis because of their accessibility and orien-
Nanga Parbat during the 1980s (Gardner and Jones, tation on Nanga Parbat. Approximately 33%, 36%,
1993), yielding denudation rates of the glacier area of and 42% of the Raikot, Buldar, and upper Rupal
4.6– 6.9 mm year
1 and 1.4 –2.1 mm year
1 for the watersheds are glaciated, respectively. Our findings,
 K. Cornwell et al. / Geomorphology 55 (2003) 25–43 27

Fig. 1. Location map of Nanga Parbat illustrating the distribution of glacial and drainage networks on the massif.
28 K. Cornwell et al. / Geomorphology 55 (2003) 25–43

when combined with comparable glacial and mass Long-term hydrograph data for the drainage basins
wasting estimates, would be expected to illuminate of the Nanga Parbat massif is only indirectly avail-
denudation conditions/rates occurring on the Nanga able. Fig. 2 illustrates the general features of hydro-
Parbat massif under modern-day climatic conditions graph data from the Astore River (drainage area of
and further our growing understanding of denudation 4040 km2) based on sediment transport analysis con-
rates in such high elevation environments. ducted by the Water and Power Development Author-
ity (WAPDA) in Pakistan during the 1970s and early
1980s at Doyian. Although the sediment yield does
2. Fluvial characteristics not equal discharge, it relates to discharge by proxy.
At this station, mean annual flow is 121.0 cm.
Discharges emanating from the high elevation The shapes that emerge from this data are charac-
basins included in this study are the result of seasonal teristics of high elevation drainage basins that are
monsoon precipitation events and glacial and snow- heavily ice and snow covered and experience consid-
melt. The annual monsoon storms that impact Nanga erable melting during the warmer summer months.
Parbat from the south from July through September Hydrograph data from the Indus (162,000 km2),
are the most dominant source of precipitation to the Kabul (67,340 km2), and Hunza (13,157 km2) Rivers
massif. A secondary source of precipitation comes (Young and Hewitt, 1990) in Fig. 3 show discharge
from westerly depressions that are the major source of patterns comparable in shape to the Astore River data.
precipitation during non-monsoonal times (generally Orographic conditions temper monsoonal influences
the winter months). Young and Hewitt (1990) sug- farther north into the Himalaya ranges (characteristic
gested that the monsoonal influence produces heavy of the Hunza and Kabul drainage systems) suggesting
precipitation in basins of Nanga Parbat primarily that glacial and snowmelt during the warmer summer
because of the massif’s geographic location in the season account for substantial volumes of discharge in
southern part of the Himalaya Range and its oro- these river systems.
graphic influence on storms tracking to the north.
Coupled with the westerly depressions that also pro- 2.1. Relief ratios
duce potentially heavy precipitation in the Himalaya
range, the Nanga Parbat massif receives an average of The primary difference observed in the drainage
over 1000 mm of precipitation annually (Ali, 1995). basin relief profiles (Fig. 4) results from incision by

Fig. 2. Sediment yield of the Astore River at the Doyian gage site as measured by the Western Area Power and Development Authority
(WAPDA). This data indirectly represents discharge conditions (by proxy) since increased discharges during the summer months correspond to
increased sediment yields.
 K. Cornwell et al. / Geomorphology 55 (2003) 25–43 29

Fig. 3. Hydrograph data from the Indus River (as it drains the Western Himalaya – Besham station) and two of its tributaries, the Kabul River
(draining the Hindu Kush Range – Warsak Station) and the Hunza River (draining portions of the Karakoram Range – Dianyor Station) (after
Young and Hewitt, 1990).

the Indus River on the north side of the massif. Rapid elevation changes over a much longer distance (>80
downcutting of the Indus River over time (Burbank km).
and Beck, 1991; Shroder and Bishop, 2000) has The distance from the glacier terminus and the
generated a relatively steep slope for north-draining confluence of the Raikot and Indus Rivers is f 13.5
basins (Raikot and Buldar) over a short distance ( < 15 km. Over that distance, the gradient of the channel
km), while the Rupal basin experiences comparable averages 0.14. About 5.5 km up from the confluence,

Fig. 4. Drainage basin relief profiles for the Raikot, Buldar, and Rupal valleys.
30 K. Cornwell et al. / Geomorphology 55 (2003) 25–43

the slope changes notably and the gradient flattens ditions. Velocity estimates were calculated by meas-
slightly from 0.16 to 0.11. This condition continues uring the time of travel of water-filled balloons over a
for about 4.5 km farther upstream until the gradient specific distance at various times throughout the day.
again becomes steeper, reaching a maximum of 0.3 The 9-in (23-cm) diameter, brightly colored, biode-
over the remaining 3.5 km to the glacier terminus. gradable balloons were filled with water (two-thirds
Buldar channel gradient (0.17) is slightly steeper capacity) and air (one-third capacity) then released
than the Raikot (0.14) and substantially steeper than into the middle of the river’s flow. When balloons had
the Rupal (0.04). The Buldar channel extends f 10.7 reached full velocity and were sampling as much of
km from the glacier terminus to the Indus confluence the river flow as possible, timing began.
and experiences an elevation change of about 1850 m. We realize that this technique does not adequately
The Rupal River, on the other hand, travels almost represent the variability of flow velocity (vertically,
83 km to its confluence with the Indus and produces a laterally, or temporally) in such dynamic settings.
much lower gradient of about 0.03. Along the channel Instead, we suggest that this approach generates a
stretches included in this study (upper 10 km), the rough mean estimate of flow velocity in the primary
gradient is about 0.04. flow path of the river when measured at the reach-
Hypsometric analysis of the study basins illustrates scale. The sometimes-violent nature of flow condi-
the nature of these high elevation, steep gradient tions and access to the active flow channel prohibit
valleys (Fig. 5). The Buldar basin has more than half utilization of other more accepted means of flow
of its mass above its median elevation and represents velocity measurements, such as current meters and
the steepest gradient within the study basins. The tracers.
Raikot basin is slightly less than Buldar conditions, We acknowledge some limitations to this techni-
while the Rupal basin has a substantial portion of its que, as the balloons generally ride in the upper
mass below its median elevation. portion of the flow column (although water-filled
balloons are mostly submerged below the surface
2.2. Velocity measurements and sample a more representative flow than objects
floating on the surface). Sometimes the balloons get
We collected velocity measurements at selected caught in eddies and small discontinuous whirlpools
localities to generate velocity profiles for comparison that occasionally eject the balloon out of the water
and calibration with computer-generated velocity con- and through the air for short distances. Considering

Fig. 5. Hysometric profiles of the Raikot, Buldar, and Rupal valleys through the study reaches.
 K. Cornwell et al. / Geomorphology 55 (2003) 25–43 31

the turbulent nature of the river system, however, tion to determine mean flow velocity and, conse-
balloon flow data seemed to produce consistent quently, to calculate discharge through each cross
results and errors associated with this approach, section.
which will likely be on the conservative (i.e., lower Field data indicate that n values are much greater on
velocity) side. high-gradient, cobble- and boulder-bed streams than
The fastest and the slowest times were discarded on low-gradient streams having similar relative rough-
out of each data set, and the remaining times were ness values (Jarrett, 1987). As gradient increases,
averaged to yield mean velocities. Velocity estimates energy losses increase as a result of wake turbulence
in the Raikot channel varied from a maximum speed and the formation of localized hydraulic jumps down-
of 2.5 m s
1 to a minimum of 2.15 m s
1 and had a stream from boulders. With few exceptions, the bank
mean value of 2.4 m s
1 F 0.14. Velocity estimates and channel materials were composed of loosely con-
from the Rupal channel varied from a minimum of solidated sediment, ranging in size from predomi-
1.67 m s
1 to a maximum of 2.26 m s
1, with a mean nantly boulders to silts and clays.
value of 2.13 m s
1 F 0.025.
The lower velocity conditions apparent in the 2.4. Raikot and Rupal rivers discharge measurements
Rupal channel likely are a result of lower gradients
(an order of magnitude less than the gradient in the Inhospitable weather conditions and native peoples
Raikot channel). Considering the level of turbulence in the Buldar valley prohibited direct measurement of
in the Raikot River, however, one would expect the discharge conditions in the channel for calibration with
velocity to be somewhat inhibited by the general WinXSPRO calculations. For these reasons, our dis-
roughness conditions of the channel. cussion of discharge conditions is focused on data
collected from the Raikot and Rupal drainage basins.
2.3. Discharge estimates In two locations along the upper portion of the
Raikot River (where conditions allowed), stream
The three basins included in this study (Raikot, depths were probed. The first location was at a small
Buldar, and Rupal) were all reasonably accessible bridge constructed over a relatively narrow (con-
during the 1996 and 1997 field seasons. Selected stricted) part of the river about 4 km below the glacier
channel sections in these basins were surveyed using terminus. The flow was too fast and turbulent for
a laser theodolite to generate cross-sectional data. Fig. probing the depths by hand with a stadia rod from the
6 outlines the relative location of the study reaches in bridge. Instead, guide ropes were attached to the
their respective drainage basins. Channel cross-sec- stadia rod and securely held upstream of the probe
tional data were entered into WinXSPRO software point (Fig. 7). This effort allowed the stadia rod to
(WEST Consultants, 1996) that manages cross-sec- remain relatively straight and not bend or get forced
tional data to generate hydraulic parameters and solve downstream by the river flow. Several probe depths
resistance equations to yield velocity, discharge, and were sounded across the channel at this location and a
shear stress estimates under various stage conditions. depth profile was constructed for the channel. The
The resistance equation selected for this analysis was cross-sectional profiles measured in this manner pro-
Jarrett’s equation (Jarrett, 1984): duced area estimates and, when combined with veloc-
ity flow measurements, stream discharge. Discharge
n ¼ 0:39 S 0:38 R
0:16 ð1Þ calculations at the bridge ranged from 29.0 cm at
maximum measured flow velocity to 7.6 cm at mini-
where n is Manning’s roughness coefficient, S is water mum measured flow velocity.
surface slope (head loss over a given length of Depth probes were also conducted near the glacier
channel), and R is hydraulic radius. terminus (about 100 m downstream) where the chan-
Jarrett’s equation yields a value of Manning’s n, nel narrows between some large boulders. Velocity
which is a function of the water surface slope and data were also obtained through this probe location to
hydraulic radius at a cross section. Resistance values evaluate stream discharge. Discharge calculations near
were then incorporated into solving Manning’s equa- the glacier terminus ranged from a maximum dis-
32 K. Cornwell et al. / Geomorphology 55 (2003) 25–43

Fig. 6. (A) Raikot and Buldar drainage basins and cross sections. (B) Rupal drainage basin and cross sections. See Fig. 1 for basin orientations
on Nanga Parbat.

charge of 20.5 (at peak measured flow velocity) to 5.4 discharge location east of the Bazhin Glacier ranged
cm (at lowest measured flow velocity). The terminus from a maximum discharge of about 48 cm (at peak
routed most of the flow through the measured section, velocity) to about 40 cm (at lowest measured veloc-
but 1 –2 cm actually flowed around the large boulders ity). Flow conditions in the Rupal River were much
that confined most of the flow. less turbulent than that of the Raikot, likely a function
Two locations along the Rupal River were also of the much higher channel gradients present in the
probed in an effort to determine channel dimensions Raikot basin.
for discharge estimates. The first locality (just east of These discharge measurements were compared to
the Bazhin Glacier) occurred where a small footbridge equivalent channel areas and discharges calculated
crossed the river and offered a straddle point across using the WinXSPRO software and are summarized
the 11.3-m-wide channel. The second locality (just in Table 1. The footnoted rows show the equivalent
west of the Tarshing Glacier) again was chosen WinXSPRO calculations using Jarrett’s equation to
because of a footbridge that spanned the 12.6-m-wide ultimately solve for velocity and discharge.
channel. Velocity data were obtained through these Tables 2 and 3 summarize WinXSPRO calculations
probe locations to evaluate stream discharge. The for cross sections in the Raikot and Rupal sections
 K. Cornwell et al. / Geomorphology 55 (2003) 25–43 33

Fig. 6 (continued ).

closest to the field calibration locales. In general, field variable flow and sediment conditions. Thus, it
discharge measurements correlated closely to WinX- becomes critical that the conditions under which a
SPRO calculations. Subsequent hydraulic estimates particular equation was developed are matched in the
and sediment yield calculations are based on dis- field study area (Nielson, 1974). Unfortunately, the
charges generated through WinXSPRO software. rivers of the Himalaya in general tend to lie outside
the descriptive characteristics under which sediment
transport equations tend to be designed.
3. Sediment transport characteristics Gravel transport equations are designed using two
primary criteria: discharge and tractive force. Dis-
Sediment transport equations are developed under charge focuses on the quantity of water, cross-sec-
experimental conditions to determine sensitivities to tional area, and slope of a river system, but some
34 K. Cornwell et al. / Geomorphology 55 (2003) 25–43

Fig. 7. Probing bedrock conditions and channel cross sections at the Raikot Bridge crossing. Guide ropes were used to stabilize the stadia rod as
dynamic flow conditions prohibited direct push approaches.

researchers (McBean and Al-Nassri, 1988) express flow type and bed material properties. Attempts to
concern over the confidence of these equations and measure or calculate bed shear stress values in moun-
the practice of determining sediment transport by tain rivers are complicated by the channel bed rough-
discharge in general. ness and the associated turbulence and velocity
Tractive force equations attempt to quantify the fluctuations—two variables that are most difficult to
shear stress of flowing water on the floor of the quantify.
channel relative to the channel slope, thereby exam- For this study, the Meyer-Peter and Muller (1948)
ining the river – riverbed interface that varies with equation (Nielson, 1974)was chosen for application

Table 1
Channel conditions and discharge calculations for Raikot and Rupal sections
Location Date Time X-sec Velocity Discharge Discharge
area (m2) (m s
1) (ft3 s
1) (m3 s
1)
Raikot Bridge (#7) 7/28/97 9:45 am 10.13 2.40 858 24.31
Raikot Bridge (#7)a 7/28/97 9:45 am 10.13 2.00 715 20.26
Glacier Portal 8/2/97 11:00 am 7.18 2.50 633 17.95
Rupal Bridge (west) (#11) 8/12/97 9:30 am 22.60 2.13 1700 48.14
Rupal Bridge (west) (#11)a 8/12/97 9:30 am 22.60 2.32 1700 52.43
Rupal Bridge (east) 8/14/97 9:25 am 23.68 2.47 2065 58.5
Tarshing Bridge 8/9/97 10:15 am 26.25 2.47 2290 64.84
a
Discharge calculations using WinXSPRO velocity calculations.
 K. Cornwell et al. / Geomorphology 55 (2003) 25–43 35

Table 2
Channel conditions and hydraulic calculations for the Raikot survey section nearest the calibration point (cross section #7)
Stage (m) Area (m2) R (m) Slope n VAVG (m s
1) Q (m3 s
1) Shear (Pa)
0.30 0.75 0.2 0.141 0.201 0.6 0.5 253.8
0.60 2.50 0.3 0.141 0.183 1.0 2.5 454.9
0.90 5.28 0.5 0.141 0.170 1.4 7.5 713.4
1.20 8.50 0.7 0.141 0.162 1.9 15.8 986.3
1.33a 10.10 0.8 0.141 0.160 2.0 20.3 1075.0
1.50 12.30 0.9 0.141 0.157 2.2 26.9 1197.0
1.80 16.61 1.1 0.141 0.152 2.6 43.0 1469.9
2.10 21.39 1.2 0.141 0.148 2.9 61.8 1675.8
2.50 28.36 1.5 0.141 0.144 3.4 97.0 2054.1
a
Comparable stream probing stage height and hydraulic parameters.

primarily because it best accommodated the discharge sediment sizes along the banks of the river channels
and sediment conditions present in the rivers of this and by following the techniques outlined by Wolman
study. Specifically, the Meyer-Peter and Muller equa- (1954) to estimate bed load conditions. Although the
tion uses a tractive force approach that applies to measured sediment (100 count) was not in the active
rivers with coarse particle sizes and high gradients. portion of the modern channel, their close proximity
The river systems in this study seem to closely match next to the channel reflected the fact that under higher
the experimental conditions used in the generation of flow conditions, these sediments were entrained and
the Meyer-Peter and Muller transport equation. Bed transported as bed load materials.
load determinations from this solution do not include
suspended sediment influences. 3.1. Suspended sediment loads
Solution of the Meyer-Peter and Muller equation
was accomplished using a program that was devel- Raikot River sediment yields were measured
oped for the U.S. Corps of Engineers by WEST throughout a 7-day time period at a location chosen
Consultants that characterizes high altitude, steep for channel accessibility and where flow conditions
mountain streams and incorporated flow, and channel were dynamic. Samples were collected at 3-h intervals
characteristics output from WinXSPRO. throughout the day (9:00 a.m., 12:00 a.m., 3:00 p.m.,
Input requirements include the D50 particle size of and 6:00 p.m.). In an attempt to further characterize
bed particles. Physical measurements of bed load diurnal variations, samples were also collected on an
sediments in these river channels were not possible hourly basis between 9:00 a.m. and 6:00 p.m. over a
due to the extremely high-energy conditions. Instead, 2-day time period and at 3-h intervals over a 33-h time
bed load characteristics were evaluated by measuring period.

Table 3
Channel conditions and hydraulic calculations for the Rupal survey section nearest the calibration point (cross section #11)
Stage (m) Area (m2) R (m) Slope n VAVG (m/s) Q (m3/s) Shear (Pa)
0.30 1.03 0.21 0.035 0.115 0.6 0.6 76.6
0.60 2.74 0.40 0.035 0.105 1.0 2.6 134.1
0.90 5.11 0.55 0.035 0.099 1.3 6.5 191.5
1.20 8.14 0.70 0.035 0.095 1.6 12.7 244.2
1.50 11.88 0.88 0.035 0.092 1.9 22.0 296.9
1.80 16.15 1.04 0.035 0.090 2.1 34.3 349.5
2.10 21.23 1.10 0.035 0.089 2.3 48.2 378.3
2.17a 22.60 1.14 0.035 0.088 2.3 52.4 391.6
2.50 29.10 1.34 0.035 0.086 2.6 76.8 454.9
a
Comparable stream probing stage height and hydraulic parameters.
36 K. Cornwell et al. / Geomorphology 55 (2003) 25–43

Suspended sediment concentrations were deter- topography (elevation 1760 – 3700 m), mean annual
mined using National Water Well Association Method temperature is 7 jC, and mean annual precipitation is
2540 D (Clesceri et al., 1989). Specifically, a sample 40 – 50 cm. The dissolved chemical constituency
50– 100 ml was collected from an active flow portion included K, Cl
, Mg+, Ca+, Na+, SO42
, SiO2, and
of the river channel. The sample volume was precisely HCO3
that yielded average dissolved concentrations
measured in a graduated cylinder and then slowly of 114.2 mg l
1 during winter and 71.8 mg l
1 in the
pulled by vacuum through a preweighed 45-Am filter. summer. As is apparent in Miller and Drever’s (1977)
The sediment-laden filter was then dried at f 100 – work, the dissolved load does not appear to vary much
105 jF (23.6 – 26.3 jC) and placed in a desiccator through seasons or flow conditions. Considering the
until cooled. The cooled samples were weighed to the water temperature conditions in the river systems
nearest 0.001 g and sediment concentrations calcu- draining Nanga Parbat (about 3.5 jC), we would not
lated in units of mg l
1. expect much variation in dissolved capacity as a
The results of these measurements are outlined in function of temperature because temperature condi-
Fig. 8. Sediment concentrations ranged from a high of tions do not exhibit much variation over time.
22,514 mg l
1 (7/28/97, 6:00 p.m.) to a low of 562 In our model of sediment yield, the dissolved
mg l
1 (7/25/97, 6:00 p.m.) with a mean concentra- sediment load plays a very minor role in sediment
tion (throwing out the highest and lowest values) of contribution. In fact, varying these concentrations by
3970 mg l
1 over this time period. Suspended sedi- an order of magnitude in either direction does not
ment loads calculated from WAPDA data from the change the sediment flux noticeably.
Doyian gage site (over a nine year span) yield an
annual maximum average suspended concentration of 3.3. Bed loads
3480 mg l
1.
The maximum concentrations of sediment in river Physical measurement of bed load sediment in the
flow appeared to coincide (as expected) with the Raikot River channel was not possible because of the
afternoon maximum temperature value. As can be extremely high-energy conditions prevalent in the
seen in Fig. 8, warm afternoons on 7/26, 7/27, and 7/ channel. Consequently, bed load concentrations were
28 resulted in high suspended sediment values, while evaluated by measuring sediment sizes along the
cloud-covered days/afternoons (7/25, 7/29, 7/30, and banks of the Raikot River channel and by conducting
7/31) produced much smaller suspended sediment a pebble count analysis (Wolman, 1954). Although
loads. the measured sediments were not in the active portion
We also noticed a direct relationship between the of the modern channel, their close proximity next to
ambient temperatures measured throughout the day the channel indicated that under higher flow condi-
and the suspended sediment load. Fig. 8 shows how tions, these sediments were entrained and transported
suspended load concentrations vary as a function of as bed load materials.
temperature (which affects ice and snow melting and, Pebble count data was incorporated into WinX-
consequently, discharge volumes). SPRO and used to solve the Meyer-Peter and Muller
sediment transport equation. This produced bed load
3.2. Dissolved sediment load rating curves (sediment in tons day
1) as a function
of discharge (Fig. 9). The rating curve generated from
Dissolved solid loads were measured in waters of Raikot data at cross section #7 produced a total
the Raikot and Buldar Rivers over the course of 7 sediment load of 11,000 tons day
1 (9980 tons
days, yielding concentrations that ranged from 40 to day
1) under bank full flow conditions (61 cm). At
60 mg l
1 in the Buldar basin and 50 to 105 mg l
1 Rupal cross section #7, rating curve data suggest a
in the Raikot basin. In comparable work in the total sediment load of 3800 tons day
1 (3450 tons
Absaroka Range in Wyoming, Miller and Drever day
1) under bank full discharge conditions of 84 cm.
(1977) determined seasonal variations of the dissolved Buldar data (cross section #2) produced a total sedi-
sediment load. The Absaroka Range is characterized ment load of 9000 tons day
1 (8160 tons day
1)
by rocks that are andesitic in composition, rugged under bank full discharge conditions of 44 cm.
 K. Cornwell et al. / Geomorphology 55 (2003) 25–43 37

Fig. 8. Suspended sediment loads and temperature data from the Raikot River over the course of 7 days (7/25 – 8/1/97).

It should be noted that the bed load estimates are in the Rupal River and 1925 Pa at bank full conditions
determined under peak annual discharge conditions, in the Raikot River.
conditions responsible for f 88% of the annual load
but only occurring over about one-third of the year. 3.4. Sediment storage and residence time
Shear stress components were also calculated
(Parker et al., 1982) in WinXSPRO runs as a function In the glaciated valleys, sediment stored in flood
of the D50 value of the bed load. Shear stresses of plains filled the parabolic groove left by the passing of
about 461 Pa were calculated for bank full discharges earlier glacial events (Harbor and Wheeler, 1992). The

Fig. 9. Bed load rating curves for the three study basins calculated from pebble count analysis in WinXSPRO software.
38 K. Cornwell et al. / Geomorphology 55 (2003) 25–43

but they happen so sporadically that most of the work

is probably done by the low magnitude, high frequency
flow conditions that typify the fluvial regime.
We quantified the total valley fill within the Raikot
and Buldar basins by generating a parabolic curve
from surveyed elevation profiles of the valley (includ-
ing high morainal deposits) perpendicular to river
flow (Fig. 10). This iterative curve methodology,
although not as precise as using bedrock position, is
a fair method of quantification, especially in the steep
regions of the Himalaya (Harbor and Wheeler, 1992)
where valley bedrock data is nonexistent.
Fig. 10. Parabolic curve generated through valley cross sections to Sediment that is not actively being moved down
generate sediment storage estimates.
the river is considered to be either temporary (in the
active flood plain) or semipermanent (fill that is
stored sediment comes from a variety of different deeper than the greatest depth of the river within the
sources, including mass wasting, fluvial, glacial, and cross section) stored sediment.
eolian actions. Once on the valley floor, however, Temporary storage volumes were calculated by
fluvial action becomes the dominant transport process constructing riverbank and flood-plain cross sections,
under the modern climatic regime. multiplying the width of the flood plain by the depth
The movement of stored sediment in the valleys is of the river at its deepest point and subtracting the
dependent on the local hydrologic conditions that vary bank full cross-sectional area (Fig. 11).
from month to month and year to year. Rare, high The sediment below the deepest portion of the
discharge events move substantially more sediment per riverbed is effectively valley fill that was generated
event than the more frequent lower discharge events, from mass wasting of over-steepened sides of the

Fig. 11. Sediment storage conditions in the Raikot and Buldar Basins. The Rupal Basin does not share the same geometric shape (tangent glacial
valleys and tributaries) as this model outlines and thus could not be accurately quantified with this approach.
 K. Cornwell et al. / Geomorphology 55 (2003) 25–43 39

Table 4 (over this 9-year record) of about 3.75  109 m3

Sediment storage conditions in the Raikot and Buldar drainage year
1. The WAPDA data indicates that 88% ( F 7%)
basins and partial conditions in the Rupal basin
of the sediment that is transported in a calendar year
Drainage basin Raikot Buldar Rupal
(and by proxy, discharge) occurs during the monsoonal
Mean Floodplain 40.7 29.5 64.8 months of June through September. Actual sediment
storage (m3 m
1)
yield data show variations in percentages ranging from
Mean Permanent 1.97  106 2.32  106 n/a
storage (m3 m
1) 96% to 76%.
Valley length-glacier 13.2 11.9 7.2a We calculated the bank full discharge capacity at
terminus-outlet (km) Raikot, Buldar, and Rupal River cross sections and
Total Floodplain 537,240 351,050 466,560 considered these discharges to account for 88% of the
storage (m3)
annual discharge and sediment yield generated
Total valley storage (m3) 2.6  1011 2.8  1011 n/a
Mean annual sediment 1.682  106 5.55  105 1.36  105 throughout a calendar year. The remaining 12% of
yield (m3 year
1) flow and sediment yield is expected to occur during the
Residence time (h) 17.6 57.6 252.4 remaining calendar months (October through May).
n/a—not applicable. Sediment transport fluxes were found for bank full
a
Valley length in upper watershed only (not to outlet). stage conditions for selected cross sections in the
study river systems. Selected cross-sections were
glacially carved valley. In these areas, more sediment is chosen based on our ability to accurately determine
added to the valley than can be effectively removed by depth of the channel from field surveys. We calculated
fluvial action. Thus, within a cross section, an overall the mean discharge during the monsoonal season and
increase over time of the elevation of the riverbed can the non-monsoonal season (discharge values changed
be expected. This condition would persist until larger- markedly) and multiplied that by the sediment trans-
scale climatic conditions reactivate these valley fill port rate (produced from the U.S. Army Corps of
sediments. Engineers software) to produce an annual sediment
Once an acceptable parabolic fit was obtained, load (Table 5). We then extrapolated that annual load
Reimann sums were used to estimate the total valley across the area of the drainage basin and determined
fill. Table 4 outlines the volume of sediment in the the maximum and minimum rate of denudation for the
various storage reservoirs calculated for the Raikot, respective basins (Table 6).
Rupal, and Buldar basins. The geographic nature of The denudation rates we calculated for the Rupal
the Rupal valley (receives drainage and sediment from drainage basin are consistent with those calculated
adjoining ranges that are not part of Nanga Parbat) from volumetric estimates of debris fans over time
does not lend itself to quantifying its sediment storage ( f 2 mm year
1 from Shroder et al., 1999) on the
capabilities to the Indus River outlet, so those values south slopes of Nanga Parbat massif (remember that
are only calculated throughout the upper watershed the Rupal values only reflect the upper-most section
and not extrapolated throughout the entire valley. All of the entire Rupal drainage network). Denudation
subsequent erosion and denudation values reported for
the Rupal valley are reported for the upper-most Table 5
section of the valley only and should not be extrapo- Data used to calculate annual sediment loads in the study drainage
lated to the entire Rupal valley. basins
River Bank full Bank full Sediment load/ Annual
3.5. Sediment yield section stage (m) discharge monsoonal sediment
(m3 s
1) season (88%) load
(m3 year
1) (m3 year
1)
No flow duration data was available on or in close
Raikot 1 2.2 28.9 4.32  105 4.91  105
proximity to the Nanga Parbat massif. A 9-year record Raikot 3 2.4 60.5 1.480  106 1.682  106
of sediment yield (suspended only) has been estab- Buldar 4 1.2 15.8 1.15  105 1.31  105
lished by WAPDA at the Doyian station along the Buldar 2 1.7 43.8 4.89  105 5.55  105
Astore River. The drainage area of the Astore River Rupal 11 2.2 61.8 5.47  104 6.22  104
at Doyian is 4040 km2, producing a mean discharge Rupal 7 2.3 84.4 1.19  105 1.36  105
40 K. Cornwell et al. / Geomorphology 55 (2003) 25–43

Table 6
Denudation rates for drainage basins in the study
Rupal Raikot Buldar
Drainage basin area (km2) 315 299 212

Min Max Min Max Min Max

Sediment load (m3 year
1) 6.2  104 1.4  105 4.9  105 1.7  106 1.3  105 5.6  105
Denudation (mm year
1) 0.2 0.4 1.6 5.6 0.6 2.6

rates from the steeper north slope of Nanga Parbat of the glacier were removed. Approximately 19,000
also are comparable with previous efforts (Gardner m3 of these sediments were redeposited along the
and Jones, 1993), which yielded basin-wide estimates drainage way, while the rest were carried downriver.
of 1.4 –2.1 mm year
1 in the Raikot basin. We recognize the significance of these low fre-
quency, high magnitude events and their ability to
move large quantities of sediment from the high
4. Floods elevation basins outlined in this study. The inconsis-
tent nature of such events and an inability to tempo-
In addition to the continual, gradual fluvial rally characterize their contribution has led us to
removal of sediment, high magnitude, low frequency recognize their occurrence but not include their sedi-
flooding events occur sporadically on the Nanga ment transport capabilities in our calculations. The
Parbat massif. Two supraglacial floods from the addition of periodic flood-transported sediments
Shaigiri Glacier in the Rupal Valley have redistrib- would be expected to increase the denudation rates
uted f 6000 m3 worth of sediment within the drain- calculated herein.
age basin (much of it being washed out into larger
drainage networks off the Nanga Parbat massif) over
the last 10 years. During non-flood years, supraglacial 5. Conclusions
meltwaters have only contributed about 100 m3 of
sediment to the transport budget. We calculated modern denudation rates on Nanga
The Rupal Valley, with its many glacier-fed tribu- Parbat that ranged from < 1 mm year
1 in the more
taries, shows much evidence of the frequency of such sluggish, lower energy upper Rupal drainage basin
events with shorelines of temporary lakes (resulting that drains the south slope of the massif to almost 6
from glacially blocked and ponded rivers), high dis- mm year
1 in the more energetic Raikot, and Buldar
charge scour channels, fields of flood-deposited drainage networks that drain portions of the north
boulders, and pendant bar deposits (Shroder et al., slope. The relatively steep slopes associated with the
1998). Raikot and the Buldar basins generate greater shear
The Tarshing Glacier blocked the Rupal River in stresses, more turbulent flow, and transport more
1851 and again in 1856, producing lakes about 2.2 km sediment than the more gently sloping Rupal reach.
long and 65 m deep before rupturing the glacier These values are consistent with denudation rates
terminus and catastrophically discharging. Water vol- on the massif calculated using other methods such as
umes of 16  106 m3 drained for up to 3 days causing debris fan accumulation rates (Shroder et al., 1999;
considerable erosion of the channel system. f 2 mm year
1 over the last 5000 year), sediment
Raikot Glacier has also produced flood events. yield from the Raikot Glacier (Gardner and Jones,
During the winter of 1994, the subglacial drainage 1993; f 4.6 –6.9 mm year
1 for the glacier area and
system of the glacier was blocked about 1.5 km 1.4 – 2.1 mm year
1 for the basin area), and the
upstream from the glacial terminus for several months. dissection and incision of glacially deposited high
Breakout thresholds were breached and more than terrace by fluvially processes (Shroder and Bishop,
45,000 m3 of sediment from the side and frontal areas 2000; f 22 F 11 mm year
1). In nearby work along
 K. Cornwell et al. / Geomorphology 55 (2003) 25–43 41

the Indus, Burbank et al. (1996) determined bedrock hundreds to thousands of square kilometers in area)
incision rates to vary from 2 to 12 mm year
1. More than the basins studied on Nanga Parbat. The effective
regionally based studies (Hewitt, 1972; Ferguson, erosion rate reported is primarily a function of glacial
1984; Hallet et al., 1996) suggested that denudation erosion, debris entrainment, and fluvial transport.
rates of more than 1 mm year
1 are not unusual in Long-term effective erosion rates >47 mm year
1
such dynamic settings. (Cai, 1994) have been reported in drainage basins in
On a broader scale, summary work by Hallet et al. SE Alaska that are comparable in area to the basins
(1996) showed much higher denudation rates (termed outlined in this study. Short-term records ( < 10 year in
effective erosion rates) of 20.31 mm year
1 (un- length) of other glaciers in SE Alaska yielded
weighted average) occurring in glaciated drainage unweighted averages of 30.39 mm year
1, with some
basins in SE Alaska where long-term (>10 year in rates exceeding 60 mm year
1 (Hunter, 1994). Fig. 12
length) records exist. In general, these glacial systems shows how the denudation rates and sediment yield
are fast moving and often much larger in area (many quantities determined in this study compared to other

Fig. 12. Global perspective of Nanga Parbat denudation rates and sediment yield conditions as compared to other glaciated mountain systems
(after Hallet et al., 1996). Circles refer to glaciers from SE Alaska, small squares to the Swiss Alps, triangles to Norway/Svalbard, and squares to
other regions including New Zealand, Asia, and Iceland.
42 K. Cornwell et al. / Geomorphology 55 (2003) 25–43

glacial drainage basins compiled in Hallet et al. surficial processes over time, the task seems over-
(1996). Specific yield conditions and denudation rates whelmingly complex indeed. This complexity has
(reported as effective erosion rate by Hallet et al., been most ably described by Shroder and Bishop
1996) calculated for the basins on Nanga Parbat rank (2000) who state that ‘‘the overall denudation of the
within the general trend of central Asian glaciers Nanga Parbat massif is a spatially and temporally
included in Hallet et al.’s (1996) compilation. Most complex mosaic of topography produced by tectonics
of the yield data from comparable basin areas in Fig. overprinted with the surficial processes of mass move-
12 come from drainage basins in SE Alaska that ment, glaciers, rivers and catastrophic floods, all
receive significantly more precipitation (building operating at different rates, through time’’.
larger glaciers) and consequently does more erosional
work (both producing and transporting sediment) than Acknowledgements
the drainage basins on Nanga Parbat, perhaps ac-
counting for the more prolific yield conditions and This work was supported in part by the National
erosion rates. Science Foundation (grants EAR-9418839 and EPS-
The denudation rates derived herein fall on the low 9720643), the Central Missouri State University
side of our initial calculations. Considering the steep Collaborative Research Grant, and the sweat and toil
relief and the energetic glacial systems on NP, we of the Nanga Parbat Geomorphology Research Group
anticipated higher denudation rates. Several reasons over the summers of 1996 and 1997. Special thanks to
may account for the discrepancy, all of which are Michael Bishop for his help with our morphometric
related to the difficulty of collecting accurate, precise, data needs and interpretations.
and reliable flow and sediment data at Nanga Parbat.
First, our field data represent conditions during a short
time frame during low flow conditions. Collecting References
sediment load data over an entire melt season would
generate a more complete picture of the fluxes inher- Ali, M., 1995. The Northern Area of Pakistan, Physical and Human
Geography. Survey of Pakistan, Pakistan.
ent in such a dynamic system and would allow for Burbank, D.W., Beck, R.A., 1991. Models of aggradation versus
more comprehensive modeling of erosion rates. Lon- progradation in the Himalayan foreland. Geologische Run-
ger sampling intervals would offer even more com- dschau 80 (3), 623 – 638.
prehensive insights. Second, we have not attempted to Burbank, D.W., Leland, J., Fielding, E., Anderson, R.S., Brozovic,
N., Reid, M.R., Duncan, C., 1996. Bedrock incision, rock uplift
model extreme events such as subglacial outburst
and threshold hillslopes in the northwestern Himalayas. Nature
floods and heavy melt/high precipitation events that 379, 505 – 510.
would generate substantial discharge and sediment Cai, J., 1994. Sediment yields, lithofacies architecture and mudrock
yields. Third, bed load probably accounts for a greater characteristics in glaciomarine environments. PhD dissertation,
proportion of the total sediment load in extraordinary Northern Illinois University, Dekalb.
steep mountain stream than in other environments, but Clesceri, L.S., Greenberg, A.E., Trussell, R.R. (Eds.), 1989. Stand-
ard Methods for the Examination of Water and Wastewater, 17th
an independent source of bed load data does not exist ed. National Water Well Association Method 2540 D. National
against which we could evaluate our estimates. Water Well Association. American Public Health Association,
Clearly, one of the challenges to developing quan- pp. 2-71 to 2-77.
titative understandings of high-elevation fluvial sys- Ferguson, R.I., 1984. Sediment load of the Hunza River. In: Miller,
tems and subsequent denudation rates is the general K.J. (Ed.), The International Karakoram Project, vol. 2. Cam-
bridge Univ. Press, Cambridge, UK, pp. 581 – 598.
paucity of discharge data available from such sites. Gardner, J.S., Jones, N.K., 1993. Sediment transport and yield
Consequently, indirect methods are commonly em- at the Raikot Glacier, Nanga Parbat, Punjab Himalaya. In:
ployed to estimate discharge conditions. These meth- Shroder Jr., J.F., (Ed.), Himalaya to the Sea. Geology, Geo-
ods, however, often assume steady, uniform flow morphology, and the Quaternary. Routledge, London, Eng-
conditions (conditions that are rare in such dynamic land, pp. 184 – 197.
Hallet, B., Hunter, L., Bogen, J., 1996. Rates of erosion and sedi-
systems), leading often to inaccurate discharge deter- ment evacuation by glaciers: a review of field data and their
minations (Jarrett, 1987). Coupled with this limitation implications. Global and Planetary Change 12, 213 – 235.
with the overall dynamic and variable nature of Harbor, J.M., Wheeler, D.A., 1992. On the mathematical descrip-
 K. Cornwell et al. / Geomorphology 55 (2003) 25–43 43

tion of glaciated valley cross-sections. Earth Surface Processes vergent shear zones of a crustal-scale pop-up structure. Geol-
and Landforms 17, 477 – 485. ogy 27, 999 – 1002.
Hewitt, K., 1972. The mountain environment and geomorphic pro- Schneider, D.A., Edwards, M.A., Kidd, W.S.F., Zeitler, P.K., Coath,
cesses. In: Slaymaker, H.O., McPherson, H.J. (Eds.), Mountain C., 1999b. Early Miocene anatexis identified in the western
Geomorphology: Geomorphological Processes in the Canadian syntaxis, Pakistan Himalaya. Earth and Planetary Science Let-
Cordillera. Tantalus Research, Vancouver, British Columbia, ters 167, 121 – 129.
No. 14, 17 – 34. Shroder, J.F., 1989. Hazards of the Himalayas. American Scientist
Hunter, L.E., 1994. Ground-line systems of modern temperate 77, 564 – 573.
glaciers and their effects on glacier stability. PhD dissertation, Shroder Jr., J.F., Bishop, M.P., 2000. Unroofing of the Nanga Parbat
Department of Geology, Northern Illinois University, Dekalb, Himalaya. In: Khan, M.A., Treloar, P.J., Searle, M.P., Jan, M.Q.
467 pp. (Eds.), Tectonics of the Nanga Parbat Syntaxis, the Western
Jarrett, R.D., 1984. Hydraulics of high-gradient streams. American Himalaya. Geological Society of London Special Publication,
Society of Civil Engineers, Journal of Hydraulics Division 110 vol. 170, pp. 163 – 179. London.
(HY11), 1519 – 1539. Shroder Jr., J.F., Bishop, M.P., Scheppy, R. 1998. Catastrophic
Jarrett, R.D., 1987. Errors in slope-area computations of peak dis- flood flushing of sediment, western Himalaya, Pakistan. In:
charges in mountain streams. Journal of Hydrology 96, 53 – 67. Kalvoda, J., Rosenfeld, C.L. (Eds.), Geomorphological Hazards
McBean, E.A., Al-Nassri, S., 1988. Uncertainty in suspended sedi- in High Mountain Areas. Kluwer Acad. Pub., pp. 27 – 48.
ment transport curves. Journal of Hydraulic Engineering 114 Shroder Jr., J.F., Scheppy, R.A., Bishop, M.P.,Denudation of small
(1), 257 – 264. alpine basins, Nanga Parbat Himalaya, Pakistan. Arctic, Antarc-
Meyer-Peter, E., Muller, R., 1948. Formulas for bed load trans- tic, and Alpine Research 31, 121 – 127.
port. In: Report on Second Meeting of the International As- WEST Consultants, 1996. WinXSPRO: A Channel Cross-Section
sociation of Hydraulic Structures Research, Stockholm, Analyzer. WEST Consultants, Carlsbad, CA.
Sweden, pp. 373 – 410. Winslow, D.M., Zeitler, P.K., Chamberlain, C.P., Hollister, L.S.,
Miller, W.R., Drever, J.I., 1977. Chemical weathering and related 1994. Direct evidence for a steep geotherm under conditions
controls on surface water chemistry in the Absaroka Mountains, of rapid denudation, Western Himalaya, Pakistan. Geology 22,
Wyoming. Geochimica et cosmochimica acta 41, 1693 – 1702. 1075 – 1078.
Nielson, D.R., 1974. Sediment transport through high mountain Wolman, M.G., 1954. A method of sampling coarse river-bed
streams of the Idaho Batholith. MS thesis, University of Idaho, material. American Geophysical Union Transactions 35 (6),
Moscow, pp. 6 – 12. 951 – 956.
Parker, G., Klingeman, P., McLean, D., 1982. Bedload and size Young, G.J., Hewitt, K., 1990. Hydrology research in the upper
distribution in paved gravel-bed streams. Journal of the Hy- Indus basin, Karakorum Himalaya, Pakistan. In: Molnar, L.
draulics Division 108(HY4) (198204), 544 – 571. (Ed.), Hydrology of Mountainous Areas. IAHS Publ., vol. 190.
Poage, M.A., Chamberlain, C.P., Craw, D., 2000. Massif-wide Wallingford, UK, pp. 139 – 152.
metamorphism and fluid evolution at Nanga Parbat, northwest- Zeitler, P.K., Sutter, J.F., Williams, I.S., Zartman, R., Tahirkheli,
ern Pakistan. American Journal of Science 300, 463 – 482. R.A.K., 1989. Geochronology and temperature history of the
Raymo, M.E., Ruddiman, W.F., 1992. Tectonic forcing of late Cen- Nanga Parbat-Haramosh massif, Pakistan. In: Malinconico,
ozoic climate. Nature 359, 117 – 122. L.L., Lillie, R.J. (Eds.), Tectonics of the Western Himalaya.
Raymo, M.E., Ruddiman, W.F., Froelich, P.N., 1988. Influence of Geological Society of America Special Paper, vol. 232. Boulder,
late Cenozoic mountain building on ocean geochemical cycles. CO, pp. 1 – 22.
Geology 16, 649 – 653. Zeitler, P.K., Chamberlain, C.P., Smith, H.A., 1993. Synchronous
Schneider, D.A., Edwards, M.A., Kidd, W.S.F., Khan, M.A., anatexis, metamorphism, and rapid denudation at Nanga Parbat,
Seeber, L., Zeitler, P.K., 1999a. Tectonics of Nanga Parbat, Pakistan Himalaya. Geology 21, 347 – 350.
western Himalaya: synkinematic plutonism within the doubly
 Geomorphology 55 (2003) 45 – 63
www.elsevier.com/locate/geomorph

Spatial distribution of sediment storage types and quantification

of valley fill deposits in an alpine basin, Reintal,
Bavarian Alps, Germany
Lothar Schrott*, Gabi Hufschmidt, Martin Hankammer,
Thomas Hoffmann, Richard Dikau
Department of Geography, University of Bonn, Meckenheimer Allee 166, 53115 Bonn, Germany
Received 9 January 2002; received in revised form 28 May 2002; accepted 10 March 2003

Abstract

Spatial patterns of sediment storage types and associated volumes using a novel approach for quantifying valley fill deposits
are presented for a small alpine catchment (17 km2) in the Bavarian Alps. The different sediment storage types were analysed
with respect to geomorphic coupling and sediment flux activity. The dominant landforms in the valley in terms of surface area
were found to be talus slopes (sheets and cones) followed by rockfall deposits and alluvial fans and plains. More than two-thirds
of the talus slopes are relict landforms, completely decoupled from the geomorphic system. Notable sediment transport is
limited to avalanche tracks, debris flows, and along floodplains. Sediment volumes were calculated using a combination of
polynomial functions of cross sections, seismic refraction, and GIS modelling. A total of, 66 seismic refraction profiles were
carried out throughout the valley for a more precise determination of sediment thicknesses and to check the bedrock data
generated from geomorphometric analysis. We calculated the overall sediment volume of the valley fill deposits to be 0.07 km3.
This corresponds to a mean sediment thickness of 23.3 m. The seismic refraction data showed that large floodplains and
sedimentation areas, which have been developed through damming effects from large rockfalls, are in general characterised by
shallow sediment thicknesses ( < 20 m). By contrast, the thickness of several talus slopes is more than twice as much. For some
locations (e.g., narrow sections of valley), the polynomial-generated cross sections resulted in overestimations of up to one
order of magnitude; whereas in sections with a moderate valley shape, the modelled cross sections are in good accordance with
the obtained seismic data. For the quantification of valley fill deposits, a combined application of bedrock data derived from
polynomials and geophysical prospecting is highly recommended.
D 2003 Elsevier B.V. All rights reserved.

Keywords: Sediment storage; Sediment volume; Geomorphometric analysis; Seismic refraction; GIS modelling; Bavarian Alps

1. Introduction

The spatial distribution of different sediment storage

types in alpine environments (such as talus sheets,
* Corresponding author. Tel.: +49-228-739098; fax: +49-228- debris cones, alluvial plains, or rockfall deposits) is
739099. the result of temporal and spatial interaction of geo-
E-mail address: [email protected] (L. Schrott). morphic processes. However, the role of sediment

0169-555X/03/$ - see front matter D 2003 Elsevier B.V. All rights reserved.
doi:10.1016/S0169-555X(03)00131-4
46 L. Schrott et al. / Geomorphology 55 (2003) 45–63

storage in alpine environments, and their geomorphic esis and the progressive systematic increase of power
coupling with specific processes, is not well under- function exponents have been discussed critically by
stood (Caine and Swanson, 1989; Trimble, 1995). The Harbor (1990). Instead of the power-law equation,
complexity of geomorphic processes in high mountain several authors proposed polynomial functions to
areas and their discontinuous spatial and temporal approximate valley cross sections (Wheeler, 1984;
character complicates the distinction of their quantita- Augustinus, 1992; James, 1996). Previous studies by
tive influence on different storage types (Dietrich and Schrott and Adams (1999, 2002) in the Dolomites
Dunne, 1978; Jones, 2000). Even a relatively homoge- showed that geomorphometric analysis for valley fill
neous landform like a talus sheet cannot simply be estimations can lead to significant overestimations, but
linked to rockfall events. Furthermore, knowledge of a validation with accurate data from seismic refraction
their evolution during Late-glacial and post-glacial surveys can improve our knowledge regarding the
times is fragmentary as data are only available for a quantification of sediment volumes in alpine basins
few locations. There is also a need for more detailed (Hoffmann and Schrott, 2002).
information about volumes of potentially mobilised The paraglacial sedimentation cycle, originally
sediments, especially in densely populated mountain introduced by Church and Ryder (1972) and subse-
regions with a high potential of natural hazards. A quently used by Church and Slaymaker (1989),
better knowledge of sediment storage may also lead to describes the magnitude of changes in sedimentation
significant improvements in process modelling and during the paraglacial period following a glaciation in
understanding landform evolution (Church and Slay- different basins in Canada. Knowledge about para-
maker, 1989; Jordan and Slaymaker, 1991). glacial sedimentation and denudation in the Alps is
The papers of Nel Caine demonstrated that the based on both back-calculations of sediment volumes
processes of sediment transfer in an alpine system can derived from large alluvial fans, valley fills, or glacio-
be understood as a cascading system with variable lacustrine deposits and modern sediment yield from
controlling factors and changing states of activity and delta growth and river load (Einsele and Hinderer,
reaction times (Caine, 1974, 1976). Early publications 1997; Müller, 1999; Hinderer, 2001). However, quan-
of Heinrich Jäckli and Anders Rapp were valuable titative measures of sediment volumes in small moun-
pioneer studies, introducing the quantification of geo- tain basins are sparse and difficult to obtain (Shroder et
morphic processes in sediment budget studies (Jäckli, al., 1999; Schrott and Adams, 2002). For the European
1957; Rapp, 1960). Alps, and especially for small alpine catchments, the
For small alpine catchments ( < 50 km 2) with extent to which sediment yields are still influenced by
numerous types of sediment storage types (debris sediments from the last glaciation is not well under-
cones, talus slopes, alluvial plains) little data are stood. Although some new findings concerning para-
available. By contrast, the importance of small rivers glacial sedimentation and denudation processes
and drainage basins in terms of sediment yield and support this model (Ballantyne, 2002), further research
sediment load has been shown by Milliman and Syvit- is needed to understand the role of sediment reservoirs
ski (1992) analysing a dataset of 280 different basins. (Müller, 1999; Hinderer, 2001).
The rather uniform trough or U-shape of many In this paper, we present preliminary results of a
alpine valleys stimulated several researchers to use research project on the role of sediment storage types
geomorphometric approaches for analysing glacial in the geomorphic system of a typical basin of the
valley developments. As one of the first researchers, northern calcearous Alps. The main objectives of the
Svensson (1961) applied a parabola to approximate the study are:
cross section of a formerly glaciated valley in northern
Sweden. Several authors have suggested that valley (i) to identify and explain the spatial distribution of
morphology progressively fits a parabolic form with sediment storage types with respect to their
increasing extent of glacial erosion (Graf, 1970; Hirano activity and geomorphic coupling;
and Aniya, 1988). Thus, the stage of valley evolution or (ii) to quantify the valley fill using a geomorpho-
a higher degree of glaciation was often modelled by the metric approach, refraction seismic techniques;
use of higher power function exponents. This hypoth- and GIS; and
 L. Schrott et al. / Geomorphology 55 (2003) 45–63 47

(iii) to evaluate the sediment storage in terms of the and Austria. The basin is located in the ‘‘Wetterstein-
paraglacial sedimentation cycle. gebirge’’ and represents one of the most prominent U-
shaped valleys in the German Alps. A thick-bedded or
This paper is thus a contribution to qualitative and massive limestone dominates the area and this pro-
quantitative interpretations of sediment storage in motes solution, karst feature appearance, and subter-
small alpine basins. ranean drainage. The considered subcatchment has a
size of 17 km2 ranging in altitude from 1050 to 2750
m asl. (Fig. 1). The total drainage basin, however,
2. Geological and geomorphological setting of the covers an area of 27 km2. The upper part of the
study area catchment has since Late-glacial times almost
decoupled from the area considered and has not been
The Reintal valley is situated in the northern part of included in this study. The upper borderline runs
the Bavarian Alps near the border between Germany almost parallel to a steep cliff and does not represent

Fig. 1. Location of study area and shaded relief of the drainage basin.
 48
L. Schrott et al. / Geomorphology 55 (2003) 45–63
Fig. 2. The distribution of sediment storage types, degree of activity and process coupling with respect to sediment input, sediment output, or both.
 L. Schrott et al. / Geomorphology 55 (2003) 45–63 49

a hydrological border. However, this borderline sep- length) (Chorley and Kennedy, 1971). The growth,
arates clearly the catchments in terms of clastic sedi- persistence, or destruction of a particular storage type
ment input into the main valley. and landform depends heavily on the processes which
The alpine character is evidenced by some rem- are responsible for the amount of sediment input or
nants of large Late-glacial moraines and a relief output. Although subystem I represents more than
exceeding 1700 m with summits above 2700 m asl. 82% of the surface of the drainage basin, covering
In several parts, the basin represents almost closed an area of almost 14 km2, the amount of sediment
sediment systems with natural sediment sinks caused storage is less important due to the dominance of steep
by the damming effects of large talus sheets and cones rock walls (>45j). Sediment storage is limited to
and rockfall deposits (Figs. 2 and 3). Thus, partially some small cirques in the upper areas of the drainage
interfingered deposits that are mostly related to active basin. Here, some shallow talus sheets and cones have
debris flows, rockfalls, and wet snow avalanches developed. In contrast, subsystems II and III (repre-
cover a main part of the valley bottom. Before rockfall senting only around 3 km2—or 18% of the catchment)
deposits, areas of sedimentation have been developed store most of the sediment generated within the catch-
(Fig. 3). Because of the negligible influence of fluvial ment. Therefore, in this study, we focus mainly on
sediment output, which is supported by the fact that in these two subsystems that are part of the mapping area
some parts of the valley only subterranean drainage (Figs. 1 and 4).
occurs, large sediment storage types such as talus Five main geomorphic processes of formation of
sheets or alluvial fans could have developed since the talus slopes and valley morphology have been
Late-glacial times (Fig. 2). The valley can be roughly observed: (i) rockfalls, (ii) debris flows, (iii) wet snow
divided into three subsystems representing a typical avalanches, (iv) episodic floods and inundations, and
landform assemblage of the sediment cascade (Fig. 4). (v) sheetwash.
Each subsystem is connected through processes and The valley bottom, as the main accumulation area,
influenced by regulators (e.g., slope angle, slope represents < 20% of the catchment surface. Never-

Fig. 3. View downvalley showing the middle part of the mapping area. The large sedimentation area in the centre was mainly caused by the
rockfall deposit in the background.
50 L. Schrott et al. / Geomorphology 55 (2003) 45–63

Fig. 4. Conceptual model of the different storage types and processes of the sediment cascade in the Reintal valley.

theless, it probably contains most of the sediment sediment storage unit was mapped along with several
mobilised and deposited during post-glacial times. attributes such as form, sediment properties, degree of
vegetation, and process coupling (Table 1). The sedi-
ment storage units were subsequently digitised for
3. Methods, techniques and modelling approaches further GIS analysis with ArcInfo and ArcView mod-
elling tools. Particular attention was paid to the
Intensive field work was carried out during the coupling and decoupling mechanisms of storage type
summers of 2000 and 2001, which included extensive and geomorphic process.
geomorphological mapping, air photo interpretation
and seismic refraction soundings. Geomorphometric
analysis was applied to test their applicability with
respect to valley cross-section modelling and to com- Table 1
Geomorphological mapping attributes
pare estimations of bedrock curvatures with obtained
results from the geophysical surveys. Attribute Description
Form curvature, slope (in degrees), micro
3.1. Geomorphological mapping and air photo forms
Sediment properties homogenous/heterogeneous, form
interpretation (surface) (rounded/angular), particle size,
orientation
A total of 126 storage units were mapped and Lithology limestone (homogeneous)
classified into eight main sediment storage types. Vegetation degree of cover, type, species, damage
For mapping purposes we used orthophotos with a by recent processes
Hydrography perennial/periodical/subterranean
resolution of 1:5000. The original orthophotos— drainage, lake with/without outlet
showing a resolution of 1:10,000—were taken in Process type, percentage of formative influence,
October 1996. The classification scheme used is activity
similar to that described by Ballantyne and Harris Geomorphic system coupled/decoupled system
(1994). To achieve a maximum of information, each Sediment storage type e.g. talus sheet, debris cone, floodplain
 L. Schrott et al. / Geomorphology 55 (2003) 45–63 51

3.2. Geomorphometric analysis a, b, and c are coefficients determined by a least-

squares fit to the empirical data (Harbor and Wheeler,
Along the valley, 32 transects were extracted from 1992).
the DEM for geomorphometric analysis (Fig. 5). In a For all transects, the lowest possible order of poly-
further step, the transects were divided into two parts: nomial was chosen for the following reasons:
the recent valley floor and the adjacent rock walls.
Because the rock wall shows a stepped profile with (i) to avoid a fitting of particular irregularities of the
steeper and flatter parts, partially caused by cirques and rock wall, which are mainly the result of
hanging valleys, the curve fitting was always based on Holocene rockfalls and weathering but do not
the first oversteepening rock wall segment, which represent the post glacial condition; and
probably represents the height of the last glacial max- (ii) to model a more moderate but not extreme
imum in the valley. We assumed that this part of the overdeepening of the valley floor.
rock wall was entirely covered during the last glacia-
tion. The length of the rock wall for the polynomial 3.3. Seismic refraction sounding
calculation varies within appr. 50 m but is of similar
magnitude (300 m). This limitation to the first rock wall Seismic refraction surveys were carried out with
segment is reasonable because the use of the entire rock 24-channel Bison Galileo equipment. A total of 66
wall would lead to curve fittings resulting in unrealistic soundings were carried out with individual profile
and very deep troughs. Furthermore, it guarantees lengths between 72 and 96 m, and geophone spacing
relatively homogenous segments for polynomial mod- between 3 and 4 m (Fig. 5; Table 2). All profiles
elling. For each transect, a polynomial function from were referenced using enlarged orthophotos (1:5000)
second to fifth order was used for modelling the cross in the field and additional GPS coordinates. The
sections. For example a quadratic equation (second- impact source was a 5-kg sledge hammer. Signals
order polynomial) has the expression were generally stacked four times for a higher signal-
to-noise ratio. Each profile was measured by 10 to 15
y ¼ a þ bx þ cx2
shots (i.e., with the impact source located every 6 to
where y and x are vertical and horizontal distances of 8 m). By means of variations in primary seismic
points on the transect from some arbitrary datum, and wave (p-wave) velocities, different layers could be

Fig. 5. Location of 32 transects and seismic refraction profiles used in this study (see also Table 2).
52 L. Schrott et al. / Geomorphology 55 (2003) 45–63

Table 2
Summary of seismic refraction survey on different storage types and velocities of the bedrock refractor
Profile Storage type Survey- Method v (m/s) Bedrock
no. distance (m)
Probableb Possiblec Mean
depth (m)
3 Alluvial fan/ 72 WFIa 2500 U 7.5
avalanche deposit
4 Debris cone 96 WFI 2400 U 5
8 Debris cone 96 Intercept >2500 U 20 – 30
9 Debris cone 96 WFI 3600 U 18
11 Talus sheet 96 Intercept U 25
12 Talus sheet 96 WFI 2850 U 18.5
13 Talus sheet 96 WFI 2910 U 23.5
14 Debris cone 96 Intercept >2500 U 15 – 30
17 Fluvial deposit 72 Intercept >2500 U 3–6
18 Talus sheet 96 WFI 4350 U 7
19 Alluvial fan 96 WFI 3550 U 11
20 Talus sheet 96 WFI 5750 U 12
22 Talus sheet 96 WFI 4950 U 11
23 Alluvial fan 96 WFI 4200 U 18
24 Alluvial fan 96 Intercept >2500 U 30 – 40
25 Alluvial fan 96 Intercept >2500 U 30 – 40
26 Alluvial fan 96 WFI 2600 U 13
27 Talus cone 72 WFI 3950 U 6.5
28 Alluvial fan 96 WFI 2400 U 7.5
29 Alluvial fan 72 WFI 2400 U 7.5
30 Floodplain 96 WFI 3050 U 10.5
31 Floodplain 96 WFI 2450 U 9.5
32 Floodplain 96 WFI 2700 U 8.5
37 Floodplain 96 Intercept >2500 U 10 – 15
39 Alluvial fan/ 96 Intercept >2500 U 10
avalanche deposit
40 Floodplain 96 WFI 2750 U 15
41 Floodplain 96 WFI 2650 U 17
44 Floodplain 96 Intercept >2500 U 16
50 Debris cone 96 WFI 2700 U 9.5
51 Debris cone 96 WFI 2800 U 26.5
56 Talus sheet 72 WFI 4400 U 10.5
57 Talus sheet 72 WFI 3700 U 8
59 Alluvial fan 96 WFI 3400 U 14
62 Talus cone 72 WFI 2700 U 10
63 Talus cone 96 WFI 2850 U 14.5
64 Avalanche deposit 72 WFI 3300 U 9.5
66 Debris cone 96 WFI 3300 U 6.5
No bedrock: v < 2200 m/s.
a
Wavefront-inversion.
b
Bedrock probable: v>2500 m/s.
c
Bedrock possible: 2200>v>2500 m/s.

identified. The depth of the different refractor hori- survey allows the interpretation of possible and
zons was calculated using the Intercept method (one- probable bedrock depths because of measured p-
dimensional) and the wavefront inversion method for wave velocities. These results were then used to test
two-dimensional interpretation (cf. Knödel et al., and validate the geomorphometric results (Fig. 6;
1997; Reynolds, 1997) (Table 2). The geophysical Table 2).
 L. Schrott et al. / Geomorphology 55 (2003) 45–63 53

Fig. 6. Flowchart showing the methodological approach for calculating sediment thicknesses and sediment volumes using DEMs and
information derived from geomorphometric cross-section analysis and seismic refraction survey.

3.4. GIS modelling approach A DEM of the bedrock (DEMbedrock)—presenting

a valley floor without sediment deposits—was gen-
An important prerequisite for the employed mod- erated using the bedrock depths of the modelled
elling approach used is the availability of a high- transects. The areas between the transects were again
resolution digital elevation model (DEM) (Fig. 6). interpolated using ArcInfo ‘topogrid’. The sediment
The DEM of the surface (DEMsurface) is based on thickness along the extracted transects and the calcu-
altitude data obtained from a photogrammetric sur- lation of the overall sediment volume was subse-
vey. Original contour intervals vary between 10 and quently derived from the difference between the two
20 m. The DEM used in this study was generated DEMs (DEMsurface  DEMbedrock). Additionally, the
using the Arc/Info algorithm ‘topogrid’ and shows a estimated depths of the seismic interpretation with
resolution of 5 m. The vertical accuracy of the probable and possible bedrock estimations (velocities
original data set is < 0.5 m. Topogrid is basically a >2200 m/s) were interpolated to create a new bedrock
discretised thin plate spline. The algorithm led to a coverage (Fig. 6; Table 2). From this coverage, cross
smoothing of data and considers the hydrography in sections with a high degree of seismic information
order to generate a continuous stream network. The were extracted and compared with the cross sections
generation of a hydrologically correct DEM is derived from geomorphometric analysis (Table 4).
ensured by automatic identification of maximum The information of the seismic survey can be used
curvature, e.g. ridges or drainage paths (Hutchinson, to check and evaluate the sediment thicknesses
1989). derived from the geomorphometric approach. Sedi-
54 L. Schrott et al. / Geomorphology 55 (2003) 45–63

Table 3 cones, and Late-glacial moraines (Table 3). Talus

Total amount and relative surface area of sediment storage types sheets are more common in the eastern and narrower
Sediment storage type Area Total part of the valley, whereas talus cones are concentrated
number
in the western and middle part of the basin. Rockfall
Talus sheet 41% 35 deposits occur frequently throughout the valley and
Talus cone 26% 24
block the axial sediment transport. Because of their
Rockfall 11% 8
Alluvial fan 9% 21 high trap efficiency, this damming effect causes the
Floodplain and fluvial 6% 9 development of large floodplains and fans. The rockfall
deposits deposits now act as separators of almost closed sedi-
Debris cone 5% 22 ment sinks (Fig. 3). Debris cones are also widespread in
Avalanche deposits 1% 5
the valley, but in general much smaller in size than the
Moraine 1% 2
large talus cones. Debris flows are frequent on north-
facing slopes where material of huge talus sheets is
ment deposits outside the main valley floor were not remobilised. Here, smaller and younger debris flows
considered in the calculation of total sediment storage are superimposed on larger and older talus sheets and
volumes because of their negligible volume. cones. They form a palimpsest of systems, where older
landforms are partially erased by younger landforms
created by processes for which older landforms operate
4. Results as a boundary condition (Chorley et al., 1984).
Visible avalanche deposits are, in general, limited
4.1. Spatial distribution of sediment storage types in to the very active avalanche tracks (recurrence inter-
the valley val V 1 year) and the deposits are interbedded with
floodplains and alluvial fans. Especially on south-
At present, the dominant landforms in the valley are facing slopes, the annually observed avalanche activ-
talus sheets followed by talus cones, rockfall deposits, ity results in 11 larger avalanche tracks eroding talus
alluvial fans and floodplains, debris and avalanche sheets and cones that are covering most of this slope.

Fig. 7. Vertical extension of the different sediment storage types.

 L. Schrott et al. / Geomorphology 55 (2003) 45–63 55

Fig. 8. Degree of activity and surface area of different sediment storage types.

On north-facing slopes, however, avalanche tracks 4.2. Activity of the geomorphic system and process
have only developed on four locations. This is coupling/decoupling
probably due to the steepness of the rock wall, which
on the other hand favours the occurrence of numer- All sediment storage units were mapped with
ous debris flows. They start on the boundary zone respect to their degree of activity and their coupling
between bedrock and adjacent talus slope, remobilis- mechanisms with geomorphic processes. Based on
ing material from older, partially inactive, or relict qualitative and semiquantitative criteria, three differ-
talus sheets and cones. At the uppermost part of the
mapping area, several alluvial fans have developed.
During snowmelt and heavy thunderstorms, debris
and fine material become frequently mobilised. This
part of the basin is characterised by a classic cirque
with a very flat valley floor. In the centre, a probably
Holocene rockfall deposit causes damming of sedi-
ments. According to eye witnesses, a separate rock-
fall occurred on the north-facing slope in 1920
(Leuchs, 1921).
The vertical extension of sediment storage ranges
from 1050 to 1650 m (Fig. 7). Two main altitudinal
levels can be differentiated that are the result of a
marked step in the longitudinal profile. In this rela-
tively narrow part of the valley, almost no sediment
could be stored. The concentration of sediment stor-
age types in the lower parts of the valley corresponds
to the increasing surface area from upper to lower
areas in the basin. The talus sheet storage type shows
the largest vertical extension and it also covers the Fig. 9. Sediment input, output, or both classified for different
greatest horizontal surface. sediment storage types.
56 L. Schrott et al. / Geomorphology 55 (2003) 45–63

ent classes of activity were distinguished and are change; in other parts of the valley, the effects of
shown in Figs. 2 and 8. geomorphic processes (debris flows, gullies) are
still very well preserved, indicating a recurrence
(i) High activity: vegetation cover 0 – 5%; during interval of at least 2 years.
the observation period of 2 years, landforms (iii) Low activity: vegetation cover 20 –90%, no
changed frequently (e.g., new gullies, debris recent impact of geomorphic processes. How-
cones) through geomorphic processes (e.g., ever, single boulders on top of well-developed
debris flows, rockfalls) causing erosion or alpine meadows or the runout of extraordinary
deposition of sediment. avalanches indicate a certain activity in these
(ii) Moderate activity: vegetation cover 5– 20%; no areas.
significant change of landforms within the
period of observation. In some areas, the impact Within the mapping area, only 21% of the surface
of geomorphic processes (e.g., only single shows a visible degree of activity. All other areas (79%
rockfall events) is not causing any observable of the mapping area) show a vegetation cover of 90–

Fig. 10. Calculated valley fill derived from geomorphometric analysis. The three selected cross sections (4, 20, 25) show shapes derived from
geomorphometric analysis and seismic refraction interpretation.
 L. Schrott et al. / Geomorphology 55 (2003) 45–63 57

100% with well-developed alpine forests. Practically 4.3. Sediment thicknesses and sediment volumes
no geomorphic impact can be inferred for time periods based on geomorphometric analysis and seismic
of decades and centuries or even longer. Fig. 8 summa- refraction surveys
rizes the degree of activity with respect to each sedi-
ment storage type. It is important to note that the talus The calculated sediment thicknesses using the
sheet storage type shows the lowest degree of activity, interpolated values of 32 polynomials show a max-
although talus sheets are widespread and the dominant imum valley fill in the uppermost part and on two
landform in terms of covered surface area. By contrast, further locations in the lower part of the valley (Fig.
the areas of avalanche deposits, which cover only 1% 10). Areas of pronounced overdeepening are always
of the total area, show the highest degree of activity— connected with relatively narrow and steep slopes,
along with the floodplains. and rock walls in the surroundings. In this case, the
The whole valley can be divided roughly into three use of polynomial functions results in relatively deep
different areas of activity (Fig. 2). The lowest part and pronounced troughs ( z 70 m). In total, however,
shows the lowest degree of activity. Inactive or relict these regions are limited to V 15% of the mapped
talus sheets cover most of the area. The observed area (Fig. 10). According to our geomorphometric
activity is limited to some smaller debris flows and to analysis, more than two-thirds of the valley is covered
the channel system of Partnach Creek. In the middle with shallow sediment mantle showing thicknesses of
part of the valley, the activity is concentrated on north- < 40 m. For 53% a sediment thickness of < 30 m and
facing slopes and a large sedimentation area, creating for 28% a thickness of < 10 m were calculated. For
a highly active floodplain. This can be explained the whole valley bottom this results in a sediment
through frequent sediment supply from the adjacent thickness of 33 m on average, which corresponds to a
rock wall and the changing braided system of Part- total sediment volume of 0.1 km3 stored in the valley.
nach Creek. Sheet flow with stone falls and rockfalls In order to check and to assess the derived cross
are frequent, and the unconsolidated material of the sections and sediment thicknesses based on the geo-
talus slopes becomes subsequently mobilised by deb- morphometric data (polynomials) they were compared
ris flows. with our seismic refraction data. The data set of our
In the upper part of the valley, the most active seismic refraction profiles allowed the validation of
landforms are alluvial fans and avalanche tracks. nine cross sections (Table 4). The range of possible
Talus slopes (sheets and cones) show, in general, a overestimations is documented by examples of mod-
moderate degree of activity. elled cross sections, and by the maps showing the
Considering the area mapped in the valley, evi- different sediment thickness derived from geomorpho-
dently only 21% of the surface area is active. These
active areas are coupled with geomorphic processes
in terms of sediment input, sediment output, or
Table 4
both. Gravitational processes influence more than
Difference between depth to bedrock estimated from polynomials
50% of the area with active storage types, 34% and geophysical soundings (refraction seimic), degree of poly-
being rockfall and 20% debris flow deposits (Fig. nomial, and maximum slope angle
2). Recent sediment transport through the basin Cross-section Maximum Degree of Slope
occurs mainly on alluvial fans and avalanche tracks, no. differencea (m) polynomial angle (j)
although these represent < 10% of the area. Most of P 3 + 25 Fifth order 25 – 30
the active talus slopes and floodplains are charac- P 4 + 100 Third order 50 – 70
terised exclusively by sediment inputs (Fig. 9). This P 5 + 70 Third order 50 – 70
implies that within the sediment cascade the greatest P 20 F0 Fourth order 30 – 45
P 21 + 60 Fifth order 45 – 55
part is currently decoupled from the sediment out- P 22 + 60 Fifth order 45 – 55
put. Currently, only some small parts of these active P 25 + 20 Fourth order 30 – 45
storage types indicate definitive sediment output, P 26 F0 Fourth order 50 – 70
visible for example on slope/channel interactions P 30 + 100 Third order 50 – 75
a
and river undercutting. Overprediction using polynomials.
58 L. Schrott et al. / Geomorphology 55 (2003) 45–63

Fig. 11. Calculated sediment thickness derived from a combined approach using geophysical data (depth of bedrock) and geomorphometric
analysis (polynomials).

metric analysis on the one hand and a combined the shallow sediment thickness of the large floodplain
approach using geophysical data and polynomials on (profiles 40, 41) in the centre of the valley ranging
the other hand (Figs. 10 and 11). The difference from only 15 to 17 m (Fig. 3; Table 2). By contrast,
between estimated depths to bedrock using polyno- the sediment thickness of several talus slopes is more
mials and cross sections with a high density of seismic than twice as much.
refraction data is shown in Table 4. The use of more
accurate seismic data results in some locations to a
significant reduction of the sediment thickness. In the 5. Discussion
centre of the valley, between the cross sections p20
and p25, the sediment thickness was reduced from 5.1. Interpretation of spatial distribution of sediment
almost 90 to < 20 m. Using the latter approach, 56% storage types with respect to process coupling
of the valley bottom is characterised by relatively
shallow sediment thicknesses of V 20 m. The interaction between topography, lithology, and
The overall corrected sediment volume is calcu- climate in mountain environments leads to the devel-
lated to be 0.07 km3 resulting in a mean sediment opment of a particular landform assemblage. The
thickness of 23.3 m. This indicates a reduction by mapping area shows clearly that talus cones and
30% (0.03 km3) taking into account seismic refraction sheets are always coupled with relatively steep rock
data (Fig. 11). The seismic refraction data show walls, whereas alluvial fans and plains are frequently
clearly that in areas with steep rock walls or relatively developed in flatter areas with fluvial activity. Ava-
narrow valley floors (V-shaped character) the use of lanche tracks and deposits can very often be linked to
polynomial equations leads to significant overestima- cirques in upper regions, and the occurrence of debris
tions of sediment thicknesses. Relatively similar flows indicates the availability of sediments and the
results were obtained in areas with a wider valley possibility of sediment concentration in couloirs (Fig.
bottom and more gentle slopes (Fig. 10). A surprising 2). The alpine cliffs of the basin are frequently
outcome according to the seismic data obtained was dissected by couloirs and broader avalanche tracks
 L. Schrott et al. / Geomorphology 55 (2003) 45–63 59

(characterised by snow accumulations lasting until a simple rockfall talus development reported from
summer) and water and debris supply especially other mountain regions with crystalline rocks, like
during rainstorms. These processes influence the form northern Scandinavia, the Canadian Rocky Moun-
of the subjacent talus (Rapp, 1960; Caine, 2001). The tains, or parts of the central Alps (Rapp, 1960; Luck-
morphological mapping of a particular landform, man, 1976; Jomelli and Francou, 2000). Because of
however, can basically imply the following problems: the relief and the extraordinary length of the rock wall
(i) several processes interact generating a less well- (>1000 m) the interaction of several processes (rock-
defined and complex landform; and (ii), equifinality, falls, debris flows, avalanches) causes the crashing of
that is, two or more processes result in one typical large blocks and boulders producing clasts of similar
landform (Thorn, 1988). In this case, it would not be size (Figs. 12 and 13).
possible to relate the sediment storage type to a
distinct process of origin. In the calcareous Alps, talus 5.2. Degree of activity and associated storage types
sheets and cones are very often not simply the result
of rockfall processes. There are other processes such Geomorphic process activity in the valley is
as sheetwash and wet snow avalanches that probably conditioned by fluctuations in climate and vegetation
play a major role in their development. Sedimento- cover since the earliest deglaciation. The main valley
logical analysis of a section and samples taken 10 and floor became ice free during Late-glacial times
15 m below the surface of a talus sheet shows a debris around 14,000 BP, whereas the upper part of the
clast with grain sizes between 0.5 and 3 cm in a sandy mapping area was probably covered by glacial ice
diamicton (Fig. 12). In fact, during heavy thunder- until the end of the Younger Dryas (Hirtlreiter,
storms a remarkable amount of sediment input 1992). We assumed that shortly after deglaciation a
through sheet flow can be observed on the boundary large part of the present-day talus slopes developed,
zone between rock wall and talus slope (Fig. 13). especially in the lowest section of the valley, which
Only a few layers of large boulders indicate periods of deglaciated at least 2000 years earlier than the upper
intense rockfall activity. This is partially in contrast to section of the valley. Many taluses in the lower part

Fig. 12. Eroded section of a large talus sheet on the north-facing slope. Layers of boulders indicate larger rockfall events. Samples were taken on
the baseline and appr. 3 m above the person.
60 L. Schrott et al. / Geomorphology 55 (2003) 45–63

The steepness and extreme relief (>1000 m) on

the north-facing slope in the middle part of the
valley favours the supply of significant quantities of
debris resulting in a higher activity (Fig. 2). Most
of the taluses in this part of the north-facing slope
are characterised by relatively unconsolidated mate-
rial, promoting the occurrence of active debris
flows. During snowmelt and intense summer pre-
cipitation, increasing sediment transport is observed
on alluvial fans, in the floodplains, and in the
channel system.

5.3. Interpretation of sediment thickness modelling

and valley fill estimations

The use of polynomials allows modelling of the

valley shape with only little data requirements. How-
ever, there are several critical points to consider. A
serious problem is basically caused by the steepness
of the rock walls and the width of the valley, which
significantly affects the curvature of the trough. The
more narrow and steep the valley forms the more
pronounced the modelled trough shape. This fre-
quently results very often in a significant overestima-
tion of depths of bedrock. Secondly, the choice of the
polynomial degree has an important influence on the
Fig. 13. Transition zone between the rock wall to the subjacent talus modelled cross section. In general, higher order poly-
slope. Sediment transport to the talus slope takes place mainly nomial functions are fitting the rock wall more accu-
through rockfall sheet flow. The size of the material corresponds to
the dominant grain size of the talus slope and varies between 2 and
rately, but model a very pronounced trough (Fig. 10).
5 cm. Even smaller irregularities in the rock wall, caused by
historical rockfalls for instance, can be fitted quite
adequately by using higher order polynomials. The
of the valley are now probably relict landforms on purpose of the modelling is, however, to estimate the
which river incision or remobilising of sediment by shape of the initial valley floor, which is now filled
debris flows has replaced sediment input as the with sediment. It is not necessarily adequate to use the
dominant mode of geomorphological activity (cf., geometry of today’s rock wall. Therefore, a lower
Ballantyne and Benn, 1994). order polynomial, fitting less pronounced recent rock
Thus, the degree of activity of sediment flux gives wall irregularities, is probably producing a more
some indication to the paraglacial slope adjustment realistic shape with a smoother and flatter trough.
and hence the temporal evolution of the valley floor. The three selected cross sections document the range
Higher process activity in the upper part of the drain- of possible bedrock estimations and show the differ-
age basin can be explained by the fact that the ences in comparison with the more accurate seismic
surrounding cirques of the hanging valleys were refraction data. Cross-section 20 shows similar results
partially glaciated until the beginning of the 20th with respect to the modelled sediment thickness, but
century. This implies still a higher availability and the fitting curve is slightly displaced. In cross-section
supply of unconsolidated sediment and can partially 25, the polynomial curve is modelling approximately
explain the increase of activity towards the upper end twice as much as the seismic data indicate; and in
of the valley. cross-section 4, an overestimation of factor 5 or higher
 L. Schrott et al. / Geomorphology 55 (2003) 45–63 61

occurs. The 32 transects used for the estimation of the them are relict landforms representing para-
total valley fill showed that good results were always glacial deposits. These landforms are de-
obtained in areas with a relatively moderate valley coupled from the present-day geomorphic
shape, whereas steep and narrow flanks are always system and currently no sediment input or
problematic. output occurs.
The seismic data show that the measured thickness (ii) At present, sediment flux activity is limited to
of sediments is relatively shallow in most areas only 21% of the valley floor. Within the basin,
throughout the valley; however, there are some talus notable sediment transport can only be ob-
sheets on the north-facing slope where the depth of served along avalanche tracks, debris flows and
bedrock could not be reached. Here, the thickness of within the floodplains.
sediments exceeds the detection limit of our seismic (iii) Rockfalls play a minor role in terms of modern
equipment used (appr. 40 to 50 m) (Fig. 12). By sediment input, but they influence greatly the
contrast, the volumes of the larger sedimentation areas development of some sediment reservoirs
associated with the dam of the rockfall are much through damming effects. During Late-glacial
smaller than expected. In these areas, we measured and Holocene periods, several locations in the
sediment thicknesses between 15 to 17 m (Fig. 3 and basin became almost closed sediment systems
cross-section 20 in Fig. 10). Investigations on talus with high trap efficiency.
slope thicknesses in the direct neighbourhood of our (iv) In qualitative terms, the geomorphic coupling
basin (Zugspitzblatt) generated from ground-penetrat- of sediment storage type and associated process
ing radar (GPR) resulted in somewhat lower values can elucidate the different degrees of activity
ranging from 5 to 13 m (Sass and Wollny, 2001). and the evolution of a particular landform.
Recent investigations on valley fillings and sedi- (v) The quantification of valley fillings in formerly
ment volumes in an alpine basin in the Dolomites/ glaciated basins is not an easy task. The
Italy demonstrate that the single use of parabolic geomorphometric approach used in this study
curves for cross-section analysis leads also to a sig- can provide a relatively quick estimation of
nificant overestimation of the bedrock depths. Com- potential sediment volumes. However, without
pared with the more accurate data based on extensive the control by more accurate data from bore-
geophysical surveys (seismic refraction and geoelec- holes or geophysical prospecting, the applica-
trical resistivity sounding), the calculated sediment tion of polynomials for valley shape modelling
volume using a parabola (exponent b = 2) was over- is speculative in its character and actual surveys
estimated by a factor of 5 (Schrott and Adams, 2002). are highly recommended. In this context, the
use of seismic refraction data allows the
validation of the geomorphometric approach.
6. Conclusions (vi) The efficiency and the degree of accuracy of
bedrock estimations using polynomials vary
In this study, promising relationships between greatly with the changing shape of the valley.
basin characteristics (e.g., valley form, local relief), Narrow valley sections with steep flanks are
geomorphic processes, and associated sediment stor- much more problematic than valley shapes
age types could be established. The quantification of with a wider valley floor and relatively mod-
valley fill deposits was assessed using a combination erate slopes. For some transects, an immense
of a geomorphometric approach applied for the whole overestimation by one order of magnitude
valley and more precise data from seismic refraction was calculated; whereas in other parts of the
soundings for some cross sections. valley, the modelled cross sections were in
A number of important implications emerge from good accordance with the obtained seismic
this study: data.
(vii) A surprising result obtained from the seismic
(i) Talus sheets are the dominant storage type in refraction data was the shallow sediment
the basin. However, more than two-thirds of thickness of the largest sedimentation area
62 L. Schrott et al. / Geomorphology 55 (2003) 45–63

(floodplain, profiles 40 and 41) in the valley Ballantyne, C.K., Harris, C., 1994. The Periglaciation of Great
with values between 15 and 17 m (Fig. 6 and Britain. Cambridge Univ. Press, Cambridge. 330 pp.
Caine, N., 1974. The geomorphic processes of the alpine environ-
Table 2). ment. In: Ives, J.D., Barry, R.G. (Eds.), Arctic and Alpine En-
(viii) The use only of bedrock data derived from vironments. Methuen, London, pp. 721 – 748.
geophysical prospecting is often not possible Caine, N., 1976. A uniform measure of subaerial erosion. Geolog-
because it is too incomplete. It may also lead to a ical Society of America Bulletin 87 (1), 137 – 140.
Caine, N., 2001. Geomorphic systems of Green Lakes Valley. In:
slight underestimation of the sediment volume
Bowman, W.D., Seastedt, T.R. (Eds.), Structure and Function of
because in some areas the bedrock cannot be an Alpine Ecosystem. Oxford Univ. Press, Oxford, pp. 45 – 74.
reached with the seismic equipment we used. Caine, N., Swanson, F.J., 1989. Geomorphic coupling of hillslope
Thus, restrictions associated with the valley and channel systems in two small mountain basins. Zeitschrift
shape, quality of digital elevation models, für Geomorphologie 33 (2), 189 – 203.
interpolation algorithms, and geophysical inter- Chorley, R.J., Kennedy, B.A., 1971. Physical Geography: A Sys-
tems Approach. Prentice-Hall International, London. 370 pp.
pretations should be considered when evaluat- Chorley, R.J., Schumm, S.A., Sudgen, D.E., 1984. Geomorphology.
ing calculated sediment volumes. Taking into Methuen, London.
account these limitations, we can conclude that a Church, M., Ryder, J.M., 1972. Paraglacial sedimentation, a con-
combined application of bedrock data generated sideration of fluvial processes conditioned by glaciation. Geo-
from polynomials and geophysical prospecting logical Society of America Bulletin 83, 3059 – 3071.
Church, M., Slaymaker, O., 1989. Disequilibrium of Holocene
is a useful modelling tool. sediment yield in glaciated British Columbia. Nature 337 (2),
452 – 454.
Dietrich, W.E., Dunne, T., 1978. Sediment budget for a small catch-
Acknowledgements ment in mountainous terrain. Zeitschrift für Geomorphologie.
Supplementband 29, 191 – 206.
Einsele, G., Hinderer, M., 1997. Terrestrial sediment yield and the
The authors thank all the members of the DFG lifetimes of reservoirs, lakes, and larger basins. Geologische
research programme ‘‘Sediment cascades in alpine Rundschau 86 (2), 288 – 310.
geosystems’’ for their cooperation and constructive Graf, W.L., 1970. The geomorphology of the glacial valley cross
discussion. Special thanks to Sandra Amlang and section. Arctic and Alpine Research 2 (4), 303 – 312.
André Niederheide for their assistance in the field. Karl Harbor, J.M., 1990. A discussion of Hirano and Aniya’s (1988,
1989) explanation of glacial-valley cross profile development.
Wörndle and his family and Charly Wehrle and his Earth Surface Processes and Landforms 15, 369 – 377.
team from the alpine Hut ‘‘Reintalanger’’ provided Harbor, J.M., Wheeler, D.A., 1992. On the mathematical descrip-
very helpful logistical support. Kirsten Hennrich’s and tion of glaciated valley cross sections. Earth Surface Processes
Chris Jenkins’ corrections and comments on the and Landforms 17, 477 – 485.
manuscript are greatly appreciated. We thank two Hinderer, M., 2001. Late Quaternary denudation of the Alps, valley
and lake fillings and modern river loads. Geodinamica Acta 14,
anonymous referees for further constructive comments 231 – 263.
on the manuscript. Financial support was provided by a Hirano, M., Aniya, M., 1988. A rational explanation of cross-pro-
grant from the Deutsche Forschungsgemeinschaft file morphology for glacial valleys and of glacial valley devel-
(DFG). opment. Earth Surface Processes and Landforms 13, 707 – 716.
Hirtlreiter, G., 1992. Spät-und postglaziale Gletscherschwankungen
im Wettersteingebirge und seiner Umgebung. Münchener Geo-
graphische Abhandlungen, Reihe B 15, 154 pp.
References Hoffmann, T., Schrott, L., 2002. Modelling sediment thickness and
rockwall retreat in an Alpine valley using 2D-seismic refraction
Augustinus, P.C., 1992. The influence of rock mass strength on (Reintal, Bavarian Alps). Zeitschrift für Geomorphologie,
glacial valley cross-profile morphometry, a case study from Suppl.-Bd. 127, 175 – 196.
the Southern Alps, New Zealand. Earth Surface Processes and Hutchinson, M.F., 1989. A new procedure for griding elevation and
Landforms 17, 39 – 51. stream line data with automatic removal of spurious pits. Journal
Ballantyne, C.K., 2002. Paraglacial geomorphology. Quaternary of Hydrology 106, 232 – 244.
Science Reviews 21, 1935 – 2017. Jäckli, H., 1957. Gegenwartsgeologie des bündnerischen Rhein-
Ballantyne, C.K., Benn, D.I., 1994. Paraglacial slope adjustment gebietes. Beiträge zur Geologie der Schweiz, Geotechnische
and resedimentation following recent glacier retreat, Fåbergstø- Serie 36, 126 pp.
len, Norway. Arctic and Alpine Research 26 (3), 255 – 269. James, L.A., 1996. Polynomial and power functions for glacial
 L. Schrott et al. / Geomorphology 55 (2003) 45–63 63

valley cross section morphology. Earth Surface Processes and kevagge and surroundings, Northern Scandinavia. Geografiska
Landforms 21, 413 – 432. Annaler 42 (2 – 3), 1 – 200.
Jomelli, V., Francou, B., 2000. Comparing the characteristics of Reynolds, J.M., 1997. An Introduction to Applied and Environ-
rockfall talus and snow avalanche landforms in an alpine envi- mental Geophysics. Wiley, New York.
ronment using a new methodological approach, Massif des Sass, O., Wollny, K., 2001. Investigations regarding alpine talus
Ecrins, French Alps. Geomorphology 35, 181 – 192. slopes using ground penetrating radar (GPR) in the Bavarian
Jones, A.P., 2000. Late quaternary sediment sources, storage and Alps, Germany. Earth Surface Processes and Landforms 26,
transfers within mountain basins using clast lithological analy- 1071 – 1086.
sis, Pineta Basin, central Pyrenees, Spain. Geomorphology 34, Schrott, L., Adams, T., 1999. An approach to the evaluation of
145 – 161. sediment storage and denudation rates in an alpine environment
Jordan, P., Slaymaker, O., 1991. Holocene sediment production in (Dolomites, Italy). Tübinger Geowissenschaftliche Arbeiten,
Lillooet River basin, British Columbia, a sediment budget ap- Series A 52, 68 – 69.
proach. Géographie Physique et Quaternaire 45 (1), 45 – 57. Schrott, L., Adams, T., 2002. Quantifying sediment storage and
Knödel, K., Krummel, H., Lange, G., 1997. Geophysik. Handbuch Holocene denudation in an Alpine basin, Dolomites, Italy. Zeits-
zur Erkundung des Untergrundes von Deponien und Altlasten, chrift für Geomorphologie. Supplementband 128, 129 – 145.
Bd., vol. 3. Springer, Berlin. Shroder, J.F., Scheppy, R.A., Bishop, M.P., 1999. Denudation of
Leuchs, K., 1921. Die Ursachen des Bergsturzes am Reintalanger small alpine basins, Nanga Parbat Himalaya, Pakistan. Arctic,
(Wettersteingebirge). Geologische Rundschau 12, 189 – 192. Antarctic and Alpine Research 31 (2), 121 – 127.
Luckman, B.H., 1976. Rockfalls and rockfall inventory data, some Svensson, H., 1961. Morphometrischer Beitrag zur Charakterisier-
observations from Surprose Valley, Jasper National Park, Can- ung von Glazialtälern. Zeitschrift für Gletscherkunde und Gla-
ada. Earth Surface Processes and Landforms 1, 287 – 298. zialgeologie 4 (61), 99 – 104.
Milliman, J.D., Syvitski, P.M., 1992. Geomorphic/tectonic control Thorn, C.E., 1988. Introduction to Theoretical Geomorphology.
of sediment discharge to the ocean, the importance of small Unwin Hyman, Boston. 247 pp.
mountainous rivers. Journal of Geology 100, 525 – 544. Trimble, S.W., 1995. Catchment sediment budgets and change. In:
Müller, B.U., 1999. Paraglacial sedimentation and denudation pro- Gurnell, A., Petts, G. (Eds.), Changing River Channels. Wiley,
cesses in an alpine valley of Switzerland. An approach to the London, pp. 201 – 215.
quantification of sediment budgets. Geodinamica Acta 12 (5), Wheeler, D.A., 1984. Using parabolas to describe the cross section
291 – 301. of glaciated valleys. Earth Surface Processes and Landforms 9,
Rapp, A., 1960. Recent development of mountain slopes in Kaer- 391 – 394.
 Geomorphology 55 (2003) 65 – 74
www.elsevier.com/locate/geomorph

Impacts of reforestation and gravel mining on the Malnant River,

Haute-Savoie, French Alps
Richard A. Marston a,*, Jean-Paul Bravard b, Tim Green c
a
School of Geology, Oklahoma State University, Stillwater, OK 74078-3031, USA
b
Faculte de Geographie, Histoire, Histoire de l’Art, Tourisme, Universite Lumiere-Lyon 2, 4, avenue Pierre-Mendes,
69676 Bron Cedex, France
c
Department of Geography and Recreation, University of Wyoming, Laramie, WY 82071-3311, USA
Received 9 March 2002; received in revised form 9 March 2003; accepted 10 March 2003

Abstract

The Malnant River is a rapidly incising river with a French name that translates as ‘‘bad creek,’’ reflecting local opinion of
the hazards from dramatic channel changes that have occurred in the last few centuries. Downcutting in the last three decades
has created severe problems for farmers in this small watershed (16 km2) as bridges are undermined, streamside roads are
threatened, and irrigation diversion structures are rendered unusable. The purpose of our study was to determine the extent and
causes of downcutting. A detailed landcover map dated 1732 revealed that forest cover had been reduced by that time to 10% of
the present-day cover. The Malnant was strongly affected by floods and debris torrents during the 18th and 19th centuries that
delivered massive amounts of sediment. During the 20th century, reforestation reduced the sediment delivery from hillslopes. In
addition, gravel extraction in the Malnant and in the Fier River (of which the Malnant is a tributary) has lowered the base level
for the river. This initiated a knickpoint that moved upstream. Weirs placed in the Malnant in 1968 were used to measure rates of
bed incision in the field. With a mean width of 4.0 m and degradation up to 36 cubic meters per meter channel length, the lower
4.5 km of the Malnant has experienced a net loss of approximately 163,000 m3 of bed material. Above the 4.5-km point on the
Malnant, bedrock controls exist that have arrested the upstream-progressing degradation.
D 2003 Elsevier Science B.V. All rights reserved.

Keywords: Forest stream; Gravel mining; France; Alps

1. Introduction has dominated in the 20th century. This incision was

stopped by large limestone blocks deposited as glacial
The Malnant has experienced dramatic channel till, replaced by lateral erosion that excavated ava-
changes over the past three centuries. Aggradation lanche debris cones (Fig. 1). Irrigation diversions
has dominated throughout most of the Holocene, were left abandoned by the channel incision. Lateral
including the Little Ice Age. However, degradation erosion led to undermining of bridges and streamside
roads (Fig. 2). In the channel, bed materials became
coarser and the alluvial water table was lowered,
* Corresponding author. Tel.: +1-405-744-9247; fax: +1-405-
744-7841.
which in turn led to destruction of habitat for a once
E-mail addresses: [email protected] (R.A. Marston), productive trout fishery (Fig. 3). The purpose of this
[email protected] (J.-P. Bravard). study was to determine the extent, timing, and causes

0169-555X/03/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0169-555X(03)00132-6
66 R.A. Marston et al. / Geomorphology 55 (2003) 65–74

Fig. 1. Incision in upper Malnant, halted by large blocks in bed material that exceed competence of stream. For scale, note person near lower
right in channel.

of channel degradation. The study has broader impli- mm, including heavy winter snows on the steep valley
cations for how changes in hillslope land use and sidewalls that frequently trigger avalanches. Precip-
geomorphic processes can reverse centuries-long itation is derived from oceanic rains and by the oro-
channel regimes from aggradation to degradation in graphic influence of the Aravis Ridge. The highest
mountain watersheds. monthly precipitation is f 190 mm in December, and
the lowest monthly precipitation is 120 mm in April.
Snow contributes in water retention with a thickness
2. Physical setting of the Fier and Malnant of 1.80 m at Thônes and 4.80 m in the Aravis Ridge
(Cholley, 1925).
The Malnant drainage covers 15.9 km2 within the The Malnant valley was excavated by glaciers
upper Fier watershed in the Bornes Range within the during the Quaternary and has been sculpted by slope
Haute Savoie of the French Alps (Fig. 4). The drain- processes since the Late Pleistocene, including mass
age is an anticlinal valley underlain by Oligocene movement from the western hogback below Couturier
limestone and sandstone that have been folded and Ridge, avalanches from the hogbacks surrounding
faulted by alpine tectonics (Figs. 5 and 6). The the valley, rockfalls from the southern end of the
watershed was glaciated during the Pleistocene and valley, and debris torrents in ravines originating above
the Little Ice Age. Valley glaciers deepened the fluvial 1500 m.
valleys, deposited till, and truncated large alluvial fans
(BRGM, 1990).
The Malnant is a 7.1-km-long tributary that drops 3. Geomorphic response to land use
from a headwater elevation of 1835 to 600 m where it
joins the Fier River, made famous as a favorite fly- 3.1. Mid-14th century to end of 19th century
fishing stream of Winston Churchill. The Malnant
contributes very little flow to the Fier but does Heavy flooding and snow avalanches from the
contribute the majority of bed load. Elevations in mid-14th century to the end of the 19th century
the Malnant range from 450 m in the valley to over created sizeable debris cones at the foot of valley
2000 m on the drainage divide. The watershed sidewalls. Large woody debris has been recovered
receives an average annual precipitation of 1500 from debris cones that yielded carbon dates of
 R.A. Marston et al. / Geomorphology 55 (2003) 65–74 67

Fig. 2. Undermining of roads (top) and bridges (bottom) by Malnant.

135 F 45 years. Severe floods were recorded in 1854, alba) and beech (Fagus silvatica) was widespread
1859, 1875, 1879, and 1899, all of which devastated from the 16th century through the early 18th century.
the hamlet of Montremont (Mougin, 1914). This hill- Forests were converted to pastures for summer graz-
slope instability may be attributed to deforestation that ing at elevations above 1500 m, to meadows for
provided wood as the primary fuel for residents of summer grazing between 1500 and 1000 m, and to
Montremont and was utilized by a local glass factory ploughed fields for cereal grains on the valley floor
in the village of Allex, active from 1755 to 1860. below 1000 m. A detailed land cover map dated 1732,
Deforestation of the original forest of white fir (Abies the so-called Sarde Map of Thônes Parish, was
68 R.A. Marston et al. / Geomorphology 55 (2003) 65–74

and that the Little Ice Age (mid-14th to the end of

the 19th centuries) was the most active period. This
was probably due to extensive forest degradation
performed since the Middle Ages, which was a
period of dense human occupation at high altitude
in a phase of climate warming until mid-14th century
(Ladurie, 1983). Rivers of the Alps experienced a
metamorphosis from meandering to braiding in
response to an increase of bed load, and this meta-
morphosis extended downstream from the inner Alps
to the piedmont areas surrounding the Alps (Bravard,
1989). However, the 19th century was not homoge-
neous since the existence of several low-flow-domi-
nated and flood-dominated periods alternated during
that century. This periodicity has been demonstrated
in the Pre-alps of Die, south of the study area. The
years 1800 – 1820, 1840 – 1842, 1850 – 1851, and
1856 – 1875 were particularly severe; and the excep-
tional flood of 1856 motivated the French Parliament
to enact the 1860’s law on montane reforestation
(Bravard, 2000).

3.2. Twentieth century

In response to the hillslope instability and associ-

ated problems for valley residents, the French govern-
ments purchased the upper 15% of the watershed
Fig. 3. Coarsening of bed material with progressive channel incision between the years of 1883 and 1903 and began a
in upper Malnant and concordant loss of aquatic habitat. program of reforestation and avalanche mitigation
(Fig. 5). On the upper slopes, structural walls were
installed and the slopes were terraced to retain snow.
studied by Moutard (1995). This map revealed that On the lower hillslopes, landowners constructed
forest cover had been reduced to 10% of the present- stonewalls and rock-filled wooden boxes to retain
day cover. avalanches before they could advance onto the valley
Mougin’s descriptions date back to a period when floor.
foresters attributed the supposed increased activity of Beginning in the 20th century, sediment supply
alpine torrents during the 19th century to human from hillslopes to the upper Malnant was reduced.
action. The French Revolution (1789 – 1799) was Without this sediment input, the Malnant began to
supposed to have increased individualism and to incise and bed load was mobilized. In our channel
have permitted an intense degradation of forests survey of the upper Malnant, the rotor from a
(Surell, 1841). This view was contested in the British plane that had crashed in the 1950s near
1920s by renowned geographers such as Lenoble the headwaters was found incorporated into the bed
(1923), who argued that forest degradation had load approximately 2 km downstream from the crash
occurred in earlier centuries. This view tended to site. Downcutting in the last three decades has
reduce blame that had been placed on the upland created severe problems for farmers in this small
peasantry. More recent works have demonstrated that watershed as bridges are undermined, streamside
the Alps have experienced several phases of roads are threatened, and irrigation diversion struc-
increased torrential activity during the Holocene tures are rendered unusable. Indeed, Malnant trans-
 R.A. Marston et al. / Geomorphology 55 (2003) 65–74 69

Fig. 4. Geologic setting of the upper Fier watershed. The Malnant enters the upper Fier from the south just downstream of Thones.

lates from French as ‘‘bad creek,’’ ‘‘. . .a native below a large block belonging to a layer of coarse
expression reflecting a secular wisdom’’ (Cholley, deposits. This alluvial layer was covered by large fir
1925). trees and riverine willows dating back to a period
when the torrent was not yet incised. Radiocarbon
dating provided the date of 135 F 45 BP. This date is
4. Field surveys of channel incision consistent with the period of increased torrential
activity described earlier.
4.1. The upper Malnant: above 820 m elevation Floods of the 20th century have triggered inci-
sion of the bed of the upper Malnant, and this
The steep upper Malnant begins at the conver- incision has exhumed large blocks. Recent activity
gence of a series of ravines and avalanche paths that of ravines, still active on the partly protected
collect water, snow, and sediment from the upper slopes, has delivered a large amount of small debris
limestone cliffs. The bed of the upper Malnant is (plates of Hauterivian marly limestones) that cov-
characterized by steep, rocky outcrops upstream of ered the coarse deposits with a 4-m-thick layer.
1100 m, by boulder and log steps between 1100 and This debris replenishes bed materials that had been
950 m, and is incised into till deposits between 950 lost by incision. Floods in the Malnant export a
and 850 m. In the section between 1100 and 950 m, portion of this material and move it downstream.
we computed that 18 out of 140 m (13%) of the drop This fine debris was excavated in the 1970s by a
in relief was controlled by log steps. The large contractor at the downstream limit of the State
woody debris originates from avalanche paths, and Forest (1020 m). This instream mining interrupted
this may reflect the increase in supplies of large the downstream transport and created a knickpoint,
woody debris following the success of afforestation which migrated upstream until blocked by large
programs. boulders. This excavation stopped 20 years ago,
At an altitude of f 950 m on the left bank of the but new deposition since that time has been rather
upper Malnant, we found a piece of wood preserved low.
70 R.A. Marston et al. / Geomorphology 55 (2003) 65–74

Fig. 5. Geomorphic map of the Malnant drainage, highlighting avalanche paths, mass movement, areas of afforestation, and the alluvial fan that
has developed where the Malnant joins the upper Fier River. Transect A – AVis shown as a geologic cross-section in Fig. 6.

4.2. The lower Malnant fan has been incised during the Holocene (undated
phase), and a new fan has been emplaced during the
The 4-km-long reach between 820 m and the Little Ice Age. The material of this late phase of the
confluence with the Fier River is comprised of gravel, fan is coarse (gravel and small boulders) if compared
cobbles, and boulders deposited during the high to the underlying material, which is composed of
energy floods of the Little Ice Age. The fining of small debris. Also, it is partly composed of urgonian
coarse sediment is linked to the reduction in bed boulders originating from the upper cliffs of the
slope, but steep tributaries delivered coarse debris watershed, while small debris of the lower layer
until the downstream end of the reach. Downstream originate from Hauterivian marly limestones with
of Bélossier Bridge (km 6.500), the Malnant River has few urgonian elements. Incision through coarse mate-
built a large alluvial fan since the Late Glacial. This rial was exacerbated when the upper layer was
 R.A. Marston et al. / Geomorphology 55 (2003) 65–74 71

Fig. 6. Geologic cross-section of the Malnant along A – AVtransect in Fig. 5.

removed and when the fine debris were reached promoted regressive erosion to the detriment of the
following bed degradation in the 1960s. Malnant River bed. Moreover, the tributaries of the
This reach has been recently degrading because of downstream reach of the Malnant have been effi-
the upstream reduction in sediment delivery and ciently controlled by the French government and are
because of gravel extraction upstream. This process almost no longer active. Degradation was measured
promoted a downstream progradation of erosion. Also on the Malnant from an altitude of 820 m (km 2.500)
gravel extraction downstream along the Fier River to the confluence of the Fier River (km 7.100 m, 595

Fig. 7. Rock spur (see arrow) suspended above the alluvial fan at mouth of Malnant, representing 7 m of incision since 1968.
72 R.A. Marston et al. / Geomorphology 55 (2003) 65–74

m). Degradation could be dated considering that bedrock outcrop in the Malnant channel bed provides
deflection devices made of concrete were installed at a local control on bedrock. Therefore, the incision
41 points along the lower 4.5 km of the Malnant in upstream of this point can be attributed directly to
1968 by the Agricultural Service of the Department of the reduction in hillslope supply of sediment. Also at
Savoy in order to trap sediment and to control strong this point, a large slope failure enters the Malnant
lateral erosion (Fig. 7). The spurs were seated into one from the valley wall to the west, resulting in local
bank and have not been successful in stopping inci- aggradation of f 1 m over a 175-m-long reach.
sion. Indeed, the rock spurs provided time markers Because the Malnant had been the dominant source
against which post-1968 incision could be measured. of bed material to the Fier and much of the sediment
We measured the change in elevation below the rock stored in the Malnant had been excavated, the Fier
spurs and the width of the stream at each point. By incised further since 1982, attaining a maximum of
considering the channel length between the rock 10 m. The pronounced knickpoint created in the Fier
spurs, it was possible to calculate the volume of bed by the mining had moved upstream to the town of
material that had been excavated (Fig. 8). Over Thônes. The knickpoint migration was stopped by
163,000 m3 have been exported from the lower 4.5 grade control structures and by paving the Fier with
km of the Malnant. This converts to a mean of 36 m3/ stones through the town.
m length of channel. Channel width varies from 4 to
18 m.
Aside from the reduction in sediment supply from 5. Discussion
hillslopes to the Malnant, two other processes have
contributed to incision. First, sediment has been The timing and causes of incision are summarized
mined at six points along the channel. Second, and in Fig. 9. During the Holocene, input of bed material
most important, the base level of the Malnant was to the upper Malnant was large but this material was
lowered and the channel has adjusted by incising. Up largely transported downstream, causing aggradation
to 7 m of bed material had been excavated from the in the lower Malnant Valley. During the Little Ice
Fier from 1918 to 1982 for the express purpose of Age, deforestation of the hillslopes overlooking the
reducing the frequency and magnitude of flooding on Malnant accelerated mass movement and snow ava-
the adjacent flood plain where commercial develop- lanches. Larger bed material was introduced and this
ment (a furniture manufacturing plant) was to be material causes aggradation throughout the upper
sited. The upstream progression of this incision ends and lower Malnant. By the middle portion of the
at a point 4.5 km from the headwaters where a 20th century, afforestation had reduced the input of
bed material to the upper Malnant. An episode of
incision began in the upper Malnant and simultane-
ously in the lower Malnant because of reduced sedi-
ment supply throughout the Fier drainage. In the late
20th century, instream mining in the upper Malnant
and in the Fier caused progressive and regressive
erosion, respectively.
The magnitude and rates of incision found in the
Fier and Malnant are comparable to those reported in
other studies. Instream mining of gravel in Cache
Creek, California, lowered the bed by 8 m (Wood-
ward-Clyde Consultants, 1976). Collins and Dunne
(1990) described how mining in the Russian River
near Healdsburg, California, in the 1950s and 1960s
led to 3 to 6 m of incision over 11 km of river.
Fig. 8. Profile of Malnant showing corresponding histogram of Kondolf and Swanson (1993) reported bridge damage
channel incision/aggradation. in Stony Creek, California, because of 5 m of bed
 R.A. Marston et al. / Geomorphology 55 (2003) 65–74 73

Fig. 9. Timing of channel changes in the Malnant.

material excavation. Rinaldi and Simon (1998) re- the upper Malnant watershed because of reforestation
ported how 1– 3 m of incision occurred in the upper and avalanche control measures. Second, gravel
Arno River because of gravel mining and construction extraction in the Fier lowered the base level for the
of two dams. Malnant and triggered channel bed adjustment that
Aggradation in the Malnant during the early Hol- progressed upstream 2.7 km. Without the supply of
ocene and Little Ice Age has been reversed in the 20th sediment from the upper Malnant, the lower reaches
century. The principle cause of incision is twofold. have experienced dramatic incision as they adjust to
First, sediment supply was reduced from hillslopes in the lower base level. These results, which combine
74 R.A. Marston et al. / Geomorphology 55 (2003) 65–74

anthropogenic and climatic factors, fit well with the California. Environmental Geology and Water Sciences 25,
results obtained by Landon et al. (1998) and Liebault 256 – 269.
Ladurie, E., 1983. Histoire du climat depuis l’An Mil. Flammarion,
et al. (1999). Paris. 254 pp.
Landon, N., Piégay, H., Bravard, J.-P., 1998. The Drôme River
incision (France): from assessment to management. Landscape
References Urban Plann. 43, 119 – 131.
Lenoble, F., 1923. La légende du déboisement des Alpes. Revue de
Géographie Alpine XI, 1 – 116.
Bravard, J.-P., 1989. La métamorphose des rivières des Alpes
Liebault, F., Clément, P., Piégay, H., Landon, N., 1999. Assessment
francßaises à la fin du Moyen-Age et à l’époque moderne. In:
of bedload delivery from tributaries: the Drôme River case,
Petit, F., Laurant, A., Pissart, A. (Eds.), Rivières: Formes, Pro-
France; Alpine, Arctic. Antarctic Res. 31, 108 – 117.
cessus, Milieu de Vie. Bulletin de la Société de Géographie de
Mougin, P., 1914. Les torrents de la Savoie. Bulletin de la Société
Liège, vol. 25, pp. 145 – 157.
d’Histoire naturelle de la Savoie, Grenoble (in 8j, 1251 pp.).
Bravard, J.-P., 2000. Le comportement hydromorphologique des
Moutard, R., 1995. L’évolution de la vulnérabilité à l’érosion des
cours d’eau au Petit Âge Glaciaire dans les Alpes francßaises et
pentes d’un bassin versant torrentiel, sous l’effet des événements
leurs piedmonts. 25èmes Journées scientifiques du GFHN, Meu-
climatiques et des interventions humaines (1730 – 1995). L’ex-
don, 28 – 29 Novembre 2000, pp. 105 – 110.
emple du Malnant dans les Bornes occidentales. Diplôme
Bureau de la Recharche Geologique et Miniere (BRGM), 1990.
d’Etudes Approfondies, vol. 1. Université J. Fourier, Grenoble.
Carte géologique Annecy-Ugine au 1/50 000e, Orléans.
49 pp.
Cholley, A., 1925. Les Préalpes de la Savoie (Genevois, Bauges) et
Rinaldi, M., Simon, A., 1998. Bed-level adjustments in the Arno
leur avant-pays. Etude de géographie régionale. A. Colin, Paris.
River, central Italy. Geomorphology 22, 57 – 71.
755 pp.
Surell, A., 1841. Etude sur les torrents des Hautes Alpes. Dunod,
Collins, B., Dunne, T., 1990. Fluvial geomorphology and river
Paris. 317 pp.
gravel mining: a guide for planners, case studies included. Spe-
Woodward-Clyde Consultants, 1976. Aggregate extraction in Yolo
cial Publication, vol. 98. California Division of Mines and Geol-
County: a study of impacts and management alternatives. Ag-
ogy, Sacramento, CA. 29 pp.
gregate Resources Advisory Committee, County of Yolo Plan-
Kondolf, G.M., Swanson, M.L., 1993. Channel adjustments to res-
ning Department.
ervoir construction and gravel extraction along Stony Creek,
 Geomorphology 55 (2003) 75 – 96
www.elsevier.com/locate/geomorph

Spatial variation in the power of mountain streams in the

Sangre de Cristo Mountains, New Mexico
Mark A. Fonstad *
Department of Geography, Southwest Texas State University, San Marcos, TX 78666-4616, USA

Received 1 February 2002; received in revised form 10 June 2002; accepted 10 March 2003

Abstract

The principle indicator of river energy expenditure, stream power, has a significant influence on many forms and process
attributes of the fluvial system, yet few basin-wide analyses of stream power variations have ever been conducted. Recent
studies hypothesize a peak in the mean stream power distribution in small (10 km2)- to intermediate (100 km2)-sized basins. To
test hypothetical stream power profiles in a high mountain setting, 129 cross-sections of stream networks within the Costilla
basin of northern New Mexico and southern Colorado were measured for channel form, local sediment conditions, and basin
characteristics. Geomorphic and hydrologic analysis of these river sites throughout the Costilla basin yielded evidence of
abundant local control over fluvial processes and forms. Within the basin, the spatial deviations of stream power from the
hypothetical patterns derived from hydraulic geometry, in some cases >200% deviation, match areas of specific geologic and
hydrogeologic control. As an alternative to traditional hydraulic descriptions of downstream channel form, a probabilistic
process – response model can incorporate local and basin-scale variables and allow more realistic feedback mechanisms than in
traditional regime theory. The probabilistic nature of this type of model also allows prediction of multiple modes of channel
adjustment, an ever-present challenge to extremal and physically based simulations.
D 2003 Elsevier Science B.V. All rights reserved.

Keywords: Stream power; Hydraulic geometry; Simultaneous-equation models; Rocky Mountains

1. Introduction This conversion is accompanied by the expenditure of

energy through geomorphic work on the river perim-
Throughout modern geomorphology, force-resist- eter, by expenditure as heat through external and
ance arguments have played a major role in explan- internal resistance, and by the transport of sediment.
ations of earth surface forms and processes (Lecce, The spatial patterns of this energy expenditure, or
1997). The energy that drives the fluvial system stream power, may be used to link stream energy and
begins as potential energy and is transformed into the spatial distribution of river forms.
kinetic energy during the process of downhill flow. With these patterns in mind, this paper focuses on
two principal research questions. How does stream
power vary geographically within a high mountain
* Tel.: +1-512-245-7809. stream basin? Also, if traditional hydraulic geometry
E-mail address: [email protected] (M.A. Fonstad). variables describing the stream power distribution do

0169-555X/03/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0169-555X(03)00133-8
76 M.A. Fonstad / Geomorphology 55 (2003) 75–96

not fully capture the geographic variability, are there ous-equation modeling allows a researcher to esti-
other variables and methods that can be used instead? mate some of these complex relationships, and is the
These questions are explored in the Costilla basin of basis for the process description of Costilla basin
northern New Mexico and southern Colorado, a streams.
mountain watershed with minimal human alteration
throughout its recent history.
In addition to the description of spatial power 2. Background
variation, quantitative systems modeling of rivers is
important for informed decision-making. Specifically, 2.1. The basis of stream power models
the ability of the policy maker to determine the effect
on a river system of various management alternatives Potential energy expenditure can be represented in
is of primary significance. Examples of this decision- many ways. Rhoads (1987) found the literature on this
making may be in river restoration plans, weighing subject confusing and inconsistent. That paper’s
management choices for dam operation, effects of semantic clarifications of stream power terminology
engineering structures in and around channels, or in are those used in this paper. Rhoads used the term
hypothesizing the stability of natural channel systems ‘‘stream power’’ only as a conceptual device, referring
to external influences such as climate change. From a to it as the time rate of energy expenditure as water
theoretical standpoint, modeling of river systems travels downstream in a channel. Because the actual
presents a difficult and complex challenge because energy expended is extremely difficult to measure,
physically based relationships between all channel various alternative measures substitute for the true
variables do not exist. The technique of simultane- stream power (Fig. 1).

Fig. 1. Channel dimensions referred to in the equations of stream power.

 M.A. Fonstad / Geomorphology 55 (2003) 75–96 77

Total stream power ( P) is the power over an entire Lawler (1992, 1995) found this quantity useful in
river reach or segment. It is defined as the integration describing particle entrainment (competence), as did
of stream powers in each infinitesimal stream length Costa (1983). Graf (1983), Magilligan (1992), and
(dx) for an entire reach Lecce (1997) found that spatial patterns of stream
Z power influenced the processes and extents of fluvial
P¼ qgQSe dx; ð1Þ erosion, transport, and sediment deposition. Lawler
x (1992) and Abernethy and Rutherfurd (1998) used
and its units are watts (W). In this equation, q is the stream power indices to describe patterns of bank
density of the water/sediment mixture, g is the accel- instability. More generally, Baker and Costa (1987),
eration due to gravity, Q is water discharge in cubic Magilligan (1992), and Wohl (2000) used stream
meters per second (cms), and Se is the energy slope of power as a measure of channel mobility thresholds.
the stream (usually approximated by either the water Knighton and Nanson (1993) related stream power
slope or the channel bed slope). The product of q and with planimetric channel pattern, and Rinaldo et
g is given the symbol c and is often assumed to be al. (1992) found mean stream power to be an
equal to 9100 N/m3. Researchers have rarely used important variable connecting evolving landscape
total stream power as a measure of energy expenditure networks and generating theoretical optimal channel
in a stream. networks.
A much more commonly used indicator of power Not all stream power is accounted for by the
is cross-sectional stream power (X), which is the variables in the above paragraphs. For instance, mean
power per unit length of a reach. Cross-sectional stream power may be lost through transverse energy
power is defined as currents such as in meander bends. Also, energy is
dissipated rapidly in changes from supercritical to
X ¼ qgQSe ; ð2Þ subcritical flow regimes that occur commonly in
where the units are watts per meter (W/m). The mountain step-pool systems. These energy losses are
importance of cross-sectional stream power was first only approximated and their importance implied in
recognized by Bagnold (1966); he used this measure most analyses of stream power variations.
in the estimation of sediment transport rates and in the The few field examinations of stream power have
prediction of stream capacity. Phillips (1989) also predicted downstream power trends through the use
found that cross-sectional stream power provided a of hydraulic geometry principles. For a specific
physically based measure of sediment transport capa- basin, the downstream increase in discharge is often
city. regarded as some power function of drainage area
Perhaps the most commonly used measure of energy such as
expenditure in stream channels is mean stream power
(x). In essence, mean stream power is the rate of energy Q ¼ kDAm ð5Þ
expenditure per unit area of the channel bed. Mean where Q is the discharge of some particular return
power is defined as interval event, and DA is the upstream drainage area
x ¼ qgRSe V ð3Þ in square kilometers (km2). The k and m coefficients
2 in this relationship can be gleaned from regional
in watts per square meter (W/m ). In this equation, R is
hydrology regression curves or distributed basin
the hydraulic radius of the channel, equal to A/P where
models. The power function also describes the down-
A is the cross-sectional area of the flow, and P is the
stream variation in width of a specified return
length of the wetted perimeter. Velocity (V) is the mean
interval event in the form
depth-averaged flow velocity in m/s. Mean power is
equivalent to X/W where W is channel width. It is also W ¼ aDAb ð6Þ
related to mean boundary shear stress (s) in N/m2 where
where the coefficients a and b in this function are
s ¼ qgRSe ; ð4Þ determined from empirical regression analysis of
multiplied by velocity (V). basin channel measurements.
78 M.A. Fonstad / Geomorphology 55 (2003) 75–96

In many hydraulic geometry equations, channel, should fragment these hypothesized energy distribu-
water, or energy slope is characterized as a power tions, even in a ‘‘natural’’ setting. This fragmentation
function of discharge in the form would disturb the expected downstream variation of
stream power and likely cause variations in channel
S ¼ gQ z ð7Þ geometry. Graf (1982, 1983), Magilligan (1992), and
which is the method espoused by Leopold and Mad- Lecce (1997) found lithologic variations to be the
dock (1953), Wolman (1955), Hack (1957), and cause of stream power variations. A smoothly varying
Miller (1958). However, exponential profiles of the energy expenditure ideal is implied in one of the more
form commonly used riparian vegetation theories, the river
continuum concept (Vannote et al., 1980). If this ideal
S ¼ aebL ð8Þ is a poor characterization of reality in a basin with
almost no human impacts, how well can it be applied
where a and b are coefficients estimated through more generally?
nonlinear regression and L is the distance downstream,
have more support from theoretical statements. Begin- 2.2. Statistical process – response models
ning with Sternberg (1875) and progressing through
Strahler (1952) and Morris and Williams (1997), The advanced statistical technique termed simul-
researchers have sought general physically based taneous-equation models addresses the principle that
explanations for the hypothesized exponential profile many variables in the fluvial system mutually inter-
of streams. The most recent models (Morris, 1993) act, and the specification of some as ‘‘independent’’
link models of sediment communition and sorting with and others as ‘‘dependent’’ will obscure some of the
the momentum equation for water flow to provide true process – response mechanisms in a river
theoretical grounding for the exponential profile (Rhoads, 1992). The causal structure of the varia-
hypothesis. bles in the model is defined by the investigator
With empirical estimates of discharge, slope, and before estimation of the coefficients, thus these
width through the hydraulic geometry functions, a models are often called ‘‘causal models.’’ A simul-
researcher can estimate downstream variations in taneous-equation model is a regression-like predic-
cross-sectional and mean stream power (Knighton, tion technique that not only has the advantage of
1999). With the hydraulic geometry equations esti- allowing mutual interactions between some varia-
mated for the basin, replacing the variables in the bles, but it also allows feedback throughout the
cross-sectional stream power equation (Eq. (2)) with specified system of equations (Berry, 1984). Addi-
those from the hydraulic geometry functions is simple, tionally, the forms of these models in principle
yielding a cross-sectional power estimate allow the inclusion of known physically based
X ¼ qgðkDAm ÞðaebDA Þ ð9Þ equations in place of statistically derived ones.
Finally, the simultaneous-equation models allow
for any desired location in the basin. Mean stream time-lagged variables and have the potential to
power is estimated in a similar fashion by dividing the describe nonlinear dynamics through time (Rhoads,
cross-sectional stream power prediction with Eq. (6), 1992).
yielding a new prediction equation The manner by which mutual interactions are
allowed in a statistical model is to specify some
x ¼ ðqgðkDAm ÞðaebDA ÞÞ=ðaDAb Þ ð10Þ
variables as completely independent, ‘‘exogenous,’’
for a particular return interval event. In both the and others as mutually adjusting, ‘‘endogenous.’’
cross-sectional and mean power cases, the down- First, normal multiple regression equations are devel-
stream power curves are nonlinear, peaking in mid- oped relating the exogenous variables to the endog-
basin reaches and then decaying into lower basins enous variables, such as the relationship between
(Fig. 2). bank type and both the width and particle sorting
A simple hypothesis driving this research is that of the channel. Then, multiple regression-like equa-
geologic changes, both lithologic and historical, tions are specified for each of the endogenous vari-
 M.A. Fonstad / Geomorphology 55 (2003) 75–96 79

Fig. 2. Example predicted downstream mean stream power curves (A) for Costilla basin experiencing an annual flow discharge ( Q1) versus a
2.33-year event ( Q2.33) with estimated channel dimension coefficients and Eq. (10); (B) for two hypothetical basins having different hydraulic
geometry trends experiencing the same annual flow discharge ( Q1).

ables that include some of the other endogenous Miller (1984) and Rhoads (1988) showed that sets of
variables. The solution of this series of equations simultaneous-equations can mimic process – response
after an external forcing mechanism requires numer- mechanisms of fluvial change. Miller’s (1984, 1991)
ical time steps after which changes in the endogenous simultaneous-equations model of Rubey’s adjustment
variables settle on new values. By invoking the hypothesis showed that channel form is more depend-
ergodic principle, investigators use spatial variance ent on the process of mutual adjustment of morpholo-
in fluvial systems to construct models of channel gic elements than is channel gradient. Rhoads (1988)
change through time. Although this is implicit in found that reasonable predictions in channel change of
bivariate hydraulic geometry and multivariate regres- semiarid mountain streams could be forecasted through
sion models of channel geometry, it is more explicit simultaneous-equation systems.
in the case of simultaneous-equation models. The use The average bankfull hydraulic geometry of allu-
of simultaneous-equation models to simulate fluvial vial channels can be defined by the simultaneous sol-
system change is an advance over bivariate and ution of nine governing or process equations (Hey,
multivariate regression because of the misspecifica- 1979). However, several of these governing equations
tion of these simpler techniques and the lack of are unknown or poorly constructed. A closed-form
feedback mechanisms between internal river varia- solution of channel geometry and changes to the
bles. These simpler models implicitly assume unidir- geometry continues to elude current mechanistic
ectional causation and thus are poor representations models. However, prediction of geometrical changes
of real-world causality structures (Hanushek and is possible through two other types of fluvial models:
Jackson, 1977). the statistical process –response models and change
80 M.A. Fonstad / Geomorphology 55 (2003) 75–96

models based on application of extremal hypotheses. the Costilla Massif, a physiographic area of extreme
Phillips (1991) found extremal hypotheses-based relief changes. The watershed spans an elevation
models to be inherently unstable and advocated stat- range of 2864– 3945 m. The mean annual temper-
istical analyses as a substitute. From a purely practical atures extend from 5 jC in the lower basin to 1.6 jC
standpoint, statistical systems models allow direct in the upper basin, and the mean annual precipitation
quantitative prediction, even when the underlying ranges from 63.5 cm at lower elevations to 88.9 cm in
physical principles are poorly known or constrained. the higher areas. Typically, snow covers the basin
Finally, in the situation where physically based mod- from November to June, with late spring snowmelt
els require new relationships, such as inclusion of the contributing to the largest runoff events. Some large
effect of bank vegetation, statistical models may floods may also be produced by late summer thunder-
empirically define which components are significant storms that are concentrated over the basin through
influences on channel change. This knowledge can orographic effects.
then be incorporated into future advances in physi- The bedrock underlying the Costilla basin is pri-
cally based models. marily Precambrian granite (Fig. 3). Tertiary intrusive
rocks border the basin in the NE. The Costilla Massif
itself is a Laramide-age thrust that compressed Pale-
3. Study area ozoic marine sediments on the eastern fringe of the
basin. The wide plateau at the lower end of the basin
Costilla basin is a high mountain watershed in the begins on the west by a large range-bounding normal
central Sangre de Cristo Mountains of New Mexico fault and by a smaller, concealed normal fault zone on
and Colorado (Fig. 3). It drains the alpine portion of the east.

Fig. 3. Location of Costilla basin study area, the sampled sites within it, and the basin geologic zones.
 M.A. Fonstad / Geomorphology 55 (2003) 75–96 81

Fig. 4. Predominant stream systems within Costilla basin: (A) step-pool, (B) pool and riffle, and (C) meander-glide.
82 M.A. Fonstad / Geomorphology 55 (2003) 75–96

The bedrock of the Costilla basin is covered in 4. Methods

many areas by late Tertiary and Quaternary units.
The lowest portion of the basin is filled by Quater- 4.1. Regional hydrology regressions
nary glacial outwash and overlying Holocene allu-
vium. The midbasin stream areas in Long Canyon, In the absence of gage records, regional hydrology
Costilla Creek #2, and Glacial Lakes valley are regression was used to predict flood magnitude and
controlled by ground and lateral moraines of Bull frequencies for the sites within the Costilla basin.
Lake and Pinedale age. In the very northernmost Nine selected high-elevation gages around the central
section of the basin, the middle elevations are Sangre de Cristo Mountains that had records of longer
topped by a mixture of colluvium and Quaternary than 20 years provided base data used in constructing
alluvium. Nearly all of the areas of glacial moraine the regional regression equations. This type of anal-
are also intermixed with zones of colluvium derived ysis first requires flood records of individual gages
from hillslope mass wasting. Small Neoglacial mor- within the region to be ranked and fitted by curves to
aines are spaced among high cirques in the upper solve for various return interval flood magnitudes.
Costilla Creek #2 basin. Then, these flood-magnitude-estimated sites are
The fluvial environments abutting high-angle entered into a regression-predicting flow magnitude
slopes in the upper elevations formed in young, typic for a given return interval flood based on the drainage
cryochrepts soils with little or no horizonation. Lower area and/or elevation of the individual gages. The
areas, particularly those areas formed into outwash chosen return interval events predicted were the 1-
and alluvium, are typic cryoboralls that have much year flow (mean annual flow), the 2-year flood, the
more soil horizonation. Local soil development is 2.33-year flow (mean annual flood), the 5-year flood,
often an effective measure of the amount of hillslope and the 10-year flood.
control over fluvial forms. Areas in the higher For Costilla basin, the upstream drainage area for
elevations with high slopes and poor soil develop- each cross-section site is known from GIS analysis of
ment contain stream systems with very large bed a digital elevation model of the basin. The calculated
load and no flood plains. This situation encourages flood discharges are probably too low in the mid-
the formation of step-pool systems in the upper ends basin, due to a more active, shallow ground-water
of Costilla Creek #2, Costilla Creek #1, and Long system; this effect is probably negligible during peak
Canyon (Fig. 4A). runoff periods, when the ground is often completely
When stream systems enter the glacial and col- saturated.
luvially dominated depositional zones of the mid-
basin, a dramatic shift occurs in the fluvial forms 4.2. Cross-sectional sampling
and processes. The very porous material of these
areas draws a percentage of the runoff into shallow To compare field measurements with the predicted
ground-water systems, reducing some of the peren- stream power curves, the field survey team made
nial flow in the streams. In addition, the down- channel form and material measurements at 129 sites
stream transition from hillslope boulders to glacial within Costilla basin. At each site, theodolite surveys
till and outwash reduces energy gradients and pro- allowed a reconstruction of the cross-sectional chan-
duces a shift from step-pool systems to pool and nel form perpendicular to the dominant direction of
riffle sequences (Fig. 4B). In the lower elevations of stream flow. Topographic maps from the U.S. Geo-
the basin, a mixture of low slopes, a large percent- logical Survey provided estimates of longitudinal
age of silts and clays in the bank material, and channel slope at each site. In areas of rapidly varied
diminished hillslope control of channel form con- flow, such as in step-pool zones, these slope estimates
tributes to the formation of classic meander sequen- will likely be incorrect, perhaps as much as several
ces (Fig. 4C). Energy expenditure in this zone is percent. However, on a basin-wide study such as this,
through elevation decrease and through transverse these errors should not destroy within-basin areal
energy gradients due to secondary currents in the comparisons, where trends from many cross-section
meander bends. sites tend to overshadow slope errors made at an
 M.A. Fonstad / Geomorphology 55 (2003) 75–96 83

individual site unless the error is more than perhaps ‘‘bankfull’’ depth of the channel in the field and
100% too high or low. through laboratory cross-section analysis also per-
Sediment surveys are necessary to solve some mitted calculation of hydraulic variables for this
equations of flow competence, flow resistance, bank level.
stability, and indices of downstream fluvial abrasion The most common method of delimiting the ‘‘bank-
and differential sorting. Because most of the fluvial full’’ level is to approximate this level as the elevation
systems in Costilla basin are bed-load dominated, of the adjacent flood plain. This method reflects the
Wolman’s (1954) method of sediment distribution historical development of fluvial ideas in lowland
measurement is applicable. In this method, each streams with substantial flood plains. In areas where
channel-bottom particle below a cross-sectional meas- there seems to be more than one flood plain level or that
uring tape is collected and measured, and the 84th no flood plain exists makes this method a poor indica-
percentile of these sizes is used in further sediment tor of the bankfull level. In these cases, researchers
calculations. have suggested other methods of bankfull delimitation,
including the average elevation of the highest surfaces
4.3. Hydraulic modeling of the channel bars, the height of the lower limit of
perennial vegetation, and the elevation at which the
The U.S. Forest Service’s WinXSPro software pack- width/depth ratio of the cross-section becomes a mini-
age (Grant et al., 1992) allowed flood parameters such mum. Williams (1978) scoured the literature for tech-
as discharge, velocity, shear stress, stream power, and niques for delimiting the ‘‘bankfull’’ level and found 11
criticality to be computed at each of the flood sites. separate techniques that were often inconsistent with
WinXSPro uses numerical integration to solve Man- each other. In this study, ‘‘bankfull’’ levels are equal to
ning’s equation in complex channels. It requires the the elevation of the major flood plain in the lower basin
input of channel topography, longitudinal slope of the and equal to the stage corresponding to a change in the
water, and some estimate of hydraulic roughness. The relation of cross-sectional area to top width in the
choice of hydraulic roughness estimation was Jarrett’s middle and upper basin zones. The latter technique
(1984) equation, because Marcus et al. (1992) found first requires that the cross-sectional surveys from the
this method to be the most accurate predictor of rough- field be entered into a computer and then visually
ness among 11 techniques in mountain streams. Jar- analyzed for the change in area to width relationship.
rett’s equation predicts Manning’s n as a function of A direct comparison of estimated bankfull dis-
slope and hydraulic radius. Although some influence of charges to discharges predicted through regional hy-
bed sediment size on the overall hydraulic roughness is drology regressions finds a wide variety of event
not accounted for by this technique, the equations allow magnitudes associated with the bankfull level (Fig.
particle measurements to be seen as independent from 5). The amount of scatter in bankfull discharges,
hydraulic variables. By including hydraulic radius in particularly in upper basin areas, makes the bankfull
the equation, the method also allows roughness to be level a poor indicator of a similar magnitude event.
essentially depth-dependent, an advance over many Flow variables and channel geometry observed at the
roughness equations. ‘‘bankfull’’ level are not directly comparable from site
To compute various flood parameters, WinXSPro to site. For these reasons, comparisons between sites
iterates Manning’s equation with flow increasing utilized return interval magnitudes from the regional
depth until the predicted cross-sectional discharge hydrology regressions rather than the bankfull dis-
equals the discharge predicted by the regional regres- charges.
sion equation. At this depth, solution of Manning’s
equation for other assorted hydraulic variables such 4.4. Conceptual process –response model and its
as velocity, shear stress, and mean stream power structural equivalent
becomes possible. Using the predicted discharges for
the Q1, Q2, Q2.33, Q5, and Q10 return interval events, The conceptual process– response model of fluvial
WinXSPro allowed solution of the hydraulic varia- change in the Costilla basin contains 11 variables,
bles for each magnitude flow. The recognition of the five of which are exogenous and six of which are
84 M.A. Fonstad / Geomorphology 55 (2003) 75–96

Fig. 5. Comparison of predicted ‘‘bankfull’’ discharges versus the discharges predicted through regional hydrology regression. The open circles
show the calculated discharge at the estimated bankfull level. The solid triangles and solid circles show the estimated 1-year and 2.33-year
discharges, respectively.

endogenous. The model is explicitly based on and a (0) indicates adjacent young Entisols. The final
Rhoads’s (1988) model system, with only minor exogenous variable, UXSMPOW, is the excess mean
alterations. This replication serves two purposes: to stream power in the next upstream section from the
test the general utility of Rhoads’s original model section under scrutiny. Excess mean stream power is
system and to be sure that the Costilla model is defined as the difference between the mean power
identified. Rhoads (1988) explicitly tested his model produced by the 1-year flow and the mean power
structure to assure identification. required to move the d80 particle at the upstream
The exogenous variables are composed of four section. This critical entrainment power is taken from
basin controls on the channel system, and a fifth is a threshold equation from Costa (1983). UXSMPOW
an indicator of dynamic sediment transport conditions is related to bed material transport.
from an upstream section through the next lower The endogenous variables in the system include
section. The first variable, AREA, is the upstream two variables of channel shape, two of particle
drainage area in square kilometers (km2). This value characteristics, and two of flow properties. Q is the
specifies the position of the cross-section within a estimated 1-year flow discharge in cms at the speci-
hydrologic network. RELIEF is the relief ratio above fied section. SHEAR is the shear stress produced by
a given section, where the ratio is equal to the the 1-year flow at the section, in N/m2. WIDTH is
subbasin relief divided by the distance of the section the top flow width, in meters, calculated for the 1-
downstream from the subbasin’s drainage divide, and year flow. SLOPE is the dimensionless rise/run
is unit dimensionless. BANK is a dummy variable that estimate from 1:24,000-scale maps. MEAN is the
indicates whether the channel section is embedded in intermediate axis length of the 80th percentile-sized
alluvial materials (0) or into colluvial, glacial, or semi particle within the section, in millimeters (mm); it is
bedrock boundaries (1). SOIL is a dummy variable a measure of the competence of the stream. Finally,
that characterizes the relative activity of the adjacent SORT is the degree of sorting of the bed particles,
channel boundary. For this variable, a (1) indicates the taken as the standard deviation of the intermediate
channel section is embedded within deep Mollisols, axis size distribution.
 M.A. Fonstad / Geomorphology 55 (2003) 75–96 85

The presence of thick, unconsolidated near-chan- Prestegaard (1983) found that resistance variables
nel materials that are often saturated to the point of and bed particle size correlated with slope in gravel-
standing water on the surface strongly suggests bed streams at bankfull stage. In examining spatial
there is an active, shallow, ground-water system distributions of rapids, Graf (1979b) found that
underlying the Costilla stream network; some stream these areas were not related to discharge changes
discharge is very likely lost through seepage in the but rather to the spatial arrangement of debris fans
mid-basin areas and regained in the lower basin. and sediment sources. In this way, fluvial processes
The rates of these losses and gains are hypothesized are locally controlled by sediments and channel
to be functions of the channel width. Therefore, boundary form and are not highly dependent on
discharge is a function of both AREA and WIDTH canyon-long processes. WIDTH is primarily a func-
in the structural model. SLOPE is both dependent tion of discharge variability, but also of SLOPE
on its position in the fluvial network (through through the energizing of bank failures during high-
competence, it is related to Q and WIDTH) and energy events. WIDTH is also related to bank
on the underlying particle distributions accounted stability, here represented by the BANK and MEAN
for by the MEAN variable and the RELIEF indices. variables.

Fig. 6. Process – response model for stream systems of Costilla basin: (A) conceptual model with hypothesized directions of causal influence,
and (B) structural model with endogenous and exogenous variables. In the conceptual model, arrows define the direction of hypothesized causal
influence, and the dashed box indicates the separation of exogenous from endogenous variables. A ‘‘ + ’’ or ‘‘  ’’ indicates the hypothesized
direction of adjustment. In the structural model, the symbols c and b refer to exogenous and endogenous relationships, respectively. The
numerical subscripts are the unique identifications for the relationships in the structural equations.
86 M.A. Fonstad / Geomorphology 55 (2003) 75–96

SHEAR at the 1-year flow level is a function of Finally, the MEAN and SORT variables are con-
the discharge and, by definition, the channel bed trolled by the adjacent bank materials and bank
slope, but also of the channel width through its activity; these quantities are represented in this
correspondence with changing water depth. Shear model by the BANK and SOIL variables. The
stress averaged throughout the water column may separation of mean particle size and particle sorting
not correlate highly with sediment transport or allows both to change during fluvial events. How-
channel geometry (Adenlof and Wohl, 1994), but ever, detailed thresholds of selective grain entrain-
it is a variable that is easy to estimate and has a ment and downstream particle sorting (Komar, 1987)
large theoretical significance. Lawler et al. (1997) are not present in this statistical model. Explicit
and Couperthwaite et al. (1998) found annual and spatial description of sediment size supply is an
event downstream bank erosion more related to effective indicator of erosion and deposition in a
spatial changes in discharge than in spatial changes river (Lane et al., 1996).
in shear stress. A simple shear stress quantity may The relationships between these system variables
not accurately describe bank erosion because the are pictured in Fig. 6A. One variable not included in
effective loosening of frost action in the banks may this investigation is the effect of large woody debris in
be more or less effective in certain areas of a river the channel system. Woody debris acts as additional
(Lawler, 1986, 1987, 1993). flow resistance and provides structural control on the

Fig. 7. Predicted versus observed mean annual flood stream power for all Costilla basin sites (A) for cross-sectional stream power and (B) for
mean stream power. For the mean annual flood ( Q2.33) return interval, the calculated coefficients for the power equations (Eqs. (9) and (10)) are
k = 0.0639, m = 0.8584, a = 0.092, and b = 0.0412, and the coefficients for the width power equation are a = 1.598 and b = 0.370.
 M.A. Fonstad / Geomorphology 55 (2003) 75–96 87

Fig. 8. Map of observed mean stream power trends in Costilla basin. In this method, graduated circles are used to represent the magnitude of the
stream power. The smallest circles represent 5 W/m2, and largest circles represent 200 W/m2.

movement of bed load throughout the steam (Adenlof coefficients in this model system appear in the struc-
and Wohl, 1994). However, quantification of woody tural process response model equations below and in
debris and its spatial arrangement is extremely diffi- Fig. 6B.
cult, especially in a large-area field setting.
The accumulated field and hydrology estimates Q ¼ a1 þ b13 WIDTH þ c11 AREA þ e1 ð11Þ
provided the input for a data matrix with 109 cases
(rows) and 11 variables (columns). All of the esti- SLOPE ¼ a2 þ b21 Q þ b23 WIDTH þ b25 MEAN
mates for the hypothesized relationships in Fig. 6A þ c22 RELIEF þ e2 ð12Þ
come from analysis of this matrix, although in prin-
ciple, some could come from theoretical arguments. WIDTH ¼ a3 þ b31 Q þ b32 SLOPE þ b35 MEAN
The postulated form of the governing structural þ c33 BANK þ e3 ð13Þ
equations is linear, but this linearity required the
initial log-normal transformation of all of the channel SHEAR ¼ a4 þ b41 Q þ b42 SLOPE
data. The result of this transformation was to normal-
ize the distribution of nearly all of the variables. The þ b43 WIDTH þ e4 ð14Þ
88 M.A. Fonstad / Geomorphology 55 (2003) 75–96

MEAN ¼ a5 þ b54 SLOPE þ c53 BANK mean power (Fig. 7B) against upstream drainage area,
and often compare these power values to those esti-
þ c54 UXSMPOW þ c55 SOIL þ e5 ð15Þ mated using the downstream hydraulic geometry.
Although they are the most frequently used technique
SORT ¼ a6 þ b64 SHEAR þ b65 MEAN þ c63 BANK for showing power variations, they may obscure the
þ c64 UXSMPOW þ c65 SOIL þ e6 ð16Þ actual downstream patterns (Lecce, 1997). Both of the
stream power lumped graphs display a large amount
of scatter in the downstream direction, and reasons for
5. Results these deviations are difficult to separate.
Rather than lumping all of the observations into
5.1. Downstream stream power trends graphs, it may be more appropriate to plot the value of
the stream power or the predicted/observed power
Graphs that group all sites in all subbasins, termed residual on a map of the basin. Although visualizing
‘‘lumped’’ graphs, display cross-sectional (Fig. 7A) or the absolute value of the power using this method is

Fig. 9. Map of predicted versus observed mean stream power residuals in Costilla basin. In the cases where the mean stream powers are
overpredicted, the graduated circles are colored black. In the cases where the mean powers are underpredicted, the circles are colored white. The
smallest circles have 0% deviation from the predicted hydraulic geometry trends, and the largest circles show 100% deviation from the predicted
trends.
 M.A. Fonstad / Geomorphology 55 (2003) 75–96 89

more difficult, the pattern of power or the power cedures, where the remaining variables are all signifi-
residuals becomes much clearer. The spatially distrib- cant at the 0.05 level. The weaker variables in some of
uted view of stream power displays a very heteroge- the constituent equations may be strengthened by
neous pattern with only a weakly defined downstream future adjustment of the model paths or the included
signal (Fig. 8). A stream power residual graph shows variables, but they also may be weak because of
a general distribution of power underestimation in the nonlinear relationships (in space and/or time) between
mass-wasting-dominated high elevation areas, over- variables that will not be simple to model in a system
estimation in the glacial-material-controlled mid- such as this simultaneous-equation equation model.
basin, and underestimation in the silt/clay-rich lower In the OLS model, SLOPE is strongly and neg-
basin (Fig. 9). atively related to Q and positively related to RELIEF,
Another indication that local conditions are inter- directions correctly postulated in the conceptual
rupting the downstream hydraulic geometry is the model. WIDTH and MEAN have a weaker relation-
explicit relationship between sediment sizes and ship with SLOPE, but their positive relationships are
stream power. If the channels of Costilla basin were in agreement with the conceptual model as well. In the
perfectly adjusted to the hydraulics within them, we case of WIDTH, only the variable Q has the direction
might expect a strong correlation between the mean hypothesized by the conceptual model, but the other
stream power and the size of the bed material. How- variables (SLOPE, MEAN, and BANK) all have very
ever, there is very little correlation between these weak correlations with WIDTH. There is a positive
variables (Fig. 10), suggesting that sediment at a site relationship between BANK predicting WIDTH, sug-
is at least to some substantial degree derived locally gesting that the colluvial and glacial materials pro-
and has not had a chance to be sorted or to undergo mote wider channels than the alluvial surfaces. While
fining through abrasion. This suggestion is strength- at first counter-intuitive, this relationship supports
ened by the initial observation that the particles are field evidence of narrow channels in the lower alluvial
often highly angular in shape. regions because of high silt/clay content of the bank
walls.
5.2. Structural equations model SHEAR exhibits all of the hypothesized directions
of response. As expected, Q is the most effective
Table 1 compares the coefficients for the model as variable in changing the shear stress, but WIDTH is
estimated through ordinary least squares (OLS) pro- also important by indirectly affecting the depth of the

Fig. 10. Relationship between Q1 mean stream power and channel bed d80 particle size for all Costilla basin sites. The coefficient of
determination (R2) of this fit is 0.078.
90 M.A. Fonstad / Geomorphology 55 (2003) 75–96

Table 1
Results of OLS statistical analysisa
Eq. (1) Eq. (4)
Q R2 = 0.925 SEE = 0.28 a = 0.05424 SHEAR R2 = 0.933 SEE = 0.18 a = 8.784
b b* SEE b b* SEE
Width  0.516  0.279 0.084 q 0.558 0.847 0.035
Area 1.492 1.175 0.057 Slope 0.973 1.341 0.033
Width  0.539  0.442 0.042

Eq. (2) Eq. (5)

2
SLOPE R = 0.727 SEE = 0.49 a =  5.206 MEAN R2 = 0.127 SEE = 0.33 a = 3.704
b b* SEE b b* SEE
q  0.585  0.644 0.104 Shear 0.159 0.306 0.056
Width 0.06348 0.038 0.116 Bank 0.04059 0.057 0.101
Mean 0.467 0.176 0.138 Uxsmpow  0.00311  0.224 0.001
Relief 0.396 0.205 0.199 Soil  0.0223  0.032 0.103

Eq. (3) Eq. (6)

2
WIDTH R = 0.445 SEE = 0.41 a = 1.305 SORT R2 = 0.598 SEE = 0.35 a =  1.504
b b* SEE b b* SEE
q 0.455 0.842 0.089 Shear 0.02607 0.032 0.062
Slope 0.04523 0.076 0.082 Mean 1.228 0.776 0.106
Mean 0.09423 0.060 0.14 Bank  0.0296  0.026 0.108
Bank 0.127 0.113 0.06
a
In this table, b is the unstandardized estimate, b* is the standardized estimate, and SEE is the standard error.

water. MEAN particle sizes also have the same An interesting aspect of the response is that the
directions of response as stated in the conceptual response curve oscillates in a more progressively
model, with section shear stress and upstream excess diminishing, or damped manner before settling on a
mean power being the most effective variables, new equilibrium (Fig. 11). The oscillations of the
respectively. The SORT index is primarily a function response variable to a system change may be nothing
of the particle size, suggesting that an increase in more than the numerical behavior of this system of
sorting will result from larger grain sizes. If particle equations. However, the oscillations may reflect some
size generally decreases in the downstream direction, realities in channel change. Graf (1977) proposed a
this result suggests that materials become less well ‘‘rate law’’ for describing change in geomorphic
sorted downstream. systems, suggesting that a system approaches a new
equilibrium state in a damped or asymptotic manner.
5.3. Effect of a hypothesized exogenous disturbance The damped oscillations that are results of the solution
of the simultaneous-equation model reflect the idea
If there is a successful estimation of the coeffi- behind the rate law. Also, if these numerical oscilla-
cients in the governing structural equations, this tions are a mirror of real fluvial behavior, then they
allows an investigator confidence to conduct ‘‘what also shed light on the formation of multiple terraces
if’’ scenarios on the channel system. For example, a during the system’s response to a single fluvial event.
channel in the lower basin may be subject to mass These change-induced oscillations may be a numer-
movement at its clay-rich sidewall. This event might ical analogue to downcutting and infilling sequences
bring a decrease in local sediment sizes. How will the of stream terrace formation. The idea, first identified
channel respond? by Schumm (1977), that a single fluvial event could
 M.A. Fonstad / Geomorphology 55 (2003) 75–96 91

Fig. 11. Modeled response of channel width to mean grain size decrease through time.

cause multiple state transitions is termed minor com- varying nor is its basis immediately recognizable.
plex response in relation to major complex response The uppermost reaches of Costilla basin have higher
because of large-scale tectonic or climatic forcing. than expected mean stream power values (Fig. 9). The
Minor terraces in rivers result from internal adjust- main reason for this is the underprediction of the
ments as dependent variables interact in fluvial sys- energy slope by the exponential slope profile. The
tems (Bull, 1990). Hey (1979) found similar erosional exponential model, although quite useful in the lower
and depositional oscillations using a qualitative proc- elevations, is a poor predictor of longitudinal knick-
ess – response model. If this numerical analogy reflects points and scarps because of glacial erosion and
complex response mechanisms, then the quantitative hillslope mass-wasting. Additionally, the channel
estimates of the oscillations should be comparable to widths in the upper basin are often overpredicted by
real-world observations. a power function, this is the result of direct hillslope
In a real river setting, the responses to external influence on lateral channel adjustments. This width
forcing mechanisms may diverge from the predicted overprediction leads to an underprediction of mean
responses due to deterministic but nonlinear effects, stream power along these segments. The forms cur-
poorly predicted stochastic noise, improper model rently present in the basin have very likely been there
specification, or inherent errors within the model for a long period of time. Aerial photos of the basin in
estimates. If the estimates of the structural equations the 1950s show very few planimetric changes in the
are poor or error-prone, such as in the prediction of stream systems of Costilla basin. Large discharges in
Costilla basin sediment sizes, the precision of the the late 1990s made very few recognizable changes in
response curves will become lower, and the estimated the channel forms, so these basin-wide patterns of
outcome of small forcing mechanisms will become stream power are not likely the result of seasonal or
more a function of the error rather than the actual episodic events.
response trend. A cursory look at Fig. 9 shows that the lowest areas
in the basin also have much higher than predicted
stream power values. Because this area has low
6. Discussion channel slopes, the reason for this pattern is not
immediately apparent. Some of the additional power
The spatial distribution of the rate of energy comes from additional surface water that resurfaces
expenditure in Costilla basin is neither smoothly from the shallow ground-water system at the ends of
92 M.A. Fonstad / Geomorphology 55 (2003) 75–96

Long Canyon and in the midsection of the lower patterns in these regions is that the surficial geology
basin. A more likely explanation for the power under- underlying these channels has very different mechan-
prediction is narrower than expected channels. ical properties from the other parts of the basin.
If banks become more cohesive (an increase in the In both regions with markedly underpredicted
silt/clay content of the banks or of the channel mean power values, large expanses of glacial and/or
perimeter) downstream, which could be associated glacio-fluvial materials underlie the stream environ-
with an increasing dominance of suspended over ments. In particular, the mid-basin of Costilla Creek
bed-load transport, width will increase more slowly #2 is underlain by a broad deposit of glacial and
and depth more rapidly than expected from the effect glacio-fluvial materials. The middle sections of Cos-
of discharge alone, giving rise to a more box-like tilla Creek #1 show the same pattern, although these
channel cross-section (Schumm, 1971; Knighton, areas are more dominated by Holocene alluviation. In
1998). This scenario is likely in the lower portion of Long Canyon, the mean stream power is not over-
Costilla basin, where the bank cohesion increases predicted by the basin equations. However, the
dramatically and walls often hold angles close to absence of large underpredictions in the middle-basin
vertical. Also, the geometry of the channels below zones, compared to the rest of Long Canyon, also
the bankfull level resembles a box more than a supports the correlation between glacial deposits and
triangle or parabola, shapes which are common in mean power overprediction.
the higher areas of the basin. The result of this change Why would mean stream power be overpredicted
in silt/clay percentage is to encourage narrower than in channels forming in glacial and glacio-fluvial
expected channels, leading to mean stream power terrains? The two explanations that best characterize
values that are higher than predicted by the basin this pattern are bank stability differences due to struc-
curves. tural properties in the local material and the removal
When the reservoir that impounds the Costilla of discharge into shallow ground-water systems. The
basin downstream of the study area was briefly middle-basin zones of the three principal valleys are
drained during 1983, engineers noted that there was gravel- and cobble-dominated channels with low silt/
almost no bed-load deposition in the reservoir; and clay contents. This decreases the bank cohesion and
even the deposition of fine silts and clays was minimal encourages wider than expected channels. Also, the
over a period of several decades (P. Mutz, New porous nature of the poorly sorted tills and outwash of
Mexico Interstate Stream Commission, personal com- these areas likely draws much of the surface water
munication, 2000). This observation supports the idea into shallow ground-water systems, reducing the dis-
that (i) suspended sediment dominates transport in the charge that would contribute to average mean stream
lowest reaches of Costilla basin; and (ii) the high power levels.
cohesion values in the channel banks are the result of While there are no direct in-situ measurements of
the silt/clay content, and they discourage bank erosion this hypothesized increase in ground-water flux, there
and sediment transport out of the basin on the decadal are several indirect indications that this processes is
scale. The high silt/clay content of the banks is very occurring. Many of the mid-basin zones have large
likely the product of both attrition of the larger hill- areas of wetland-like, near-channel environments, in
slope and glacial materials in the upper and middle many cases having such saturated soils that the sur-
zones and the selective transport of these fines during face is often covered with pools of standing water.
brief storms. Bankfull channel areas are considerably smaller in
The most striking pattern in Fig. 9 is the large these areas, again suggesting that stream discharges
overprediction of mean stream power in some of the are smaller in these middle zones. Reconnaissance
mid-basin valleys. These valleys do not have a con- measurements of active instream discharge also
sistent vegetation pattern that controls bank stability showed a decrease that requires greater ground-water
and channel widths. They likely do not have a differ- flux in the mid-basin areas to satisfy the conservation
ent channel disturbance regime, such as increased of water mass. This shallow ground-water re-emerges
hillslope mass wasting, from the lower basin. The into Costilla Creeks #1 and #2 in the lowest end of the
most obvious and direct explanation for the power basin.
 M.A. Fonstad / Geomorphology 55 (2003) 75–96 93

An imperfect but suitable analogy exists between stream to adjust to a minimum energy expenditure
the fragmentation of stream power by geologic tran- rate, then the question arises, why would a mountain
sitions and the effect of humans in fragmenting river basin have a curved mean stream power distribution
systems. Direct hillslope control of width adjustments rather than an equal expenditure rate throughout? This
in the high basins is analogous to artificial narrowing seeming paradox between the mountain power curves
of rivers through channelization. In the areas of and the minimum energy expenditure hypothesis may
Costilla basin with glacial and glacio-fluvial material be resolved by assuming that the river landscape is
controlling stream morphology, the channel system is dominated by the minimum energy principle over
analogous to unchannelized reaches. Also, these same long time spans (landscape evolution spans; Rinaldo
areas lose water in much the same way as would be et al., 1992) and perhaps short time spans (Knighton,
expected in areas with water diversions, and the loss 1998). However, the minimum energy principle might
of mean stream power may also be analogous to this not be a controlling mechanism during time lengths
human-induced situation. Because many mountain where contingent processes such as glacial advance
areas have a diverse surficial geology and a complex and accelerated mass wasting control the stream
geologic history, they too are likely to have natural processes and landforms.
fragmentation of their stream systems, both in energy In Costilla basin, the recognition that power does
expenditure rates and morphology. River fragmenta- not linearly increase downstream and that the peak
tion is not just a human consequence, and river ideas stream power distributions occur in different areas
stemming from the assumption that natural river land- depending on the magnitude of the fluvial event
scapes are not locally or even extensively fragmented allows for the nonlinear distribution of sediment in
may lead to errors in understanding and prediction. flood plains. In this view, the general decrease in
There exists, however, another possibility for the sediment delivery ratios downstream is not in contra-
spatial distribution of observed stream power devia- diction to a general downstream stream power in-
tions from the predictive curve. If the extremal hypo- crease suggested by the hydraulic geometry and by
thesis of minimum energy expenditure (Chang, 1980) Lecce (1997). Instead, the downstream power often
creates equilibrium channel forms at a landscape decreases throughout most of a basin, especially
scale, then changes in the downstream power curve during low-magnitude events. Additionally, irregular-
will adjust the channel to reduce the curve toward a ities in the structure of power distributions (such as in
straight line rather than the empirical curves predicted high elevation glaciated areas) remove power from the
from the observed stream data. In Costilla basin, this river and allow for deposition in upstream areas.
would occur by adjusting the power distributions to be Although no direct data support such a stream
less than predicted in the mid-basin areas and greater power/sediment delivery ratio correlation in Costilla
than predicted in the upper and lower basin (again see basin, this seems to be a reasonable explanation.
Fig. 9). Because the geologic regions in Costilla occur In light of the inability of strict hydraulic geometry
in the same general areas as the hypothesized stream functions to describe fluvial processes, additional
power adjustments, separating the two possibilities is variables and interactions can be represented through
quite difficult. simultaneous-equation models. However, two major
One possible refutation of the minimum energy limitations of using statistically analyzed process –
extremal hypothesis in this case is the general mean response models pose significant difficulties. The first
power heterogeneity over short (< 1 km) reaches. If is the absence of known fluvial thresholds, the most
the stream channels were really equilibrium forms and notable being the lack of competence thresholds in
supposing that spatial autocorrelation rules apply in the increase or decrease in sediment sizes during
channel systems, then power within these small dis- channel changes. This absence gives the user the
tances should have less variability than power distri- erroneous impression that internal change always
butions between separated basin zones, a possibility occurs even under small external control changes.
not supported by Figs. 7– 9. Other lacking thresholds include the failure of banks
If the patterns of stream power residuals (Fig. 9) at a specified angle, the meandering to braiding
result from geology rather than a tendency of the threshold (Van den Berg, 1995), the processes
94 M.A. Fonstad / Geomorphology 55 (2003) 75–96

occurring near regime shifts, the thresholds associ- with mutual adjustment, including the grain size
ated with vegetation stability (Hupp and Osterkamp, increase.
1996), and channel downcutting (Graf, 1979a). The success of statistical simultaneous-equation
Threshold issues can be even more problematic; models in forecasting single-strand channel change
Alexander and Bull (1981) and Buffington and suggests an extension that should allow transitions
Montgomery (1997) discussed the ways in which from single- to multiple-strand streams or visa-versa.
real sediment transport may not correlate with well- Currently, the field data required in building models
known, empirically based thresholds. of this new type do not exist. However, field data
The second major limitation is the absence of collection in areas of both single and multiple strands
direct temporal knowledge within the model. The or laboratory experiments should allow for rapid
assumption in using such models is that all of the development of this class of models.
internal variables mutually adjust at the same rate to Assuming that a change in a single channel can be
continuously maintain the dynamic equilibrium speci- understood as a simultaneous-equation system, then
fied in the structural equations. Not only are the individual channels within a multiple-strand stream
mutual adjustments regarded as changing at the same could be individually modeled with a separate simul-
rate, but that rate is also not known in real temporal taneous-equation system for each strand. The transi-
scales. The output is in numerical ‘‘time steps’’ that tion from single- to multiple-strand streams should be
may or may not mirror real-world time. represented as a discriminant function, such as the one
Statistical governing equations may be improved based on stream power as suggested by Van den Berg
by including nonlinear, higher-order terms. However, (1995). Multiple regression would then allow the
the feedback processes in a simultaneous-equations prediction of how many channels exist in the multi-
model may allow small nonlinear terms to create ple-strand stream. To decide whether a channel returns
dramatic numeric instability through time. This phe- to a single channel or remains in a multiple channel
nomenon is a statistical analogy to deterministic chaos state would require computing a set of average values
that may be present in real-world geomorphic systems for each of the stands in the multiple-strand stream.
(Malanson et al., 1992). However, the sensitivity to For instance, if the discrimination of single from
initial conditions makes the model extremely unstable multiple channels is a function of both mean stream
to any such nonlinear inclusions, and the model power and median bed material size, then the average
‘‘blows up.’’ Under what conditions a system with mean stream power and median particle size should be
nonlinear terms is stable and when it ‘‘blows up’’ is calculated for all of the strands.
not well understood.
Immediate possibilities exist for extending the
simultaneous-equation model approach. The first is 7. Conclusions
to incorporate known thresholds into the model,
whether these thresholds are known through obser- The primary research questions addressed in this
vation or through theoretical arguments. To do this, report were (i) how does stream power vary geograph-
the model is interconnected not only by the governing ically within a high mountain stream basin? And (ii) if
structural equations but also by logical hierarchies of traditional hydraulic geometry variables describing the
the discriminant threshold equations. For example, if stream power distribution do not fully capture the
an external change to the river system suggested an geographic variability, are there other variables and
increase in particle sizes, a logical comparison would methods that can be used instead? Analysis of
be invoked comparing the estimated shear stresses observed stream power in Costilla basin, as well as
with those necessary to move the particle sizes that mapping of stream-adjacent environmental settings,
the model says should move. If the threshold equation supports some general conclusions. First, stream
is not satisfied and the particle should not move, then power varies nonlinearly in the downstream direction,
the systems model is returned to a state that only and this nonlinearity is augmented by highly hetero-
transports the previous size particle. If the threshold geneous power patterns within subbasin zones. The
equation is satisfied, the model is allowed to proceed major deviations of observed power from predicted
 M.A. Fonstad / Geomorphology 55 (2003) 75–96 95

trends are likely the result of stream fragmentation by a subalpine stream of the Colorado Rocky Mountains. U.S.A.
geologic diversity. This type of fragmentation is par- Arctic and Alpine Research 26 (1), 77 – 85.
Alexander, D., Bull, W.B., 1981. Threshold of critical power in
tially analogous to human alteration of the fluvial streams: discussion and reply. GSA Bulletin 92, 310 – 312.
system. The position and magnitude of the power Bagnold, R.A., 1966. An Approach to the Sediment Transport
trends in Costilla basin change depending on the Problem from General Physics. USGS Professional Paper, vol.
magnitude of a fluvial event. These changes, along 422-J. Washington, DC.
Baker, V.R., Costa, J.E., 1987. Flood power. In: Mayer, L., Nash, D.
with the measured nonlinearity in spatial power pat-
(Eds.), Catastrophic Flooding. Allen and Unwin, Winchester,
terns, may allow the decrease in the sediment delivery pp. 1 – 21.
ratio in the downstream direction. The method of Berry, W.D., 1984. Nonrecursive Causal Models. Quantitative Ap-
mapping observed mean stream power versus pre- plications in the Social Sciences, vol. 37. Sage Publications,
dicted mean stream power residuals is a more powerful Newbury Park, CA. 95 pp.
way of finding areas of common stream behavior than Buffington, J.M., Montgomery, D.R., 1997. A systematic analysis
of eight decades of incipient motion studies, with special refer-
with the more commonly used lumped graph method. ence to gravel-bedded rivers. Water Resources Research 33 (8),
The observed deficiencies in spatial description of 1993 – 2029.
power through simple hydraulic geometry statements Bull, W.B., 1990. Stream-terrace genesis: implications for soil de-
were improved through solution of a structural equa- velopment. Geomorphology 3, 351 – 367.
tion model. All of the primary directions of calculated Chang, H.H., 1980. Geometry of gravel streams. Journal of the
Hydraulics Division. Proceedings of the American Society of
response were those hypothesized in the earlier con- Civil Engineers 106, 1443 – 1456.
ceptual model of channel change. In this respect, the Costa, J.E., 1983. Paleohydraulic reconstruction of flash-flood
qualitative predictions of Rhoads’s (1988) model are peaks from boulder deposits in the Colorado Front Range.
generalizable to the perennial streams of the southern GSA Bulletin 94, 986 – 1004.
Couperthwaite, J.S., Mitchell, S.B., West, J.R., Lawler, D.M., 1998.
Rocky Mountain region.
Cohesive sediment dynamics on an inter-tidal bank on the tidal
Trent, UK. Marine Pollution Bulletin 37 (3 – 7), 144 – 154.
Graf, W.L., 1977. The rate law in geomorphology. American Jour-
Acknowledgements nal of Science 277, 178 – 191.
Graf, W.L., 1979a. The development of montane arroyos and gul-
lies. Earth Surface Processes 4, 1 – 14.
This project was funded by three primary sources:
Graf, W.L., 1979b. Rapids in canyon rivers. Journal of Geology 87,
a National Science Foundation Doctoral Dissertation 533 – 551.
Improvement Grant, a Sigma Xi grant for data Graf, W.L., 1982. Spatial variation of fluvial processes in semi-arid
visualization, and the Vermejo Park Ranch riparian lands. In: Thorn, C.E. (Ed.), Space and Time in Geomorphol-
baseline project managed by David Vackar and Bob ogy. Allen and Unwin, Boston, MA, pp. 193 – 217.
Graf, W.L., 1983. Variability of sediment removal in a semiarid
Ohmart. Margaret Ferranti, Dustin Long, and Shadin
watershed. Water Resources Research 19, 643 – 652.
Dias provided invaluable field assistance during the Grant, G.E., Ducal, J.E., Koerper, G.J., Fogg, J.L., 1992. XSPRO: a
summer of 1998. Will Graf, Ron Dorn, Bob Balling, channel cross-section analyzer. U.S. Department of the Interior
and Ramon Arrowsmith offered critical and insightful Technical, 387.
discussions on the methods and implications of this Hack, J.T., 1957. Studies of Longitudinal Stream Profiles in Virgin-
ia and Maryland. USGS Professional Paper, vol. 294-B. Wash-
research. Karen Fonstad graciously produced several
ington, DC.
of the cartographic figures. Steve Walsh and an Hanushek, E.S., Jackson, J.E., 1977. Statistical Methods for Social
anonymous reviewer gave insightful comments that Scientists. Academic Press, New York.
significantly improved the manuscript. Hey, R.D., 1979. Dynamic process – response model of river chan-
nel development. Earth Surface Processes 4, 59 – 72.
Hupp, C.R., Osterkamp, W.R., 1996. Riparian vegetation and flu-
vial geomorphic processes. Geomorphology 14, 277 – 295.
References Jarrett, R.D., 1984. Hydraulics of high-gradient streams. Journal of
Hydraulic Engineering 110 (11), 1519 – 1539.
Abernethy, B., Rutherfurd, I.D., 1998. Where along a river’s length Knighton, A.D., 1998. Fluvial Forms And Processes: A New Per-
will vegetation most effectively stabilise stream banks? Geo- spective. Wiley, New York. 383 pp.
morphology 23, 55 – 75. Knighton, A.D., 1999. Downstream variation in stream power. Geo-
Adenlof, K.A., Wohl, E.E., 1994. Controls on bedload movement in morphology 29, 293 – 306.
96 M.A. Fonstad / Geomorphology 55 (2003) 75–96

Knighton, A.D., Nanson, G.C., 1993. Anastomosis and the contin- Miller, T.K., 1991. A model of stream channel adjustment: assess-
uum of channel pattern. Earth Surface Processes and Landforms ment of Rubey’s hypothesis. Journal of Geology 99, 699 – 710.
18, 613 – 625. Morris, P.H., 1993. Two-dimensional model for subaerial deposi-
Komar, P.D., 1987. Selective entrainment by a current from a bed of tion of mine tailings slurry. Transaction of the Institution of
mixed sizes: a reanalysis. Journal of Sedimentary Petrology 57 Mining and Metallurgy, Series A 102, 181 – 187.
(2), 203 – 211. Morris, P.H., Williams, D.J., 1997. Exponential longitudinal pro-
Lane, S.N., Richards, K.S., Chandler, J.H., 1996. Discharge and files of streams. Earth Surface Processes and Landforms 22,
sediment supply controls on erosion and deposition in a dynam- 143 – 163.
ic alluvial channel. Geomorphology 15, 1 – 15. Phillips, J.D., 1989. Fluvial sediment storage in wetlands. Water
Lawler, D.M., 1986. River bank erosion and the influence of frost: Resources Bulletin 25, 867 – 873.
as statistical examination. Transactions of the Institute of British Phillips, J.D., 1991. Multiple modes of adjustment in unstable river
Geographers 11, 227 – 242. channel cross-sections. Journal of Hydrology 123, 39 – 49.
Lawler, D.M., 1987. Bank erosion and frost action: an example Prestegaard, K.L., 1983. Variables influence water-surface slopes
from south Wales. In: Gardiner, V. (Ed.), International Geomor- in gravel-bed streams at bankfull stage. GSA Bulletin 94,
phology. Wiley, Chichester, UK, pp. 575 – 590. 673 – 678.
Lawler, D.M., 1992. Process dominance in bank erosion systems. Rhoads, B.L., 1987. Stream power terminology. Professional Geog-
In: Carling, P.A., Petts, G.E. (Eds.), Lowland Floodplain Riv- rapher 39 (2), 189 – 195.
ers: Geomorphological Perspectives. Wiley, New York, NY, Rhoads, B.L., 1988. Mutual adjustments between process and form
pp. 117 – 143. in a desert mountain fluvial system. Annals of the Association of
Lawler, D.M., 1993. Needle ice processes and sediment mobiliza- American Geographers 78 (2), 271 – 287.
tion on river banks: the River Ilston, West Glamorgan, UK. Rhoads, B.L., 1992. Statistical models of fluvial systems. Geomor-
Journal of Hydrology 150, 81 – 114. phology 5, 433 – 455.
Lawler, D.M., 1995. The impact of scale on the processes of chan- Rinaldo, A., Rodriguez-Iturbe, I., Rigon, R., Bras, R.L., Ijjasz-Vas-
nel-side sediment supply: a conceptual model. IAHS Publica- quez, E., Marani, A., 1992. Minimum energy and fractal struc-
tion, vol. 226. IAHS Press, Wallingford, UK, pp. 175 – 184. tures of drainage networks. Water Resources Research 28 (9),
Lawler, D.M., Couperthwaite, J.S., Bull, L.J., Harris, N.M., 1997. 2183 – 2195.
Bank erosion events and processes in the Upper Severn basin. Schumm, S.A., 1971. Fluvial geomorphology: the historical perspec-
Hydrology and Earth System Sciences 1 (3), 523 – 534. tive. In: Shen, H.W. (Ed.), River Mechanics, vol. I, pp. 4.1 – 4.30.
Lecce, S.A., 1997. Nonlinear downstream changes in stream power Schumm, S.A., 1977. Fluvial System. Wiley, New York.
on Wisconsin’s Blue River. Annals of the Association of Amer- Sternberg, H., 1875. Untersuchungen uber langen-und querprofil
ican Geographers 78 (3), 471 – 486. geschiebefuhrende flusse. Zeitschrift fur Bauwes 25, 483 – 506.
Leopold, L.B., Maddock, T., 1953. The Hydraulic Geometry of Strahler, A.N., 1952. Dynamic basis of geomorphology. GSA Bul-
Stream Channels and Some Physiographic Implications. USGS letin 63, 923 – 938.
Professional Paper, vol. 252. Washington, DC. Van den Berg, J.H., 1995. Prediction of alluvial channel pattern of
Magilligan, F.J., 1992. Thresholds and the spatial variability of flood perennial rivers. Geomorphology 12 (4), 259 – 279.
power during extreme floods. Geomorphology 5, 373 – 390. Vannote, R.L., Minshall, G.W., Cummins, K.W., Sedell, J.R., Cush-
Malanson, G.P., Butler, D.R., Georgakakos, K.P., 1992. Nonequili- ing, C.E., 1980. The river continuum concept. Candian Journal
brium geomorphic processes and deterministic chaos. Geomor- of Fish Aquatic Science 37, 130 – 137.
phology 5, 311 – 322. Williams, G.P., 1978. Bank-full discharge of rivers. Water Resour-
Marcus, W.A., Roberts, K., Harvey, L., Tackman, G., 1992. An ces Research 14 (6), 1141 – 1154.
evaluation of methods for estimating manning’s n in small Wohl, E., 2000. Mountain Rivers. American Geophysical Union,
mountain streams. Mountain Research and Development 12 Washington, DC. 320 pp.
(3), 227 – 239. Wolman, M.G., 1954. A method of sampling coarse river-bed ma-
Miller, J.P., 1958. High Mountain Streams: Effects of Geology on terial. Transaction of the American Geophysical Union 35 (6),
Channel Characteristics and Bed Material. New Mexico Bureau 951 – 956.
of Mines Memoir 4, Socorro. Wolman, M.G., 1955. The Natural Channel of Brandywine Creek,
Miller, T.K., 1984. A system model of stream-channel shape and Pennsylvania. USGS Professional Paper, vol. 271.
size. GSA Bulletin 95, 237 – 241.
 Geomorphology 55 (2003) 97 – 109
www.elsevier.com/locate/geomorph

Local scouring and morphological adjustments in steep channels

with check-dam sequences
Mario A. Lenzi *, Francesco Comiti
Department of Land and Agroforest Environment, University of Padova, Agripolis, Via Romea 16, 35020 Legnaro (PD), Padua, Italy
Received 6 December 2001; received in revised form 18 August 2002; accepted 10 March 2003

Abstract

This paper describes bed profile and grain size distribution adjustments in a mountain river (Maso di Spinelle River, Italian
Alps) stabilized by a sequence of boulder check-dams. The control works were originally designed to simulate the geometry of
natural step-pool channels, where tumbling flow is the dominant hydraulic regime. Local scouring downstream of 29 drop
structures is analysed through the use of nondimensional parameters where maximum scour depth and scour length are
normalised to the drop height. Prior laboratory data reveal a pattern similar to field scours, where complex interactions occur
between drop height, critical flow depth, and step spacing. The linkage between scour length and depth is also discussed. There
seems to be a maximum step height for impinging jets that is approximately twice the drop height; this maximum may explain
the upper limit of the steepness factor found in high-gradient step-pool streams. If such a maximum upper limit is confirmed by
further studies, this may aid designs of foundation heights for transverse control works in steep channels.
D 2003 Elsevier Science B.V. All rights reserved.

Keywords: Boulder check-dams; Step-pool; Scouring; Morphological adjustments; Mountain rivers; Italian Alps

1. Introduction slope) ideally representing a dynamic equilibrium

between bed scouring and aggradation. Different
High-gradient streams often exhibit a naturally methods evaluate such ultimate bed slopes, including
formed step-pool architecture, which likely represents empirical relationships (Woolhiser and Lenz, 1965;
self-adjustment of the stream towards higher bed Della Lucia and Fattorelli, 1981) and analytical
stability (Lenzi, 2001). Accordingly, staircase-like approaches based on incipient motion criteria (Gess-
sequences of check-dams or bed sills are widely used ler, 1970; Ferro and Giordano, 1988; Julien and
for channel bed stabilisation to inhibit incision. Wargadalam, 1995; Porto and Gessler, 1999; Gaudio
Traditionally, the principle governing prevention of et al., 2000).
channel erosion is the reduction of the actual longi- The ultimate channel slope has to be evaluated
tudinal channel slope S to a lower value Seq (ultimate because structure heights z and numbers n are
designed to provide a total vertical drop for the reach
length Lr equal to:
* Corresponding author. Tel.: +39-49-827-2675; fax: +39-49-
827-2686.
E-mail address: [email protected] (M.A. Lenzi). n
z ¼ ðS  Seq ÞLr ð1Þ

0169-555X/03/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0169-555X(03)00134-X
98 M.A. Lenzi, F. Comiti / Geomorphology 55 (2003) 97–109

Because number of works relates—inversely—to the metric course of the stream. One should also bear in
spacing L between them, one design parameter limits mind local changes in gradient, widening, and sec-
all others. tions that are most suitable for lateral keying (Lenzi
In spite of its theoretical importance, equilibrium and D’Agostino, 1998). Similarly, in assigning
slope itself is a very ‘‘tricky’’ concept because of the wavelength, moving within a certain range of varia-
complexity of the bed load rate – water discharge tion for L is better than attempting to keep a fixed
relationship with regard to different return intervals. distance between works. It is worth noting that the
Moreover, nonuniform grain size distributions make use of Eq. (2) for determining the distance between
reliable determination of Seq difficult because analyt- the transverse works implies that the bed between
ical approaches require the evaluation of the critical two steps—without considering the scour hole—
shear stress. must be at least flat or with a more or less reverse
However, when spacing between transverse works slope.
is made short enough, an ultimate slope as commonly Taking into account the existence of a scour depth,
intended disappears (Comiti et al., 2001; Lenzi et al., the bedslope downstream of the scour hole approx-
2003) and flow dynamics tends to resemble natural imates zero if pool length is roughly greater than two-
step-pool hydraulics. The regular hydraulic pattern thirds of the distance L. In this case, jet diffusion
established in step-pool systems, referred to as tum- downstream of each drop dominates flow hydrody-
bling flow (Whittaker, 1987; Grant, 1997), is charac- namics (Comiti et al., 2001; Lenzi et al., 2003). Such a
terised by an energetic dissipation largely due to spill geometry is nevertheless the actual step-pool config-
resistance generated by diffusion of impinging jets uration, and the concept of an ‘‘equilibrium’’ slope no
causing roller eddies (Wohl, 2000; Wohl and Thomp- longer has meaning.
son, 2000). The risk of work’s failure because of undermining
In many high-gradient streams (S = 0.05 –0.20), a must be carefully taken into account in both tradi-
step-pool morphology may represent an ideal ecolog- tional control works and in geomorphologically based
ical objective. Therefore, geometric features associ- interventions described above. Undermining is com-
ated with step-pools can guide designs of boulder- monly caused by local scouring because of the
made transverse works. Lenzi (2002) describes the impinging jet arising from each drop structure (Mason
types of stream restoration measures in high-gradient and Arumugam, 1985; Bormann and Julien, 1991;
streams. In accordance with the largest grain sizes in Stein and Julien, 1994).
the stream (D90, D100), the structure height H includ- Laboratory studies (Comiti et al., 2001; Lenzi et
ing the foundations can be calculated, maintaining an al., 2002, 2003) previously established main equili-
average relation H/D90 = 2 and keeping it within the brium (i.e., long-term) dimensions of local scour-
range H/D90 = 1 –4 (Lenzi, 1999; Lenzi et al., 2000). ings at bed sills in steep channels. Because field
The distance L between works or artificial steps data were totally lacking, a field survey has been
can be estimated by applying the maximum flow carried out in a case-study river in order to validate
resistance criterion, as determined by Abrahams et flume results.
al. (1995) in the laboratory. These authors have This paper summarises results obtained from field
observed that on a step-pool reach, the condition of measurements to address the following topics:
lowest mean flow velocity (and therefore of maximum
global resistance to flow) occurs when there is a (i) evaluation of morphological and sedimentolog-
regular spacing and the following geometric condition ical adjustments in a case-study river, where
holds: boulder check-dams stabilised the streambed;
(ii) analysis of maximum depth and length of scour
H=L holes within a sequence of transverse structures,
1V V2 ð2Þ
S in relation to the geometrical and hydraulic
characteristics of the system;
The criterion, however, should be applied in a (iii) comparison of field data with laboratory results;
flexible way, trying to follow altimetric and plani- and
 M.A. Lenzi, F. Comiti / Geomorphology 55 (2003) 97–109 99

(iv) geomorphological implications for natural step- Maso di Spinelle’s bed, especially the middle –
pool channels. lower part, experienced considerable incision and
deepening during the flood of 3 –4 November 1966
with a return interval 100 < Tr < 200 years. The aver-
2. Study area, field survey, and laboratory tests age bed width increased from 6– 7 to 35 –50 m after
the flood, while the bed profile decreased by about
2.1. Field case-study description 10– 15 m in some reaches. Similar results, although
lower in magnitude, accompanied the flood event of 2
In 1996 and 1997, the Torrent Control Agency October 1993 (50 < Tr < 75 years).
(TCA) of the Autonomous Province of Trento (Italy), After the 1993 flood, further incision was pre-
with the scientific support of the University of Padua, vented by using boulder check-dams that should
carried out hydraulic control works on selected rea- reproduce natural step-pool structures. TCA con-
ches of a mountain river, the Maso di Spinelle, using structed both uncemented boulder check-dams and
transverse boulder works to protect channel against boulder check-dams strengthened with cement.
incision. The Maso di Spinelle forms, along with the The stability of the interventions described above
Maso di Calamento, the Maso River, a tributary that was actually tested during the final stages of their
flows into the Brenta River (NE Italy, Fig. 1). For a construction, in October 1996, when a flood with a
complete description of the basin, see D’Agostino et maximum discharge around 43 m3 s 1 and a return
al. (1997), Lenzi et al. (2000), or Lenzi (2002). time of f 10 years occurred. On 7 October 1998,
Degraded deposits of Quaternary age dominate the another flood with a peak discharge of 52 m3 s 1 and
channel, particularly in the lower reach. Talus often a return interval of 20 –25 years put the completed
covers moraine and fluvioglacial deposits close to the control works through more rigorous testing. The
banks. Moraine deposits lead to large boulders of behaviour of both check-dams strengthened with
various shapes intermixing with a high percentile of concrete and uncemented boulder steps was satisfac-
sandy and sandy-gravelly materials on stream banks. tory during both of these floods.
Some of these boulders, whose volumes range to Detailed field surveys were carried out in Septem-
hundreds of cubic metres, caused huge flow diversions ber 2000 along two reaches stabilised with cemented
during 1966 and 1993 floods, leading to bank slides boulder check-dams separated by a block stone ramp.
(Sonda, 1998). The channel width in both rests between 20 and 25 m,
with an average of 23.5 m.
The upstream 300-m-long reach (R1, Fig. 2a)
presents a mean gradient of about S = 0.125, whereas
the shorter—90 m—downstream reach (R2, Fig. 2b)
is somewhat steeper (S = 0.159). A total station-posi-
tioning system was used to measure thalweg points in
order to generate a longitudinal profile of both scour
holes and intervening bed spans (Fig. 3a and b).
Twenty-two cemented boulder check-dams were
installed along reach R1, with spacing ranging from
9.2 to 23.8 m; 220 points were surveyed, and 21 scour
holes analysed. As for the shorter reach R2, 72 points
were measured; spacing between check-dams ranged
from 5.3 to 18.8 m, and eight scour holes were
considered.
For each scour hole, the drop height, maximum
scour depth, and scour length were calculated from
Fig. 1. Location map and watershed divide of the Maso di Spinelle the profiles. Scour length is defined as the distance
River. between the upstream step and a typical ‘‘cusp’’
100 M.A. Lenzi, F. Comiti / Geomorphology 55 (2003) 97–109

Fig. 2. Photos of the surveyed reaches in the Maso di Spinelle River: (a) Reach R1, featuring a mean gradient of 0.125; the closest artificial step
is the No. 16 (see profile in Fig. 3a). (b) Reach R2, featuring a mean gradient of 0.159; the closest artificial step in No. 28 (see profile in Fig. 3b).

downstream. In fact, a well-defined linear berm of The actual water level difference responsible for
coarse clasts (0.7 –0.8 m in diameter) characterised the scour development (i.e., the air-surrounded verti-
the downstream end of pools below artificial steps. cal distance) is reasonably assumed to correspond to
Such a berm likely represents the zone where roller’s the drop height zs. This neglects the fact that water
turbulent stress due to jet diffusion abruptly calms depth at the pool end (i.e., the control depth down-
down and hydrodynamics tends towards a linear flow. stream) might differ from the nearly critical water
Moreover, the berm likely corresponds to the sedi- depth that is achieved at the step’s crest. However,
ment mound reported by authors describing local such an assumption was already proved in laboratory
scouring below free overfall (Breusers and Raudkivi, tests (Gaudio et al., 2000; Lenzi et al., 2002).
1991) and to flow-separation berms downstream of Water discharge can be assumed to be equal in the
hydraulic jumps described by Allen (1984) and Carl- two reaches because there are no tributaries from the
ing (1995). When the distance between two check- valley sides. For an assessment of the actual dis-
dams was short compared to the pool length, berms charge that shaped the scour holes, see Section 2.2.
were not so conspicuous. As for the critical flow depth, which is a determinant
Table 1 summarises scour hole characteristics. parameter in local scouring (Comiti et al., 2001),
Maximum scour depth H refers to the difference in interestingly, the channel width in R2 is slightly nar-
height between the check-dam crest and the deepest rower; and therefore this reach is characterised by a
point in the pool; Z is the difference in height between slightly larger flow depth than R1. This might partly
one check-dam and one downstream; zs is the actual justify the fact that most R2 points displace a bit
drop height with respect to the pool end; ls is the scour higher (i.e., larger nondimensional depth and length,
length, and L is the distance between the upstream and Fig. 6a and b).
the downstream check-dams (Fig. 4). Before construction of control works, surface bed
In some cases, zs and Z are equal because the material was sampled at three sites with a modified
spacing L is so short that the bed affected by the scour version (Lenzi, 1992) of the ‘‘grid-by-number’’ me-
hole almost extended to the downstream check-dam. thod (Wolman, 1954). An averaged grain size distri-
Otherwise, the two vertical parameters differ because bution was obtained to use it—appropriately scaled—
an ‘‘equilibrium’’ slope establishes downstream of the in the laboratory tests described in Section 2.2.
scour hole. The value of such a slope, calculated from In September 2001, two other samplings were
the profile, appears to be rather constant in R1 (around performed following the same methodology. The
0.04– 0.05). number of clasts measured was 133 in R1 and 95 in
 M.A. Lenzi, F. Comiti / Geomorphology 55 (2003) 97–109 101

alluvium reproduced in scale the surface mixture of

the Maso di Spinelle River. It was highly nonuni-
form, featuring a geometric standard deviation rg ¼
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
D84 =D16 ¼ 5:38 and ranging from coarse sand to
small cobbles. Wooden sills were inserted into the bed
to simulate stable steps; two spacing values were tested
(1 and 0.5 m) for all three initial slopes.
Maximum scour depth and scour length were
measured at equilibrium conditions. In Series 1 tests
(spacing 1 m), an equilibrium slope Seq was attained
downstream of the scour hole; and this was evaluated
from the bed profiles. Once the equilibrium slope was
obtained, the drop height was calculated. For Series 2
(spacing 0.5 m), the distance between the sills was too
short to permit the development of a uniform reach,
and thus the drop heights were calculated by the
profiles. A complete description of the laboratory

Table 1
Local scouring dimensions in the Maso di Spinelle River
Reach No. H (m) zs (m) Z (m) ls (m) L (m)
R1 1 2.53 1.86 1.86 6.20 10.10
2 2.09 1.41 1.41 5.90 14.30
3 1.57 1.20 1.20 5.00 11.55
4 2.43 1.30 1.30 7.15 20.85
5 3.10 1.84 1.84 6.80 9.65
6 2.35 1.10 1.10 7.95 17.95
7 2.05 1.16 1.16 5.45 12.25
Fig. 3. Longitudinal profiles of the surveyed reaches: (a) reach R1 8 2.47 1.09 1.09 6.20 20.50
after construction (1997) and present (2001); (b) reach R2 at present 9 2.66 2.14 2.14 4.40 9.60
(2001). The numbers identify the location of the 29 analysed scour 10 3.17 2.50 2.50 7.30 13.20
holes. 11 2.17 1.38 1.38 11.80 16.50
12 2.80 1.81 1.81 6.60 10.75
13 2.48 1.49 1.49 5.35 15.05
14 1.72 0.79 0.79 4.00 12.95
R2. The sample grid did not cover pool sites. How- 15 2.08 1.20 1.20 4.60 12.85
ever, the pool bottom resulted in being mainly arm- 16 2.05 0.86 0.86 5.80 23.80
ored by boulders ranging from 0.8 to 1 m. All grain 17 2.20 1.42 1.42 6.45 11.55
size distributions are shown in Fig. 5. 18 2.90 1.46 1.46 7.30 21.80
19 2.73 1.85 1.85 7.40 12.30
20 3.80 2.32 2.32 7.80 11.40
2.2. Laboratory tests 21 3.46 2.46 2.46 6.20 9.20
R2 22 3.60 2.56 2.56 5.00 8.00
The experimental research was carried out prior to 23 3.61 2.13 2.13 7.40 18.85
the field survey at HR Wallingford (UK) in the 24 2.99 2.12 2.12 5.40 8.55
25 2.25 0.97 0.97 4.60 8.00
Sloping Sediment Duct Facility. Three different initial
26 2.44 2.12 2.12 3.80 13.20
slopes of 0.078, 0.114, and 0.148 were adopted, 27 2.05 1.10 1.10 5.50 9.05
testing a wide range of flow rates for each slope value 28 2.14 1.38 1.38 5.70 7.30
without sediment feeding (clear water runs) for a total 29 1.01 0.47 0.47 3.10 5.30
of 26 runs. The grain size distribution of the bed For the symbols see text in Section 2.1 or Fig. 4.
102 M.A. Lenzi, F. Comiti / Geomorphology 55 (2003) 97–109

Fig. 4. Definition sketch of the measured variables for local scouring dimensions. The flow is from left to right, and the boulder check-dams are
shaded in grey. H is the maximum scour depth, ls is the scour length, zs is the drop height, Z is the elevation drop between two adjacent check-
dams. On the left, the situation where the works are closely spaced and zs corresponds to Z is showed; on the right, where the spacing increases,
the drop height zs is smaller than the drop Z.

experiments can be found in Comiti (1999), Comiti et clear relationships both for the nondimensional scour
al. (2001), and Lenzi et al. (2003). length and depth. Nonetheless, two different nondi-
In order to analyse the laboratory and field data mensional lengths and depths were actually observed,
together, the scour hole dimensions measured in the depending on the spacing L, in certain ranges of such
Maso di Spinelle River were assumed to be caused by ratio.
the October 1998 flood because no other events have
occurred along the channel since. For that flood event a
peak discharge of 52 m3 s 1 was estimated (return 3. Results and discussion
period of 20 – 25 years), which implies (using the aver-
age channel width for both R1 and R2) a critical water 3.1. Evolution of bed morphology and surface grain
depth hc at the steps’ crests of about 0.8 m. This para- size distribution
meter is needed because it was shown (Comiti et al.,
2001; Lenzi et al., 2003) that the ratio between the The control works were constructed implementing
critical water depth and the drop height (hc/zs; a sort of a (H/L)/S ratio of 1.1– 1.3, thus a weak reverse bed
‘‘relative depth’’ to the drop dimension) highlights slope was assured between most boulder check-dams,
as shown in Fig. 3a (1997 profile). First, this section
compares the changes in the bed profile that occurred
from 1997 to 2001 in the reach R1.
Where artificial steps are long-spaced (for example
Nj8, 13, 16), local scouring formed a well-defined
hole, deepening the former bed level. Downstream of
scour holes, prevalent depositional dynamics (fav-
oured by the large channel width, too) raised the level
and formed a sloping bed with a gradient of around
0.04 – 0.05. In the shorter spacings, however, erosive
processes have prevailed, lowering the bed only in
connection with scour hole development. At some
locations, the pool depth has not increased signifi-
cantly, indicating that initial geometry was already in
a general equilibrium with later flow rates. Berms
described in Section 2.2 are clearly distinguishable in
profiles as abrupt changes (a sort of ‘‘cusp’’) in bed
inclination at the end of pools.
Fig. 5. Surface grain size distributions of the Maso di Spinelle: Although not shown by the profiles, lateral and
before control works (A, B, C) and at present (R1 and R2). median bars occurred mostly in the long-spaced spans,
 M.A. Lenzi, F. Comiti / Geomorphology 55 (2003) 97–109 103

as can be seen in Fig. 2a. In such cases, water at low

discharge flows within one or two narrower channels.
Step-pool beds formed naturally along most of these;
see particularly Nos. 4 and 14. Median bar formation
was probably triggered by upstream step irregularities
that caused differential jet flow concentration, in turn
affecting local flow rates below the drop.
The original profile for reach R2 is not available.
However, because of short distances between steps, an
evolution similar to R1 short-spaced steps likely
occurred. R2 reach lacked bars.
As for the surface grain size distribution, Fig. 5
shows a large decrease in particle size in the upper and
median parts of the curve. The general reduction has
not developed for the finest material (up to D20, rough-
ly) because most of the large boulders found in the
Maso River bed had been used for building the artificial
steps. Thus, the previous coarsest clasts with diameters
up to 4 m (Table 2) are now ‘‘grouped’’ as boulder
check-dams. In sum, the ‘‘shift’’ of grain size distribu-
tion cannot be seen as a natural evolution pattern.
The fining of bed material contains important
consequences for enhancing sediment transport, if
hydraulic conditions remain unaltered. On the other
hand, lower bed slopes between steps downstream of
the pools should reduce flow competence, counter-
acting this fining effect.

3.2. Local scouring characteristics

In Fig. 6a, ratios H/zs (hereafter called relative scour

depths) plot against the ratio zs/L between the drop
height zs and the spacing L. The parameter zs/L roughly Fig. 6. Nondimensional dimensions of scour holes plotted versus the
measures the potential energy lost by the falling jet in ratio zs/L: (a) maximum scour depth; (b) scour length.
the span length fixed by two check-dams. When this
parameter increases, the relative scour depth is reduced, declining from values around 2.5 to about 1.3. The
initial general decreasing trend up to zs/L c 0.1 seems
Table 2 to reach an asymptote at an H/zs value of about 1.4. The
Some characteristic grain sizes of past (A, B, C) and present (R1, residual scatter is probably due in part to the well-
R2) surface bed material of the Maso di Spinelle River
graded nature of the grain size distribution. This easily
A B C R1 R2 leads to larger boulders occasionally being present at
(1996) (1996) (1996) (2001) (2001)
the pool bottom. Uneven planimetry of step crests—
Dmax (mm)a 4050 2990 2005 800 1000 causing higher jet flow concentration—might be an
D90 (mm)b 1783 1448 1218 304 274
D50 (mm)b 338 416 430 111 68
additional source of variability.
D16 (mm)b 48.5 37 90 37 23 The same pattern is shown by relative scour
a
Dmax is the largest boulder size. lengths, ls/zs (Fig. 6b). Here, the declining trend
b
D90, D50, and D16 are the diameters for which the 90%, 50%, accentuates; the ratio moves from 8 to 2 without
and 16% of the particles are finer, respectively. asymptotic behaviour.
104 M.A. Lenzi, F. Comiti / Geomorphology 55 (2003) 97–109

The scour holes that plot outside of the points’ accelerates due to gravity. Second, it determines the
clouds in both Fig. 6a and b (Nos. 9, 11, 26) are impinging angle, which in turn influences scour hole
noteworthy. These originated by those check-dams dimensions (Rajaratnam, 1981; Bormann and Julien,
that resemble (because of the positioning of boulders) 1991; Stein and Julien, 1994). Further studies (Little
a short, steep ramp where flow energy dissipates and and Murphey, 1982; Wu and Rajaratnam, 1998)
deflects more irregularly than in the other steps. In addressed the existence of different jet regimes at
these places scour depth and length are no longer drops, which depend essentially on the relative impor-
determined only by the parameter zs/L. Complex tance between the drop height and flow depth. Highly
energy dissipation patterns make it difficult to analyse differing flow fields downstream of the drop corre-
these local scourings. As for scour No. 25 in Fig. 6a, spond at different regimes. Finally, jet aeration (shown
no clear explanations for its far larger relative depth to be involved in scouring effectiveness) is also
are available. It might, however, be due either to a determined by drop height (Mason and Arumugam,
somewhat different local flow field (e.g., caused by 1985; Mason, 1989).
the pool’s boulders) or to an upstream flow concen- As for the distance L between the steps, its con-
tration that determined a thicker free jet. tribution to scouring dynamics seems fairly complex
In Fig. 7, the mutual relationship between the scour and still poorly understood. For bed sills (i.e., struc-
depth and length is displayed where both dimensions tures with negligible initial heights over the channel
are normalised to the drop height. A consistent, bed), spacing primarily determines the dimension of
seemingly linear trend might exist; this highlights the drop height. In fact, Gaudio et al. (2000) proposed
the tight relationship between vertical and longitudi-
nal dimensions of scour holes that will be addressed in zs ¼ ðS  Seq ÞL ð3Þ
Section 3.3.
Considering the rather consistent results described where S and Seq stand for the initial and the equili-
above, parameter zs/L might synthesise both the role brium channel slope, respectively. In check-dam sys-
of drop height on jet characteristics and the impor- tems, spacing only partly contributes to drop height,
tance of spacing on scouring interference. Drop height which is mostly due to the initial structure elevation.
affects local scouring by free jets in various ways. Nevertheless, after events with major erosive or dep-
First, it represents the distance along which the jet ositional dynamics, the drop may be altered substan-
tially, even if only for a period of time.
However, the laboratory results presented here
showed that when the spacing L is short enough the
scouring dynamics becomes heavily affected by the
presence of the downstream sill. A type of ‘‘interfer-
ence’’ onset renders the scouring process less effective
(Comiti et al., 2001; Lenzi et al., 2003). The ‘‘inter-
fering scourings’’ displayed much reduced maximum
depths and scour lengths. Such ‘‘inhibited’’ scours
were characterised by ratios ls/L greater than about 0.4
and never larger than 0.67 even for very high flow
rates, showing a roughly asymptotic behaviour. The
reduction of scour lengths was also tentatively
inferred to be the primary controlling factor on shal-
lower scour depths.

3.3. Comparison with laboratory results

Fig. 7. Relationship between nondimensional maximum scour depth Nondimensional maximum scour depths and
and length; a consistent trend is apparent. lengths are plotted in Fig. 8a and b, respectively.
 M.A. Lenzi, F. Comiti / Geomorphology 55 (2003) 97–109 105

description for ‘‘noninterference’’ conditions given

in previously cited papers (Comiti et al., 2001; Lenzi
et al., 2003).
A clear distinction between ‘‘noninterfering’’ and
‘‘interfering’’ conditions was not possible even for the
flume tests for the lowest hc/zs values ( < 0.5– 0.6). At
these small ratios, falling jets tend to become more
vertical with smaller scour lengths. Thus, even shorter
spacings are required for the onset of interference
effects. Field data having large relative depth (>1) are
lacking because only scour hole No. 29 is available.
Nevertheless, such a point apparently confirms the
existence in nature of a mechanism that inhibits local
scouring because of downstream control exerted by a
nonerodible sill point.
For the flume data, a unique relationship between
the parameter hc/L and the ratio ls/L was discovered to
exist. Field measurements in Fig. 9 show a much
larger scatter that may be due to the influence of drop
height, perhaps negligible in the laboratory research.
Moreover, field data present an upper value for ls/L of
about 0.7 –0.8, whereas flume points appear to tend
toward an asymptote around 0.6 –0.7 (Fig. 9).
After a certain stage, the energy available might be
enough to break such upper limits of ratios ls/L, and
this could be linked to the threshold between stable
and unstable tumbling flow as reported by Whittaker

Fig. 8. Nondimensional dimensions of scour holes (both field and

flume data) plotted versus relative critical flow depth (hc/z): (a)
maximum scour depth; (b) scour length. The overlap between field
and Series 1 flume points is rather good.

The independent parameter is the ratio hc/zs (Comiti et

al., 2001; Lenzi et al., 2003). A general overlap does
exist which indicates the soundness of the nondimen-
sional approach.
Several field points fall along the trend delineated
by Series 1 laboratory tests, which were referred to as
‘‘noninterfering scours’’. Almost all of these points
indicate scour holes that are characterised by a long Fig. 9. Scour length as a proportion of the spacing L in relation to
spacing and, therefore, by an equilibrium slope down- the ratio hc/L; field data generally plot above flume points, and
stream (see profile in Fig. 3a). This matches the present a larger scatter.
106 M.A. Lenzi, F. Comiti / Geomorphology 55 (2003) 97–109

(1987). In Whittaker (1987) flume experiments, unsta- Therefore, the idea that an intrinsic geometry of
ble tumbling flow (i.e., no longer well-defined scour scour holes might be a basic aspect in local scouring
holes, flow instabilities, slug waves) takes place at a (Franke, 1968; Whittaker, 1987; Gaudio et al., 2000;
value of hc/L between 0.11 and 0.14 depending on the Lenzi et al., in review) seems to be valuable and might
gradient (from 0.098 to 0.248). Some of this paper’s prove to be important in the understanding of fluvial
flume data fall within Whittaker’s range without forms that are connected to diffusing jets such as step-
showing signs of instability. The much finer material pool structures.
used by Whittaker (D90 = 0.0049 m) might be the
reason for the unstable tumbling flow he observed. 3.4. Geomorphological implications for step-pool
As for the scour holes’ geometry, Lenzi et al. channels
(2003) have already shown for the laboratory test—
both interference and noninterference conditions—the In analysing step-pool bed forms, Abrahams et al.
consistence of the ratio ls/(s + h), where s is the (1995) first introduced the parameter c=(H/L)/S, called
‘‘residual’’ scour depth calculated as H  zs and h is the steepness factor, where H is step height, L step
tailwater depth above the scour hole. A precise wavelength, and S reach gradient. They showed that
evaluation of tailwater depths in the field was very parameter c ranged between 1 and 2, with an average
difficult, so the critical water depth is assumed in Fig. value of 1.5, in all surveyed step-pool streams. Abra-
10 as a rough estimation, just to permit a comparison hams et al. also carried out laboratory tests with fixed
between flume and Maso di Spinelle data. The agree- weirs and found that maximum flow resistance (i.e.,
ment is rather good and indicates that most scours least mean velocity) occurs for a regular configuration
have ratios ranging between 3 and 4.5, with no of steps having values of c within the same range.
influence from the geometry of the system summar- They then concluded that step-pool geometry max-
ised by the zs/L parameter. The scatter may be due imizes resistance to flow and therefore leads to a more
differences between actual tailwater and critical depth, stable bed morphology.
which is slightly variable for each scour hole depend- Lenzi (2001) analysed the bed evolution in the Rio
ing on the downstream control. For flume data only, a Cordon, a small creek with step-pool morphology in
ratio of f 3 was found using the tailwater depth. the eastern Alps. Lenzi’s results indicated that max-
imum flow resistance conditions were gradually
reached after a series of ordinary flood events. Thus,
for real step-pool channels, maximum flow resistance
extremal hypothesis can be interpreted as the final
stage of a process of progressive bed adjustment
mainly due to pool scouring and bed armouring. In
the Rio Cordon, an extraordinary flood with unlimited
sediment supply caused a sudden decrease in the
steepness factor to 0.79 as a result of sediment burial
and increased step wavelength. The scouring pro-
cesses within pools during several subsequent ordi-
nary floods raised the parameter c back to a value of
1.33.
The upper limit for the steepness factor may be
related to local scouring maximum depths. Fig. 11
shows the ratio H/Z measured in the Maso di
Spinelle plotted against ls/L; the drop Z is the
difference in level between two artificial steps (see
Fig. 10. Geometry of the scour hole (with s = H  zs) as an invariant sketch in Fig. 4).
of the system; the tailwater depth (not known for field data) is The former parameter corresponds to the steepness
approximated by the critical depth. factor when a mean channel gradient S is converted
 M.A. Lenzi, F. Comiti / Geomorphology 55 (2003) 97–109 107

and those of Chartrand and Whiting (2000, Fig. 6)

showed a clear trend indicating that for gradients
larger than 0.07 –0.08 the ratios H/L fall within the
range 1 –2. At these slopes—and at ordinary flood
stages responsible for most pool scouring—a general
passage from a surface to an impinging jet regime
might be possible. The jet regime threshold should be
linked to the combination of higher flow depths and
smaller coarse particle sizes that determine the drop
heights (Chin, 1999). Such a finding may also be
useful, if confirmed by further studies, as a guideline
for designing foundation heights for transverse control
works in steep channels.

4. Conclusions
Fig. 11. The steepness factor c = H/Z plotted versus the ratio ls/L; the
range (1 < c < 2) proposed by Abrahams et al. (1995) for natural
Local scouring characteristics (maximum depth
step-pool channels is highlighted.
and length) downstream of 29 drop structures in a
steep mountain river were analysed and compared to
into a specific ratio between drop and length for each
flume data obtained by experiments in the same
step. In fact,
nonuniform sediment mixture appropriately scaled.
H=L H=L H The results are consistent and indicate that drop
c¼ ¼ ¼ : ð4Þ height, flow depth, and step spacing affect scouring
S Z=L Z
dynamics in a complex way, confirming at least in
The graph shows that only two scour holes present H/ part results previously obtained in laboratory tests.
Z slightly >2, whereas the average value is 1.47, very At least two nondimensional parameters turned out
close to the 1.5 claimed by Abrahams et al. (1995). to be involved: the ratio of critical flow depth to step
The water discharge that scoured the Maso di spacing and the ratio of drop height to step spacing.
Spinelle’s pools was pretty large (estimated return However, one more parameter is probably needed to
period 20 – 25 years), but one could wonder about account for different grain size distributions. The
the effects of a still heavier flow rate. As long as geometry of scour holes was found to be approxi-
unstable tumbling flow is not triggered (see Section mately invariant.
3.3), scour length maintains an upper limit fixed by An explanation for the steepness factor’s upper
the distance L, for whatever flow rate. From Fig. 11, value, around 2, measured in high-gradient (>0.08)
scour depth H can be inferred to not increase beyond a step-pool streams has been proposed referring to the
value of roughly two times the drop height zs = Z evidence of a maximum scouring depth occurring
when ls/L = 1. within the impinging jet regime. At lower slopes,
Such a pattern might be valid only for a tumbling step-pool geometry might be characterised by non-
flow geometry corresponding to the impinging jet dimensionally deeper pools because of a possible
regime that generated our data. For surface jet higher scouring efficiency occurring at surface jet
regimes, in contrast, the overall hydrodynamics differs regimes.
substantially (Little and Murphey, 1982; Wu and
Rajaratnam, 1998) and such an altered flow field
could lead to higher nondimensional scour depths. Acknowledgements
This is consistent with the step-pool channel’s steep-
ness factor, which is rarely higher than 2 for steep Field activity was supported partly by the
channels (over about 0.07– 0.08). Chin’s (1999) data University of Padova, Research Project 2001 –2002
108 M.A. Lenzi, F. Comiti / Geomorphology 55 (2003) 97–109

‘‘Hydraulic control and restoration of mountain Ferro, V., Giordano, G., 1988. Formulazioni e campi di applica-
streams according to morphological and environ- zione del criterio del moto incipiente per la deduzione della
penzenza di equilibrio di un corso d’acqua. Idrotecnica 3,
mental criteria’’, and partly by the European Union, 253 – 262 (in Italian).
‘‘Damocles’’ (Debrisfall Assessment in Mountain Franke, P.G., 1968. Erosione a valle di traverse fluviali. L’Energia
Catchments for Local End-users) project, contract elettrica 6, 445 – 446 (in Italian).
no. EVG1-CT-1999-00007. Funding for laboratory Gaudio, R., Marion, A., Bovolin, V., 2000. Morphological effects
of bed sills in degrading rivers. Journal of Hydraulic Research,
experiments was provided by the Commission of the
IAHR 38 (2), 89 – 96.
European Communities as part of the Programme Gessler, J., 1970. Self stabilization tendencies of alluvial channels.
‘‘Training and Mobility of Researchers-Access to Journal of the Waterways, Harbors and Coastal Engineering
Large Scale Facilities’’ under contract no. ERB- Division, ASCE 99, 235 – 249.
FMGE-CT95-0082. The authors thank Mr. Luca Mao Grant, G., 1997. Critical flow constrains flow hydraulics in mobile-
and Mr. Javier Loza Lozano for helping in field bed streams: a new hypothesis. Water Resources Research 33 (2),
349 – 358.
surveys and in elaborating the longitudinal profiles. Julien, P.Y., Wargadalam, J., 1995. Alluvial channel geometry:
Ms. Katherine Ackerley is also acknowledged for the theory and applications. Journal of Hydraulic Engineering,
English text revision. ASCE 121 (4), 312 – 325.
Lenzi, M.A., 1992. Campionamento ed analisi di materiale d’alveo
con componenti grossolane. Il bacino attrezzato del Rio Cor-
don. Quaderni di Ricerca n. 13. Regione Veneto, Venezia, Ita-
References lia, pp. 159 – 178. In Italian.
Lenzi, M.A., 1999. Morfologı́a y estabilidad de las secuencias en
Abrahams, A.D., Li, G., Atkinson, J.F., 1995. Step-pool streams: escalones en los torrentes alpinos de elevada pendiente. Ingen-
adjustment to maximum flow resistance. Water Resources Re- ierı́a del Agua 6, 151 – 162 (in Spanish).
search 31 (10), 2593 – 2602. Lenzi, M.A., 2001. Step-pool evolution in the Rio Cordon,
Allen, J.R.L., 1984. Sedimentary Structures: Their Character and Northeastern Italy. Earth Surface Processes and Landforms
Physical Basis, Vol. 1, Development in Sedimentology 30A. 26, 991 – 1008.
Elsevier, Amsterdam. Lenzi, M.A., 2002. Stream bed stabilization using boulder check
Bormann, N., Julien, P.Y., 1991. Scour downstream of grade-control dams that mimic step-pool morphology features in Northern
structures. Journal of Hydraulic Engineering 117 (5), 579 – 594. Italy. Geomorphology 45, 243 – 260.
Breusers, H.N.C., Raudkivi, A.J., 1991. Scouring. Hydraulic Struc- Lenzi, M.A., D’Agostino, V., 1998. Dinamica dei torrenti con mor-
tures Design Manual, IAHR. A.A. Balkema, Rotterdam, The fologia a gradinata e interventi di sistemazione dell’alveo. Qua-
Netherlands. derni di Idronomia Montana 17, 37 – 56 (in Italian).
Carling, P.A., 1995. Flow-separation berms downstream of a Lenzi, M.A., D’Agostino, V., Sonda, D., 2000. Ricostruzione Mor-
hydraulic jump in a bedrock channel. Geomorphology 11, fologica e Recupero Ambientale dei Torrenti. Ed. Bios, Cosenza,
245 – 253. Italy. In Italian.
Chartrand, S.M., Whiting, P.J., 2000. Alluvial architecture in head- Lenzi, M.A., Marion, A., Comiti, F., Gaudio, R., 2002. Local scour-
water streams with special emphasis on step-pool topography. ing in low and high gradient streams at bed sills. Journal of
Earth Surface Processes and Landforms 25, 583 – 600. Hydraulic Research, IAHR 40 (6), 731 – 739.
Chin, A., 1999. The morphologic structure of step-pools in moun- Lenzi, M.A., Marion, A., Comiti, F., 2003. Interference processes
tain streams. Geomorphology 27, 191 – 204. on scouring at bed sills. Earth Surface Processes and Landforms
Comiti, F., 1999. Studio sperimentale dei processi di scavo a valle 28, 99 – 110.
di soglie di fondo. MS thesis, University of Padova, Italy. In Little, W.C., Murphey, J.B., 1982. Model study of low drop grade
Italian. control structures. Journal of Hydraulic Engineering, ASCE 108
Comiti, F., Lenzi, M.A., Marion, A., 2001. Interference of erosion- (10), 1132 – 1146.
control structure in mountain streams. Proc. 3rd International Mason, P.J., 1989. Effects of air entrainment on plunge pool scour.
Symposium on Environmental Hydraulics, ISEH, Tempe, AZ, Journal of Hydraulic Engineering, ASCE 115 (3), 385 – 399.
USA. Mason, P.J., Arumugam, K., 1985. Free jet scour below dams and
D’Agostino, V., Cerato, M., Da Re, F., Lenzi, M.A., 1997. La flip buckets. Journal of Hydraulic Engineering, ASCE 111 (2),
sistemazione idraulica dei torrenti con briglie in massi. Dendro- 220 – 235.
natura 2, 39 – 52 (in Italian). Porto, P., Gessler, J., 1999. Ultimate bed slope in Calabrian streams
Della Lucia, D., Fattorelli, S., 1981. Nuovo Metodo per la Stima upstream of check dams: field study. Journal of Hydraulic En-
della Pendenza dopo la Sistemazione dei Torrenti del Trentino. gineering, ASCE 125 (12), 1231 – 1242.
Proc. Inter. Conf, Problemi idraulici nell’assetto territoriale della Rajaratnam, N., 1981. Erosion by plane turbulent jets. Journal of
montagna, vol. F. Istituto di Idraulica Agraria, Universita’ degli Hydraulic Research, IAHR 19 (4), 339 – 358.
Studi di Milano, Milano, Italia, pp. 1 – 13. In Italian. Sonda, D., 1998. Applicazione di criteri morfologici nella sistema-
 M.A. Lenzi, F. Comiti / Geomorphology 55 (2003) 97–109 109

zione dei torrenti: il caso della Val Campelle (TN). MS thesis, small step-pool channel. Earth Surface Processes and Landforms
University of Padova, Italy. In Italian. 25, 353 – 367.
Stein, O.R., Julien, P.Y., 1994. Sediment concentration below free Wolman, M.G., 1954. A method of sampling coarse river-bed ma-
overfall. Journal of Hydraulic Engineering, ASCE 120 (9), terial. Transactions-American Geophysical Union 35, 951 – 956.
1044 – 1053. Woolhiser, D.A., Lenz, A.T., 1965. Channel gradient above gully-
Whittaker, J.G., 1987. Sediment transport in step-pool streams. In: control structures. Journal of Hydraulic Engineering, ASCE 91
Thorne, C.R., Bathurst, J.C., Hey, R.D. (Eds.), Sediment Trans- (3), 165 – 187.
port in Gravel Bed Rivers. Wiley, New York, pp. 545 – 570. Wu, S., Rajaratnam, N., 1998. Impinging jet and surface flow re-
Wohl, E.E., 2000. Mountain rivers. Water Resource Monograph 14. gimes at drops. Journal of Hydraulic Research, IAHR 36 (1),
Washington DC, USA, p. 320. 69 – 74.
Wohl, E.E., Thompson, D.M., 2000. Velocity characteristics along a
 Geomorphology 55 (2003) 111 – 124
www.elsevier.com/locate/geomorph

Predicting fine sediment dynamics along a pool-riffle

mountain channel
Sara Rathburn *, Ellen Wohl 1
Department of Earth Resources, Colorado State University, Ft. Collins, CO 80523, USA

Received 27 November 2001; received in revised form 4 June 2002; accepted 10 March 2003

Abstract

Fine sediment dynamics in mountain rivers are of concern because of implications for aquatic habitat, channel stability, and
downstream sediment yields. Many mountain river systems have episodic fine sediment transport because of infrequent, point-
source sediment inputs from landslides; basin instability triggered by land uses such as logging; or infrequent mobilization of the
coarse surface layer in channels. Dam removal, which is now more likely along mountain rivers, may also provide a substantial
fine sediment input to downstream channel reaches.
Fine sediment storage in the interstices of spawning gravels and within pools along mountain rivers is of particular interest
because of impacts to aquatic organisms. In this study we focus on sediment dynamics within pools of the North Fork Poudre
River in Colorado as an example of the processes controlling fine sediment deposition, storage, and transport within laterally
constricted pools. The 1996 release of f 7000 m3 of silt-to gravel-sized sediment from a reservoir on the North Fork provided
an opportunity to develop a field data set of fine sediment dynamics and to test the predictions of three different one- or two-
dimensional sediment transport and hydraulic models against the field observations.
The models were calibrated against quantitative measurements of pool scour and fill. One-dimensional HEC-6 results
indicate that robust simulations yield the greatest agreement between predicted and measured pool bed elevation change.
Model calibration on two pools and validation on one pool indicate that at least 58% of observed bed changes after the
sediment release were predicted by HEC-6. Modeling accuracy using quasi-two-dimensional GSTARS 2.0 was considerably
more variable, and no pool-wide trends were obtained. The two-dimensional model RMA2 substantially improved the
representation of eddy pool hydraulics within a compound pool of the North Fork. Results from the hydraulic modeling,
coupled with bed load and total load computations, delineate areas of scour and deposition which are consistent with
observations in the field.
A conceptual model of sediment delivery and storage for laterally confined pools suggests that persistent deposition of fine
sediment within eddies distal from the sediment source may result from sediment releases. The original loss of channel capacity
facilitated additional deposition within eddies as sediment within upstream proximal pools became mobilized. At high
discharges, the development of a strong shear zone prevents degradation of sediment deposits within the eddy. Central portions
of these proximal pools may clear according to existing models, whereas deposition within recirculating zones may be long-term.

* Corresponding author. Tel.: +1-970-491-6956; fax: +1-970-491-6307.

E-mail addresses: [email protected] (S. Rathburn), [email protected] (E. Wohl).
1
Tel.: +1-970-491-5298; fax: +1-970-491-5298.

0169-555X/03/$ - see front matter D 2003 Elsevier B.V. All rights reserved.
doi:10.1016/S0169-555X(03)00135-1
112 S. Rathburn, E. Wohl / Geomorphology 55 (2003) 111–124

Water managers could use these models to estimate minimum pool volume for overwinter habitat and residence time of pool
sediment.
D 2003 Elsevier B.V. All rights reserved.

Keywords: Mountain channels; Pools; Sediment release

1. Introduction daries. The channel bed and banks are likely composed
of bedrock or very coarse clasts that are mobilized
Excess sediment from natural or human-induced infrequently (Grant et al., 1990; Blizard and Wohl,
perturbations may substantially alter sediment trans- 1998). The higher channel-bed gradients, narrower
port and channel morphology along mountain rivers. In valley bottoms, and resistant boundaries imply that
many mountainous regions, rivers have relatively low mountain rivers are less likely to move laterally as a
rates of fine sediment transport because of limited result of a sediment influx than are lower gradient, less
sediment supply associated with a coarse bed surface confined rivers (Wohl, 2000). Second, mountain rivers
layer (Wohl, 2000). Hillslope instability or other point tend to have hydraulically rough boundaries, highly
sources of sediment may alter this condition and turbulent flow, and many localized zones of flow
temporarily overwhelm the river’s transport capacity. separation that change substantially as a function of
In much of the world, such point sources of sediment changing stage (Bathurst, 1988; Jarrett, 1991; Furbish,
are increasingly likely to be associated with dams 1993). Consequently, mountain rivers may store
(Williams and Wolman, 1984; Collier et al., 1996). excess fine sediment in relatively small deposits that
Aging reservoirs may contain large volumes of sedi- become stable or unstable in relation to stage (Harvey
ment that are periodically released during dam oper- et al., 1993; Ergenzinger et al., 1994; Schmidt and
ations (Wohl and Cenderelli, 2000), or ultimately Gintz, 1995; Lisle and Hilton, 1999). Finally, moun-
released as a result of dam decommissioning (Simons tain rivers often have fairly small flood plains, so
and Simons, 1991; Doyle et al., 2003). Abrupt sedi- flood-plain storage is not a particularly important
ment releases result from a variety of sources and may component of sediment dynamics (Wohl, 2000). Riv-
cause pool infilling, fining of the channel-bed sub- ers with snowmelt- and glaciermelt-dominated flow
strate, formation of lateral bars or tributary mouth regimes, in particular, may not even have overbank
deposits, and aggradation and loss of conveyance flow during many years. Because of these character-
(Lisle, 1982; Anthony, 1987; Schmidt, 1990; Madej istics, mountain rivers are likely to store excess fine
and Ozaki, 1996; Montgomery et al., 1999; Wohl and sediment in zones of flow separation associated with
Cenderelli, 2000). These changes in channel morphol- either channel irregularities (such as tributary mouths
ogy are likely to adversely impact fish populations by or pool margins) or with smaller, low-velocity zones
covering spawning gravels, reducing pool overwinter caused by bed roughness (e.g., in the lee of boulders or
and resting habitat, and altering macroinvertebrate in gravel interstices).
habitat and community composition (Gray and Ward, The coarse-grained channel boundaries and spa-
1982; Zuellig et al., 2002). Understanding fine sedi- tially and temporally complex flow hydraulics make it
ment dynamics within mountain rivers becomes critical difficult to quantitatively predict mountain channel
in order to design sediment releases and flow regimes response to excess fine sediment. Standard sediment
that minimize adverse sediment impacts to channel transport models such as HEC-6 cannot account for
morphology and aquatic communities. changing sediment supply, for example, because they
Problems associated with abrupt increases in fine assume constant supply within each time step. One-
sediment supply occur along all types of rivers. Moun- dimensional models (e.g., HEC-6) also cannot account
tain rivers, however, have some unique characteristics for either zones of flow separation, where the majority
that govern both channel response to the sediment and of sediment storage occurs, or for differential scour
our ability to predict that response (Wohl, 2000). First, and deposition over the width of a cross section. As a
mountain rivers tend to have resistant channel boun- result, the most widely used sediment transport models
 S. Rathburn, E. Wohl / Geomorphology 55 (2003) 111–124 113

are poorly equipped to deal with fine sediment dynam-

ics along mountain rivers.
In this paper, we use a case study to examine
several methods of predicting mountain channel
response to excess fine sediment. We focus on fine
sediment dynamics within pools because, in pool-riffle
mountain channels, pools are likely to provide the
most substantial and persistent storage sites for excess
fine sediment (Lisle, 1982; Lisle and Hilton, 1992).
We evaluate the ability of three standard flow and
sediment transport models to simulate sediment
dynamics observed during the months following a
reservoir-sediment release along the North Fork Cache
la Poudre River in northern Colorado. The primary
objective of the study was to determine the feasibility
of using each of these models to predict fine sediment
dynamics along pool-riffle mountain channels. We
also developed a conceptual model of mountain river
response to fine sediment that may help resource
managers to understand possible channel responses
and to evaluate options preceding or following a Fig. 1. Location map of the North Fork Poudre River study site and
sediment release. modeling reaches. Roman numerals indicate modeling reaches with
cross section locations shown as dashed lines perpendicular to the
channel. Cross sections 12 – 15 define the Tick Pool, with cross
sections 12 and 14 in the upstream and downstream eddies of the
2. Study area compound pool, respectively.

The North Fork Cache la Poudre River drains an

area of f 900 km2 at the study site (Fig. 1). The river a manner that drives the central flow of water through
has a snowmelt-dominated flow regime that varies the constriction. The smaller downstream eddy occurs
from winter base flows as low as 0.1 m3/s to spring at the point of flow separation where the channel again
peak flows up to 40 m3/s. Flow is partially regulated by expands beyond the constriction.
Halligan Dam, which is immediately upstream from In late September 1996, approximately 7000 m3 of
the study site. At the study site, the North Fork has gravel-to clay-size sediment was released from Halli-
incised a 160-m-deep canyon into Precambrian Silver gan Reservoir. Flow through the dam was shut off
Plume granite (Tweto, 1979). The channel has well- during the later stages of the sediment release, result-
developed, pool-riffle sequences and an average gra- ing in extensive deposition of fine sediment for more
dient of 0.011. Bedrock is exposed discontinuously than 12 km downstream. The fine sediment filled the
along the channel bed and banks, which are locally interstices of the channel bed along riffles and runs.
covered by or composed of a thin veneer of cobble-to Pools were filled with sediment in proportion to their
boulder-size clasts. Minimal sand and finer sediments distance downstream from the dam. At 0.5 km down-
are present along the channel. No tributary inflow is stream, pools were completely filled with fine sedi-
present along the length of the study area. ment up to 3.5 m deep. At 3.2 km downstream, pools
Pools along the North Fork Poudre River and many lost about 50% of their pre-release volume, and the
other mountain rivers are compound pools with a volume loss declined to about 30% by 4.9 km down-
strong lateral constriction and associated upstream stream. Sediment filling the pools also became pro-
and downstream eddies (Figs. 1 and 2). The upstream gressively finer downstream, from predominantly sand
eddy is created because of backwater ponding upstream and silt at 0.5 km to predominantly silt and clay at 4.9
from a constriction, which elevates the water surface in km (Wohl and Cenderelli, 2000).
114 S. Rathburn, E. Wohl / Geomorphology 55 (2003) 111–124

Fig. 2. (A) Velocity magnitude plot of the Tick Pool for low-flow simulation (4.05 m3/s) in RMA2. Color contours are in m/s. Arrows are a fixed
length and are not scaled to the magnitude of the velocity. (B) Contour plot of velocity magnitude within the Tick Pool for the high-flow
simulation (10.1 m3/s) in RMA2. Velocity gradient was calculated between the upstream eddy and the main flow at lines labeled v1 and v2.
 S. Rathburn, E. Wohl / Geomorphology 55 (2003) 111–124 115

Complete mortality of the aquatic community along time step and takes into account the effects of sediment
the North Fork below Halligan Dam occurred as a gradation. The user inputs channel geometry based on
result of the sediment release. An estimated 4000 surveyed channel cross sections; discharge at the
rainbow trout, brown trout, and white suckers were upstream boundary, represented as a series of steady,
killed by the sediment release, effectively destroying uniform flows; and inflowing sediment load, sediment
the noted wild trout population within The Nature rating curve, and the gradation of bed material. Default
Conservancy’s Phantom Canyon Preserve, which options are available for input data. Alternatively,
occupies 10 km along the river corridor downstream options such as one of 14 sediment transport equations
from Halligan Dam. may be selected. Output includes bed elevation
changes and sizes of sediment moving at each cross
section. HEC-6 simulates uniform changes in bed
3. Methods and prior results elevation over the entire channel width and has no
provisions for simulating lateral channel changes such
Following the sediment release, a 5-km-long study as meander migration or lateral changes in bed slope.
area was established downstream from the dam (Fig. GSTARS 2.0 is a quasi two-dimensional model
1). Initial surveys of channel geometry and sampling provided by the U.S. Bureau of Reclamation that
of sediment deposits were conducted during October utilizes a stream tube approach to accommodate differ-
1996, while flow remained shut down. During the ential scour and deposition over the width of a cross
subsequent snowmelt hydrograph (February – August section (Yang et al., 1998). Stream tubes are concep-
1997), repeat channel surveys, hydraulic measure- tual tube-like surfaces with boundaries defined by
ments, and suspended and bed load sampling were streamlines. Hydraulic parameters and sediment rout-
used to characterize channel response. As summarized ing computations are made for each stream tube,
in Wohl and Cenderelli (2000), the deposited sediment allowing the position and width of each tube to change.
was progressively scoured from the upstream and then GSTARS uses governing equations similar to those of
the downstream pools. Initial sediment reworking in HEC-6, except that GSTARS incorporates the momen-
the pools created a deep, narrow thalweg scoured to tum equation in backwater computations when the
the original pool bed, with additional sediment depo- flow regime changes between subcritical and super-
sition in lateral eddies. Continued reworking restored critical. Input is similar to that for HEC-6 but offers a
80 –90% of the original volume of both upstream and broader range of options with few default choices built
downstream pools by the end of the snowmelt season. into the model. Output includes vertical and lateral
The channel became supply-limited with respect to changes in bed elevation and sizes of sediment moving
finer size fractions (clay to medium sand) first at in each stream tube.
upstream and then at downstream sites, and eventually The results of the HEC-6 and GSTARS simulations
became supply-limited with respect to coarse sand to are discussed fully in Rathburn and Wohl (2001) and
pebbles. Bed load transport rates at a site were strongly summarized herein. Two scenarios representing a low,
linked to the depletion of sediment stored in upstream short (3.4 m3/s for 1 month) and a high, extended (10
pools. m3/s for 6 months) discharge were modeled. Within
The field data set was used as a standard against each scenario, a simulation using default options was
which to calibrate and verify sediment transport and run to represent limited data availability and a robust
hydraulic models. Two hydraulic models, HEC-6 and simulation was performed that utilized the entire set of
GSTARS 2.0, are discussed briefly here, and at length field data as input. The models were calibrated against
in Rathburn and Wohl (2001). HEC-6 is a one-dimen- pool bed elevations as obtained during repeat field
sional model that predicts scour and deposition within surveys. HEC-6 results indicate that long-term, robust
rivers and reservoirs (U.S. Army Corps of Engineers, simulations yield the closest agreement between pre-
1998a). The governing equations in HEC-6 include the dicted and measured pool bed elevation change.
energy equation, and conservation of mass for water Greater than 50% (range 56 –77%) of the actual scour
and sediment. The model assumes that sediment sup- and deposition within three pools was modeled using
ply and demand are satisfied within each reach at each HEC-6. Modeling accuracy using GSTARS 2.0 was
116 S. Rathburn, E. Wohl / Geomorphology 55 (2003) 111–124

considerably lower (range—12% to 92%) and more (d50 = 0.092 mm). The critical value of the Shields
variable. Version 2.0 of GSTARS does not cope with parameter s*c corresponds to the beginning of particle
stratified beds, such as a coarse layer of cobbles and motion, and was determined from the van Rijn (1984)
boulders overlain by fine-grained reservoir sediment. sediment transport relationship as follows:
The final model evaluated was RMA2, a two-
dimensional finite element model that computes s*c i0:14D0:64
* ð2Þ
water-surface elevations and depth-averaged horizon-
tal velocity components for free-surface turbulent flow where: D* = the dimensionless particle diameter, deter-
(Donnell et al., 1997). RMA2 is a numerical model that mined from.
divides the flow domain into discrete but not necessa-   13
rily uniform elements. The original RMA2 was devel- ðG  1Þg
D * ¼ ds ð3Þ
oped at Water Resources Engineers for the U.S. Army m2
Corps of Engineers, and later enhanced by King and
where: ds = a selected sediment diameter, the d50 of bed
Norton at Resource Management Associates (RMA)
material (m); G = specific gravity of the particle; g =
(King, 1990). The model uses iterative numerical
acceleration due to gravity (m/s2); m = kinematic vis-
approximation techniques to approach a convergent
cosity of the fluid (m2/s).
solution to the nonlinear mathematical expressions that
Critical shear stress for the sand diameter sampled
describe two-dimensional flow. Numerical models of
was then determined by
this type are based on a vertically integrated form of the
full three-dimensional Reynolds-averaged Navier – sc is*c ðG  1Þqgd50 ð4Þ
Stokes equations for turbulent flow. Input includes
channel geometry used to build the finite element where: s*c = critical Shields parameter (dimension-
mesh; material properties and boundary conditions less); G = specific gravity of the particle; q = density
(eddy viscosity and roughness coefficient) for each of the fluid (kg/m3); g = acceleration due to gravity
element; and water-surface elevations calculated using (m/s2); d50 = median particle diameter (m).
HEC-RAS (U.S. Army Corps of Engineers, 1998b). As The threshold of incipient motion was then deter-
with HEC-6 and GSTARS, low and high discharge mined for flow on a plane bed under turbulent flow
scenarios were simulated using RMA2, but modeling over a hydraulically rough boundary by comparing the
focused on one 70-m-long pool, informally named the ratio of so/sc. The ratio of these values was used to
Tick Pool (Fig. 1), which spans cross sections 12– 15. develop an index of particle stability (Cluer, 1997) that
Results of the RMA2 modeling provided depth and could be applied to each element in order to identify
depth-averaged velocity for each node in the finite probable areas of scour, transport, or deposition.
element mesh over the low and high discharges simu- Finally, patterns of sediment mobility were examined
lated. These output values were used to calculate bed by calculating transport rates using the hydraulic out-
shear stress (in Pa) throughout the Tick Pool using the put from RMA2 with the Schoklitsch equation and the
following equation (from Richardson et al., 1990; Yang (1973) equation for sand.
Julien, 1995):

qu2 4. RMA2 model results and discussion

so ¼    ð1Þ
yo 2
5:75log 12:27
ks RMA2 outputs of flow depth and velocity vector
and magnitude are generally consistent with staff-gage
where: q = density of the fluid (kg/m3); u = depth- data and surface patterns of flow observed in the pool
averaged velocity from the RMA2 results (m/s); at low and high discharges via automated repeat
yo = flow depth from RMA2 results (m); ks = height photographs (Rathburn, 2001). A thread of high veloc-
of roughness elements, or the d50 of bed sediment (m). ity exists along the pool thalweg, with flow separation
Bed shear stress was compared to the critical value and recirculation zones on either side of the thalweg
of shear stress at incipient motion for very fine sand (Fig. 2). At the higher discharge, the eddy cell size is
 S. Rathburn, E. Wohl / Geomorphology 55 (2003) 111–124 117

Table 1 Because centripetal acceleration is directed at the

Hydraulic parameters of the Tick Pool from RMA2 simulations center of an eddy, this acceleration can be used as an
Discharge Velocity gradient Centripetal indicator of eddy strength. Centripetal acceleration,
(m3/s) across eddy shear acceleration (m/s2)
and thus eddy strength, also increases with dis-
zone (m/s m)
charge.
4.01 v1 = 0.620; v2 = 0.345 ac1 = 0.014; ac2 = 0.009
As the discharge more than doubled between the
10.1 v1 = 0.779; v2 = 0.407 ac1 = 0.049; ac2 = 0.046
low and high flows, the upstream eddy expanded in
Location of velocity gradient (v1 and v2) and centripetal acceleration
(ac1 and ac2) calculations are shown on Fig. 2b.
size by over 0.5 m, the maximum velocity of the eddy
doubled, and the centripetal acceleration increased
larger for all recirculating zones within the pool. The threefold (Table 1). If eddy strength translates directly
velocity gradient of the shear zone between each eddy into sediment entrainment and transport potential, then
and the thalweg increases with discharge. Centripetal sediment persistence within eddies following a sedi-
acceleration of the eddy is a measure of the magnitude ment release becomes most effective at combinations
of the acceleration of a particle of water moving in the of high velocity over a small eddy radius, which
eddy, and is calculated in m/s2 by: should correspond to an eddy geometry that minimizes
the initial volume of sediment storage. At combina-
m2 tions of lower velocity and large eddy radius, the eddy
ac ¼ ð5Þ
r strength will be lower, possibly indicating greater ease
where: m = maximum linear velocity on the outward of sediment removal because of lower eddy strength
edge of the eddy (m/s); r = radius of the eddy (m). via centripetal acceleration.

Fig. 3. Dimensionless particle stability, to/tc, for the high-flow simulation. Values greater than 1 indicate particle mobility, values equal to 1
indicate incipient motion, and values less than 1 indicate particle stability. Lines across the channel denote cross sections 12, 13 and 14 of the
Tick Pool.
118 S. Rathburn, E. Wohl / Geomorphology 55 (2003) 111–124

Results from the particle stability index (Fig. 3) ment wave (Lisle, 1982; Pickup et al., 1983; Madej
indicate that areas of scour coincide with areas of high and Ozaki, 1996) and dispersal in place (Sutherland et
velocity and high boundary shear stress, such as along al., 1998). We use these studies, and our observations
the thalweg, whereas deposition occurs in areas of low along the North Fork Poudre River, to develop models
velocity and low boundary shear stress, such as within of reach-scale and width-scale channel response to a
recirculating eddies. This index is overly simplified, sediment pulse.
predicting only particle motion or stability, rather than We define a reach as a segment of consistent channel
allowing for the simultaneous transport, aggradation, geometry that is at least several channel widths in
and degradation that accompany flow. Given this length. Rivers exhibit three dominant reach-scale
simplification, the patterns delineated using the index responses to an increase in sediment load, either (i)
(Fig. 3) generally correspond to those observed in the transporting the introduced load, (ii) aggrading por-
field. tions of the channel, or (iii) degrading or incising the
Additionally, the RMA2 output was used for sedi- deposited sediment (Fig. 4). For each sediment path-
ment transport calculations, as a basis for comparison way there is a hierarchy by which sediment is parti-
with actual measured loads. Bed load transport capaci- tioned into various channel components, depending on
ties for the Tick Pool as predicted by the Schoklitsch sediment supply or the order of events. In order to use
equation are on the same order of magnitude as the Fig. 4, channel reaches subject to sediment loading
rates predicted by bed load transport equations used in must be delineated. The primary criteria by which
the one-dimensional models, but are probably more channel reaches are distinguished for purposes of
accurate because friction slope values from the model- evaluating channel response to a sediment release are
ing results are used in the two-dimensional transport channel gradient, followed by channel complexity,
calculations. The predicted capacities from RMA2 are defined as the amount of flow separation and recircu-
an order of magnitude higher than those measured, lation. Rivers with bed gradients >0.02 are considered
suggesting that the pool was supply-limited under steep-gradient (Jarrett, 1992) and will generally
most conditions, and that flow separation within the respond to sediment influxes by transporting the major-
Tick Pool eddy partially trapped sediment. ity of the imposed load. However, steep-gradient chan-
nels may possess large, low-velocity areas lateral to the
channel where flow separation and recirculation
5. Conceptual model of pool sedimentation change the sediment transport patterns. In these cases,
patterns quantifying the degree to which the recirculating zones
offset the overall transport tendency of a steep-gradient
Conceptual models offer an alternative to the time- reach is important.
and labor-intensive modeling efforts required for one- Transport of sediment occurs in steep reaches
and two-dimensional numerical simulations of flow because effective boundary shear stress s0 is greater
and sediment transport. Our intent in developing a than the critical shear stress sc required to initiate
conceptual model of pool sedimentation patterns is to transport for a specified grain size (Fig. 4). Actual
provide potential users, such as water managers, with a quantification of s0 at the reach scale will require
qualitative understanding of the factors that govern approximation of mean conditions. Boundary shear
sediment erosion and deposition within pools along stress, for example, is a function of bed gradient and
mountain rivers. Using channel characteristics ob- flow depth. These must be averaged across space
served in the field and information about the sediment (either at the cross-sectional or reach level) and
contained in the release, managers may be able to potentially across a range of discharges. Critical shear
assess processes operating at various length and time stress can only be applied to a single grain size, which
scales and to predict recovery processes and rates for a is assumed to be representative of the entire grain-size
river. distribution.
Previous studies of point-sediment influences to In the second pathway of response (Fig. 4), aggra-
mountain rivers have documented varied channel dation occurs where the channel gradient shallows or
responses, including downstream translation of a sedi- where abrupt channel expansions create flow separa-
 S. Rathburn, E. Wohl / Geomorphology 55 (2003) 111–124 119

Fig. 4. Reach-scale conceptual model of channel response to a sediment release. Transport, aggradation, and degradation represent three pathways
of channel response to a sediment release, where so is boundary shear stress, and sc is critical shear stress for incipient particle motion.

tion and zones of reduced flow velocity and deposi- critical shear stress required to move sediment (Fig.
tion. In general, patterns of aggradation are controlled 4). The degradation pathway may not always follow
by characteristics of the sediment release; water dis- sequentially after aggradation. If the introduced sedi-
charge accompanying the release; and channel geom- ment supply is high and the discharge is sufficient to
etry, such that boundary shear is less than critical transport the load, then degradation of the channel may
shear stress needed for the initiation of particle occur as the second major pathway of response with
motion. Initial channel aggradation occurs in pools, aggradation occurring simultaneously in low velocity
with numerous studies documenting preferential fill- areas.
ing of pools resulting from large increases in sediment The initial response during the degradational phase
load (Lisle, 1982; Madej and Ozaki, 1996; Wohl and is incision of the channel thalweg, followed by pool
Cenderelli, 2000). Pool deposition can effectively excavation and overall channel bed erosion (Fig. 4
create a more uniform reach gradient and flow depth moving from #1 first response to #3 last response). The
and can enhance bed mobility because the finer- degradation phase along the North Fork Poudre River
grained sediment delivered to the channel decreases occurred mainly as selective transport, in that sediment
bed roughness. Simultaneous with deposition in was entrained as a function of grain size. Smaller clasts
pools, if sediment supply remains high (Fig. 4 moving were more readily mobilized from the bed surface and
from #1 to #4, low to high sediment load), lateral bars were preferentially mobile at low flow (Wohl and
develop, often in low-velocity areas adjacent to the Cenderelli, 2000). Sustained selective transport with-
banks of the channel. Aggradation of marginal bars out a constant supply of sediment leads to the develop-
along the North Fork Poudre River occurred adjacent ment of a stable armor layer (Sutherland, 1987)
to pools, where channel expansion created recirculat- through progressive winnowing of fine material from
ing eddies. the bed.
In general, channel aggradation tends to over- Other aspects of the sediment pulse figure prom-
steepen channel slopes. With a reduction or cessation inently in the model of reach-scale response to a large
in the sediment supply and with the release of clear influx of sediment. These include the grain-size dis-
water, existing deposits are scoured and degradation tribution of the sediment comprising the pulse and the
occurs representing the third pathway of response. sediment release volume. Sediment moving as bed
During degradation, bed shear stress exceeds the load is generally deposited closer to the release source
120 S. Rathburn, E. Wohl / Geomorphology 55 (2003) 111–124

than is suspended sediment. In addition, the overall

sediment volume decreases downstream.
Recognizing the complex nature of the sediment
pathways and the concurrence of the responses in time
and space is important. Schumm’s (1973) complex
response model for the stages of channel adjustment
following channel incision also applies as a model of
channel aggradation and degradation on the North
Fork Poudre system following the reservoir sediment
release (Fig. 4). Transport, aggradation, and degrada-
tion can occur concurrently within a reach, depending
on the channel geometry and boundary roughness, the
flow regime, and the properties of the released sedi-
ment. This is demonstrated along the North Fork
Poudre (where channel behavior alternated between
aggradation and degradation both downstream, cross
stream, and with time) as sediment from upstream
pools was mobilized and transported into downstream,
more distal pools.
Because pools are such critical storage and release
points for sediment, a width-scale conceptual model
focused on processes governing sediment movement
within eddies in pools is useful for management deci-
Fig. 5. Composite cross section through the Tick Pool showing
sions related to the reestablishment of fish habitat. Fig. aggradation and degradation of reservoir sediment over the 1996 –
5 illustrates the response of compound pools along the 1997 hydrograph. Arrows indicate perspective is from upstream to
North Fork Poudre River following the 1996 sediment downstream through the compound pool.
release. Initial deposition formed a veneer of essen-
tially uniform depth over the pool bottom. A stepped,
preliminary discharge during February – March 1997 most of the fine sediment in transport (Fig. 6a). As
eroded a thalweg along the upstream portion of the pool discharge increases, the reattachment point begins to
(cross section 12, Fig. 1) and the constriction, but also migrate downstream, eventually becoming anchored
deposited sediment on the original marginal deposits on the bedrock constriction, thereby increasing the
and across the downstream portion of the pool (cross size of the recirculation zone. The upper limit in size
section 14, Fig. 1). Original deposition within the pool of the recirculation zone is attained when the reat-
elevated the channel bottom, causing a loss of channel tachment point becomes fixed on the bedrock out-
capacity. As a result, sediment mobilized from crop. At moderate discharges, a strong, prominent
upstream pools (more proximal to the sediment source) eddy fence or shear zone develops, and the trajectory
during the preliminary discharge was transported and of inflowing water becomes oriented directly down-
deposited into lateral pool areas that were otherwise not stream rather than into the upstream eddy (Fig. 6b).
available under normal discharges. Subsequent sus- Sediment that was deposited at lower discharges is
tained high flows largely removed the sediment from not flushed out of the eddy at higher flows because of
most portions of the pool, leaving only smaller the strong recirculating flow and pronounced eddy
marginal deposits. fence.
These processes can also be examined in terms of Sediment removal from the pool can be explained
flow patterns within the pool (Fig. 6). At low flows, by Thompson’s (1997; Thompson et al., 1998; Thomp-
sediment deposition is limited to small recirculating son and Hoffman, 1999) model of pool flushing. Sedi-
areas on the shoreward side of the eddy because the ment deposited in steep-sided marginal bars within
angle of inflowing water into the upstream pool keeps pools may slump into the pool thalweg during waning
 S. Rathburn, E. Wohl / Geomorphology 55 (2003) 111–124 121

6. Conclusions

This research represents the first evaluation of

existing numerical models against a complete data
set from a sediment release on a mountain river. The
North Fork Poudre River exhibits the unique charac-
teristics of mountain rivers that influence channel
response to a sediment release including a fixed bed
with limited lateral mobility, turbulent flow with
strong flow separation, stage-dependent sediment stor-
age, and minimal overbank flow and sediment depo-
sition. Our work demonstrates the importance of pool
dynamics (channel geometry and sediment transport
processes) in understanding and predicting sediment
storage and mobility following an influx of sediment
to mountain rivers.
The results of the numerical modeling indicate that
one- and semi-two-dimensional sediment transport
models may be useful predictive tools for determining
pool recovery following a reservoir sediment release.
Of the two models evaluated in this study, HEC-6
model predictions of pool bed change agree with field
measurements of pool scour and deposition to a much
Fig. 6. Conceptual model of sediment transport and eddy flow for
laterally confined compound pools using the Tick Pool as an
greater degree than GSTARS 2.0 (Table 2). Pool-wide
example. At low discharge (A), deposition is restricted to small trends of predicted versus observed aggradation and
recirculation zones in upstream and downstream eddies. As discharge degradation for three pools along the North Fork
increases (B), the reattachment point (RP) moves downstream and Poudre were 58% (on average) for the long-term,
becomes fixed on the bedrock outcrop. The separation point (SP) snowmelt runoff simulation using HEC-6 and Yang’s
does not migrate with higher discharges, preventing low-flow
deposits from being eroded. Arrows indicating velocity are schematic
(1973) sand transport equation. These results indicate
and not to scale. that more than half of observed bed changes after a
sediment release would be predicted by HEC-6. The
predictive utility of HEC-6 as applied to the North
stages of flow as the velocity gradient between the Fork Poudre River is moderate to high, in that esti-
thalweg and the marginal eddies declines. With an mates of fish survival by water resources managers
increase in discharge, the slumped sediment is scoured following a reservoir sediment release would be based
out of the thalweg by the central jet of high velocity on knowledge of pool depth recovery of >50%. In
flow. In this way, central portions of pools may have a addition, HEC-6 simulations have a utility for predict-
mechanism for self-clearing, given the proximity to the ing flushing flows directed at pool recovery. Model
constriction, whereas the recirculating zones have the simulations could be run until the desired pool recov-
potential for persistent sediment storage in the shore- ery was achieved, and the associated flow could be
ward portion of the eddies. Sedimentation within com- requested as the minimum flushing discharge for sedi-
pound pools of mountain rivers occurs as transient ment mobilization and transport.
sediment storage and transfer such that proximal pools Lateral variations in velocity are important for an
are sources of sediment for more distal pools, and distal understanding of channel dynamics, especially where
pools may be long-term sinks of sediment depending flow separation and recirculating eddies are significant
on pool geometry. The deposits from a sediment pulse components of the flow field, as in bedrock channels
can be viewed as sediment in transit, although the with irregular boundaries. The two-dimensional
movement is intermittent. hydraulic modeling with RMA2 was designed to over-
122 S. Rathburn, E. Wohl / Geomorphology 55 (2003) 111–124

Table 2
Comparison of numerical models based on certain criteria by which potential users might evaluate the models for applicability to a given
sediment release situation
HEC-6 GSTARS 2.0 RMA2
Data requirements moderate – high moderate – high high
Expertise moderate moderate – high very high
Results >50% accuracy, pool-wide trend 13 – 90% accuracy, no replicated low-velocity areas
pool-wide trends well, general flow field for low
and high discharge
Advantages cross section based, default cross section-based, semi- nodal hydraulic parameters,
options, sediment transport model two-dimensional, sediment visual display of output
transport model
Limitations purely one-dimensional, limited few default options, not suited hydraulic model only, outstrips
transport formulas for stratified beds calibration data, calibration
data difficult to collect
Predictive ability for North moderate – good low moderate
Fork application
Predictive utility for pool moderate – good low low
habitat restoration

come the limitations of a purely one- or semi-two- laterally constricted pools helps to fill in deficiencies
dimensional model. The two-dimensional modeling in the predictions of the numerical models. Reach-
substantially improves the delineation of eddy pool scale sediment transport drives the main sediment
flow recirculation patterns. The cross-stream compo- delivery system, and width-scale processes drive the
nent of flow is accounted for, and modeling simula- pool-specific redistribution of sediment. The combi-
tions of low and high discharge for the North Fork nation of numerical and conceptual models provides
Poudre River show well-developed recirculating zones the most robust understanding of sediment issues
of flow that are in broad agreement with timed photo- surrounding eddy pools.
graphs and field measurements of depth and velocity Additional research pertaining to sediment releases
within the study pool. However, inferred sediment into mountain rivers would benefit from numerical
transport did not correspond well to field-measured models that are specifically designed for systems with
patterns of erosion and deposition within eddy pools. steep gradients, complicated flow structures created
The main limitation in using RMA2 to infer sed- by irregular bed topography and geometry, and fine
imentation patterns is the inability to accurately rep- sediment transport in an originally coarse-grained
resent the processes of sediment transport using a system. The main challenge of modeling is an accu-
hydraulic model. Disparities between the resolution rate representation of transport processes of fine bed
of the field data on sedimentation patterns and the material in a steep-gradient, cobble-boulder channel.
results of the sediment transport rates predicted using Any future modeling would also benefit from an
hydraulic modeling output severely limit the extent to increased spatial domain to include a treatment of
which field and modeling results can be compared. If sediment routing of a translating and dispersing slug
broad hydraulic features, such as low velocity areas, of sediment. Improved accuracy of flow and sediment
are used to make predictions of sediment deposition transport modeling also requires more detailed field
and erosion, then the agreement between field and data on a scale commensurate with the model results.
modeling data is greatly improved. In the end, the Although the acquisition of field data poses logistical
predictive ability of the two-dimensional model is problems during high flow, flume experiments may
moderate for low velocity areas of the pools. However, serve as an adequate analog provided that the con-
the predictive ability of RMA2 for this application ditions at the field site can be replicated. Given the
along the North Fork Poudre River is low (Table 2). recent interest in dam removal and the associated
A qualitative understanding of the processes asso- uncertainties of sedimentation hazards, a successful
ciated with sediment scour and deposition within solution to the problem of routing sediment released
 S. Rathburn, E. Wohl / Geomorphology 55 (2003) 111–124 123

from a reservoir has important practical and scientific In: Kirkby, M.J. (Ed.), Process Models and Theoretical Geo-
value. When and how sediment that accumulates morphology. Wiley, Chichester, pp. 141 – 158.
Furbish, D.J., 1993. Flow structure in a bouldery mountain stream
behind a dam can be released so as to limit impacts with complex bed topography. Water Resources Research 29,
to downstream aquatic ecosystems is highly useful 2249 – 2263.
knowledge, especially as reservoirs age and as demand Grant, G.E., Swanson, F.J., Wolman, M.G., 1990. Pattern and origin
for regulated flows increases. of stepped-bed morphology in high-gradient streams, western
Cascades, Oregon. Geological Society of America Bulletin
102, 340 – 352.
Gray, L.J., Ward, J.V., 1982. Effects of sediment releases from
Acknowledgements a reservoir on stream macroinvertebrates. Hydrobiologia 96,
177 – 184.
This research was supported by funding from Trout Harvey, M.D., Mussetter, R.A., Wick, E.J., 1993. A physical proc-
Unlimited, the U.S. Bureau of Reclamation, the ess – biological response model for spawning habitat formation
for the endangered Colorado squawfish. Rivers 4, 114 – 131.
Colorado Water Conservation Board, and NSF grant
Jarrett, R.D., 1991. Wading measurements of vertical velocity pro-
CMS-9727061. Numerous individuals helped with the files. Geomorphology 4, 243 – 247.
original field data collection and the modeling, Jarrett, R.D., 1992. Hydraulics of mountain rivers. In: Yen,
including Dan Cenderelli, Doug Thompson, Heather B.C. (Ed.), Channel Flow Resistance: Centennial of Man-
Knight, Jason Alexander, Allison Thornton, Francisco ning’s Formula. Water Resources Publications, Littleton, CO,
pp. 287 – 298.
Simoes, and Lyle Zevenbergen. The Nature Conserv-
Julien, P.Y., 1995. Erosion and Sedimentation. Cambridge Univ.
ancy and the Colorado Division of Wildlife provided Press, New York, NY.
logistical support, and two anonymous reviewers King, I.P., 1990. Program Documentation: RMA2-A Two-Dimen-
provided helpful reviews on a draft of this manuscript. sional Finite Element Model for Flow in Estuaries and Streams.
We thank them all. Ver 4.3. Resource Management Associates, Lafayette, CA.
Lisle, T.E., 1982. Effects of aggradation and degradation on riffle-
pool morphology in natural gravel channels, northwestern Cal-
ifornia. Water Resources Research 18, 1643 – 1651.
References Lisle, T.E., Hilton, S.J., 1992. The volume of fine sediment in
pools: an index of sediment supply in gravel-bed streams. Water
Anthony, D.J., 1987. Stage dependent channel adjustments in a Resources Bulletin 28, 371 – 383.
meandering river, Fall River, Colorado. MS thesis, Colorado Lisle, T.E., Hilton, S.J., 1999. Fine bed material in pools of natural
State University, Ft. Collins. gravel bed channels. Water Resources Research 35, 1291 – 1304.
Bathurst, J.C., 1988. Velocity profile in high-gradient, boulder-bed Madej, M.A., Ozaki, V., 1996. Channel response to sediment wave
channels. Proceed. International Assoc. Hydraul. Res. Interna- propagation and movement, Redwood Creek, California, USA.
tional Conf. on Fluvial Hydraulics ’88. Interntnl. Assoc. Hy- Earth Surface Processes and Landforms 21, 911 – 927.
draul. Research (IAHR), Budapest, Hungary, pp. 29 – 34. Montgomery, D.R., Panfil, M.S., Hayes, S.K., 1999. Channel-bed
Blizard, C.R., Wohl, E.E., 1998. Relationships between hydraulic mobility response to extreme sediment loading at Mount Pina-
variables and bedload transport in a subalpine channel, Colorado tubo. Geology 3, 271 – 274.
Rocky Mountains, USA. Geomorphology 22, 359 – 371. Pickup, G., Higgins, R.J., Grant, I., 1983. Modelling sediment
Cluer, B.L., 1997. Eddy bar responses to the sediment dynamics of transport as a moving wave—the transfer and deposition of
pool-riffle environments. PhD dissertation, Colorado State Uni- mining waste. Journal of Hydrology 60, 281 – 301.
versity, Ft. Collins. Rathburn, S.L., 2001. Modeling pool sediment dynamics in a
Collier, M.P., Webb, R.H., Schmidt, J.C., 1996. Dams and rivers: a mountain river. PhD dissertation, Colorado State University,
primer on the downstream effects of dams. U.S. Geological Ft. Collins.
Survey Circular 1126 (94 pp.). Rathburn, S.L., Wohl, E.E., 2001. One-dimensional sediment trans-
Donnell, B.D., Finnie, J.L., Letter Jr., J.V., McAnally Jr., W.H., port modeling of pool recovery along a mountain channel after a
Roig, L.C., Thomas, W.A., 1997. Users Guide to RMA2 WES. reservoir sediment release. Regulated Rivers: Research and
Ver 4.3. U.S. Army Corps of Engineers, Waterways Experiment Management 17, 251 – 273.
Station Hydraulics Laboratory, Vicksburg, MS. Richardson, E.V., Simons, D.B., Julien, P.Y., 1990. Highways in the
Doyle, M.W., Stanley, E.H., Harbor, J.M., 2003. Predicting channel river environment, U.S. Dept. of Transportation, Federal High-
response to dam removal using geomorphic analogies. Journal ways Administration.
of the American Water Resources Association 38, 1567 – 1579. Schmidt, J.C., 1990. Recirculating flow and sedimentation in the
Ergenzinger, P., De Jong, C., Christaller, G., 1994. Interrelation- Colorado River in Grand Canyon, Arizona. Journal of Geology
ships between bedload transfer and river-bed adjustment 98, 709 – 724.
in mountain rivers: an example from Squaw Creek, Montana. Schmidt, K.-H., Gintz, D., 1995. Results of bedload tracer experi-
124 S. Rathburn, E. Wohl / Geomorphology 55 (2003) 111–124

ments in a mountain river. In: Hickin, E.J. (Ed.), River Geo- U.S. Army Corps of Engineers, 1998a. HEC-6 Scour and Deposi-
morphology. Wiley, Chichester, pp. 37 – 54. tion in Rivers and Reservoirs User’s Manual. Ver 4.1. Hydro-
Schumm, S.A., 1973. Geomorphic thresholds and the complex re- logic Engineering Center, Davis, CA.
sponse of drainage systems. In: Morisawa, M.E. (Ed.), Fluvial U.S. Army Corps of Engineers, 1998b. HEC-RAS River Analysis
Geomorphology. Publications in Geomorphology, State Univer- System User’s Manual. Ver 2.2. Hydrologic Engineering Center,
sity of New York, Binghamton, pp. 299 – 310. Davis, CA.
Simons, R.K., Simons, D.B., 1991. Sediment problems associated van Rijn, L.C., 1984. Sediment transport: Part III. Bedforms and
with dam removal: Muskegon River, Michigan. Proceedings of alluvial roughness. ASCE Journal of Hydraulics Division 10
the 1991 National Conference American Society of Civil Engi- (12).
neers. ASCE, New York, pp. 680 – 685. Williams, G.P., Wolman, M.G., 1984. Downstream effects of dams
Sutherland, D.G., 1987. Static armour layers by selective erosion. In: on alluvial rivers. U.S. Geological Survey Professional Paper
Thorne, C.R., Bathurst, J.C., Hey, R.D. (Eds.), Sediment Trans- 1286. 83 pp.
port in Gravel-bed Rivers. Wiley, Chichester, UK, pp. 243 – 267. Wohl, E.E., 2000. Mountain Rivers. Water Resources Monograph,
Sutherland, D.G., Hansler, M.E., Hilton, S.J., Lisle, T.E., 1998. vol. 14. American Geophysical Union Press, Washington, DC.
Sediment wave evolution and channel morphologic changes Wohl, E.E., Cenderelli, D.A., 2000. Sediment deposition and trans-
resulting from a landslide, Navarro River, northwestern Califor- port patterns following a reservoir sediment release. Water Re-
nia. Geological Society of America Abstracts with Programs 29, sources Research 36, 319 – 333.
A-360. Yang, C.T., 1973. Incipient motion and sediment transport. Journal
Thompson, D.M., 1997. Hydraulics and pool geometry. PhD dis- of Hydraulics Division 99, 1679 – 1703.
sertation, Colorado State University, Ft. Collins. Yang, C.T., Trevino, M.A., Simoes, F.J., 1998. User’s Manual for
Thompson, D.M., Hoffman, K.S., 1999. Pool dimensions in coarse- Generalized Stream Tube Model for Alluvial River Simulation.
grained, New England channels. Geological Society of America Ver 2.0 (GSTARS 2.0). U.S. Bureau of Reclamation Technical
Annual Meeting, Abstracts with Programs 31, A-49. Service Group, Sedimentation and River Hydraulics Group,
Thompson, D.M., Nelson, J.M., Wohl, E.E., 1998. Interactions be- Denver, CO.
tween pool geometry and hydraulics. Water Resources Research Zuellig, R.E., Kondratieff, B.C., Rhodes, H.A., 2002. Benthos re-
34, 3673 – 3681. covery after an episodic sediment release into a Colorado Rocky
Tweto, O., 1979. Geologic Map of Colorado. U.S. Geological Sur- Mountain river. Western North American Naturalist 62, 59 – 72.
vey Misc. Invest. Map G77115, Denver, CO.
 Geomorphology 55 (2003) 125 – 137
www.elsevier.com/locate/geomorph

The geomorphic significance of step–pools in mountain streams

Anne Chin *
Department of Geography, College of Geosciences, Texas A&M University, College Station, TX 77843-3147, USA
Received 16 April 2002; received in revised form 6 October 2002; accepted 10 March 2003

Abstract

This paper develops a conceptual model to define the changing role and significance of step – pools as energy dissipators in
steep mountain channels. Although energy dissipation is a significant function of step – pools, this role changes over time with
variations in discharge. Steps are effective in reducing stream energy at low flows, but their effectiveness diminishes with
increasing stage. Accordingly, the manner in which energy dissipation occurs also varies. With increasing flows, spill resistance
gives way to a dominance in form and grain resistance. The conceptual model for the changing role of step – pools is illustrated
with data from hydraulic analysis and modeling of step – pools in the Santa Monica Mountains of California. The model points
to the importance of the size of steps in determining their role in energy dissipation and in their interactions with channel
hydraulics. The model offers a new articulation for the geomorphic significance of step – pools in mountain streams, and it
serves as a useful template for a more complete understanding of step – pools over a longer time scale.
D 2003 Elsevier Science B.V. All rights reserved.

Keywords: Step – pool; Energy dissipation; Mountain stream geomorphology

1. Introduction forested streams, steps are commonly composed of

logs mixed with vegetative debris (Marston, 1982;
Steps and pools are ubiquitous bed forms in Curran and Wohl, in press).
mountain stream channels, occurring where gradients Step – pools serve a fundamental role in river sys-
exceed f 2% and materials are in the gravel to tems because they provide hydraulic resistance (Abra-
boulder size range (Grant et al., 1990; Montgomery hams et al., 1995). Steps induce water to flow over and
and Buffington, 1997; Wohl, 2000a). Coarse particles through the large roughness elements and plunge into
spanning the channel width create steps (Hayward, the pool below, promoting tumbling flow (Peterson and
1980), which alternate with finer sediments in pools to Mohanty, 1960) where much of the flow’s kinetic
produce a characteristic, repetitive sequence of bed energy is dissipated by roller eddies. Steps also cause
forms (Chin, 2002) (Fig. 1) with a stepped longitudi- a distinct vertical drop in the water surface elevation as
nal profile resembling a staircase. The step – pool water flows from step to pool. Through vertical fall,
morphology similarly develops in bedrock channels steps decrease the amount of potential energy that
(Duckson and Duckson, 1995, 2001; Wohl, 2000b). In otherwise would be available for conversion to a
longitudinal component of kinetic energy used for
* Tel.: +1-979-845-7141; fax: +1-979-862-4487. erosion and sediment transport (Ashida et al., 1976;
E-mail address: [email protected] (A. Chin). Marston, 1982). In these ways, steps provide the ability

0169-555X/03/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0169-555X(03)00136-3
126 A. Chin / Geomorphology 55 (2003) 125–137

Fig. 1. Step – pool sequences in the Santa Monica Mountains, California.

to counterbalance steep gradients, thereby preventing energy-dissipating features in the fluvial system.
excessive erosion and channel degradation (Heede, Although early work has recognized step – pools as
1981). The role of step – pools is especially important important energy-dissipating mechanisms in steep
in confined mountain streams that prohibit lateral slopes (i.e., Heede, 1972; Church and Jones, 1982;
adjustments and energy dissipation by meandering Chin, 1989), this issue could be investigated explic-
and braiding (Chin, 1989, 2002). itly, and the implications for the changing role of
Despite their common occurrence and functional step – pools in the evolution of the fluvial system
importance, step – pools have received little attention in could be explored. The role of step – pools changes
fluvial geomorphological research compared to riffle– over time because the effectiveness of fall obstruc-
pools. Recent work has generated important new data tions varies with river stage. Steps are most effective
from around the world (e.g., Chartrand and Whiting, in reducing stream energy at low flows when vertical
2000; Wohl and Thompson, 2000; Lenzi, 2001; Zim- fall is most pronounced. For example, steps account
mermann and Church, 2001; Lee and Ferguson, 2002), for 40 – 100% of the total drop in water surface
but the specific formative processes are not completely elevation in channels of Colorado, Arizona, and
understood, and the significance of step – pools in the Washington (Heede, 1972, 1981; Curran and Wohl,
broader fluvial system has not been fully articulated. in press). However, the effectiveness of energy reduc-
Because upland step – pool channels are linked to tion by steps is greatly impaired at increasingly high
fluvial responses in large downstream basins (Mont- flows (Heede, 1972; Hayward, 1980; Marston, 1982;
gomery and Buffington, 1997) and because step – pool Whittaker and Davies, 1982). Energy loss due to
units are important riparian ecosystems (Scheuerlein, individual steps is diminished at high flows because
1999), increased knowledge of step – pools is critical to the water surface profile and energy line flatten
the understanding of the overall functioning of the river (Stuve, 1990; Leopold, 1994; Zimmermann and
system. As urbanization increasingly encroaches upon Church, 2001) and flow resistance decreases corre-
mountain fronts in response to population growth spondingly (Beven et al., 1979; Lee and Ferguson,
(Chin and Gregory, 2001), improved understanding 2002). However, although the importance of docu-
of step – pools could also contribute to approaches to menting energy dissipation at higher stages has been
the design and management of steep channels. recognized (Marston, 1982), critical data have been
This paper develops a conceptual model for the lacking to enable a more complete picture of energy
changing role and significance of step – pools as dissipation by step – pools, owing to the difficulties of
 A. Chin / Geomorphology 55 (2003) 125–137 127

making process measurements in rugged and often bed forms and roughness elements that impart form
densely vegetated terrains (Chin, 1989). and grain resistance. Thus, defining these two points
The role of step – pools also changes over time would form the basis for a model of the changing role
because the manner in which flow resistance is and significance of step – pools in the fluvial system
imparted by steps varies with stage. At low flows, over a longer time scale. Such a model acknowledges
when the staircase-like step structure is most pro- the progressive development of step – pools in
nounced and water plunges from step to pool, spill between large, channel-forming events that break
resistance is the dominant energy-dissipating mecha- down the sequences (Lenzi, 2001). This model there-
nism in step – pool channels (Abrahams et al., 1995), fore applies to step – pool systems that are hydrauli-
approaching as much as 80– 90% (Curran and Wohl, cally controlled and capable of being submerged and
in press). However, as stage increases and the step – restructured at high flows. Examples include the Rio
pool shape becomes increasingly hidden, spill resist- Cordon of Italy (Lenzi et al., 1997; Lenzi, 2001), the
ance gives way to a dominance in form resistance and Lainbach River of Germany (Ergenzinger, 1992;
then to grain resistance. Thus, the relative dominance Gintz et al., 1996), and streams in the Santa Monica
of the energy-dissipating mechanism by steps changes Mountains of the United States (Chin, 1998, 1999a)
when considered over time with variations in dis- where large floods have been observed to submerge
charge. Although the stage-dependency nature of and mobilize steps.
roughness conditions has been noted (Ergenzinger, For the remainder of the paper, the term ‘‘energy
1992; de Jong and Ergenzinger, 1998; Lee and Fer- dissipator’’ is used as in an engineering sense
guson, 2002) and attempts at partitioning flow resist- (Vischer, 1995), in that when a step functions like a
ance in step –pool channels during low flow have baffle or fall obstruction and induces free water fall, it
been made (Curran and Wohl, in press), further is an energy dissipator. However, when a step is
consideration of energy dissipation over a range of submerged and vertical fall is no longer present, I
flows would reveal additional insights for step –pools will refer to it as a ‘‘roughness element’’ (Bathurst,
over a longer time scale. 1987). Using this terminology, submergence repre-
Herein, I outline a conceptual model to define the sents a point when steps cease to function as energy
changing role and significance of step – pools as dissipators and behave simply as roughness elements
energy dissipators in the fluvial system. The model in the stream channel, like any other.
is built upon answering four specific research ques-
tions. First, how stable are step – pool sequences?
Second, how effective are step – pools in reducing 2. Study area and methods
potential energy through vertical falls? Third, how
does potential energy dissipation vary with increasing 2.1. Theoretical structure
flow? Fourth, at what stage do step –pools become
submerged? This investigation considers the low flow step –
Identifying the points of instability and submer- pool form and process as a middle member of the
gence are key to the development of the conceptual range of possible spectrum of states in the evolution of
model. The point of instability is important because it the fluvial system (Fig. 2). By working backward and
marks a change in the fundamental role of steps in the forward from this central state, the channel-forming
river system (Chin, 1998). When steps are stable, they flow can be inferred and the effects of the step –pool
are independent variables that regulate surrounding morphology on stream energy dissipation can be
flow, channel hydraulics, and energy dissipation; once determined. Hydraulic reconstructions estimate the
steps are mobilized, they become a dependent variable threshold of step –pool mobility; direct field measure-
that adjusts to prevailing flow and energy conditions. ments and hydraulic modeling allow energy dissipa-
Similarly, the point of submergence marks a major tion to be evaluated. Energy dissipation with
shift in the dominant role of step – pools in the channel increasing flow permits identification of the threshold
system, from functioning as fall obstructions that discharge at which step – pools become ineffective
induce spill resistance, to behaving more like other energy-dissipating features. Although this threshold
128 A. Chin / Geomorphology 55 (2003) 125–137

Fig. 2. Theoretical structure for investigating step – pools. Low flow represents a middle member of the range of possible spectrum of states in
the evolution of the fluvial system.

discharge does not necessarily have to equal the study reaches. The specific algorithm used was devel-
channel-forming flow, presumably, at some high- oped by Costa (1983) for small, steep mountain
magnitude flow, the step breaks down and channel streams. A series of computations uses particle size
adjustment occurs. Therefore, this theoretical diagram as the independent variable to determine the velocity
represents one complete life cycle of a step – pool and flow depth necessary to mobilize coarse step
sequence, and the task is to define the dimensions particles:
of the diagram.
0:487
v ¼ 0:18 d ð1Þ
2.2. Field sites from Santa Monica Mountains
where d is the average size of the five largest
Streams from the Santa Monica Mountains of particles (mm) and v is the average velocity (m
southern California are used to define the fre- s 1). Depth is determined from a family of equa-
quency – magnitude dimensions of the conceptual tions that defines Fig. 7 in Costa (1983) (i.e., for
model for the changing role of step – pools. The Santa
Monica Mountains were the site of previous studies
on the stability (Chin, 1998), origin (Chin, 1999a), Table 1
morphology (Chin, 1999b), and periodicity (Chin, Study reaches in the Santa Monica Mountains
2002) of step –pool systems. This analysis uses data Reach Step – Length Slope Channel Step
from the hydraulic reconstructions of step –pool mobi- pool (m) (m/m) width sizea
lity reported in Chin (1998). Thirteen of the original sequences (m) (mm)
15 study reaches from Big Sycamore Creek and Cold Cold Creek Preserve 48 186 0.115 2.5 490
Creek are selected for further analysis. These reaches Stunt 42 169 0.063 2.5 417
contain well-developed step – pools. They vary from Helsley 45 198 0.038 2.9 313
Bobcat 39 221 0.019 3.2 365
0.02 to 0.12 in slope and from 125 to 270 m in length Jude 30 225 0.033 5.6 550
(Table 1). For more descriptions of the study reaches Monte 27 218 0.022 6.6 403
and the Santa Monica Mountains, see Chin (1998, Big Canyon 37 210 0.096 3.5 493
1999a,b, 2002). Sycamore Klein 33 170 0.050 2.2 380
Creek Laughlin 29 160 0.047 2.4 405
Scott 38 270 0.061 3.6 519
2.3. Hydraulic reconstructions Bridge 40 190 0.036 5.5 461
Overlook 28 125 0.024 3.6 294
Working backwards from the low-flow state (Fig. Wood 28 193 0.017 4.7 305
2), hydraulic reconstructions were used to estimate the a
Calculated by averaging the b-axis of the five largest rocks
high-magnitude flow needed to mobilize steps in the comprising each step.
 A. Chin / Geomorphology 55 (2003) 125–137 129

slope 0.005, D = 0.012d 0.872 where D is the average a horizontal datum. Thus, PE per unit mass is directly
depth (m); for slope 0.010, D = 0.005d 0.788). Com- proportional to height or elevation:
puted flow depths were evaluated in the field
against high water marks and flood debris, visible PE=mah ð3Þ
at about half of the study reaches, which suggested Potential energy dissipation caused by steps is the
that such depths were reasonable estimates for the ratio of the cumulative change of water surface
Santa Monica Mountains (Chin, 1999a). Discharge elevation in vertical falls (h) to the total stream relief,
was computed from cross-sectional surveys and or change in water surface elevation (Fig. 3):
then related to a flow frequency based on regional
relationships (Young and Cruff, 1967). More back- PE losssteps ¼ h=total stream relief ð4Þ
ground for this methodology is detailed in Chin
(1998), which also contains the complete set of This ratio, commonly used for log steps and organic
data and results for the 15 study reaches in the debris (e.g., Heede, 1972, 1981; Keller and Swanson,
Santa Monica Mountains. Threshold discharges 1979), expresses the potential energy reduced by steps
computed with this method are conservative esti- that otherwise would be available for conversion to
mates (Grant et al., 1990; Scheuerlein, 1999; Zim- kinetic energy and sediment transport. Marston (1982)
mermann and Church, 2001) because such factors provides more details of energy transformations in
as particle imbrication and interaction are unac- stream channels.
counted for. To evaluate potential energy dissipation by step –
pools over a range of flows, the approach was to work
2.4. Energy dissipation forward from the central low-flow state to the high-
magnitude discharge where steps are submerged and
An assessment of the significance of step – pools in are no longer effective as energy dissipators (Fig. 2).
stream energy dissipation is to consider potential This threshold marks the shift from the role of steps in
energy in a channel reach: inducing free water fall to form roughness taking a
greater importance. A series of water surface profiles
PE ¼ mgh ð2Þ were constructed in order to evaluate energy dissipa-
tion and to identify the point of submergence. Field
where PE is potential energy; m is mass; g is accel- surveys provided data for the low-flow profile, but
eration of gravity; and h is height (or elevation) above similar data could not be obtained at higher flows.

Fig. 3. Potential energy dissipation by steps.

130 A. Chin / Geomorphology 55 (2003) 125–137

Instead, water surface profiles were modeled for channels, HEC-2 allowed another glimpse of high-
higher flows using the step-backwater HEC-2 pro- magnitude flow conditions where field measurements
gram of the U.S. Army Corps of Engineers (1990). were nearly impossible (Keller and Florsheim, 1993;
Carling and Wood, 1994). Similar applications to
2.5. Hydraulic modeling pool – riffle reaches (Keller and Florsheim, 1993)
and bedrock channels (Wohl et al., 1999) yielded
The HEC-2 model calculates water surface profiles good results.
based on the solution of the one-dimensional energy The modeling was performed for a 30-m portion of
equation. It is the original version of HEC-RAS, the Stunt Reach in Cold Creek, 1 of the 13 study reaches
River Analysis System program (U.S. Army Corps of (Fig. 1; Table 1). Seven step –pool sequences com-
Engineers, 1998). The newer HEC-RAS model pro- prise this reach, named Stunt Subreach. The reach is
vides several improved hydraulic features over HEC-2 comparatively simple hydraulically, a desirable trait
(Brunner and Piper, 1994), such as alternative channel for modeling (O’Connor and Webb, 1988). The
subdivision for conveyance calculations and mixed accessibility of the reach also facilitates data gathering
flow regime calculations, but the two programs are for model calibration and verification. Basic input
fundamentally the same. For example, comparison of data for HEC-2 are surveyed channel geometry (cross
water surface elevations calculated with HEC-2 and sections and length of channel between consecutive
those with HEC-RAS at f 2000 cross sections cross sections), initial stage and discharge, flow
showed that, for the 10% chance flood, 73.1% of regime, and the energy loss coefficients (Manning’s
the cross sections had identical water surface eleva- n and expansion/contraction coefficients). A total of
tions and 96.9% were within F 0.006 m (Bonner et 24 surveyed channel cross sections represented Stunt
al., 1994). Similarly, for the 1% chance flood, 70.1% Subreach; these were taken at every major break in
of the cross sections showed no difference and 95.8% slope. Five flow profiles were generated; these served
were within F 0.006 m. as starting points for indicating trends in energy loss
Because HEC-2 is a one-dimensional model for by steps at higher flows.
steady, gradually varied flow, the assumptions are The remainder of the paper addresses the four
difficult to meet for mountain streams. Even so, the stated research questions. I then explore the implica-
focus on modeling higher flows with HEC-2 in this tions of these findings for the changing role of step –
application makes the program a reasonable choice. pools, followed by the development of the conceptual
For example, although HEC-2 often gives inaccurate model outlining the significance of step – pools as
predictions of hydraulic parameters during low flows energy dissipators in mountain streams.
(Miller and Wenzel, 1985), these errors decline with
increasing stage (Carling and Wood, 1994; Keller and
Florsheim, 1993) because bed-generated turbulence 3. Mobility, energy dissipation, and the changing
tends to be less important at higher flows (Jarrett, role of step –pools
1984). Also, although numerous local supercritical
transitions probably occur in step – pool channels, 3.1. Threshold of step mobility
modeling flow as subcritical nevertheless gives good
results (O’Connor and Webb, 1988) because subcrit- As reported in Chin (1998), results of the hydraulic
ical flow has been largely reported for mountain calculations indicate that step – pools in the study
streams owing to extreme turbulence (Heede, 1972; reaches of the Santa Monica Mountains are mobilized
Jarrett, 1984; Lopez and Falcon, 1999). Therefore, by discharges ranging from 0.6 to 295.5 m3 s 1,
because the complex hydraulics associated with depending on the size of the particles comprising
mountain channels are not likely to be represented the steps. The larger the particle sizes, the greater
fully by any model currently available, simple the flows required. The calculated discharge values
hydraulic programs provide useful approximations correspond to recurrence intervals of about 2 to 200
for these environments (Lopez and Falcon, 1999). years. Thus, for steps consisting of 100- to 200-mm
As a first attempt to model flow through step – pool rocks, movement occurs on the order of 2 to 15 years.
 A. Chin / Geomorphology 55 (2003) 125–137 131

On the other hand, only the largest steps, composed of Table 2

1-m boulders or larger, remain stable for periods of Potential energy dissipation at low flow
100 to 200 years. Most steps (median rock size of Reach Total Cumulative PE
relief height of dissipation
about 400 mm) restructure every 15 to 50 years on
(m) steps (m) (%)
average.
Cold Creek Preserve 20.88 18.63 89
These results are applied to Stunt Subreach for
Stunt 10.43 10.21 98
developing the conceptual model for step – pools. Helsley 6.63 5.65 85
Steps in Stunt Subreach are composed of clast sizes Bobcat 4.27 3.79 89
ranging from 300- to 792-mm. Thus, they are mobi- Jude 7.35 6.74 92
lized by flows with magnitudes ranging from f 7.1 Monte 4.21 3.47 82
All reaches 53.77 48.49 90
to 43.5 m3 s 1 (Fig. 4), depending on the particle
Big Canyon 8.81 7.82 89
sizes comprising each step. These critical discharges Sycamore Klein 7.55 7.38 98
correspond to frequencies of about 11 to 62 years in Creek Laughlin 16.02 16.02 100
the Santa Monica Mountains (Chin, 1998). In Fig. 4, Scott 20.32 20.12 99
the smallest step (Step A, average 300-mm rocks) and Bridge 6.98 6.83 98
Overlook 2.74 2.19 80
the largest step (Step B, average 792-mm rocks) are
Wood 2.85 2.70 95
highlighted for the purpose of illustrating the con- All reaches 65.27 63.25 97
ceptual model that follows.

3.2. Energy dissipation at low flow Although as a group, steps are effective in
reducing nearly all the elevation losses in study
Potential energy dissipation by step –pools in the reaches, a large proportion of the losses are accom-
study reaches of the Santa Monica Mountain plished by relatively few individuals. For example,
streams is nearly complete at low flow (Table 2). for Stunt Subreach (Fig. 4), Step B alone accounts
In Cold Creek, steps account for 82 – 98% of the for 17.8% of the elevation drops in this reach,
total elevation loss in the study reaches (90% for compared to only 1.6% for Step A. Thus, to the
all reaches combined). Similarly, the cumulative extent that step sizes vary within a given reach,
height of steps nearly equals the total drop in large steps in the Santa Monica Mountains play
elevation in the Big Sycamore Creek reaches more prominently in potential energy dissipation.
(ratio = 80 –100%; 97% for all reaches combined). Because in hydraulically controlled reaches such as
These values indicate that, through vertical fall, those in the Santa Monica Mountains, step height is
steps are effective in reducing flow energy that dependent upon the particle sizes comprising the
otherwise might be available for bed and bank step (Egashira and Ashida, 1991; Chin, 1999b;
erosion. Chartrand and Whiting, 2000), potential energy

Fig. 4. Critical flow for initiating motion of step particles, Stunt Subreach, Cold Creek.
132 A. Chin / Geomorphology 55 (2003) 125–137

reduction is expected to vary with step particle size Table 3

accordingly. Energy dissipation with increasing flows, Stunt Subreach
Dischargea Total drop Cumulative drop PE reduction
(cm) of water surface of water surface by steps (%)
3.3. Dissipation with increasing flows
elevation (m) elevation at steps (m)
0.05 1.85 1.66 90
Field survey and hydraulic modeling generated
0.30 1.66 1.35 81
five water surface profiles for Stunt Subreach (Fig. 2.50 1.51 1.01 67
5). They included the range of flows capable of 17.63 1.39 0.37 27
mobilizing steps in this reach: 0.05, 0.3, 2.5, 17.63, a
The discharge of 51.93 cm is not included because a flow of
and 51.93 m3 s 1. These discharges correspond to that magnitude is expected to break down the series of steps in this
flow frequencies ranging from < 1 to about 72 reach.
years in the Santa Monica Mountains (Chin,
1998). Visual inspection indicates that the low-flow potential energy dissipation by steps reduces to 67%
profile follows the step – pool bed configuration (Table 3), attributable to the larger steps only. The
closely, but the profiles depart with increasing high smaller steps, when submerged, therefore become
flows. The profiles for the two highest flows (17.63 ineffective as energy dissipators at higher flows.
and 51.93 cm) show evidence of the two large At the two highest flows, it is unclear whether
pools only, as the intervening smaller ones are vertical drops in the water surface profiles are caused
entirely submerged. This submergence is analogous directly by the steps underneath. This is probably the
to the drowning out of weirs or log jams. case for the discharge of 17.63 cm because flow is
Analysis of the reconstructed water surface profiles contained within the channel and is insufficient to
(Fig. 5) shows a trend of decreasing energy losses mobilize the larger steps in Stunt Subreach (Fig. 4).
with increasing discharge (Table 3). At the low flow The bed profile remains largely intact. However, the
of 0.05 m3s 1, because the water surface profile individual large steps contribute to only 27% of the
follows the staircase-like bed configuration closely total loss in water elevation at this flow (Table 3).
and creates a dominant vertical component to the On the other hand, the high-magnitude flow of 51.93
flow, potential energy dissipation by steps is nearly cm is best interpreted as flood waves because of the
complete at 90%. At a flow of 0.30 m3 s 1, as two of large depths and flow that spill overbank. According
the smallest steps are drowned out, the remaining to the hydraulic calculations (Chin, 1998), a flow of
steps are left to account for 81% of the total drop in this magnitude is expected to break down this series
water surface elevation in the reach. At 2.5 m3 s 1, of step – pools in Stunt Subreach (Fig. 4). Thus, not
because submergence occurs for the entire series of only are these steps ineffective in energy dissipation
small steps in the central portion of Stunt Subreach, at this discharge, the entire channel bed becomes a

Fig. 5. Water surface profiles from hydraulic modeling.

 A. Chin / Geomorphology 55 (2003) 125–137 133

dependent variable that adjusts to flow and energy tive, channel restructuring occurs. This general idea
conditions. can be refined to consider the two energy-dissipating
mechanisms by steps: potential energy loss through
3.4. Point of submergence—defining the changing vertical falls, which approximates spill resistance, and
role of step – pools kinetic energy dissipation by form and grain rough-
ness. As potential energy reduction diminishes at high
When steps are submerged, they lose their distinct flows (Table 3), energy is available for conversion to
property of inducing vertical fall. More potential kinetic energy and sediment transport, which become
energy is then available for conversion to kinetic more important, until flows are sufficient to move
energy and sediment transport; spill resistance gives boulders and reform the channel. Thus, a general
way to a dominance in form and grain roughness in model for the significance of step –pools needs to
energy dissipation. Thus, at the point of submergence, incorporate the relative dominance of potential and
steps no longer function as energy dissipators; they kinetic energy dissipation. Fig. 6 illustrates this
become more like other bed forms and roughness model, which is an expansion of the right side of
elements in the channel system. They are analogous to the theoretical diagram of Fig. 2.
pool – riffle sequences when submerged.
In Stunt Subreach (Fig. 5), the point of submer- 4.2. Application to specific test cases
gence varies for each individual step because of
variations in their sizes. At the low flow, each of the To define its dimensions, the general model is
seven steps contributes to the total elevation loss in applied to the two steps highlighted earlier in Stunt
this reach. Therefore, all individual steps participate as Subreach, Step A and Step B (Fig. 4). They are the
energy dissipators that induce spill resistance, even smallest and largest steps in the reach, respectively,
though the extent varies between individual steps offering greatest contrast. Step A and Step B are
because of varying sizes. As steps become increas- composed of 300- and 793-mm particles, respectively.
ingly drowned at higher flows, an increasing number Working backward from the low-flow form (Fig.
of steps cease to function as energy dissipators, and 7A), hydraulic calculations showed that a step of this
they increasingly become more like other bed forms in size (300-mm) would require flows of about 7.1 m3
stream channels. Thus, at the 2.50 cm flow, three of s 1 to mobilize. This represents the channel-forming
the smaller steps in the central portion of the reach flow and corresponds to a recurrence interval of about
become bed forms that impart form and grain rough- 11 years in Stunt Subreach. Working forward with
ness. At the highest flows, only the largest individual increasing discharge, analysis of water surface profiles
steps remain as energy dissipators. Because large indicated that this small step would become sub-
steps are submerged less frequently, they serve more merged and cease functioning as an energy dissipator
prominent roles as energy dissipators over a longer
time scale. For small steps, their role changes readily
when they are drowned during smaller floods.

4. A conceptual model for step –pools

4.1. Toward a general model for step – pools

The point of submergence defined above can be

connected with the channel-forming flow determined
by the hydraulic analysis to develop a conceptual
model describing the function of step –pool systems.
It was originally suggested (Fig. 2) that, as flow
increases and energy dissipation ceases to be effec- Fig. 6. A general model for energy dissipation by step – pools.
134 A. Chin / Geomorphology 55 (2003) 125–137

Fig. 7. Conceptual model for step – pools, Stunt Subreach. (A) 300-mm rocks; (B) 792-mm rocks.

at a flow of about 0.3 m3s 1, or every year on 4.3. A conceptual model for steps of varying sizes
average. Therefore, when flow exceeds 0.3 m3 s 1,
the step behaves more like a roughness element. A series of nested diagrams for varying step sizes
Channel hydraulics are analogous to those of pool – results in a conceptual model defining the changing
riffle channels and include kinetic energy dissipation role of step – pools as energy dissipators (Fig. 8). Each
and sediment transport, until flow reaches a critical of the boxes represents a step of a certain size,
magnitude of about 7.1 m3 s 1. Presumably, at this arbitrarily chosen from 300- to 800-mm. The time
point, the step breaks down and reforms, and the scale is linear in the diagram, so the length of the box
whole cycle begins again. indicates the average life span for each step. The
The same explanation applies for the larger step, vertical dashed lines separate the zones characterized
composed of 792-mm rocks (Fig. 7B). The difference is by potential and kinetic energy dissipation. Therefore,
the relative dominance of potential versus kinetic it identifies the point at which a step makes that
energy dissipation. In this case, the larger step has a transition, or shift, from functioning as an energy
longer life span, with restructuring expected to occur dissipator to behaving like a roughness element in
once every 62 years on average in Stunt Subreach. the river system.
Because submergence occurs at relatively high flows, A point that emerges clearly is how the size of a step
the step behaves as an energy dissipator over a much dictates its stability and its role in energy dissipation.
larger range of flows and as a roughness element First, the stability of steps obviously depends on their
relatively seldom. Thus, the zone dominated by kinetic size as well as flow conditions. Small steps composed
energy dissipation is much smaller. of 300-mm rocks are unstable in comparison, breaking
 A. Chin / Geomorphology 55 (2003) 125–137 135

Fig. 8. Conceptual model for steps of varying sizes.

down every 11 years on average in Stunt Subreach. On of the elevation losses in the study reaches in the Santa
the other hand, large steps on the order of 800 mm are Monica Mountains. Through vertical falls, steps reduce
structures that remain in place for periods of over 60 the potential energy that otherwise might be available
years. Second, the role of steps in energy dissipation for conversion into a longitudinal component of work,
also changes according to their size. Within their life thereby offsetting steep gradients. Third, energy dis-
spans, the relative dominance of potential energy dis- sipation by steps diminishes at increasing flows. As
sipation increases with increasing size of the step. This step – pools become increasingly drowned at high
is clearly shown by the shifting of the dashed lines to flows, spill resistance gives way to a dominance in
the right within the boxes in Fig. 8, so that large steps form resistance and, to a lesser extent, grain resistance.
would function as energy dissipators through a large Thus, the manner in which energy dissipation occurs
proportion of the time, whereas small steps would varies with discharge. Fourth, the point at which steps
behave more as roughness elements much of the time. become submerged marks a transition in their role as
energy dissipators to roughness elements in the fluvial
system. Because submergence occurs more readily for
5. Summary and conclusions small steps, these steps function more as roughness
elements over a longer time scale, whereas large steps
Analysis of step – pools in streams in the Santa serve more prominent roles as energy dissipators.
Monica Mountains of California yields answers to the These findings permit the development of a con-
four stated research questions for this paper and ceptual model that defines the changing role and
insights into the significance of step – pools as energy significance of step – pools in the larger fluvial system.
dissipators in the fluvial system. First, step –pools in The model articulates how the function of step – pools
streams in the Santa Monica Mountains are adjustable changes over time with variations in discharge. It also
bed forms that are capable of being restructured. The points to the importance of the step size in determining
flows required to destabilize step – pool sequences its role in energy dissipation and in its interactions with
depend on the particle sizes comprising steps. For steps channel hydraulics. Small steps often function as
in Stunt Subreach, with clast sizes ranging from 300 to roughness elements that impart form and grain rough-
800 mm, these flows have recurrence intervals of about ness, rather than as energy dissipators that induce spill
11 to 62 years. Second, steps are effective energy resistance. These steps regulate channel hydraulics
dissipators at low flows. Steps account for 80– 100% comparatively seldom, but they are important as
136 A. Chin / Geomorphology 55 (2003) 125–137

dependent variables that adjust to the prevailing flow References

and energy conditions. On the other hand, the role of
energy dissipation becomes primary for large steps. Abrahams, A.D., Li, G., Atkinson, J.F., 1995. Step – pool streams:
These steps function predominantly as energy dissipa- adjustment to maximum flow resistance. Water Resources Re-
search 31 (10), 2593 – 2602.
tors during the evolution of the fluvial system. There-
Ashida, K., Takahashi, T., Sawada, T., 1976. Sediment yield and
fore, they maintain an independent status throughout transport on a mountainous small watershed. Disaster Preven-
much of the time, regulating and controlling flow and tion Research Institute Bulletin, Kyoto University 26 (240),
channel hydraulics. 119 – 144.
The model developed in this paper has flow magni- Bathurst, J.C., 1987. Flow resistance in boulder-bed streams. In:
Hey, R.D., Bathurst, J.C., Thorne, C.R. (Eds.), Gravel-Bed Riv-
tudes and time dimensions that apply to Stunt Reach in
ers. Wiley, Chichester, UK, pp. 403 – 465.
Cold Creek. However, the model is generally applica- Beven, K., Gilman, K., Newson, M., 1979. Flow and flow routing in
ble if frequency –magnitude relations are defined for upland channel Networks. Hydrological Sciences 24, 303 – 325.
other areas and if the boxes and separations in the Bonner, V., Brunner, G., Jensen, M., 1994. HEC river analysis
model are interpreted as zones rather than clear boun- system (HEC-RAS). In: Cotroneo, G.V., Rumer, R.R. (Eds.),
Hydraulic Engineering ’94, vol. 1. American Society of Civil
daries. The model offers a new articulation for the Engineers, New York, pp. 376 – 380.
geomorphic significance of step – pools in mountain Brunner, G.W., Piper, S.S., 1994. Improved hydraulic features of
streams, and it serves as a useful template for a more the HEC river analysis system (HEC-RAS). In: Cotroneo, G.V.,
complete understanding of step – pools over a longer Rumer, R.R. (Eds.), Hydraulic Engineering ’94, vol. 1. Ameri-
time scale. Because step – pools characterize mountain can Society of Civil Engineers, New York, pp. 502 – 506.
Carling, P.A., Wood, N., 1994. Simulation of flow over pool – riffle
areas that cover a sizable portion of the earth surface,
topography: a consideration of the velocity-reversal hypothesis.
increased knowledge of step – pools is important in the Earth Surface Processes and Landforms 19, 319 – 332.
broader understanding of mountain geomorphology Chartrand, S.M., Whiting, P.J., 2000. Alluvial architecture in head-
and earth surface systems. water streams with special emphasis on step – pool topography.
Earth Surface Processes and Landforms 25, 583 – 600.
Chin, A., 1989. Step – pools in stream channels. Progress in Phys-
ical Geography 13 (3), 391 – 408.
Acknowledgements Chin, A., 1998. On the stability of step – pool mountain streams.
The Journal of Geology 106, 59 – 69.
The Association of American Geographers, Arizona Chin, A., 1999a. On the origin of step – pool sequences in mountain
State University (Office of the Vice President for streams. Geophysical Research Letters 26 (2), 231 – 234.
Chin, A., 1999b. The morphologic structure of step – pools in
Research), and Texas A&M University (Office of the
mountain streams. Geomorphology 27 (3 – 4), 191 – 204.
Vice President for Research) supported portions of the Chin, A., 2002. The periodic nature of step – pool mountain
data collection and analysis contained in this paper. The streams. American Journal of Science 302 (2), 144 – 167.
following individuals provided critiques to various Chin, A., Gregory, K.J., 2001. Urbanization and adjustment of
drafts of this manuscript and helpful theoretical ephemeral stream channels. Association of American Geogra-
phers Annals 91 (4), 595 – 608.
discussions: Will Graf (University of South Carolina),
Church, M., Jones, D., 1982. Channel bars in gravel-bar rivers. In:
Linda O’Hirok (California State University), Bernie Hey, R.D., Bathurst, J.C., Thorne, C.R. (Eds.), Gravel-Bed Riv-
Bauer (University of Southern California), John ers. Wiley, Chichester, UK, pp. 291 – 338.
Wolcott (University of Southern California), Doug Costa, J.E., 1983. Paleohydraulic reconstruction of flash-flood
Sherman (Texas A&M University), Judy Haschen- peaks from boulder deposits in the Colorado Front Range. Geo-
logical Society of America Bulletin 94, 986 – 1004.
burger (University of Auckland), and Ken Gregory
Curran, J.H., Wohl, E.E., 2003. Large woody debris and flow re-
(University of London). Richard Dixon (Southwest sistance in step – pool channels, Cascade Range, Washington.
Texas State University) and an anonymous reviewer Geomorphology 51, 141 – 157.
made further helpful comments. Lei Wang, Paul de Jong, C., Ergenzinger, P., 1998. Dynamic roughness, sediment
Rindfleisch, and Zengwang Xu (Texas A&M Univer- transport and flow structures in a mountain stream. In: Klinge-
man, P.C., Beschta, R.L., Komar, P.D., Bradley, J.B. (Eds.),
sity) were research assistants; Barbara Trapido-Lurie
Gravel Bed Rivers in the Environment. Water Resources Pub-
(Arizona State University) generously offered techni- lications, Highlands Ranch, CO, pp. 39 – 60.
cal expertise. Field assistants are too numerous to Duckson, D.W., Duckson, L.J., 1995. Morphology of bedrock step
name, but I am grateful to all of them. pool systems. Water Resources Bulletin 31 (1), 43 – 51.
 A. Chin / Geomorphology 55 (2003) 125–137 137

Duckson, D.W., Duckson, L.J., 2001. Channel bed steps and pool Miller, B.A., Wenzel, H.G., 1985. Analysis and simulation of low
shapes along Soda Creek, Three Sisters Wilderness, Oregon. flow hydraulics. Journal of Hydraulic Engineering 111 (12),
Geomorphology 38, 267 – 279. 1429 – 1446.
Egashira, S., Ashida, K., 1991. Flow resistance and sediment trans- Montgomery, D.R., Buffington, J.M., 1997. Channel-reach mor-
portation in streams with step – pool bed morphology. In: Arma- phology in mountain drainage basins. Geological Society of
nini, A., Di Silvio, G. (Eds.), Fluvial Hydraulics of Mountain America Bulletin 109 (5), 596 – 611.
Regions. Lecture Notes in Earth Sciences, vol. 37. Springer- O’Connor, J.E., Webb, R.H., 1988. Hydraulic modeling for paleo-
Verlag, New York, pp. 45 – 58. flood analysis. In: Baker, V.R., Kochel, R.C., Patton, P.C. (Eds.),
Ergenzinger, P., 1992. River bed adjustments in a step – pool sys- Flood Geomorphology. Wiley, New York, pp. 393 – 402.
tem: Lainbach, Upper Bavaria. In: Hey, R.D., Billi, P., Thorne, Peterson, D.F., Mohanty, P.K., 1960. Flume studies of flow in steep,
C.R., Tacconi, P. (Eds.), Dynamics of Gravel Bed Rivers. Wiley, rough channels. Journal of the Hydraulics Division, ASCE 86,
Chichester, UK, pp. 415 – 430. 55 – 79.
Gintz, D., Hassan, M.A., Schmidt, K.-H., 1996. Frequency and Scheuerlein, H., 1999. Morphological dynamics of step – pool sys-
magnitude of bed load transport in a mountain river. Earth Sur- tems in mountain streams and their importance for riparian eco-
face Processes and Landforms 21, 433 – 445. systems. In: Jayawardena, A.W., Lee, J.M., Wang, Z.Y. (Eds.),
Grant, G.E., Swanson, F.J., Wolman, M.G., 1990. Pattern and origin River Sedimentation. Balkema, Rotterdam, The Netherlands,
of stepped-bed morphology in high-gradient streams, Western pp. 205 – 210.
Cascades, Oregon. Geological Society of America Bulletin 102, Stuve, P.E., 1990. Spatial and temporal variation of flow resistance
340 – 352. in an Alpine river. In: Lang, H., Musy, A. (Eds.), Hydrology in
Hayward, J.A., 1980. Hydrology and stream sediments from Tor- Mountainous Regions: I. Hydrological Measurements. The
lesse stream catchment. Special Publication, vol. 17. Tussock Water Cycle, vol. 193. International Association of Hydrologi-
Grasslands and Mountain Lands Institute, Lincoln College, New cal Sciences, Wallingford, UK, pp. 307 – 314.
Zealand. 236 pp. U.S. Army Corps of Engineers, 1990. HEC-2 Water Surface Pro-
Heede, B.H., 1972. Influences of a forest on the hydraulic ge- files Users Manual. Hydrologic Engineering Center, Davis, CA.
ometry of two mountain streams. Water Resources Bulletin 8 U.S. Army Corps of Engineers, 1998. HEC-RAS River Analysis
(3), 523 – 530. System Users Manual. Hydrologic Engineering Center, Davis,
Heede, B.H., 1981. Dynamics of selected mountain streams in the CA.
western United States of America. Zeitschrift fur Geomorpho- Vischer, D.L., 1995. Types of energy dissipators. In: Vischer, D.L.,
logie 25 (1), 17 – 32. Hager, W.H. (Eds.), Energy Dissipators. IAHR Hydraulic Struc-
Jarrett, R.D., 1984. Hydraulics of high-gradient streams. Journal of tures Design Manual, vol. 9. Balkema, Rotterdam, The Nether-
Hydraulic Engineering 110 (11), 1519 – 1539. lands, pp. 9 – 23.
Keller, E.A., Florsheim, J.L., 1993. Velocity-reversal hypothesis: a Whittaker, J.G., Davies, T.R.H., 1982. Erosion and sediment trans-
model approach. Earth Surface Processes and Landforms 18, port processes in step – pool torrents. In: Walling, D.E. (Ed.),
733 – 740. Recent Developments in the Explanation and Prediction of
Keller, E.A., Swanson, F.J., 1979. Effects of large organic material Erosion and Sediment Yield. Proc. Exeter Symp, vol. 137.
on channel form and fluvial processes. Earth Surface Processes International Association of Hydrological Sciences, Paris, pp.
4, 361 – 380. 99 – 104.
Lee, A.J., Ferguson, R.I., 2002. Velocity and flow resistance in Wohl, E.E., 2000a. Mountain Rivers. Water Resources Monograph
step – pool streams. Geomorphology 46, 59 – 71. Series, vol. 14. American Geophysical Union, Washington, DC.
Lenzi, M.A., 2001. Step – pool evolution in the Rio Cordon, 320 pp.
Northeastern Italy. Earth Surface Processes and Landforms Wohl, E.E., 2000b. Substrate influences on step – pool sequences in
26, 991 – 1008. the Christopher Creek drainage, Arizona. The Journal of Geology
Lenzi, M.A., Billi, P., D’Agostino, V., 1997. Effects of an extremely 108, 121 – 129.
large flood on the bed of a steep mountain stream. In: Sam, S.Y., Wohl, E.E., Thompson, D.M., 2000. Velocity characteristics along a
Langendoen, E.J., Shields, F.D. (Eds.), Proceedings of the Con- small step – pool channel. Earth Surface Processes and Land-
ference on Management of Landscapes Disturbed by Channel forms 25, 353 – 367.
Incision: Stabilization, Rehabilitation, Restoration. Oxford, Mis- Wohl, E.E., Thompson, D.M., Miller, A.J., 1999. Canyons with
sissippi, pp. 1061 – 1066. undulating walls. Geological Society of America Bulletin 111
Leopold, L.B., 1994. A View of the River. Harvard Univ. Press, (7), 949 – 959.
Cambridge, MA. 298 pp. Young, L.E., Cruff, R.W., 1967. Magnitude and frequency of floods
Lopez, J.L., Falcon, M.A., 1999. Calculation of bed changes in in the United States, Part 11, Pacific Slope Basins in California.
mountain streams. Journal of Hydraulic Engineering 125 (3), U.S. Geological Survey Water Supply Paper, vol. 1685. U.S.
263 – 270. Government Printing Office, Washington, DC.
Marston, R.A., 1982. The geomorphic significance of log steps in Zimmermann, A., Church, M., 2001. Channel morphology, gradient
forest streams. Association of American Geographers Annals profiles and bed stresses during flood in a step – pool channel.
72, 99 – 108. Geomorphology 40, 311 – 327.
 Geomorphology 55 (2003) 139 – 154
www.elsevier.com/locate/geomorph

Equal-mobility bed load transport in a small, step-pool channel

in the Ouachita Mountains
Daniel A. Marion a,*, Frank Weirich b
a
Southern Research Station, USDA Forest Service, Oxford, MS 38655-4915, USA
b
Geoscience Department, University of Iowa, Iowa City, IA 52242, USA
Received 4 December 2001; received in revised form 2 July 2002; accepted 10 March 2003

Abstract

Equal-mobility transport (EMT) of bed load is more evident than size-selective transport during near-bankfull flow events in
a small, step-pool channel in the Ouachita Mountains of central Arkansas. Bed load transport modes were studied by simulating
five separate runoff events with peak discharges between 0.25 and 1.34 m3/s (1.0- to 1.6-year recurrence intervals) in a natural
channel using controlled releases from a storage tank. EMT occurrence was investigated using four different bed load
relationships suggested by previous research. With each of these approaches, the relationship of a given bed load characteristic
(Dmax, distribution percentile, displacement distance and skewness) to some independent factor (s*, c s and grain size) was
assessed to determine which transport mode was evident. Regression models derived using combinations of these four
relationships with different datasets provide seven separate tests. Five of the seven tests indicate that EMT occurred or was
predominant. Several reasons may explain the apparent contradictory results, but the confounding effects of changes in the
structural arrangements of bed material prior to or during the events seem particularly important.
Published by Elsevier Science B.V.

Keywords: Bed load transport; Equal mobility; Size-selective; Step-pool channel; Bed material size; Bankfull streamflow

1. Introduction (Andrews, 1983). It challenges the traditional, size-

selective transport (SST) theory (Shields, 1936) that
The equal-mobility transport (EMT) theory has holds that bed load size is directly proportional to
been advanced to explain how variability in bed displacement forces (e.g., shear stress, discharge,
material sizes and exposure of individual grains affects stream power).
bed load entrainment and grain size (Parker et al., Past field testing for the occurrence of EMT has
1982). The theory states that for a channel bed com- largely been done in gravel-bed channels where the
posed of heterogeneous grain sizes, the larger grains largest grains are cobble-sized or smaller (e.g., An-
are more easily entrained than smaller ones because drews, 1983; Ashworth and Ferguson, 1989; Komar
they are relatively more exposed to lift and drag forces and Shih, 1992). Step-pool channels differ from these
channels in that they exhibit a wider range of bed ma-
* Corresponding author. Tel.: +1-662-234-2744x248; fax: +1-
terial sizes (typically sands and finer to small boulders
662-234-8318. and larger; Heede, 1972; Laronne and Carson, 1976;
E-mail address: [email protected] (D.A. Marion). Best and Keller, 1986; Grant et al., 1990; Ergenzinger,

0169-555X/03/$ - see front matter. Published by Elsevier Science B.V.

doi:10.1016/S0169-555X(03)00137-5
140 D.A. Marion, F. Weirich / Geomorphology 55 (2003) 139–154

1992; Ryan, 1994) and a stepped longitudinal structure. 2. Study site description and methods
Only Blizard and Wohl (1998) have considered
whether EMT occurs in step-pool channels. While they We tested the bed load transport mode using data
concluded there was a general trend toward EMT in the collected during five simulated flow events in a step-
step-pool channel they studied in the Colorado Rocky pool reach located within the Ouachita Mountains of
Mountains, they also found evidence suggesting that central Arkansas.
SST occurred.
The objective of this study is to assess whether 2.1. Study site description
EMT or SST occurs in a step-pool channel during near-
bankfull streamflow events. Four different testing The study reach is located on an unnamed tributary
methods are used to evaluate which bed load transport of Little Bear Creek, hereafter referred to as Toots
mode occurs. The current study was undertaken as part Creek (Figs. 1 and 2). Toots Creek has environmental
of a larger investigation into how bed load and channel characteristics typical of the region. Annual precip-
morphology interact in a typical step-pool channel in itation averages 130 cm, occurring almost entirely as
the Ouachita Mountains (Marion, 2001b). rain; and streamflows are ephemeral to intermittent.

Fig. 1. Location of the Toots Creek study area in the Ouachita Mountains of Arkansas. Data sources: conterminous USA (Environmental
Systems Research Institute, 1983); physiographic sections (Fenneman and Johnson, 1946); Toots Creek area contours, roads, streams and water
bodies (Ouachita National Forest).
 D.A. Marion, F. Weirich / Geomorphology 55 (2003) 139–154 141

Fig. 2. Toots Creek study reach, local geology and water storage tank. Geology compiled from field mapping on 18 January 2001 by Daniel A.
Marion and Charles G. Stone.

The catchment area above the study reach is 39 ha, shortleaf pine (Pinus echinata Mill.) overstory and a
overall relief is 140 m, and hillslopes range from 15 to mixed hardwood understory including white oak
30%. Vegetation is predominantly composed of a (Quercus alba L.), red oak (Quercus rubra L.) and
142 D.A. Marion, F. Weirich / Geomorphology 55 (2003) 139–154

Fig. 3. Toots Creek study reach looking upstream from just above Bridge 4.

various hickories (Carya spp.) (Marion and Malanson, structures, where cobble-sized and larger grains lean
in press). against one another, dipping upstream and strongly
The 96-m study reach has banks composed of resisting individual grain entrainment. Open struc-
mixed colluvial and alluvial deposits. Bankfull widths tures, where grains are loose and have minimal over-
average 4.2 m and the channel has a weighted (by lap, account for the smallest portion of the study
channel length) mean gradient of 8.8% (Fig. 3). The reach. Here, grains are generally smaller, better sorted
reach contains segments representing four different and occur in relatively small patches within step
hydraulic unit types (Grant et al., 1990): step, cascade, treads (Marion and Weirich, 1999).
rapid and riffle (Fig. 2). The reach-wide, grain-size
distribution ranges from silts to large boulders. Grain
sizes within hydraulic unit types display the same
relative size relationships reported by Grant et al.
(1990): cascades are coarser than rapids which are
coarser than riffles (Fig. 4).
The channel bed is almost entirely composed of
sandstone clasts with only very small amounts of
shale and quartz. Grain sphericity is predominantly
bladed to very elongate (after Sneed and Folk, 1958),
while particle roundness is mostly subangular to
subrounded. Across the bed surface, neighboring
grains exhibit all three of the structural arrangements
originally described by Laronne and Carson (1976).
Closed structures occur over most of the study reach
Fig. 4. Surface grain-size distributions for hydraulic unit types
wherein grains are in close contact and voids between (Grant et al., 1990) within the Toots Creek study reach. Grain sizes
the large clasts are infilled by smaller grains. Most of determined by pebble count of grains >1.0 mm. The single-step
the particles forming step risers exhibit imbricate segment is excluded due to the small number of clasts present.
 D.A. Marion, F. Weirich / Geomorphology 55 (2003) 139–154 143

Coarse (diameter>10 cm) woody debris loading on five consecutive days and sequenced so that Event
was small within the study reach and appeared to have 1 had the smallest peak flow while Event 5 had the
very little effect on bedload transport. The Toots largest (Fig. 6). Bankfull discharge was estimated to
Creek volume (2.45 m3/100 m) is very similar to the be 1.11 m3/s (1.4-year recurrence interval) from
mean of 2.2 m3/100 m computed from data at 27 other channel features and computed hydraulic values.
sites in the Ouachita Mountains (Fowler, 1994), but Bed load was assessed in two ways. During each
differs markedly in both amount and influence on event, bed load transport rate was sampled at Bridge 2
sediment routing from those reported for step-pool (Fig. 2) using a 73-mm Helley – Smith sampler.
channels elsewhere (e.g., Heede, 1972; Megahan and Bridge 2 occurs within a riffle hydraulic unit type
Nowlin, 1976; Nakamura and Swanson, 1993). (Grant et al., 1990) that has a D50 of 104 and 51 mm
for the bed surface and subsurface, respectively.
2.2. Experimental methods Three 1-min samples were taken at three fixed loca-
tions across the bridge and were composited into one
Five individual flow events with peak discharges sample for analysis. Composite samples were gener-
ranging from 0.25 to 1.34 m3/s (1.0- to 1.6-year ally taken at 4– 5-min intervals, depending on the
recurrence intervals) were simulated. All flow events event duration. Discharge and other hydraulic char-
were created using controlled releases from a storage acteristics associated with the bed load data were
tank (Fig. 5) (see Marion and Weirich, 1997, for determined at Bridge 2 using methods described in
details on system operation). Events were produced Marion (2001b).

Fig. 5. Water storage tank and flume used to simulate near-bankfull events at Toots Creek.
144 D.A. Marion, F. Weirich / Geomorphology 55 (2003) 139–154

Fig. 6. Bed load transport rates and discharges used during the July 1995 experiments at Toots Creek. Bar widths indicate the 3.0-min time
interval over which bed load was measured during Events 1 – 5.
 D.A. Marion, F. Weirich / Geomorphology 55 (2003) 139–154 145

Bed load displacement distance was determined class containing the Dmax as his measure of Dmax size,
using 79 bed particles between 34 and 380 mm (b- whereas we used the bed load D95. In addition, follow-
axis) in size. Tracers were initially installed along ing Ashworth and Ferguson (1989), we used the bed
eight cross-sections evenly spaced within the central surface D50 (D50b) rather than the subsurface D50 as
portion of the study reach (Fig. 2). At each section, 6– was done by Andrews (1983) because the lack of
10 grains were randomly selected in proportion to the extensive pavement breakdown during the five events
reach-wide, grain-size distribution, painted white and suggested that the bed surface was the primary bed load
replaced to approximate their original orientation and source and a more appropriate gauge of grain exposure.
packing. All tracers were sandstone and had the same Thus, the relationship we examined was s*aDc 95/D50b.
shape characteristics as noted above. Their initial
locations fell within a large variety of microsite 2.3.2. Bed load displacement distance vs. grain size
situations (e.g., within pocket pools, adjacent to larger A second approach is based upon the assumption
grains, in clusters of similar size grains or within finer that if SST occurs then bed load displacement distance
grained bed patches). Tracer position was measured decreases with increasing grain size; i.e., an inverse
by tape after each event and used to compute displace- linear relationship should be evident (Ashworth and
ment distances. Ferguson, 1989). Thus, in a linear regression between
During the five events, bed load moved under distance and grain size, b1 should be negative if SST
Phase 1 or Phase 2 (after Warburton, 1992) condi- occurs and near zero (i.e., no relationship) if EMT
tions. Dmax sizes ranged from 74 to 180 mm and unit occurs. Using the tagged rock data for each event,
bed load transport rates ranged from 3.5  10 6 to linear regressions were computed for both grain weight
3.3  10 2 kg/s/m (Fig. 6) (Marion, 2001a). Marion and b-axis length (two measures of size) against
(2001a) showed that these rates span the lower to displacement distance. In addition, both size metrics
middle portion of the overall range defined by five were regressed against cumulative displacement over
previous studies of step-pool channels. all five events. Event 3 data were not considered
because only two tracers moved during this event.
2.3. Testing methods
2.3.3. Grain size of selected bed load distribution
Four approaches have been previously advanced or percentiles vs. shear stress
suggested for determining whether EMT occurs. In a third approach, if grain size increases with
increasing shear stress (s), as predicted by SST, then
2.3.1. Critical dimensionless shear stress vs. relative a positive relationship should be evident in the regres-
bed material grain size sion model (i.e., b1>0) for any given bed load distribu-
According to Parker et al. (1982), an inverse rela- tion percentile (e.g., D50) as a function of s (Komar and
tionship (i.e., regression slope coefficient [b1] = 1) Shih, 1992). Conversely, if b1 = 0, then equal-mobility
should exist between critical dimensionless shear stress is defined as occurring. We derived models for the D05,
(s*)
c and maximum bed load size (Dmax) when EMT D16, D50, D84 and D95 bed load percentiles against s.
occurs. Andrews (1983) tested this using data from All models were tested using a general linear model
natural streams by regressing s*c against Dmax/D50sub, (GLM) to assess both the effect of s and event magni-
where D50sub is the median subsurface grain size. D50 is tude in case the latter affected the s –Dx relationship.
used to standardize Dmax for bed material size differ-
ences between different locations. This ratio is assumed 2.3.4. Bed load grain-size skewness vs. shear stress
to reflect the relative exposure of a grain within the Lastly, if SST occurs, then a progressive increase in
supporting matrix. Andrews found a highly significant Dmax size with increasing s must also occur (Komar
inverse relationship between Dmax/D50sub and s*. c This and Shih, 1992). The latter would cause bed load
analysis assumes that the dimensionless shear stress grain-size skewness to increase in magnitude with
computed when each bed load sample was taken is the increasing s (i.e., b1>0). EMT would be apparent if
critical value necessary to entrain the Dmax measured. b1 V 0. We derived a s – skewness model and tested
Andrews used the geometric mean of the half-u size whether b1>0. As with the previous test, we also used
146 D.A. Marion, F. Weirich / Geomorphology 55 (2003) 139–154

a GLM to assess whether the s– skewness relationship magnitude ( P = 0.46) so data from all five events
was affected by event magnitude. were used together. Two models were tested: one
With each of these approaches, the relationship of a based on all data and one excluding the one case
given bed load characteristic to some independent where D95>32 mm (Fig. 7). The case in question has a
forcing factor was assessed to determine which trans- very large leverage value, thus, a large effect on the
port mode was evident. The nature of each test deter- model slope, and falls within the size range in which
mined whether the proposed hypothesis was testing for Helley –Smith accuracy is most questionable.
the presence of EMT or SST. The first method des- In a similar analysis, Ashworth and Ferguson
cribed above tested for EMT (Ha: b1 = 1), while the (1989) excluded all cases where s*>1.0
c on the assump-
other three tested for SST (Ha: b1>0). Notably, all of tion that gravel-size grains could be suspended under
these approaches were originally developed using data such conditions and thus escape sampling. However,
from gravel-bed channels that lacked the significant an analysis of suspended sediment data during all five
structural roughness features, such as boulder clusters events showed no evidence of significant gravel sus-
and steps, that are common in step-pool channels. pension (Marion, 2001b). Therefore, these cases were
included in the present analysis.
The probability that b1 p 1 is quite low (0.04) for
3. Results both models, indicating that EMT occurred. However,
while both model slopes are significant (b1 p 0) at the
Results from the tests differed in which transport 0.05 level, both models also have very low R2 values
they indicated was operational during near-bankfull (0.33 and 0.18) and relatively large standard errors for
events. these slope coefficients. Thus, while the occurrence of
EMT is more evident than SST (where b1 p 1), it is
3.1. s*c vs. Dmax/D50b not conclusively demonstrated by these models.

This test indicated that EMT occurred when data 3.2. Tracer displacement distance vs. grain size
from all events were considered together. However,
confidence in this conclusion is not high. The s*c – Initial results indicate that SST occurred in Event 1
Dmax/D50b relationship does not vary with event but not in Events 2, 4 and 5 (see Table 1). Transport

Fig. 7. Critical dimensionless shear stress as a function of relative bed load grain size during all five experimental events at Toots Creek.
 D.A. Marion, F. Weirich / Geomorphology 55 (2003) 139–154 147

Table 1
Tracer recovery rates and slopes for regression models of displacement distance vs. grain size at Toots Creek (data are for Events 1 – 5 and
cumulative displacement)
Event No. of Recovery No. tracers Slope coefficients
tracers rate (%) displaced Using all data Excluding outliers
b-axis Weight b-axis Weight
1 79 100 16 0.66** 0.18** N/A N/A
2 79 100 7 0.01 0.06 0.68 0.40**
3 77 97 2 N/A N/A N/A N/A
4 77 97 10 0.15 0.03 0.43 N/A
5 77 97 15 0.63 0.49 1.00* 0.46**
Cumulative 77 97 30 0.13 0.18 N/A N/A
N/A = not applicable. Number of outliers excluded is 0, 1, 1 and 3 for Events 1, 2, 4 and 5, respectively.
* Probability of regression slope z 0 is V 0.10.
** Probability of regression slope z 0 is V 0.05.

distance is not related to bed load grain size for Event 4 or for cumulative distance over all five events.
Events 2, 4 and 5 but is related for Event 1. All The difference between these two sets of results
individual events and cumulative displacement do suggests that larger sample sizes may be needed for
have models with b1 < 0, but only for Event 1 is the this test to produce a more certain answer.
coefficient significant at the 0.05 level. Results are Schmidt and Ergenzinger (1992) observed grain
very similar for both b-axis length and grain mass. weight and transport distance relationships similar to
Therefore, only the relationship with grain mass is those at Toots Creek in the Lainbach, a step-pool
discussed below. stream in southern Germany. They measured signifi-
Despite the initial results, the majority of the data cant negative correlations between grain weight and
for Events 2, 4 and 5 do appear to show a clear trend travel distance in two of three separate events. How-
(see Fig. 8). Further analysis indicates that significant ever, they found that at best, they could only explain
relationships can be derived for Events 2 and 5 if a 33% of the travel distance variance even when includ-
limited number of samples are excluded. These ing a factor for grain shape along with weight. There-
possible outliers are identified in Fig. 8. For Event fore, they concluded that there was only a weak
2, exclusion of one possible outlier produces a tendency for travel distance to decrease as weight
significant model. The excluded case has both high increased. These findings are similar to Toots Creek in
leverage and a large residual value, thus, it has a that relationship strength varies significantly by event,
large effect on both b1 and its probability of not being that general trends are evident, but that these trends
less than zero. For Event 5, exclusion of three are only weakly expressed.
possible outliers also results in a substantially differ-
ent model. The excluded cases all have very large 3.3. Bed load Dx vs. s
residual values, causing the original mean square
error term (0.50) to be twice that for Event 4 and This test produces differing results depending on
over seven times those for Events 1 and 2. The how event data are combined. SST is not evident for
revised model has a b1 with a P value of 0.03 and any except perhaps the largest percentile when data
mean square error of 0.14. No improvement is from all events are considered together. Probabilities
achieved for Event 4 by excluding the two possible are >0.29 for all percentiles except D95 where P =
outliers. 0.10. Such a dataset is similar to that used by Komar
Thus, SST is not consistently evident, but is indi- and Shih (1992) in that it is composed of observations
cated in particular events. SST is most apparent for from multiple events.
Event 1, but Events 2 and 5 may also exhibit such However, a GLM test indicates that the s – Dx
transport. In contrast, SST cannot be discerned for relationship varies between Events 1 – 3 and 4 – 5.
148 D.A. Marion, F. Weirich / Geomorphology 55 (2003) 139–154

Fig. 8. Tracer travel distance vs. grain mass for individual events and total distance over all events. Possible outliers denoted with dot symbol.
Regression models are based on all data (solid line) and with possible outliers excluded (dashed).

SST is evident during Events 4 – 5, but not during percentiles at the 0.05 level, but R2 values are not very
Events 1 – 3. The plot in Fig. 9 shows the difference high (see Table 2).
between event groupings for bed load D50 and is Shear stress appears to have no effect on bed load
typical of the relationships exhibited between s and size during Events 1 – 3, thus, EMT is indicated. All
the other percentiles. For Events 4– 5, all but one of model slopes test equal to zero. Large variability
the b1 values are greater than zero for the five size across the observed s range causes this lack of any
 D.A. Marion, F. Weirich / Geomorphology 55 (2003) 139–154 149

coefficients or R2 values are reported). They found

significant models only at particular sites: one site for
the Dmax and D84 sizes, and another for the D16 size.
Interestingly, they found no significant relationship for
D50, the percentile showing the weakest relationship
for Events 4 –5 at Toots Creek. They found that at-a-
point data produced significant regressions at more
sites and for more size percentiles at any given site
than did mean values. Their data were taken through-
out one entire spring runoff season, not individual
events like at Toots Creek.
The relationship between s and bed load grain size
evident at Toots Creek is different from that docu-
mented by Shih and Komar (1990). Toots Creek
Fig. 9. Bed load D50 size variation with shear stress. Model for data
models are shown in Fig. 10 along with the models
taken from Events 1 – 3 is statistically different than that for Events
4 – 5. determined by Shih and Komar for bed load D50 and
D90 from the gravel-bed stream they studied. The
models indicate that much smaller grain sizes moved
relationship. Fig. 9 illustrates this for Events 1 –3 and at Toots Creek than at Oak Creek for a given s. For
is typical of the variability in all other grain-size example, when s equals 30 N/m2, the expected D50 at
percentiles with shear stress. Oak Creek is 3.0 mm while that at Toots Creek is 0.45
Blizard and Wohl (1998) found s was very effective mm, roughly a tenfold decrease in size. Toots Creek
in explaining bed load size variation at East St. Louis models also have lower increase rates for grain size in
Creek. They derived significant (a = 0.05) regressions comparison to comparable Oak Creek percentiles (see
using both at-a-point and cross-sectional mean values. Fig. 10). Bed material size is finer at Oak Creek with a
Using the latter (same as used for Toots Creek), they D50 of 60 and 20 mm for the surface and subsurface,
obtained significant regressions between s and three of respectively. At Bridge 2, the corresponding values
the four size metrics they tested: Dmax, D84 and D16 (no are 104 and 51 mm. Thus, one possible explanation
for the different responses might be that larger bed

Table 2
Regression statistics for models of bed load size (mm) at given
percentiles as a function of bed shear stress (N/m2) during Events 4
and 5 at Toots Creek Bridge 2a
Bed load b0 b1 Standard 95% confidence R2
3
percentile (  10 ) error of b1 limits on b1
Lower Upper
D05 8.67 0.72** 0.36 0.16 1.60 0.40
D16 9.89 0.86** 0.40 0.11 1.83 0.44
D50 5.01 1.31* 0.74 0.51 3.13 0.34
D84 3.70 1.51** 0.67 0.12 3.14 0.46
D95 7.28 1.56** 0.54 0.24 2.88 0.58
a
Data are from Helley – Smith samples taken at Toots Creek
Bridge 2. Regressions were computed using log-transformed data
for both variables. However, values listed below have been
transformed back to the original units. Thus, all models are of the
form Dx = b0sb1, where Dx = bed load percentile, b0 and b1 = re- Fig. 10. Comparison of bed load grain size – shear stress relationship
gression coefficients and s = shear stress (N/m2). *Probability of b1 between Toots Creek and Oak Creek. Plot shows models of grain-
V 0 is V 0.10. **Probability of b1 V 0 is V 0.05. For all models, size changes with shear stress for given grain-size distribution
sample size = 8 and df = 6. percentiles.
150 D.A. Marion, F. Weirich / Geomorphology 55 (2003) 139–154

tests are based upon different assumptions. They fall

into three types: those utilizing a representation of Dmax
(sc* test), those utilizing just the larger particles (tracer
displacement test) and those utilizing the entire bed
load size distribution (percentile size and skewness
tests). Komar and Shih (1992) showed that sc* and

Table 3
Summary of hypothesis tests to determine whether size-selective or
equal-mobility bed load transport is exhibited at Toots Creek during
the five experimental events
Model and slope Result Conclusion
condition testeda (probability
[ P] under H0)
Model: tracer Using all data, Cannot reject H0;
travel distance P < 0.03 for EMT indicated in
vs. grain weight. Event 1; P > 0.28 most events.
Ha: b1 < 0 for Events 2, 4, 5
(i.e., SST occurs). and cumulative
Event 3 not for all events.
Fig. 11. Bed load skewness variation with shear stress during all five tested due to Excluding outliers, Reject H0;
Toots Creek events. Line shows the regression model of the trend. insufficient data. P < 0.03 for SST indicated
Events 1, 2 and 5; in most events.
P > 0.45 for Event 4
grains are relatively more exposed at Oak Creek and cumulative for
all events.
because the surrounding bed matrix contains smaller Model: bed load Using combined data Cannot reject H0;
sizes (Andrews, 1983). percentile grain from all events, EMT indicated.
size vs. shear P > 0.29 for all size
3.4. Bed load grain-size skewness vs. s stress for D05, D16, percentiles except
D50, D84, D95. D95 where P = 0.10.
Ha: b1 > 0 Using data for Cannot reject H0;
According to the skewness vs. s test, EMT occurs. (i.e., SST occurs). Events 1 – 3, P > 0.32 EMT indicated.
The probability is 0.32 that b1 V 0 using data from all for all size percentiles.
events. During the five experimental events, skewness Using data for Reject H0;
remains essentially constant over the s range observed Events 4 – 5, P < 0.05 SST indicated.
(no differences between events) (see Fig. 11). for all size percentiles
except D50 where
P = 0.06.
Model: bed load Using combined Cannot reject H0;
4. Discussion grain-size data from all events, EMT indicated.
skewness vs. P = 0.32.
Results are summarized in Table 3 for all of the shear stress
magnitude.
tests discussed above. Five of the seven tests support Ha: b1 > 0
the conclusion that EMT occurred or was predomi- (i.e., SST occurs).
nant. Interestingly, EMT is indicated in all four tests Model: relative Using combined data Reject H0;
when data from all events are used together (cumu- bed load D95 where D95 V 32 mm EMT indicated.
lative displacement for tracers). grain size vs. from all events,
dimensionless P = 0.04.
shear stress.
4.1. Explanation of results Ha: b1 = 1
(i.e., EMT occurs).
Several explanations may account for the seemingly a
b1 = slope of regression model. Dx = x-percentile of bed load
contradictory results between different tests. First, the grain-size distribution.
 D.A. Marion, F. Weirich / Geomorphology 55 (2003) 139–154 151

percentile size tests can produce contradictory results A marked change in bed material size composition
using the same data. Komar and Shih offered a com- prior to Events 4 and 5 could also explain their
pelling argument that the theoretical basis of the sc* test apparent SST. Kuhnle (1992) showed that EMT from
leads to the contradictory predictions that Dmax changes the two separate components of a bimodal bed dis-
rapidly with small changes in sc* and that bed load tribution could produce an overall SST relationship.
grain-size distributions do not change. However, the He found this to occur in a gravel-bed, pool-riffle
sc* – Dmax relationship has been widely reported for channel during relatively low-energy peak flows like
gravel bed (see Komar and Shih, 1992 for a recent those used in this study. However, this explanation
citation list) and recently for step-pool channels (Ryan, does not seem relevant for Toots Creek. While the
1994), so it is included here. subsurface layer in the study reach does have a
Second, the tracer displacement tests supporting bimodal size distribution, the surface does not. Be-
SST are somewhat suspect. Event 1 results might be cause extensive pavement breakup was not indicated
questioned because Event 1 was the first event after for most of the bed load samples, then the surface
tracer replacement. Despite efforts to reestablish their should have been the primary bed load source. While
original position, tracer exposure may have been either accelerated bank erosion or sediment release
somewhat higher than under natural conditions or in from the breakdown of a step might cause transient
subsequent events. Increased exposure, especially of changes in bed material size distribution, post-event
smaller tracers, would increase the likelihood of SST. observations and measurements showed that neither
In addition, results implying SST for Events 2 and 5 occurred.
are uncertain because they are only significant when The occurrence of ‘‘bed waves’’ could affect the
possible outliers are excluded. variability observed in all of the relationships tested.
Third, bed load transport modes may be stage A bed wave is an increase in sediment storage at a
dependent. Komar and Shih (1992) proposed that given location over time or relative to adjacent up-
modes shift from ‘‘apparent’’ equal-mobility at low stream and downstream channel areas (after Hoey,
stages to size-selective at higher stages then back to 1992). Their occurrence and downstream movement
equal-mobility at still higher stages. Based on their in step-pool channels have been directly observed in
analysis of data from a gravel-bed channel, Komar and the field (Hayward, 1980), indicated by intra- and
Shih concluded that at stages where tractive forces interevent surveys (Ashida et al., 1981; Ergenzinger,
were sufficient to mobilize the bed (i.e., when most of 1992; Warburton, 1992; Gintz et al., 1996; Blizard
the grain sizes within the bed are present in the bed and Wohl, 1998) and demonstrated in flume studies
load) that SST occurred. This would correspond with (Whittaker and Davies, 1982; Whittaker, 1987). Peri-
Phase 2 transport in a step-pool channel (Warburton, odic oscillations in bed load transport rates at a
1992). At lower stages, the bed pavement would not be particular site (‘‘bed load pulses’’; Hoey, 1992) have
fully mobilized (Phase 1 transport) and Komar and been attributed to the presence of bed waves (Hay-
Shih reasoned that random fluctuations in bedload size ward, 1980; Whittaker and Davies, 1982; Whittaker,
rather than bed material size variability would cause 1987; Blizard and Wohl, 1998).
EMT to be evident. Komar and Shih further predicted The passage of bed waves through the study reach
that the transport mode switches back to EMT at some could have affected the bed load relationships derived
high stage when bed load size distribution matches that for this study in a variety of ways. The sampled Dmax
of the bed material (Phase 3). A shift is indicated at could have been larger due to increased protrusion of
Toots Creek where Events 1 –3, with relatively low the largest clasts within a matrix of smaller clasts in a
stresses, exhibited equal-mobility in the Dx vs. s test, bed wave. Conversely, Dmax could have been smaller
while Events 4 – 5, with generally higher stresses, through burial of the largest clasts by smaller ones.
exhibited SST. Fig. 8 provides an example of this dif- Tracer displacements could have been similarly af-
ference for bed load D50 grains. However, if a mode fected if tracer grains were caught up in a bed wave.
change did occur, then why similar differences be- The preferential entrainment or hiding of larger grains
tween events are not evident in the bed load skewness could also affect bed load size distributions, thereby
vs. s results is unclear. altering either percentile sizes or skewness. Additional
152 D.A. Marion, F. Weirich / Geomorphology 55 (2003) 139–154

effects can also be imagined. Clearly, the influence of (Andrews, 1983). Both of these relate to how avail-
bed waves could introduce increased variability to the able a grain is for entrainment.
observed bed load relationships and possibly affect Tracer studies have determined that other factors
the test results. also influence sediment availability in step-pool chan-
Bed waves were not directly observed within Toots nels. The structural arrangement of bed grains has also
Creek during any of the five events. However, bed load been shown to affect how available they are for
pulses did occur during the Event 1 recession flow (Fig. entrainment. Laronne and Carson (1976) found grains
6), suggesting bed waves may have been present during within open or infilled structures to be more mobile
that time. During the period between 50 and 200 min, than grains in tight (vertically infilled and imbricate)
discharge was fairly constant, yet bed load transport structures. Within the latter, not only hiding but also
rates oscillated between 2.1  10 5 and 2.1  10 4 the packing of the interlocked grains further restricts
kg/s/m. The sample numbers during Events 2 – 5 are their mobility. The disruption of tight bed structures
insufficient to judge whether bed load pulses occurred. would be expected to increase sediment availability
If bed waves did occur, they could only have resulted and increase particle mobility. This response was
from in-channel sources as no out-channel sediment observed by Gintz et al. (1996) after a large event
inputs occurred during the experiments. We assessed obliterated an existing step-pool pattern.
whether bed waves may have affected the test results by Thus, the disruption of structural arrangements
excluding the suspect samples, reanalyzing the data within the bed during certain events might explain
and repeating those tests which previously utilized the the inconsistent expression of EMT. As noted earlier,
suspect samples. the removal and reinstallation of tracers prior to Event 1
With one possible exception, the interpretations of could have disturbed the local bed organization and
the test results remain the same even when possible thereby increased tracer mobility. Tractive forces dur-
bed wave samples are excluded. The one exception is ing the higher flows in Events 4 and 5 may have been
for the sc* –D95/D50b relationship where P = 0.08 for sufficient to dislodge some larger pavement clasts,
the test statistic in the new model. Therefore, the expose previously hidden grains and generally increase
rejection of the H0 (i.e., SST occurs) is less certain sediment availability to the point that SST was man-
than for the original analysis ( P = 0.04), but still ifested.
plausible. Test statistic probabilities increase some- Grain shape is another factor that has been shown
what in magnitude in all of the other tests, too, but this to affect particle mobility in step-pool channels.
acts to increase confidence in the original test inter- Schmidt and Ergenzinger (1992) found that rod-
pretations. The tracer displacement data could not be shaped grains had the greatest mobility while disc-
reassessed as there is no way to determine which shaped ones had the least. Laronne and Carson (1976)
tracers might have been affected by possible bed observed that disc-shaped grains were more prone to
waves during Event 1. Interestingly, Event 1 is the being incorporated in tight structural arrangements
only event in which tracer data clearly exhibit SST that limited their movement. Differences in grain
(Table 3). We conclude from these additional tests that shape between sites might explain variations in the
bed waves did not confound the original test results. degree to which EMT is apparent, with EMT being
Changes in sediment availability prior to or during more evident at sites where grain shapes are predom-
a given event could either promote or inhibit SST. inantly disc-shaped. However, grain shape does not
EMT can be thought of as the result when factors explain differences between events at a single site like
other than an individual grain’s size influence its Toots Creek where the grain-shape distribution re-
entrainment potential. Past EMT work has focused mains constant.
on the size of neighboring bed material relative to a
given particle’s size as the primary factor influencing 4.2. Comparison to past work
entrainment potential (Andrews, 1983; Wiberg and
Smith, 1987). Neighboring clast size affects both the Blizard and Wohl (1998) also found a general trend
exposure of a grain to tractive forces and the force towards EMT for a step-pool channel of similar size
necessary to roll it over the adjacent downstream clast and under similar flow conditions as those at Toots
 D.A. Marion, F. Weirich / Geomorphology 55 (2003) 139–154 153

Creek. While they found several instances of signifi- and Shih (1992). If so, then finding ways to predict
cant positive relationships between some bed load when the mode shifts occur should greatly improve
percentile sizes and s (indicative of SST), they con- our ability to predict bed load transport in step-pool as
cluded based on bed load size skewness and the lack well as gravel-bed channels. Clearly, additional work
of overall coarsening of bed load size distributions is needed to confirm these patterns and, if real, predict
with increased s that equal-mobility was evident. their operation.
Thus, their results are very similar to this study in
that different relationships indicated different trans-
port modes, but equal-mobility seemed most apparent. Acknowledgements

We thank the two anonymous reviewers for their

5. Conclusions constructive comments on this paper, and the many
Forest Service and University of Iowa personnel who
Altogether, the results, while not unequivocal, lend provided assistance during this project. We also thank
greater support to the conclusion that equal-mobility the editors of this special issue for inviting us to
was the predominant transport mode during the near- submit this paper. The following organizations
bankfull flow events used in this study. Five of the provided funding for this research: the USDA Forest
seven tests support the conclusion that EMT occurred Service, Southern Research Station and Stream
or was predominant. Some of the tests indicating SST Systems Technology Center; the Ouachita National
are suspect because they require excluding possible Forest; the National Council of the Paper Industry for
outliers from the analysis. The remaining tests indi- Air and Stream Improvement; and the Iowa Institute
cating SST can be explained by plausible changes in of Hydraulic Research at the University of Iowa.
sediment availability (through changes in the struc-
tural arrangements of bed grains) that promote SST
occurrence. In general, our results strongly indicate References
that bed size heterogeneity significantly affects bed
load entrainment and size in step-pool channels. They Andrews, E.D., 1983. Entrainment of gravel from naturally sorted
also suggest that changes in the structural arrange- riverbed material. Geological Society of America Bulletin 94,
ments of bed material prior to or during events may 1225 – 1231.
Ashida, K., Takahashi, T., Sawada, T., 1981. Sediment yield and
confound this effect.
transport on a mountainous small watershed. Bulletin of the
Our results seem to fit within the conceptual model Disaster Prevention Research Institute 26, 119 – 144.
of bed load transport proposed by Komar and Shih Ashworth, P.J., Ferguson, R.I., 1989. Size-selective entrainment of
(1992) for gravel-bed channels. The apparent trans- bed load in gravel bed streams. Water Resources Research 25,
port mode appears to shift depending upon the level of 627 – 634.
pavement mobility present. These shifts in transport Best, D.W., Keller, E.A., 1986. Sediment storage and routing in a
steep boulder-bed, rock-controlled channel. In: DeVries, J.
mode correspond to the three different phases of bed (Ed.), Proceedings of the Chaparral Ecosystems Research Con-
load transport identified by Warburton (1992) for step- ference. California Water Resources Center Report, vol. 62.
pool channels. In the case of Toots Creek, the shifts University of California at Davis, Davis, CA, pp. 45 – 55.
are only evident in the s– Dx relationship, and addi- Blizard, C.R., Wohl, E.E., 1998. Relationships between hydraulic
tional data will be required to confirm these shifts. variables and bed load transport in a subalpine channel, Colo-
rado Rocky Mountains, U.S.A. Geomorphology 22, 359 – 371.
When our work is combined with previous efforts, Environmental Systems Research Institute, 1983. States (digital
certain consistent patterns in bed load transport begin coverage). Spheroid Projection. Environmental Systems Re-
to emerge for step-pool channels. Entrainment does search Institute, Redlands, CA.
appear to be significantly affected by neighboring Ergenzinger, P., 1992. Riverbed adjustments in a step-pool system:
clast size as well as other structural characteristics of Lainbach, Upper Bavaria. In: Billi, P., Hey, R.D., Thorne, C.R.,
Tacconi, P. (Eds.), Dynamics of Gravel-Bed Rivers. Wiley, Chi-
the bed. The mutual interactions of these factors may chester, UK, pp. 415 – 429.
well change at certain force thresholds, producing Fenneman, N.M., Johnson, D.W., 1946. Physical divisions of the
shifts between EMT and SST as proposed by Komar United States (Physiography). U.S. Geological Survey, Washing-
154 D.A. Marion, F. Weirich / Geomorphology 55 (2003) 139–154

ton, DC, 1:7,000,000 scale. Original map in polyconic projection Marion, D.A., Weirich, F., 1999. Fine-grained bed patch response to
and digital coverage available on-line from U.S. Geological Sur- near-bankfull flows in a step-pool channel. In: Olsen, D.S.,
vey in Albers Equal Area Projection at http://water.usgs.gov/ Potyondy, J.P. (Eds.), Wildland Hydrology, Proceedings of the
lookup/getspatial?physio. Specialty Conference, Bozeman, MT. American Water Resour-
Fowler, W.P., 1994. Woody debris dynamics in zero order streams ces Association, Herndon, VA, pp. 93 – 100.
of the Ouachita National Forest: preliminary findings. In: Baker, Megahan, W.F., Nowlin, R.A., 1976. Sediment storage in channels
J.B. (Compiler), Proceedings of the Symposium on Ecosystem draining small forested watersheds in the mountains of central
Management Research in the Ouachita Mountains: Pretreatment Idaho. Proceedings of the Third Federal Inter-Agency Sedimen-
Conditions and Preliminary Findings, Hot Springs, AR. General tation Conference, Denver, CO, pp. 4-115 – 4-126. Water Re-
Technical Report SO-112. USDA. Forest Service, Southern For- source Council, Sedimentation Committee, place of publication
est Research Station, New Orleans, LA, pp. 182 – 185. unknown.
Gintz, D., Hassan, M., Schmidt, K., 1996. Frequency and magni- Nakamura, F., Swanson, F.J., 1993. Effects of coarse woody debris
tude of bedload transport in a mountain river. Earth Surface on morphology and sediment storage of a mountain stream
Processes and Landforms 21, 433 – 445. system in Western Oregon. Earth Surface Processes and Land-
Grant, G.E., Swanson, F.J., Wolman, M.G., 1990. Pattern and origin forms 18, 43 – 61.
of stepped-bed morphology in high-gradient streams, Western Ouachita National Forest. Contours, roads, streams, and water (dig-
Cascades, Oregon. Geological Society of America Bulletin 102, ital coverages). Stateplane projection. Ouachita National Forest,
340 – 352. Hot Springs, AR.
Hayward, J.A., 1980. Hydrology and stream sediments in a moun- Parker, G., Klingeman, P.C., McLean, D.G., 1982. Bedload and size
tain catchment. Tussock Grasslands and Mountain Lands Insti- distribution in paved gravel-bed streams. American Society of
tute. Special Publication, vol. 17. Lincoln College, Canterbury, Civil Engineers, Proceedings, Journal of the Hydraulics Divi-
New Zealand. 236 pp. sion 108 (HY4), 544 – 571.
Heede, B.H., 1972. Flow and channel characteristics of two high Ryan, S.E., 1994. Effects of transbasin diversion on flow regime,
mountain streams. Research Paper RM-96. US Department of bed load transport, and channel morphology in Colorado moun-
Agriculture, Forest Service, Rocky Mountain Forest and Range tain streams. PhD dissertation, University of Colorado, Boulder,
Experiment Station, Fort Collins, CO. 12 pp. CO.
Hoey, T., 1992. Temporal variations in bedload transport rates and Schmidt, K.H., Ergenzinger, P., 1992. Bedload entrainment, travel
sediment storage in gravel-bed rivers. Progress in Physical lengths, step lengths, rest periods-studied with passive (iron,
Geography 16 (3), 319 – 338. magnetic) and active (radio) tracer techniques. Earth Surface
Komar, P.D., Shih, S., 1992. Equal mobility versus changing bed Processes and Landforms 17, 147 – 165.
load grain sizes in gravel-bed stream. In: Billi, P., Hey, R.D., Shields, I.A., 1936. Application of similarity principles and turbu-
Thorne, C.R., Tacconi, P. (Eds.), Dynamics of Gravel-Bed Riv- lence research to bed-load movement. Berlin. Translated by Ott,
ers. Wiley, Chichester, UK, pp. 73 – 93. W.P. and van Uchelen, J.C. Publication no. 167. California
Kuhnle, R.A., 1992. Fractional transport rates of bed load on Good- Institute of Technology, Hydraulics Laboratory, Pasedena, CA,
win Creek. In: Billi, P., Hey, R.D., Thorne, C.R., Tacconi, P. 36 pp. In German.
(Eds.), Dynamics of Gravel-Bed Rivers. Wiley, Chichester, UK, Shih, S., Komar, P.D., 1990. Differential bedload transport rates in a
pp. 141 – 155. gravel-bed stream: a grain-size distribution approach. Earth Sur-
Laronne, J.B., Carson, M.A., 1976. Interrelationships between bed face Processes and Landforms 15, 539 – 552.
morphology and bed-material transport for a small, gravel-bed Sneed, E.D., Folk, R.L., 1958. Pebbles in the lower Colorado River,
channel. Sedimentology 23, 67 – 85. Texas, a study in particle morphogenesis. Journal of Geology
Marion, D.A., 2001a. Bed load transport rates at near-bankfull 66, 114 – 150.
flows in a step-pool channel. Proceedings of the Seventh Fed- Warburton, J., 1992. Observations of bed load transport and channel
eral Interagency Sedimentation Conference, Reno, NV. U.S. bed changes in a proglacial mountain stream. Arctic and Alpine
Subcommittee on Sedimentation, Washington, D.C., USA, Research 24 (3), 195 – 203.
pp. III-32 – III-39. Whittaker, J.G., 1987. Sediment transport in step-pool streams. In:
Marion, D.A., 2001b. Field experiments on channel morphology Thorne, C.R., Bathurst, J.C., Hey, R.D. (Eds.), Sediment Trans-
and bedload interactions at near-bankfull flows in a small, step- port in Gravel-Bed Rivers. Wiley, Chichester, pp. 545 – 570.
pool stream in the Ouachita Mountains of Arkansas. PhD dis- Whittaker, J.G., Davies, T.R.H., 1982. Erosion an sediment trans-
sertation, University of Iowa, Iowa City. port processes in step-pool torrents. In: Walling, D.E. (Ed.),
Marion, D., Malanson, G., in press. Ordination of wood vegetation Recent Developments in the Explanation and Prediction of Ero-
in a Ouachita National Forest watershed. In: Guldin, J.M. (Ed.), sion and Sediment Yield: Proceedings of the Exeter symposium.
Proceedings of the Symposium on Ecosystem Management Re- IAHS Publication, vol. 137. International Association of Hydro-
search in the Ouachita and Ozark Mountains, Hot Springs, AR. logical Sciences, Wallingford, UK, pp. 99 – 104.
US Department of Agriculture, Forest Service, Southern Re- Wiberg, P.L., Smith, J.D., 1987. Calculations of the critical shear
search Station, Asheville, NC. stress for motion of uniform and heterogeneous sediments.
Marion, D.A., Weirich, F., 1997. Simulating storm events. Environ- Water Resources Research 23 (8), 1471 – 1480.
mental Testing and Analysis 6 (4), 15 – 16.
 Geomorphology 55 (2003) 155 – 171
www.elsevier.com/locate/geomorph

Boulder weathering and erosion associated with a wildfire,

Sierra Ancha Mountains, Arizona
Ronald I. Dorn *
Geography Department, Arizona State University, Tempe, AZ 85287-0104, USA

Received 1 November 2001; received in revised form 13 June 2002; accepted 10 March 2003

Abstract

An April – May 2000 ‘‘Coon Creek Fire’’ burned f 37.5 km2 of the Sierra Ancha Mountains, 32.3 km miles north of Globe,
AZ—including 25 sandstone and 19 diorite boulders surveyed in 1989 and resurveyed after the burn, after the summer 2000
monsoon season, and after the winter 2001 season. When viewed from the perspective of cumulative eroded area, both
sandstone and diorite displayed bimodal patterns with 79% of sandstone boulder area and 93% of diorite boulder area
undergoing either no fire-induced erosion or fire-induced erosion >76 mm. When stretched over cumulative boulder areas,
erosion due to the fire averaged >26 mm for sandstone and >42 mm for diorite. Post-fire erosion from thunderstorm summer
rains averaged < 1 mm for 5 diorite and 1 mm for 10 sandstone boulders. While only a single diorite boulder eroded an average
of 1.2 mm after the winter, winter erosion removed an average of 5.5 mm from 14 sandstone boulders. Thus, fire appears to
increase a rock’s susceptibility to post-fire weathering and erosion processes, as predicted by Goudie et al. [Earth Surf. Process.
Landf. 17 (1992) 605]. In contrast to experimental research indicating the importance of size in fire weathering, no statistically
significant relationship exists between erosion and boulder height or boulder surface area—a result similar to Zimmerman et al.
[Quat. Res. 42 (1994) 255]. These data exclude 12 original sites and 85 boulders at sites impacted by the fire that could not be
relocated, with a reasonable cause for the lack of relocation being boulder obliteration by the fire. Given evidence from 10Be
and 26Al cosmogenic nuclides [Earth Planet. Sci. Lett. 186 (2001) 269] supporting the importance of boulders in controlling
evolution of nonglaciated, bouldered landscapes [Geol. Soc. Amer. Bull. 76 (1965) 1165], fire obliteration of boulders could be
an important process driving drainage evolution in nonglaciated mountains.
D 2003 Elsevier Science B.V. All rights reserved.

Keywords: Weathering; Fire; Sierra Ancha Mountains; Arizona; Erosion; Landscape evolution; Boulder; Geomorphology; Landforms

1. Introduction enabled by the Columbian genocide (Diamond, 1997),

encompasses the right conditions for wildfire (Pyne,
Fire and mountains go together, especially in the 1998). Mountainous public lands with a long history
North American west (Pyne, 1998). The collective of fire-suppression (Pyne, 1982; Kaye and Swetnam,
North American vision of mountains as wilderness, 1999) host abundant vegetative fuel (Wadleigh and
Jenkins, 1996). Fire only requires rhythmic wetting
* Fax: +1-480-965-8313. and drying (Swetnam and Betancourt, 1998), sparked
E-mail address: [email protected] (R.I. Dorn). by lightning or anthropogenically (Bird, 1995).

0169-555X/03/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0169-555X(03)00138-7
156 R.I. Dorn / Geomorphology 55 (2003) 155–171

From a pyrogenic perspective, geomorphology Viewed from a longer time perspective, fire events
influences mountain wildfires at a variety of spatial often end up being critical agents of Holocene debris
scales (Brown et al., 2001). At a large scale of hill slope destabilization (Marion et al., 1995). Analyses of
plots, slopes create moisture gradients that vary fuel tree fire scars provide excellent means of reconstruct-
and fire patterns (Bridge and Johnson, 2000). Fires in ing fire frequencies (Allen et al., 1998; Swetnam and
pre-European Wyoming, for example, tended to Betancourt, 1990). Even lacking tree rings, pedoan-
spread down a slope (Baker and Kipfmueller, 2001). thracological methods permit analyses of longer-term
Bare rock cover is particularly important in a variety fire histories (Carcaillet, 1998; Carcaillet and Thinon,
of mountain contexts (McDonald et al., 1996; Don- 1996). Maceral, sedimentological, and organic geo-
negan and Rebertus, 1999; Moody and Meentemeyer, chemistry analyses of organic remains in soil, regolith,
2001). Topography and moisture variability at the and lacustrine deposits also facilitate understanding of
watershed scale (Bian and Walsh, 1993) influences ancient fires (Siffedine et al., 1994; Lichtfouse et al.,
wildfire behavior (Heyerdahl et al., 2001). Of course, 1997; Lichtfouse, 1999; Edwards et al., 2000; Laird
at the megageomorphic scale, orography generates and Campbell, 2000).
moisture, thus facilitating growth of fuel (Allen et Admitting the aforementioned is only a sampling
al., 1998). of the literature, fire comprises a relatively small
From a geomorphic perspective, fire modifies a component of research on mountain geomorphology.
variety of mountain geomorphic processes and rates Similarly, fire is not a significant part of the even
and at a variety of spatial scales (Swanson, 1981). smaller rock and mineral weathering literature in
Locally, fire alters pedogenesis and permafrost (Power mountain geomorphology. Over the last quarter cen-
and Power, 1995; Mazhitova, 2000), increases rates of tury, for example, no mountain weathering paper,
soil erosion substantially (Morris and Moses, 1987; cited more than 20 times in science citation and social
Shakesby et al., 1993; Prosser and Williams, 1998) science citation indices as of this penning, concerned
plays a role in mountain slope asymmetry (Selkirk et fire (Thorn, 1976, 1979, 1980; Chinn, 1981; Mat-
al., 2001), and utilizes efficient pre-existing pathways thews and Shakesby, 1983; McGreevy and Whalley,
of mountain sediment transport. Debris flows, for 1983, 1985; Dixon et al., 1984; Mahaney et al., 1988;
example, are important agents of mountain sediment McCarroll, 1989; Drever and Zobrist, 1992; Gislason
transport (Butler and Walsh, 1994). Using debris-flow et al., 1996; Raymo, 1994; Condie et al., 1995).
pathways, fires alter debris erosion and deposition The literature on fire as a weathering process is, by
(Soons, 1994; Cannon et al., 2001). Fire-induced itself, still underdeveloped. When it comes to field
episodic sediment erosion even modifies loci and observations, geomorphologists have long known of
timing of aggradation at distant lowland locations the importance of fire in rock and mineral weathering
(Germanoskiy and Miller, 1995). Considered from a (Blackwelder, 1927; Emery, 1944). However, the
summary perspective of different mountain sites literature suffers from a paucity of data and hence
around the Mediterranean, fires list among the most consists of anecdotal observations (Ollier, 1983; Bir-
important contemporary erosional influences—on par keland, 1984, p. 64; Bierman and Gillespie, 1991;
with grazing (Cerda, 1998). Dragovich, 1993). Consider just the interface between
At an interface of geomorphology and biogeogra- fire and epilithic organisms. While the weathering
phy in mountains, Butler (2001) highlights two key literature on organisms matured enough to host excel-
mountain geomorphic processes that produce disturb- lent review papers (Viles, 1995), very little is known
ance corridors: snow-avalanche paths and debris about how fire influences biochemical activity of
flows. Both corridors maintain feedbacks with moun- lithobionts (Garty, 1992). The state of field research
tain wildfire. Avalanche paths form important fuel is such that quantitative observations on a single
breaks (Malanson and Butler, 1984; Suffling, 1993), sagebrush range fire justifies publication of rare
while wildfires greatly increase frequency and mag- insights into rates of boulder erosion (Zimmerman et
nitude of geomorphic activity in debris-flow corridors al., 1994). When averaged over the entire exposed
(Cleveland, 1977; Wohl and Pearthree, 1991; Soons, boulder surface area, Zimmerman et al. (1994) found
1994; Meyer et al., 1995). that erosion rates averaged 0.9 mm, median erosion
 R.I. Dorn / Geomorphology 55 (2003) 155–171 157

was 0.4 mm, and the range varied from < 0.1 to 6.1 ifornia, Berkeley) to get to know a new region by
mm. surveying field sites, upon arriving at Arizona State
Archaeologists, ecologists, and experimental geo- University, I surveyed rock surfaces in a wide variety
morphologists dominate attempts to develop theories of different settings around Arizona—thinking I
on the effects of fire on rock weathering. Archaeolo- would return decades later to detect small amounts
gists make important field observations on surface of erosion in only a small percentage of surveyed rock
effects of fire (Kelly and McCarthy, 2001) and also faces. One of the field sites, however, experienced a
develop general models of thermolithofractography as major forest fire in April – May 2000—offering a
a means to characterize and interpret ‘‘fire cracked’’ unique chance to resurvey fire-impacted boulders.
rock (House and Smith, 1975; McFarland, 1977; Because the literature contained the hypothesis that
Kritzer, 1995; Rapp et al., 1999; Wilson, 1999). field outcrops should have a greater degree of sus-
Ecologists acquire data on the importance of fire- ceptibility to weathering and erosion processes after a
related weathering in nutrient cycles such as potas- fire (Goudie et al., 1992), I also monitored boulder
sium (Brais et al., 2000) and phosphorus (Brundrett et erosion after the fire.
al., 1996). Working in the long tradition (Tarr, 1915)
of controlled experimental (McGreevy and Whalley,
1985; McGreevy et al., 2000; Warke and Smith, 1998) 2. Field setting
and field (Warke et al., 1996) weathering studies,
experiments on fire (Goudie et al., 1992; Allison Mountain ranges in the SW United States host
and Goudie, 1994; Nealson, 1995; Allison and Bris- repeated forest fires. No data are available, unfortu-
tow, 1999) reveal substantive insights. For example, nately, on the last time the ‘‘Coon Creek’’ site burned
erosion of rocks in response to fire depends heavily on in the Sierra Ancha Mountains. In regional terms,
rock physical properties and varies with different rock however, fire-scar analysis reveals periodic high fire
types, inversely with rock size and positively with decades (e.g., 1740-80 and 1830-60 in Fig. 1) that
water content. occur during rapid switching from extreme wet to dry
Given this research interface of mountains, fire, years in the southwest (Swetnam and Betancourt,
and rock weathering, this paper rests in the tradition of 1998). Thus, prior fire events undoubtedly influenced
providing a rare dataset of field observations. Follow- the study site.
ing advice given by Bob Sharp (California Institute of The Workman Creek SNOTEL site in the lower end
Technology) and Luna Leopold (University of Cal- of the study area recorded winter snowfall since 1983.

Fig. 1. Summary of fire-scar chronologies of 27 mountain ranges of the SW United States, modified from Figs. 9 to 12 in Allen et al. (1998).
This chronology does not include the study site, but derives from the surrounding region.
158 R.I. Dorn / Geomorphology 55 (2003) 155–171

The long-term winter average is 635 mm (25 in.) of would have started near crests and moved slowly
precipitable water. During the winter season preceding downhill, this fire started near the base of the drainage
the burn, the SNOTEL site received only 20% of basin. The fire moved upslope rapidly with rising
normal (http://www.aracnet.com/cgi-usr/cpacheco/ flames. Winds of 6– 12 m/s helped produce a large,
cdbs_tbl)—indicating conditions were extraordinarily hot, and destructive fire. Ground temperatures are not
dry in spring 2000. known.
The Coon Creek fire started at a campfire on April In 1989, I surveyed 129 boulders in 20 sites in the
26, 2000. The April –May fire burned approximately burned area. Thirty-seven boulders were also sur-
37.5 km2 in the Sierra Ancha Mountains, Tonto veyed at five sites not burned in the fire. Surveyed
National Forest, 32.3-km miles north of Globe, AZ. boulders range from 2070 to 2290 m on slopes
In contrast to a natural lightning-strike fire, which hosting ponderosa pine (Pinus ponderosa), Gambel

Fig. 2. Coon Creek burn study area, Sierra Ancha Mountains, central Arizona. Numbers designate clusters of surveyed boulders and correspond
with tables, figures, and in-text discussions. For example, Boulder 18.6 in Fig. 3 refers to boulder 6 from site 18.
 R.I. Dorn / Geomorphology 55 (2003) 155–171 159

oak (Quercus gambelii), mixed-conifer woodland season and again in May 2001 after the winter snow
with blue spruce (Picea pungens), Douglas-fir (Pseu- season.
udotsuga menziesii var. glauca), white fir (Abies con- Summer thunderstorm precipitation values are not
color), corkbark fir (A. lasiocarpa var. arizonica), recorded at the SNOTEL site in the surveyed basin;
southwestern white pine (P. strobiformis), and quak- however, at drier Globe, ‘‘Arizona Monsoon’’ rains
ing aspen (Populus tremulodies). were extensive. The Globe area received more than
In June 2000, I relocated 25 sandstone boulders 300 mm of precipitation in Summer 2000. Because
and 10 diorite boulders at eight sites (Fig. 2). The 44 Coon Creek is higher and more susceptible to mon-
refound boulders were resurveyed twice more, in soon precipitation, totals were likely much higher. The
October 2000 after the summer monsoon precipitation Workman Creek SNOTEL site had a slightly above

Fig. 3. Example of two extremes in erosion behavior within a few meters of each other at site 18 (see Fig. 2). (A) Diorite boulder 18.6 (50 mm
across) did not experience erosion after the fire, and its epilithics still coat the boulder. (B) Diorite boulder 18.15 (1500 mm across) lost all but
10% of its original surface.
160 R.I. Dorn / Geomorphology 55 (2003) 155–171

average snowfall of 685 mm of precipitable water water content within the boulders, at least around the
from October 2000 through April 2001—the winter nails.
after the fire (ftp://ftp.wcc.nrcs.usda.gov/data/snow/ In order to understand the possible influence of
snotel/reports/water_year/arizona/10s01s.gif). nails as a monitoring device, an experiment was
conducted on diorite and sandstone clasts removed
from the field area. Blow torches heated 24 diorite and
3. Methods 24 sandstone clasts that were spheroidal in shape with
diameters of about 10 cm. Three cm-long nails were
All 37 boulders surveyed at five sites outside the emplaced with epoxy in half of the clasts (12 diorite,
burned area in 1989 had a complete cover of epilithic 12 sandstone). Half of the clasts (6 diorite, 6 sand-
lichens and other lithobionts. Similarly, 129 boulders stone) soaked in distilled water for 1 h prior to
surveyed at 20 sites in the burn area in 1989 had a torching, while 12 clasts (6 diorite, 6 sandstone) air
complete or almost complete cover of epilithic dried for 24 h prior to torching. Blow torches heated
organisms (e.g., Fig. 3) with the original intention rocks until the backsides (away from the torch)
of monitoring the effects of weathering over a period reached a temperature of approximately 400 jC as
of decades. Most of the sites occur along broad ridge measured by thermocouples and verified by optical
lines. Selected boulders had no clear orientation. microscope examination of commercial temperature
With permission of the Forest Service, holes were pellets that melt at specific temperatures, in this case
drilled into boulders with 76-mm steel nails em- melting at 371 jC but not 427 jC.
placed with epoxy to establish baselines for future
surveys.
My efforts after the April – May 2000 fire failed to 4. Results
relocate 85 boulders at 12 of the original sites. Since
GPS technology was not available during the original Only two of the diorite and sandstone boulders
survey, relocation was to be aided by additional outside of the fire-impacted region experienced
surveying with respect to particular trees and land- detectable erosion. Obviously, erosion and weathering
forms such as slope breaks and drainage features. occurs under the epilithic cover. However, those
With no success, I attempted relocation at each sub- changes are the subject of a study emphasizing micro-
sequent visit to the site after the monsoon season in scopic techniques and beyond the scope of this paper.
October 2000 and after the snowfall season in May
2001. Although metal detectors found nails at sites
with missing boulders, the presence of nails is cer- Table 1
Field observations on Boulder 1.2 (sandstone), at about 2070 m, a
tainly not conclusive evidence for complete boulder boulder with a maximum height of 500 mm and a surface area of
obliteration at these nonrelocated sites. 1,560,000 mm2 (italic, bold, and underline formatting helps track
Scanning electron microscopy also analyzed rock what happened over time to the different surfaces)
samples collected in the field before the fire and After April – May After monsoon After winter
samples of the host rock and spalled fragments col- fire (June (October 2000 (May 2001 resurvey)
lected after the fire. Wavelength and energy dispersive resurvey) resurvey)
X-ray spectrometry supplemented secondary electron Surface Area in Surface Area in Surface Area in
microscopy and backscattered electron microscopy erosion mm2 erosion mm2 erosion mm2
imaging techniques. >76 312,000 >76 312,000 >76 312,000
Employing 76-mm-long nails, while facilitating 32 468,000 32 468,000 32 156,000
resurveys, resulted in several important limitations >76 312,000
0 780,000 0 312,000 0 312,000
for this study. First, many spalls exceeded nail length. 5 468,000 5 312,000
In these cases, 76 mm is an obvious minimum amount 8 156,000
of boulder erosion. Second, the steel and epoxy may Average after Average after Average after
have altered internal thermal gradients and could bias fire = 24.8 mm monsoon winter = 35.4 mm
results. Third, nail emplacement could have increased season = 26.3 mm
 R.I. Dorn / Geomorphology 55 (2003) 155–171 161

Table 2 soon-season spall. Also after the winter, 30% of the

Erosion of diorite boulders, averaged over the entire exposed boulder that had a post-fire surface loss of 32 mm then
boulder surface area (blank entries indicate no change; entries are
sorted from least to most average erosion over the year of spalled a large-enough block to erode a nail and
monitoring) exceed the 76-mm monitoring depth.
Site.Boulder Boulder Boulder After After After Rather than present a series of similar case studies
number height Area fire monsoon winter for all 44 boulders, I followed Zimmerman et al.
(mm) (mm2)
11.6a 500 300,000 0.0
11.9 600 275,000 0.0 Table 3
18.20 1000 900,000 0.0 Erosion of sandstone boulders, averaged over the entire exposed
18.6 600 340,000 0.0 1.2 boulder surface area (blank entries indicate no change; entries are
14.1 1450 1,150,000 4.6 5.0 sorted from least to most average erosion over the year of
14.13 690 627,000 7.9 8.7 monitoring)
17.2 700 840,000 9.1 9.7 Site.Boulder Boulder Boulder After After After
17.12 300 675,000 22.7 number height area fire monsoon winter
17.3 300 240,000 24.8 25.0 (mm) (mm2)
18.1 300 865,800 60.8
10.3a 500 160,000 0.0
11.3 1950 1,500,000 68.4
14.9 1250 1,350,000 0.0
18.15 400 1,350,000 68.4
6.15 400 325,000 0.0 0.7
11.12 200 202,500 75.8
6.11 3450 850,000 5.5 5.8
11.15 250 152,000 76.0
3.9 1100 1,680,000 3.8 5.3 6.7
11.16 300 105,000 76.0
14.10 2000 3,120,000 7.8 8.3
11.2 400 360,000 76.0
14.12 820 1,054,000 8.8 9.7
17.1 600 1,760,000 76.0
6.6 800 440,000 10.5
17.6 600 1,260,000 76.0
14.3 600 1,960,000 12.0
18.25 300 1,200,000 76.0
14.6 1300 2,310,000 12.2 12.4 12.9
1.4 300 1,166,300 19.2 19.6 20.2
Number of 19 19 19
6.12 650 921,200 23.2 24.3
boulders
3.11 400 495,000 24.8 24.8 26.1
Number of 15 5 1
10.7 650 330,000 21.5 30.4
boulders
6.16 1100 1,125,000 23.3 31.3
eroded
3.4 400 990,000 3.9 5.0 34.0
Percentage of 79 26 5
10.10 800 1,320,000 33.6 35.1
boulders
1.2 500 1,560,000 24.8 26.3 35.4
changed
b 10.12 1050 560,000 24.2 24.8 38.6
Average change >53.2 0.5 1
6.5 300 270,000 47.4 49.2 50.1
Median change 68.4 0.5
10.4 600 525,000 56.7
a
Site 11, boulder 6. 3.12 350 840,000 61.2 62.4
b
Because the maximum amount that could be recorded was 76 1.13 300 317,400 76.0
mm, this is a minimum average change. 1.3 150 917,900 76.0
10.11 1100 1,150,000 76.0

The two unfired sandstone boulders that eroded Number of 25 25 25

boulders
appear to have been chipped mechanically by bullets. Number of 22 10 14
Each boulder in the area impacted by the Coon boulders
Creek wildfire experienced its own complex erosional eroded
history. Consider, for example, Boulder 1.2 in Table 1. Percentage of 88 40 56
After the fire, f 50% of this sandstone boulder boulders
changed
retained its epilithic surface cover, albeit charred in Average change >39.4b 1.0 5.5
places. A portion of this section of the boulder then Median change 24.8 1.0 1.2
eroded in relatively thin 5-mm sheets after the mon- a
Site 10, boulder 3.
soon season. Then after the winter, the surface b
Because the maximum amount that could be recorded was 76
changed again: another 3 mm eroded from the mon- mm, this is a minimum average change.
162 R.I. Dorn / Geomorphology 55 (2003) 155–171

(1994) in averaging erosion over each exposed boulder fragmented. Only 5 non-soaked diorite clasts did not
surface area. Thus, for Boulder 1.2 (Table 1), two- break apart, although millimeter-scale surface spalling
thirds of the net boulder erosion took place as a result took place on these 5. These results suggest that
of the fire and subsequent cooling. Very little change wetting and lithology may be more important in
took place during the monsoon season. However, one eroding clasts than the presence of nails.
third of this boulder’s erosion occurred during the
winter.
Table 2 summarizes erosion history of the 19 5. Discussion
resurveyed diorite boulders. Almost 80% of these
diorite boulders experienced some fire erosion, with The results of this study generate several distinct
almost a third averaging more erosion than the nail points of discussion relevant to different aspects of the
penetration of 76 mm. After the monsoon, only a prior literature.
quarter of the diorite boulders eroded with an average
of < 1 mm. Only one diorite boulder experienced 5.1. Bimodal erosion
additional spalling after the winter.
Table 3 summarizes a much more active erosion The June 2000 survey after the fire revealed a
history for 25 resurveyed sandstone boulders. All but strongly bimodal pattern to boulder erosion. Instead of
three sandstone boulders experienced some fire ero- aggregating data by each separate boulder (cf. Tables
sion, and one of those three eroded after the winter. 2 and 3 and Zimmerman et al., 1994), Figs. 4 and 5
Like the diorite, erosion rates dropped precipitously aggregate spalling thickness by the total area of a
after the fire. However, 40% of the boulders eroded an given spall thickness. Viewed in this cumulative
average of 1 mm during the monsoon season, with fashion, more than half of the area of all of the
more than 50% eroding an average of 5.5 mm during sandstone boulders did not erode after the fire, but a
the winter. little more than 20% of the sandstone surface area
The blow torch experiment resulted in fragmenta- eroded >76 mm (Fig. 4). Diorite boulders displayed
tion of 19 of the 24 clasts into smaller centimeter and even stronger bimodal behavior, but in a reverse
subcentimeter-sized pieces. All 12 of the water-soaked pattern (Fig. 5): >60% of the surface area eroded
clasts fragmented (3 diorite with nail, 3 sandstone >76 mm, while almost a third of the diorite surface did
with nail, 3 diorite without nail, 3 sandstone without not erode after the fire. Fig. 3 photographically
nail). Of the air-dried clasts, 4 with nail (3 sandstone, illustrates these extremes of no erosion and extensive
1 diorite) and 3 without nail (3 sandstone, 0 diorite) erosion.

Fig. 4. Erosion of sandstone boulder area, graphed by spalling depth for all sandstone boulders resurveyed after the burn.
 R.I. Dorn / Geomorphology 55 (2003) 155–171 163

boulder fractures (Fig. 6), rock coatings that could

be seen on open spalled rock faces after the fire.
Laminar calcrete is common in arid and semiarid
settings (Goudie, 1996), but it also occurs in humid
contexts (Reams, 1990). Carbonate responds differ-
ently to thermal stress than silicate minerals (Gaffey et
al., 1991), and organic material in laminar calcrete
(Chitale, 1986; Klappa, 1979) further changes the
response to firing. Thus, the laminar calcrete coatings
may have facilitated diorite spalling along these
internal fractures.
The style by which rock surfaces weather and
erode has implications for interpretation of cosmo-
genic nuclides that build up in situ in rocks. Although
Fig. 5. Erosion of diorite boulder area, graphed by spalling depth for the current perspective (Hancock et al., 1999; Heim-
all diorite boulders resurveyed after the burn. sath et al., 1999; Jackson et al., 1999) assumes that
rock erosion or soil production from bedrock behaves
The general bimodal pattern for diorite does not in a fashion that can be readily modeled with simple
change substantially after the monsoon and winter modeling assumptions, such as 5-mm erosion per
seasons. The pattern for sandstone also remains bimo- 1000 years, these bimodal findings reveal the oppo-
dal after summer and winter, but changes slightly. site. Some boulders are clearly susceptible to fire-
Total sandstone surface area with zero erosion dec- related weathering, and others are not. A wide degree
lines to 43% after the monsoon and 38% after the of irregular behaviors exists in rock weathering, with
winter, and total surface area with erosion >76 mm the need for cosmogenic models to interface with
increased to 22% after the monsoon and 27% after the scalable weathering conceptual models (Trudgill,
winter. 2000). The point is that weathering research can help
I am not sure why diorite tended to erode in the generate a realistic, field-contextualized, weathering
fashion of spalling fairly thick blocks. One possible error term that does not now exist in any reported
explanation is the presence of laminar calcrete in cosmogenic dating result.

Fig. 6. Humid laminar calcrete lining fractures in diorite boulders, as seen in this backscattered electron microscope image of the calcrete
surrounded by plagioclase.
164 R.I. Dorn / Geomorphology 55 (2003) 155–171

5.2. Why did sandstone erode more in winter? internal quartz weakness (Pope, 1995b) or perhaps
internal fractures in quartz (Pope, 1995a) aided frac-
According to laboratory research, field outcrops turing of quartz grains.
should have a greater degree of susceptibility to In contrast, sandstone spalling faces examined after
weathering and erosion processes after a fire (Goudie the winter were dominated by breakage along bedding
et al., 1992). Field data for the weathering of sand- planes, grain boundaries, and especially cutting across
stone boulders (Table 3) supports this experimental weathered feldspar (Fig. 8). The general condition
insight. More than half of the sandstone boulders seen in Fig. 8 is based on SEM examination of eight
eroded an average of 5.5 mm during the winter. Four comparative samples from June 2000, October 2000,
sandstone boulders that did not erode after the and May 2001 collections. Feldspar discongruent
monsoon eroded after the winter. Five sandstone dissolution probably did not occur during the fire
boulders experienced substantial additional losses event; measurements of feldspar weathering over time
after the winter, including Boulder 1.2 exemplified in different circumstances (Dorn and Brady, 1995;
in Table 1. Gordon, 1996) indicate that the swiss-cheese appear-
Initial fire erosion revealed a pattern of erosion that ance in Fig. 8 likely takes tens of thousands of years
cut across bedding planes and individual minerals. to develop. Thus, observed internal feldspar weath-
Microscopic observation in the field and laboratory ering likely developed in situ long before the fire,
frequently revealed individual quartz grains breaking perhaps even during a prior rock cycle.
by conchoidal fracturing (Fig. 7). At first glance, this Warke and Smith (1994) emphasized that ‘‘in-
did not make sense. Quartz expands approximately herited’’ weathering can greatly influence weathering
four times more than the feldspar in the sandstone behavior, and inherited weathering may indeed be
(Winkler, 1975), suggesting that post-fire breakage very important in explaining the erosion behavior at
should occur along grain boundaries. I speculate that Coon Creek. In this case, inherited porosity may
have helped support hygroscopic and capillary water,
that in turn helped produce micro-environmental
conditions (cf. Warke and Smith, 1998) prompting
fracturing along feldspar-rich bedding planes. Exper-
imental studies revealed the importance of moisture
in fired rocks (Allison and Goudie, 1994), suggesting
that the fire may have helped weaken previously
weathered feldspar grains—making them more sus-
ceptible to post-fire weathering processes such as
winter-time frost weathering (McGreevy and Whal-
ley, 1985).
Inherited weathering may also play a role in the
limited diorite erosion experienced during the mon-
soon and winter seasons. Only one of the diorite
boulders eroded during the winter, and only an aver-
age 1.2 mm. A comparison of the face of a fire spall
with the face of the winter spall (Fig. 9) revealed the
fire spall to be cut across minerals that were not very
weathered. In contrast, winter spalling cut grains with
substantial internal porosity (Fig. 9). These SEM
insights are anecdotal because I only examined spal-
ling from one winter boulder and four fire-spalled
Fig. 7. Erosional processes from the fire event include internal
breakage of quartz grains, as evidenced by the conchoidal fracture diorite boulders. Still, these diorite observations indi-
pattern in this quartz grain from Boulder 1.2 collected after the fire cate the importance of ‘‘inherited’’ weathering (cf.
in June 2000. Warke and Smith, 1994).
 R.I. Dorn / Geomorphology 55 (2003) 155–171 165

Fig. 8. Spalled grain from sandstone Boulder 1.2 collected after the winter season in May 2001. The less-weathered upper part of the image is
quartz, and the heavily porous lower part of the image is plagioclase feldspar. The spall face cut across the bedding plane of heavily weathered
plagioclase.

5.3. Does size matter? tance of more internal water in pore spaces closer to
the ground. But considered overall, size was not a
Size should matter in a boulder’s erosional re- significant factor at the Coon Creek burn, just like size
sponse to wildfire. Changes in rock modulus of did not matter on a glacial moraine in Wyoming
elasticity should generate differences in weathering (Zimmerman et al., 1994).
and erosion. Experimental studies reveal strong cor-
relations between rock size and the amount of spal- 5.4. Lost boulders, slope armoring, and climate
ling; the percentage of rock mass eroded in one study change
was proportionally higher for larger rocks, explained
by differences in temperature gradients (Nealson, My failure to relocate two-thirds of the originally
1995). Similarly, some field research also revealed surveyed boulders could be attributed to incomplete
that size matters, with fire rounding larger boulders care in taking field notes in 1989, complete erosion of
the most (Hellborg, 1995). boulders at a site, or some combination. Based on
My data, however, did not reveal any statistically finding nails with metal detectors at locales where the
significant relationship between boulder height or boulders should have been, my subjective opinion is
boulder surface area and erosion. Results that are that fire obliterated many, if not most, of these
not statistically significant reveal a weak negative boulders (Fig. 11). The upper frame shows a half-
relationship between boulder height and erosion rates decimated boulder. The lower frame shows a site that
for sandstone and diorite, where higher boulders tend formerly hosted boulders.
to experience proportionally less erosion than bould- Although the blow torch experiment was not
ers closer to the ground (Fig. 10). conclusive regarding the importance of nail em-
I have no explanation for size not mattering, except placement in boulder fragmentation, results suggest
to stress the obvious that each boulder has a long and that the presence of water and the role of lithology
individual history of prior weathering. That ground may be more important than the occurrence of an
huggers eroded more could reflect the importance of epoxied nail. Birkeland (1984, p. 64) did not use
prior weathering at the ground surface or the impor- nails in a qualitative analysis of fire weathering,
166 R.I. Dorn / Geomorphology 55 (2003) 155–171

Fig. 9. Comparison of backscattered electron microscope images of (A) diorite boulders only eroded from the fire (from Boulder 17.12) with (B)
the single monitored diorite boulder experiencing winter erosion (from Boulder 18.6). Note the greater degree of internal weathering in (B).
Hygroscopic and perhaps capillary water stored in the pores may have promoted wintertime frost weathering of Boulder 18.6.

Fig. 10. Strongest relationship between erosion and boulder size, in this case between sandstone boulder height and erosion averaged over the
boulder surface. This relationship is not statistically significant, with a p value of 0.15.
 R.I. Dorn / Geomorphology 55 (2003) 155–171 167

Fig. 11. The potential for fire shattering of entire boulders. (A) This image illustrates shattering of half of a 2-m diameter sandstone boulder near
site 14. (B) This image illustrates the position of site 12, creating a humorous condition for field assistants brought to help survey fire-shattered
boulders.

and yet he observed that fires are able to turn If fires do obliterate boulders, producing clasts
boulders into rubble. Thus, fire may be a key agent more readily transported by slope and fluvial pro-
in pre-paration of 10– 20-cm diameter clasts for flu- cesses, then wildfires become extraordinarily impor-
vial transport. tant agents of landscape evolution. The importance of
168 R.I. Dorn / Geomorphology 55 (2003) 155–171

boulders influencing slope evolution (Wahrhaftig, ered by monitoring the effects of controlled burns on
1965; Abrahams et al., 1985; Allison and Higgitt, carefully planted and monitored rock samples might
1998) received new support from a study of 26Al and allow researchers to link experimental studies and
10
Be nuclides (Granger et al., 2001). Granitic boulders serendipitous studies of wildfires.
and bedrock emerge as fine sediment erodes rapidly;
because boulders erode more slowly than do regolith-
covered catchments, boulders control overall rates of Acknowledgements
erosion and drive morphogenesis (Granger et al.,
2001). If agents, such as fire, can obliterate boulders Thanks to Tonto National Forest for permission to
in a single mountain wildfire event, then wildfires conduct the surveys, Arizona State University for
have the potential to be key limiting agents in the facilities support, and D. Butler, D. Marston, and an
geomorphic evolution of bouldered slopes—essen- anonymous reviewer for comments.
tially allowing drainages to develop more rapidly in
fired locations.
Processes of wildfire erosion of boulders are not References
limited to currently forested mountains. Consider, for
example, the granitic pediment and inselberg moun- Abrahams, A.D., Parsons, A.J., Hirsch, P.J., 1985. Hillslope gra-
tain landscapes of the Mojave and Sonoran Deserts dient – particle size relations: evidence of the formation of debris
(Abrahams et al., 1985; Oberlander, 1989). The Hol- slopes by hydraulic processes in the Mojave Desert. Journal of
ocene is a hyper-arid anomaly in the western USA Geology 93, 347 – 357.
Abrahams, A.D., Parsons, A.J., Luk, S., 1988. Hydrologic and sedi-
(Smith et al., 1983). Woody plants were far more ment responses to simulated rainfall on desert hillslopes in
abundant in wetter phases of the Quaternary (Spauld- southern Arizona. Catena 15, 103 – 117.
ing, 1990; McAuliffe and Van Devender, 1998). Abrahams, A.D., Soltyka, N., Parsons, A.J., Hirsch, P.J., 1990.
Wildfire, thus, may have been an important geomor- Fabric analysis of a desert debris slope: Bell Mountain, Califor-
phic agent in desert geomorphology, creating locally nia. Journal of Geology 98, 264 – 272.
Allen, C., Swetnam, T., Betancourt, J., 1998. Landscape changes in
important patterns of hillslope erosion through the southwestern United States: techniques, long-term data sets,
degrading boulders into sizes more compatible with and trends. In: Sisk, T.D. (Ed.), Perspectives on the Land Use
hydraulic transport (Abrahams et al., 1988, 1990; History of North America: A Context for Understanding Our
Allison and Higgitt, 1998). Changing Environment. U.S. Geological Survey Biological Re-
sources Division, Biological Science Report USGS/BRD/BSR-
1998-0003. See also http://biology.usgs.gov/luhna/chap9.html.
U.S. Geological Survey, Reston, VA, p. 104.
6. Concluding argument for controlled burn Allison, R.J., Bristow, G.E., 1999. The effects of fire on rock
studies weathering: some further considerations of laboratory experi-
mental simulation. Earth Surface Processes and Landforms 24,
This study was made possible by the serendipity of 707 – 713.
Allison, R.J., Goudie, A.S., 1994. The effects of fire on rock weath-
a major wildfire in a location of a prior survey of rock ering: an experimental study. In: Robinson, D.A., Williams,
surfaces. Results highlight both agreements and com- R.B.G. (Eds.), Rock Weathering and Landform Evolution. Wi-
plexities in comparisons between field results and ley, Chichester, UK, pp. 41 – 56.
experimental research on the effects of fire on rock Allison, R.J., Higgitt, D.L., 1998. Slope form and associations with
weathering. ground boulder cover in arid environments, northeast Jordan.
Catena 33, 47 – 74.
One scaling solution to link wildfires and labora- Baker, W.L., Kipfmueller, K.F., 2001. Spatial ecology of pre-Euro-
tory studies would be to institute controlled studies of American fires in a southern rocky mountain subalpine forest
rock weathering at planned burns. Controlled burns do landscape. Professional Geographer 53, 248 – 262.
not have the intensity or longevity of wildfires. How- Bian, L., Walsh, S.J., 1993. Scale dependencies of vegetation and
ever, since we are all in the very early stages of this topography in a mountainous environment of Montana. Profes-
sional Geographer 45, 1 – 11.
science, currently driven by data generated from Bierman, P., Gillespie, A., 1991. Range fires: a significant factor in
‘‘endpoints’’ of intense wildfires and controlled exper- exposure-age determination and geomorphic surface evolution.
imental research, the sort of intermediate data gath- Geology 19, 641 – 644.
 R.I. Dorn / Geomorphology 55 (2003) 155–171 169

Bird, M.I., 1995. Fire, prehistoric humanity, and the environment. Dixon, J.C., Thorn, C.E., Darmody, R.G., 1984. Chemical weath-
Interdisciplinary Science Reviews 20, 141 – 154. ering processes on the Vantage Peak nunatak, Juneau Icefield,
Birkeland, P.W., 1984. Soils and Geomorphology. Oxford Univ. southern Alaska. Physical Geography 5, 111 – 131.
Press, New York. Donnegan, J.A., Rebertus, A.J., 1999. Rates and mechanisms of
Blackwelder, E., 1927. Fire as an agent in rock weathering. Journal subalpine forest succession along an environmental gradient.
of Geology 35, 134 – 140. Ecology 80, 1370 – 1384.
Brais, S., David, P., Ouimet, R., 2000. Impacts of wild fire severity Dorn, R.I., Brady, P.V., 1995. Rock-based measurement of temper-
and salvage harvesting on the nutrient balance of jack pine and ature-dependent plagioclase weathering. Geochimica et Cosmo-
black spruce boreal stands. Forest Ecology and Management chimica Acta 59, 2847 – 2852.
137, 231 – 243. Dragovich, D., 1993. Fire-accelerated boulder weathering in the
Bridge, S., Johnson, E., 2000. Geomorphic principles of terrain Pilbara, Western-Australia. Zeitschrift für Geomorphologie N.F.
organization and vegetation gradients. Journal of Vegetation 37, 295 – 307.
Science 11, 57 – 70. Drever, J.I., Zobrist, J., 1992. Chemical weathering of silicate rocks
Brown, P.M., Kaye, M.W., Huckaby, L.S., Baisan, C.H., 2001. Fire as a function of elevation in the southern Swiss Alps. Geochi-
history along environmental gradients in the Sacramento Moun- mica et Cosmochimica Acta 56, 3209 – 3216.
tains, New Mexico: influences of local patterns and regional Edwards, K.J., Whittington, G., Tipping, R., 2000. The incidence of
processes. Ecoscience 8, 115 – 126. microscopic charcoal in late glacial deposits. Palaeogeography,
Brundrett, M.C., Ashwath, N., Jasper, D.A., 1996. Mycorrhizas Palaeoclimatology, Palaeoecology 164, 247 – 262.
in the Kakadu region of tropical Australia: 2. Propagules of Emery, K., 1944. Brush fires and rock exfoliation. American Jour-
mycorrhizal fungi in disturbed habitats. Plant and Soil 184, nal of Science 242, 506 – 508.
173 – 184. Gaffey, S.J., Kulak, J.J., Bronnimann, C.E., 1991. Effects of drying,
Butler, D.R., 2001. Geomorphic process—disturbance corridors: a heating, annealing and roasting on carbonate skeletal material
variation on a principle of landscape ecology. Progress in Phys- with geochemical and diagenetic implications. Geochimica et
ical Geography 25, 237 – 248. Cosmochimica Acta 55, 1627 – 1640.
Butler, D.R., Walsh, S.J., 1994. Site characteristics of debris flows Garty, J., 1992. The postfire recovery of rock-inhabiting algae,
and their relationship to alpine treeline. Physical Geography 15, microfungi, and lichens. Canadian Journal of Botany - Revue
181 – 199. Canadienne de Botanique 70, 301 – 312.
Cannon, S.H., Kirkham, R.M., Parise, N., 2001. Wildfire-related Germanoskiy, D., Miller, J.R., 1995. Geomorphic response to wild-
debris-flow initiation processes, Storm King Mountain, Colora- fire in an arid watershed, Crow Canyon, Nevada. Physical
do. Geomorphology 39, 171 – 188. Geography 16, 243 – 256.
Carcaillet, C., 1998. A spatially precise study of Holocene fire Gislason, S.R., Arnorsson, S., Armannsson, H., 1996. Chemical
history, climate and human impact within the Maurienne valley, weathering of basalt in southwest Iceland: effects of runoff,
North French Alps. Journal of Ecology 86, 384 – 396. age of rocks and vegetative/glacial cover. American Journal of
Carcaillet, C., Thinon, M., 1996. Pedoanthracological contribution Science 296, 837 – 907.
to the study of the evolution of the upper treeline in the Mauri- Gordon, S., 1996. Assessing the effects of temperature on plagio-
enne valley (North French Alps): methodology and preliminary clase weathering, Iztaccihuatl, Mexico. Masters Thesis, Geog-
data. Review of Palaeobotany and Palynology 91, 399 – 416. raphy, Arizona State University, Tempe.
Cerda, A., 1998. Relationships between climate and soil hydrolog- Goudie, A.S., 1996. Organic agency in calcrete development. Jour-
ical and erosional characteristics along climatic gradients in nal of Arid Environments 32, 103 – 110.
Mediterranean limestone areas. Geomorphology 25, 123 – 134. Goudie, A.S., Allison, R.J., McClaren, S.J., 1992. The relations
Chinn, T.J.H., 1981. Use of rock weathering-rind thickness for between modulus of elasticity and temperature in the context
Holocene absolute age-dating in New Zealand. Arctic and Al- of the experimental simulation of rock weathering by fire. Earth
pine Research 13, 33 – 45. Surface Processes and Landforms 17, 605 – 615.
Chitale, J.D., 1986. Study of petrography and internal structures Granger, D.E., Riebe, C.S., Kirchner, J.W., Finkel, R.C., 2001.
in calcretes of West Texas and New Mexico (Microtextures, Modulation of erosion on steep granitic slopes by boulder ar-
Caliche). PhD dissertation, Geosciences, Texas Tech University, moring, as revealed by cosmogenic Al-26 and Be-10. Earth and
Lubbock. Planetary Science Letters 186, 269 – 281.
Cleveland, G.B., 1977. Analysis of erosion following the Marble Hancock, G.S., Anderson, R.S., Chadwick, O.A., Finkel, R.C.,
Cone fire, Big Sur basin, Monterey County, California. Cali- 1999. Dating fluvial terraces with Be-10 and Al-26 profiles:
fornia Division of Mines and Geology Open File Report 77- application to the Wind River, Wyoming. Geomorphology 27,
12, 1 – 11. 41 – 60.
Condie, K.C., Dengate, J., Cullers, R.L., 1995. Behavior of rare- Heimsath, A.M., Dietrich, W.E., Nishiizumi, K., Finkel, R.C., 1999.
earth elements in a paleoweathering profile on granodiorite in Cosmogenic nuclides, topography, and the spatial variation of
the Front Range, Colorado, USA. Geochimica et Cosmochimica soil depth. Geomorphology 27, 151 – 172.
Acta 59, 279 – 294. Hellborg, C., 1995. Fire weathering on Isterklev Moor. Bachelor’s
Diamond, J., 1997. Guns, Germs, and Steel: The Fates of Human Thesis, Geography, Göteborg University, Göteborg, Sweden.
Societies. Norton, New York. Heyerdahl, E.K., Brubaker, L.B., Agee, J.K., 2001. Spatial controls
170 R.I. Dorn / Geomorphology 55 (2003) 155–171

of historical fire regimes: a multiscale example from the interior Jotunheimen, Southern Norway. Arctic and Alpine Research 21,
west, USA. Ecology 82, 660 – 678. 268 – 275.
House, J.H., Smith, J.W., 1975. Experiments in the replication of McDonald, D.J., Cowling, R.M., Boucher, C., 1996. Vegetation –
fire-cracked rock. Arkansas Archaeological Survey Research environment relationships on a species-rich coastal mountain
Series [The Cache River Archaeological Project, assembled by range in the fynbos biome (South Africa). Vegetatio 123,
M.B. Schiffer and J.H. House], vol. 8. University of Arkansas, 165 – 182.
Fayetteville, AR, pp. 75 – 80. McFarland, P., 1977. Experiments in the firing and breaking of
Jackson, L.E., Phillips, F.M., Little, E.C., 1999. Cosmogenic Cl-36 rocks. Calgary Archaeologist 5, 31 – 33.
dating of the maximum limit of the Laurentide Ice Sheet in McGreevy, J.P., Whalley, W.B., 1983. The geomorphic significance
southwestern Alberta. Canadian Journal of Earth Sciences 36, of rock temperature variations in cold environments. A discus-
1347 – 1356. sion. Arctic and Alpine Research 14, 157 – 162.
Kaye, M.W., Swetnam, T.W., 1999. An assessment of fire, climate, McGreevy, J.P., Whalley, W.B., 1985. Rock moisture and frost
and Apache history in the Sacramento Mountains, New Mexico. weathering under natural and experimental conditions. A com-
Physical Geography 20, 305 – 330. parative discussion. Arctic and Alpine Research 17, 337 – 346.
Kelly, R., McCarthy, D.F., 2001. Effects of fire on rock art. Amer- McGreevy, J.P., Warke, P.A., Smith, B.J., 2000. Controls on stone
ican Indian Rock Art 27, 169 – 176. temperatures and the benefits of interdisciplinary exchange.
Klappa, C.F., 1979. Lichen stromatolites: criterion for subaerial Journal of the American Institute for Conservation 39, 259 – 274.
exposure and a mechanism for the formation of laminar cal- Meyer, G.A., Wells, S.G., Jull, A.J.T., 1995. Fire and alluvial chro-
cretes (caliche). Journal of Sedimentary Petrology 49, 387 – 400. nology in Yellowstone National Park: climatic and intrinsic con-
Kritzer, K., 1995. Thermolithofractography: a comparative analysis trols on Holocene geomorphic processes. Geological Society of
of cracked rock from an archaeological site and cracked rock America Bulletin 107, 1211 – 1230.
from a culturally sterile sera. MS Thesis, Anthropology, Ball Moody, A., Meentemeyer, R.K., 2001. Environmental factors influ-
State University, Muncie, IN. encing spatial patterns of shrub diversity in chaparral, Santa
Laird, L.D., Campbell, I.D., 2000. High resolution palaeofire sig- Ynez Mountains, California. Journal of Vegetation Science 12,
nals from Christina Lake, Alberta: a comparison of the charcoal 41 – 52.
signals extracted by two different methods. Palaeogeography, Morris, S.E., Moses, T., 1987. Forest-fire and the natural soil-ero-
Palaeoclimatology, Palaeoecology 164, 111 – 123. sion regime in the Colorado Front Range. Annals of the Asso-
Lichtfouse, E., 1999. Temporal pools of individual organic substan- ciation of American Geographers 77, 245 – 254.
ces in soil. Analusis 27, 442 – 444. Nealson, E., 1995. Fire as a geomorphic agent in rock weathering:
Lichtfouse, E., Bardoux, G., Mariotti, A., Balesdent, J., Ballentine, the effect of rock size on weathering efficiency under simulated
D.C., Macko, S.A., 1997. Molecular, C-13, and C-14 evidence forest fire conditions. Masters Thesis, Geography, University of
for the allochthonous and ancient origin of C-16 – C-18 n-al- Iowa, Iowa City.
kanes in modern soils. Geochimica et Cosmochimica Acta 61, Oberlander, T.M., 1989. Slope and pediment systems. In: Thomas,
1891 – 1898. D.S.G. (Ed.), Arid Zone Geomorphology. Belhaven Press, Lon-
Mahaney, W.C., Vortisch, W., Julig, P., 1988. Relative differences don, pp. 56 – 84.
between glacially crushed quartz transported by mountain and Ollier, C.D., 1983. Fire and rock breakdown. Zeitschrift für Geo-
continental ice—some examples. American Journal of Science morphologie 27, 363 – 374.
288, 810 – 826. Pope, G.A., 1995a. Internal weathering in quartz grains. Physical
Malanson, G.P., Butler, D.R., 1984. Avalanche paths as fuel breaks. Geography 16, 315 – 338.
Implications for fire management. Journal of Environmental Pope, G.A., 1995b. Newly discovered submicron-scale weathering
Management 19 (3), 229 – 238. in quartz: geographical implications. Professional Geographer
Marion, J., Filion, L., Hetu, B., 1995. The Holocene development 47, 375 – 387.
of a debris slope in subarctic Quebec, Canada. Holocene 5, Power, G., Power, M., 1995. Ecotones and fluvial regimes in Arctic
409 – 419. environments. Hydrobiologia 303 (1 – 3), 111 – 124.
Matthews, J.A., Shakesby, R.A., 1983. The status of the Little Ice Prosser, I.P., Williams, L., 1998. The effect of wildfire on runoff
Age in southern Norway. Relative-age dating of Neoglacial and erosion in native Eucalyptus forest. Hydrological Processes
moraines with Schmidt hammer and lichenometry. Boreas 13, 12, 251 – 265.
333 – 346. Pyne, S.J., 1982. Fire in America: A Cultural History of Wildland
Mazhitova, G.G., 2000. Pyrogenic dynamics of permafrost-af- and Rural Fire. Princeton Univ. Press, Princeton, NJ.
fected soils in the Kolyma Upland. Eurasian Soil Science 33, Pyne, S.J., 1998. Pyre on the mountain. In: Rothman, H.K. (Ed.),
542 – 551. The Second Opening of the West. University of Arizona, Tuc-
McAuliffe, J.R., Van Devender, T.R., 1998. A 22,000-year record son, pp. 38 – 52.
of vegetation change in the north-central Sonoran Desert. Rapp, G.J., Balescu, S., Lamothe, M., 1999. The identification of
Palaeogeography, Palaeoclimatology, Palaeoecology 141, granitic fire-cracked rocks using luminescence of alkali feld-
253 – 275. spars. American Antiquity 64, 71 – 78.
McCarroll, D., 1989. Potential and limitations of the Schmidt ham- Raymo, M.E., 1994. The Himalayas, organic carbon burial, and
mer for relative-age dating. Field-tests on neoglacial moraines, climate in the Miocene. Paleoceanography 9, 399 – 404.
 R.I. Dorn / Geomorphology 55 (2003) 155–171 171

Reams, M.W., 1990. Stromatolitic humid climate carbonates: a va- throw on the cause of the disaggregation of granite. Economic
riety of calcrete. In: Douglas, L.A. (Ed.), Soil Micromorphol- Geology 10, 348 – 367.
ogy: A Basic and Applied Science. Elsevier, Amsterdam, Thorn, C.E., 1976. Quantitative evaluation of nivation in Colorado
pp. 395 – 400. Front Range. Geological Society of America Bulletin 87,
Selkirk, P.M., Adamson, D.A., Downing, A.J., 2001. Landform and 1169 – 1178.
vegetation change in the Greaves Creek Basin: an asymmetric Thorn, C.E., 1979. Bedrock freeze – thaw weathering regime in an
hanging valley in the Blue Mountains, New South Wales. Aus- alpine environment, Colorado Front Range. Earth Surface Pro-
tralian Geographer 32, 45 – 75. cesses and Landforms 4, 211 – 228.
Shakesby, R., Coelho, C., Ferreira, A., Terry, J., Walsh, R., 1993. Thorn, C.E., 1980. Nivation. An arctic – alpine comparison and re-
Wildfire impacts on soil-erosion and hydrology in wet Mediter- appraisal. Journal of Glaciology 25, 109 – 124.
ranean forest, Portugal. International Journal of Wildland Fire 3, Trudgill, S.T., 2000. Weathering overview—measurement and mod-
95 – 110. elling. Zeitschrift für Geomorphology Supplementband 120,
Siffedine, A., Bertrand, P., Fournier, M., Martin, L., Servant, M., 187 – 193.
Soubies, F., Suguio, K., Turcq, B., 1994. The lacustrine organic Viles, H., 1995. Ecological perspectives on rock surface weather-
sedimentation in tropical humid environment (Carajas, eastern ing: towards a conceptual model. Geomorphology 13, 21 – 35.
Amazonia, Brazil)—relationship with climatic changes during Wadleigh, L., Jenkins, M.J., 1996. Fire frequency and the vegeta-
the last 60,000 years BP. Bulletin de la Societe Geologique de tive mosaic of a spruce-fir forest in northern Utah. Great Basin
France 165 (6), 613 – 621. Naturalist 56 (1), 28 – 37.
Smith, G.I., Barczak, V., Moulton, G., Liddicoat, C., 1983. Core Wahrhaftig, C., 1965. Stepped topography of the southern Sierra
KM-3, a surface to bedrock record of Late Cenozoic sedimen- Nevada. Geological Society of America Bulletin 76, 1165 – 1190.
tation in Searles Valley, California. U.S. Geological Survey Pro- Warke, P.A., Smith, B.J., 1994. Inheritance effects on the efficacy
fessional Paper, vol. 1256. Washington, DC, pp. 1 – 24. of salt weathering mechanisms in thermally cycled granite
Soons, J.M., 1994. Changes in geomorphic environments in Canter- blocks: a simulation experiment. In: Bell, E., Cooper, T.P.
bury during the Aranuian. New Zealand Journal of Botany 32 (Eds.), Granite Conservation and Weathering. Trinity College,
(3), 365 – 372. Dublin, pp. 19 – 27.
Spaulding, W.G., 1990. Vegetation and climatic development of the Warke, P.A., Smith, B.J., 1998. Effects of direct and indirect heating
Mojave Desert: the last glacial maximum to the present. In: on the validity of rock weathering simulation studies and dura-
Betancourt, J.L., Van Devender, T.R., Martin, P.S. (Eds.), Pack- bility tests. Geomorphology 22, 347 – 357.
rat Middens. University of Arizona Press, Tucson, pp. 166 – 199. Warke, P.A., Smith, B.J., Magee, R.W., 1996. Thermal response
Suffling, R., 1993. Induction of vertical zones in sub-alpine valley characteristics of stone: implications for weathering of soiled
forests by avalanche-formed fuel breaks. Landscape Ecology 8 surfaces in urban environments. Earth Surface Processes and
(2), 127 – 138. Landforms 21, 295 – 306.
Swanson, F.J., 1981. Fire and geomorphic processes. In: Mooney, Wilson, D.C., 1999. The experimental reduction of rock in a Camas
M.A., Bonnicksen, T.M., Christensen, N.L., Lotan, J.E., Oven: towards an understanding of the behavioral significance
Reiners, W.A. (Eds.), Fire Regimes and Ecosystem Properties. of fire-cracked rock. Archaeology in Washington 7, 81 – 89.
U.S. Department of Agriculture General Technical Report WO- Winkler, E.M., 1975. Stone: Properties, Durability in Man’s Envi-
26, Washington, DC, pp. 401 – 420. ronment. Springer, Berlin. 230 pp.
Swetnam, T.W., Betancourt, J., 1990. Fire – southern oscillation Wohl, E.E., Pearthree, P.P., 1991. Debris flows as geomorphic
relations in the Southwestern United States. Science 249, agents in the Huachuca Mountains of southeastern Arizona.
1017 – 1020. Geomorphology 4, 273 – 292.
Swetnam, T.W., Betancourt, J.L., 1998. Mesoscale disturbance and Zimmerman, S.G., Evenson, E.B., Gosse, J.C., Erskine, C.P., 1994.
ecological response to decadal climatic variability in the Amer- Extensive boulder erosion resulting from a range fire on the
ican Southwest. Journal of Climate 11, 3128 – 3147. type-Pinedale moraines, Fremont Lake, Wyoming. Quaternary
Tarr, W.A., 1915. A study of some heating tests and the light they Research 42, 255 – 265.
 Geomorphology 55 (2003) 173 – 202
www.elsevier.com/locate/geomorph

Influence of substrate on the distribution of the Hawaiian

Silversword (Argyroxiphium sandwicense DC.)
in Haleakala (Maui, HI)
Francisco L. Pérez *
Department of Geography, University of Texas, Austin, TX 78712-1098, USA
Received 16 November 2001; received in revised form 30 May 2002; accepted 10 March 2003

Abstract

The spatial patterns of the Hawaiian silversword (Argyroxiphium sandwicense DC.) were studied in Haleakala (Maui, HI).
The silversword is a ‘‘giant’’ rosette plant nearly brought to extinction by human impact and goat browsing during the 1920s,
but stern protection has resulted in the resurgence of plant populations. Silversword regeneration is occurring vigorously in soils
with surficial layers of volcaniclastic fragments. Ten sites with sizable silversword populations and the associated substrates
were examined in Haleakala’s crater between 2175 and 2755 m. At each site, the population structure of 100 plants was
determined along wandering-quarter transects, which limit sampling bias. Substrates where rosettes were rooted and the size of
the biggest stone fragments upslope and downslope from the plant’s base were determined. Volcaniclastic substrates were
examined with photographs along 12-m-long transverse transects. Fine-debris samples (gravel and pebbles) were analyzed by
mechanical sifting of surficial fragments gathered from 15  15 cm miniplots.
Volcaniclastic fragments in silversword areas have different sources. (i) Most were ejected during eruptions yielding
pyroclastic materials (ash to blocks) that built cinder cones on Haleakala’s crater. (ii) On steep slopes, weathering of (olivine
alkali) basalt outcrops and subsequent mass wasting of rock fragments contributed a mantle of clasts to slopes below. (iii) On
some cinder cones, welded spatter was produced during late eruptive stages; this coalesced into ruff-like agglutinate layers
around crater rims, called Pohaku-o-Hanalei (‘‘wreath of stones’’). Agglutinate deposits weather gradually, supplying clasts that
roll downhill and accumulate on cone flanks. (iv) On weathered aa lava flows, stone fragments and rocky crags abound along
eroding lava-flow ridges, associated with vegetation; but the intervening troughs, covered with fine ash and cinder, remain bare.
Silversword population structures indicate healthy regeneration. All populations display a typical pattern of decreasing
silversword numbers with increasing size ( = age). Nearly 65% of the rosettes are rooted near cobbles (5 – 10 cm long) or
pebbles (2.5 – 5 cm), and f 23% grow at the base of blocks (>10 cm) or outcrops; 9% of the plants are on gravel ( < 2.5 cm),
but only 1% have germinated on sandy soil without granules or stones; 2% of the silverswords grow on organic litter of dead
rosettes. Plants on six sites showed noticeable differences in clast size between their upslope and downslope sides; particles are
significantly larger upslope from rosettes. This suggests that successful seedlings germinate below clasts, which deflect finer
downslope-moving sediment, thus protecting young plants from burial by descending debris. Statistical tests indicate a few
blocks above plants may have been stopped by adult rosettes. Such a relationship underscores the dynamic nature of steep
cinder and rubble (talus) slopes, regularly affected by dry debris slides, frost creep, and other mass-wasting processes. Slope
instability is also evinced by (i) common bending and deformation of silversword roots—particularly in seedlings—which trail

* Fax: +1-512-471-5049.
E-mail address: [email protected] (F.L. Pérez).

0169-555X/03/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0169-555X(03)00139-9
174 F.L. Pérez / Geomorphology 55 (2003) 173–202

upslope; (ii) pronounced downhill tilting and asymmetry of tall rosettes on steep gradients; (iii) presence of accumulated fine-
debris steps (9.9 – 13.8 cm thick) upslope from plants and blocks; (iv) ubiquitous fine-earth flags exposed below embedded
blocks, which intercept descending debris and deflect it laterally; (v) widespread miniature sorted stripes.
D 2003 Elsevier Science B.V. All rights reserved.

Keywords: Cinder cones; Clasts; Slope processes; Soil properties; Talus slopes; Volcaniclastics

1. Introduction not much else has been ascertained since Alexander’s

comments regarding the substrate requirements of this
1.1. Geomorphic and geoecological significance of plant. In his landmark thesis, Kobayashi (1973) noted
volcaniclastic covers rosettes grew best on pyroclastic ejecta and that high
rates of silversword reproduction occurred on sites
Over the past few years, several publications have with a complete cover of tabular fragments 2.5– 5.0
focused on the influence of coarse particles on soil cm thick. Several authors have also briefly mentioned
properties (see thorough reviews by Poesen and the connection of ‘‘volcanic gravels’’ and ‘‘cinder’’
Lavee, 1994a,b), but few works have dealt with the with the rosettes (Ruhle, 1959; Siegel et al., 1970;
effects of volcaniclastic materials on substrates or on Bruegmann, 1995). I recently showed (Pérez, 2000)
the plants that occupy them. Anthropologists have how Haleakala’s soils insulated by a f 5-cm-thick
tried for decades to elucidate the role of a pyroclastic mantle of volcaniclastic sediments remained moist for
cover (a.k.a. gravel mulch; Weaver, 1919) in amelio- up to 3 weeks; while the same soils, if not covered by
rating substrate conditions. Layers of volcaniclastic clastic fragments, became totally dry after just 3 to 6
ejecta have been utilized as an aid to farming in days.
hyperarid areas by different cultures in the past,
including Native Americans after eruptions at Sunset 1.2. Recovery of the Hawaiian silversword at
Crater, AZ f 1071 A.D. (Maule, 1963; Colton, 1965; Haleakala
Doolittle, 1998, 2000), and farmers in Lanzarote
(Canary Islands) following vast pyroclastic eruptions The Hawaiian silversword (Argyroxiphium sand-
in 1730 – 1736 (Dinkins, 1969; Araña and López, wicense DC.) is the most spectacular plant in Halea-
1974). Several controlled field (Othieno, 1980; Oth- kala’s crater. This ‘‘giant’’ rosette plant has a single,
ieno and Ahn, 1980; Pérez, 1998a) and lab experi- thin (5– 8 cm) woody stem that remains covered by
ments (Groenevelt et al., 1989; Pérez, 2000) have the dry, shriveled, but strongly attached leaves the
recently investigated the efficiency of volcanic scoria plant produces; a dense, broad crown of long, narrow,
layers in reducing evaporation from, or altering the extremely shiny, new leaves caps the plant (see Fig.
temperature of, underlying soil layers. 15). The silversword is monocarpic and dies after
Most visitors at Haleakala cannot fail to notice the flowering once in the dry season (June –September),
presence of widespread volcaniclastic layers blanket- when it produces an attractive, 90- to 240-cm-tall stalk
ing the crater floor. In 1841, the first scientific party to that may contain as many as 500 purple flower heads.
venture down into the crater—part of the U.S. Explor- Several nineteenth century reports from Haleakala
ing Expedition (Wilkes, 1845, p. 254)—noted with (Missionary Herald, 1829; Wilkes, 1845; Alexander,
curiosity the black volcanic ‘‘gravel’’ formed by 1870) indicated that numerous rosettes grew on the
‘‘small angular pieces of cellular lava. . . (with) irreg- outer mountain flanks and the crater floor; cinder
ular cavities’’ on the crater. Alexander (1870, p. 45) cones and some slopes were so thickly covered by
appears to have been the first to notice the close silverswords that the landscape seemed to be ‘‘bathed
relationship of such substrates with the Hawaiian in moonlight’’ (Ruhle, 1959; Kobayashi, 1973; Mack,
silversword, and he wrote that ‘‘. . .myriads of silver- 1984). The silversword is found now above 2100 m
swords. . . are most abundant over the volcanic sand and only inside the crater—protected within Halea-
(i.e., gravel) in the central part of the crater.’’ Sadly, kala National Park (HNP)—where it is restricted to
 F.L. Pérez / Geomorphology 55 (2003) 173–202 175

cinder cones, loose volcaniclastic substrates, and display and to generally assess the success of plant
narrow ledges on inaccessible palis (cliffs) along the regeneration during the recent past.
crater’s rim (Ruhle, 1959; Kobayashi, 1973). Hiker
vandalism and browsing by feral goats and cattle
quickly reduced the numbers of silverswords between 2. The study area
the late 1800s and early 1900s (Yocom, 1967); by the
1920s, the plant was nearly extinct (Medeiros and Maui, the second largest island of the Hawaiian
Loope, 1994). Since the 1930s, stiff regulations archipelago, is located at 20j50VN. and 156j20VW.
coupled with eviction and massive extermination of The youngest and tallest part of the island, east Maui,
goats resulted in recovery of rosette populations. Stone reaches 3055 m elevation at Haleakala (Fig. 1). At the
and Loope (1987) indicated that over 20,000 goats mountaintop is Haleakala’s crater, a large erosional
were removed from HNP after 1916; Yocom’s (1967) depression f 12  4 km and >800 m deep. This
statistics suggested even greater numbers. In 1986, a miscalled ‘‘crater’’ was actually formed as two streams
goat-proof fence encircling the crater (!) was com- rapidly incised and expanded their upper basins on
pleted; as of mid-1992, only about 12 goats survived opposite sides of the island; the two valley heads
within the fenced area (Loope and Medeiros, 1994). finally merged through the Ko’olau (N) and Kau-po
Silversword regeneration appears to have been swift. (S) Gaps, creating a single, vast depression across the
In 1935, an estimated 4000 rosettes were left in summit (Stearns, 1942; Macdonald et al., 1983). The
Haleakala (Bruegmann, 1995); and by 1963 about crater floor, from 2000 to 2600 m, is occupied by
10,000 plants existed. Repeated censuses on Ka- several post-erosional lava flows and huge cinder
Moa-o-Pele cinder cone (Kobayashi, 1973, p. 10) cones (Fig. 2) that erupted during the last period of
showed a net increase in plant numbers of 53% during renewed volcanism (Hana Formation) (Macdonald,
a 27-year period (1935 – 1962). The valuable surveys 1978); these cinder cones seem to have developed
of Kobayashi (1973, p. 36) indicated a rise to 43,000 during the past 3000 to 5000 years. Spelling of all
plants by 1972. In 1986, f 50,000 rosettes were Hawaiian geographical or other native names follows
tallied (Rundel and Witter, 1994); and by 1991, silver- Pukui et al. (1974, 1992).
sword populations had expanded to almost 65,000 The Hawaiian Islands are influenced by the NE
plants (Bruegmann, 1995). Despite this successful trade winds; these create a spatially variable pattern
resurgence in a little over 50 years, the Haleakala of precipitation due to orographic effects; east
silversword is considered ecologically fragile and has Maui’s windward flank receives up to 9800 mm,
been officially listed as a ‘‘threatened’’ species since while some leeward areas get f 250 mm (Noguchi
1991 (U.S. Fish and Wildlife Service, 1992). et al., 1987). The uplands of Maui are arid, due to
The goals of this study are to (i) analyze in detail isolation from marine moisture sources caused by a
the particle-size distribution of both coarse and fine subsidence inversion that forms f 70% of the time
surficial volcaniclastic sediment on 10 sites with at 1200 –2400 m (Blumenstock and Price, 1967).
sizable populations of silverswords in Haleakala; (ii) High mountains that penetrate the inversion layer
determine, after censusing along transects of 100 suppress upward flow and cause trade wind circu-
silverswords per site, the type of substrate where each lation to divide and move around high peaks (Leo-
plant was rooted and the size of the largest stone pold, 1949). Highland skies remain persistently clear
fragment adjacent to plants’ bases, both upslope and and cloudless, allowing high insolation during the
downslope from them; (iii) examine the vertical day but quick longwave loss into the thin air during
stratification and variation in thickness of surficial the night (Giambelluca and Nullet, 1991). Air at
volcaniclastic layers, as well as of a representative soil high elevations is extremely dry; relative humidity
profile below them; (iv) briefly describe the pattern (RH) is normally < 40%, and often drops to 5 –10%
and extent of root deformation in silversword seed- (Whiteaker, 1983; Pérez, 2000, personal observa-
lings—presumably caused by soil creep—on steep tion). Precipitation in the summit is variable both in
slopes; and (v) study the overall population structure time and space; National Park Headquarters (2143
(size-class distribution) that the silverswords currently m) registers an average of f 1300 mm/year and a
 176
F.L. Pérez / Geomorphology 55 (2003) 173–202
Fig. 1. Location maps of the study area. (a) Hawaiian archipelago; the island of Maui is shaded in black. Latitude (N) and longitude (W) are shown. (b) Island of Maui. Scale is in
kilometers, altitude in meters; contour interval is 305 m ( = 1000 ft). Stippled area in east Maui indicates location of Haleakala National Park. Base map: State of Hawai’i, Principal
Islands, U.S. Geological Survey (1971), scale 1:500,000. (c) Topographic map of the western portion of Haleakala’s crater (near summit). Scale is in kilometers, altitude in meters;
contour interval is 61 m ( = 200 ft). National Park boundaries are indicated by straight dash – dot lines; road is shown with a thick dash line, trails with a thin dotted line. Study sites are
identified by dark circles and numbers. Names refer to the main cinder cones on the crater. Base maps: Kilohana and Luala’i-lua Hills (U.S. Geological Survey, 1983), scale 1:24,000.
Site 57 is located about 700 m to the right of the map’s edge.
 F.L. Pérez / Geomorphology 55 (2003) 173–202 177

Fig. 2. General view toward the north of Haleakala’s crater, from the upper section of the Ke-one-he’ehe’e (Sliding Sands) trail. The photo,
taken during early morning hours, shows only a few clouds intruding upon the crater through the Ko’olau Gap (flat area on the upper left),
which is covered by extensive lava flows. Hanakauhi peak is seen on the upper right, among the growing clouds. Two large cinder cones (Ka-
ma’o-li’i, on the left, and Pu’u-o-Maui, on the right) appear in the middle ground. Some silversword rosettes (site 4) cover the steep slope on the
foreground, which lies on the south wall of Haleakala’s crater. July 25, 2001 (Hk—01.35d).

pronounced seasonality (Yocom, 1967; Kobayashi, may surpass 33 jC, but surface soil temperatures at
1973). About 75% of the rain comes between noon often climb to 45 –48 jC in dark, bare ground;
October and April, but < 75 mm/month fall during and air RH can drop down to 6 –9% (Pérez, 2000,
the drier summer period (May to September). In unpublished data). Freezing can occur any time of
addition, the area experiences a broad, short-term, the year in the crater, where 121 –187 freeze – thaw
yearly variation; some years (e.g., 1980) receive annual (nightly) cycles occur at the ground surface
about three times the annual mean, while others (Noguchi et al., 1987).
(e.g., 1983) may reach less than a third (Leuschner Vegetation in Haleakala varies swiftly with eleva-
and Schulte, 1991). Rain records for the crater are tion; increasing aridity over the inversion level gen-
poor; the summit receives < 500 mm/year, but some erates a drought-induced timberline at f 2200 m,
crater areas may get < 130 mm/year (Yocom, 1967). where a thick subalpine scrub gives way to the
Different sources (Kobayashi, 1973; Noguchi et al., sparsely vegetated alpine desert (Whiteaker, 1983;
1987; Giambelluca and Nullet, 1991) have indicated Leuschner and Schulte, 1991). Because of the spatial
that the study sites might receive between f 600 rainfall patterns, the summit and the western half of
(at the summit) and 1500 mm/year (site 57, at low the crater lie within this alpine desert, which is
elevation, near the Ko’olau Gap) (Fig. 1c). Like associated with some 35 vascular plant species
other tropical mountains, Haleakala shows a narrow (Ruhle, 1959; Yocom, 1967). Common plants include
annual temperature range of just 3.8 jC (Whiteaker, native grasses, such as piliuka (Trisetum glomeratum),
1983), but the daily range is about 20 jC (Rundel Deschampsia nubigena, and Agrostis sandwicensis.
and Witter, 1994). Air maxima at the study sites The dominant shrubs on the crater are kupaoa
178 F.L. Pérez / Geomorphology 55 (2003) 173–202

(Dubautia menziesii), pukiawe (Styphelia tameia- 3. Research methods

meiae), and catchfly (Silene hawaiiensis) (Wagner et
al., 1990); however, the ’ahinahina (Haleakala silver- 3.1. Field sampling
sword) remains the most conspicuous plant in the
crater. Ten sites occupied by large populations of silver-
Substrates in Haleakala’s crater have been scarcely sword rosettes were investigated in Haleakala’s crater;
studied. Soils there are described as porous, coarse, fieldwork was conducted during summer (July –
poor, with low organic matter, high infiltration rates, August) in 1996, 1998, and 2001. Site altitude was
and meager water-storage capacity; such character- determined with an altimeter—probable error: F 10
istics cause edaphic aridity and frequent soil-water m—and cross-checked on USGS topographic maps.
stress to vegetation and do not provide a firm substrate Several measurements of slope angle were obtained
for rooting (Ruhle, 1959; Leuschner and Schulte, with a clinometer; aspect was evaluated with a com-
1991). The shallow, sandy, weakly developed profiles pass. The total number of plants was estimated at each
developed over loose cinders or lava were broadly site by approximate plant counting during several
classified by Cline (1955) as Regosols and Lithosols, visits; categories reported are: A = < 200; B = 200–
respectively. Similar volcanic soils of Maui’s high- 500; C = 500 – 1000; D=>1000 plants (Table 1).
lands are now catalogued as either Inceptisols or Approximate percentage of the ground covered by
Andisols (Soil Survey Staff, 1975, 1994; Leuschner stone fragments was visually assessed over large areas
and Schulte, 1991). Noguchi et al. (1987) determined at all sites; four cover classes (1 = < 35%, 2 = 35–
the texture of two samples from the Sliding Sands 70%, 3 = 70– 90%, 4=>90% clasts) were identified.
trail, with only 6 –8% fines ( < 0.063 mm) and were At each site, 100 silversword rosettes were cen-
often covered by a 1 –3 cm layer of vesicular gravel. I sused along two to four intersecting wandering-quar-
recently examined soils at Haleakala on sites 6 and 7 ter (WQ) transects (Catana, 1963). A plant was
(Fig. 1) (Pérez, 2000, 2001). Soils are coarse, as they randomly chosen at the edge of the population; the
contain f 93% sand and little to no clay; organic transect extended from there to the nearest plant
matter is normally just 1.7%. Surface soils are, how- within the silversword cluster until 25 – 50 plants
ever, greatly enriched by the addition of organic had been surveyed. The process was repeated until
material from dead rosettes, which double organic 100 plants were examined at each site. The resulting
levels to 3.4%. This raises the water-holding capacity zigzagging shape of the WQ transect strongly limits
from 23% to 43% and adds a substantial amount of sampling bias during fieldwork, as researchers are not
nutrients, particularly C, N, Ca+ +, Mg+ +, and K+ to free to choose the plants that are included (Mueller-
soils beneath dead silverswords. Walker and Powell Dombois and Ellenberg, 1974; Pérez, 1994). Data
(1999) recently also studied several soil properties in obtained for each plant included distance between
populations of Hawaiian silverswords in the island of successive rosettes along transects, plant height (ex-
Hawai’i. cluding inflorescence, if any present), crown diameter,
Kobayashi (1973) examined one sample each of reproductive stage, substrate type where each silver-
red and black cinders and noted that the clasts had sword was rooted, and size (a axis) of the largest stone
been only slightly affected by weathering and chemi- fragment adjacent to the plant’s base, both upslope
cally differed little from each other or from the and downslope from it (Pérez, 1994). The thickness of
bedrock. Such variation in cinder color is actually small debris piles accumulated upslope from rosettes
caused by iron oxidation during and after cinder was estimated at eight sites with a ruler and a
cone formation (Macdonald, 1967). Accordingly, carpenter’s level, used to determine the height of the
Kobayashi (1973, p. 75) ascribed greater ‘‘. . . impor- miniature steps to the nearest 0.5 cm. Four dead,
tance (to) physical rather than chemical differences desiccated seedlings were examined on steep slopes
of the substrate in determining the distribution of at sites 2 and 11; roots were carefully traced (A, in
plants.’’ I was unable to locate additional relevant 1998) or photographed (B, C, D, in 2001) to evaluate
soil information or profile descriptions for Haleaka- patterns and degree of root deformation presumably
la’s crater. caused by soil creep.
 F.L. Pérez / Geomorphology 55 (2003) 173–202 179

Table 1
General characteristics of 10 clast-covered sites with Hawaiian silverswords, sampled within Haleakala’s crater
Site characteristics Sampling sitesa
1 2 3 4 5 6 7 10 11 57
b
Altitude (m) 2665 2725 2755 2610 2625 2505 2510 2415 2355 2175
Aspect (j)b 15 22 20 29 43 19 70 163 342 44
Slope angle (j)b 26 12 27 25 5 8 14 20 24 9
Clast coverc 3 2 4 4 4 2 4 3 1 1
Rosette numberd A B A B A B B D D C
Geologye hb-ta hco-ta hc-ta hc-ta hb-ta ha-hb hb hc hc hm
Topographyf Stslo Geslo Stslo Stslo Geblo Geabl Geblo Ccocr Ccofl Lafla
a
Site numbers follow Kobayashi (1973); see Fig. 1 for general site locations.
b
Altitude, aspect, and slope angle indicated are site averages.
c
Approximate percentage of the ground covered by clasts was assessed in the field; key: 1 = < 35%; 2 = 35 – 70%; 3 = 70 – 90%; 4=>90%.
d
Number of silversword rosettes in each population estimated in the field; key: A = < 200 plants; B = 200 – 500 plants; C = 500 – 1000
plants; D=>1000 plants.
e
General site geology follows Macdonald (1978); consult text for description of geology units and of site topographic types.
f
Key to site topography: Stslo = steep (talus) slopes on crater flank below weathering lava outcrops; Geslo = gentle slope, below weathering
aa lava outcrops; Geblo = gentle slope, mantled with volcaniclastic blocks; Geabl = gentle slope, blanketed with windblown ash and with
scattered volcaniclastic blocks; Ccocr = cinder cone, inner crater slopes, covered by different volcaniclastic fragments; Ccofl = cinder cone, outer
flank, more sparsely covered by different volcaniclastic fragments; Lafla = weathering basalt lava flows with intervening troughs mantled by
fine ash and cinder. Consult text for additional details.

Average size of clast fragments exposed at the revealed no statistical difference between the techni-
ground surface was also measured by photographic ques (Caine, 1969). The vertical structure of clastic
sieving at 10 sites with rosette populations. In layers was noted at several sites; the depth of a
addition, four plots (1f, 4f, 6f, 57f) contiguous with surficial openwork layer of lava cobbles and blocks
the silversword sites, devoid of vegetation (plots 1f, was measured with a ruler next to 100 plants
6f, 57f) or with scarce plant cover (4f), and with inspected at site 3, where the mantle of volcaniclastic
finer exposed soils were also studied. Plots 4f, 6f, fragments was particularly thick and continuous.
57f were examined by photosieving; as grains on Although the size of large clasts can be efficiently
plot 1f were exceedingly small, they were sampled estimated by photosieving, accurate assessment of
on a 15  15 cm miniplot and then analyzed by smaller particles requires direct sampling and separa-
sifting (see below). A 100  50 cm graduated wood- tion by mechanical sifting (Pérez, 1989, 2000). Rep-
en frame, divided into two square segments each resentative samples of fine volcaniclastic sediment
0.25 m2 in area, was laid on the ground along 12-m- (gravel with some pebbles) were collected from the
long, continuous, linear transects parallel to slope sites by gathering all surficial fragments on a 15  15
contours and z 12 photos were sequentially taken cm miniplot (Pérez, 2000) randomly selected at each
for subsequent analysis (Fig. 3). To minimize undue site; samples were later mechanically sifted in the lab.
shading by larger clasts or nearby plants, photo- A soil profile was thoroughly examined down to 65
transects were taken at noontime and only during cm; to minimize site disturbance, a deeply trenched
sunny days with clear skies. Similar photographic section of the Sliding Sands trail was inspected after a
methods have been widely used in recent years to fresh soil face was exposed by scraping 6 –8 cm of
study particle size on different kinds of slopes, soil from the eroded, steep trailside. This location
including talus (Caine, 1969; Brückl et al., 1974; allowed extensive lateral tracing of soil horizons.
Carniel and Scheidegger, 1974; Pérez, 1989, 1993, Profile description follows the U.S. Comprehensive
1998a,b), periglacial rubble slopes (Church et al., System (Soil Survey Staff, 1975, 1994); however, a
1979; Iwata, 1983), and desert slopes (Abrahams et brief description in the European ORSTOM system
al., 1986). Comparison of results with those obtained (Segalen, 1984) is also included for broader interna-
by more traditional in situ clast measurement has tional interpretation of results.
180 F.L. Pérez / Geomorphology 55 (2003) 173–202

Fig. 3. Representative photographic plot (#12) along a photosieving transect on site 3, 2755 m. The graduated wooden frame (100 cm long, 50
cm wide) is divided into two equal segments of 0.25 m2. This plot is mostly covered by cobbles (particles 5 – 10 cm long). July 31, 2001 (Hk—
01.157d).

3.2. Laboratory analysis were assessed by digital CUHUN analyzer. Available

phosphorus was extracted with an HCl/NH4F solution
Clastic samples were air-dried for a week, oven- and determined in a photoelectric colorimeter with
dried for 48 h at 105 jC, and weighed. Particle-size ammonium molybdate (Olsen and Sommers, 1982).
distribution of all samples was analyzed in detail by Exchangeable bases (Ca, Mg, K) were obtained by
hand sifting through a 15-mesh sequence (50, 45, atomic-absorption spectrophotometry from ammo-
37.5, 31.5, 25, 22.4, 19, 16, 12.5, 9.5, 8, 6.3, 4.75, nium acetate extracts at pH 7.0; cation-exchange
3.35, 2 mm); finer soil grains contained within clastic capacity (CEC) was calculated after converting ppm
samples were separated with a 9-mesh series (1.4, 1, cation contents to cmol+ kg 1 (Liegel et al., 1980).
0.7, 0.5, 0.355, 0.25, 0.125, 0.09, 0.063, 0.038 mm).
Color for fine volcaniclastic materials and soil-profile 3.3. Sedimentological techniques and statistical treat-
specimens was determined in dry samples with the ment of data
Munsell Soil Color Charts (1992). Particle density
was estimated for 10 small pyroclastic particles by Images from the photographic transects were pro-
displacement of water (Blake and Hartge, 1982). Soil jected on a screen with a printed grid (5-cm grid
bulk density was calculated by filling a 35 cm3 plastic spacing; cf. Pérez, 1993) of numbered dots. On each
cylinder with oven-dried soil and weighing three photo, 10 dots were arbitrarily chosen using random
replicates per sample (Pérez, 2000). Soil structure number tables to select their X/Y coordinates. The
and consistence (dry and moist levels) were assessed longest (a) axis of the particles beneath the dots was
for profile samples following Soil Survey Staff (1994) meticulously measured without replacement (i.e.,
methods. Soil organic matter was evaluated by loss on clasts were not remeasured once sampled). This
ignition at 375 jC (Ball, 1964), as this temperature method is more accurate than sampling with replace-
does not affect fine grain ( V 0.063 mm) content. Soil ment (Dunkerley, 1996), as it partially offsets the bias
pH was measured with a glass electrode potentiometer toward large stones that have a disproportionately
in a 2:1 paste (water/soil) using three replicates per high chance of lying under a grid (Leopold, 1970).
sample. Organic carbon and total nitrogen percentages If a stone was overlapped by others or a shadow
 F.L. Pérez / Geomorphology 55 (2003) 173–202 181

obscured it, a new grid coordinate was chosen. A total distributions were plotted on log paper and numerical
of 120 particles (10 stones  12 photos) was sampled indices derived in the phi (/) scale (Folk, 1980); /
for each site, for a grand total of 1200 data points. units were converted to mm and are reported. Two
Particles were grouped in four classes: blocks (B), measures of central tendency were computed: the
with a mean size (a axis)>10 cm; cobbles (C) = 5 – 10 median particle size (D50 or /50) and the graphic
cm; pebbles (P) = 2.5 – 5 cm; gravel (G) = V 2.5 cm mean particle size (Mz). The first index is the 50th
(Iwata, 1983; Pérez, 1998a). Resolution sets practical percentile diameter, which identifies the most abun-
limits to the minimum clast size than can be accu- dant particle size in a sample (Inman, 1952); because
rately measured by photosieving. Some authors (Dun- it has been widely used in other sedimentological
kerley, 1996, p. 574) have indicated that grains as studies (Corey and Kemper, 1968; Pérez, 1998a,
small as 2 mm can be counted; I set the minimum 2000), this index can be useful for comparison pur-
particle size that could be reasonably identified on poses. Mz is calculated as Mz=(/16 + /50 +/84)/3,
photos to f 5– 10 mm. However, direct examination where /16 is the / size at percentile 16 (the point in
of the substrate along plant transects permitted a more the curve where 16% of the particles are smaller) and
precise division of the finest texture (i.e., surface type) /84 is the point at which 84% of the grains are smaller
type ( V 25 mm) into gravel—areas with a thin veneer (Folk, 1966); this index is more adequate for bimodal
of coarse granular material—and finer sandy soil (S), and/or sharply skewed sediments. Grain sorting was
where such granules were absent and maximum grain determined with /r, a dispersion measure computed
size was f 2 mm (Pérez, 1998a). as /r=(/84 /16)/2 and reported in / units (Inman,
Photosieving data were compared among sites with 1952; Pérez, 2000).
the percentage of the ground surface covered by
different particle classes and also with the average
stone size per site, calculated as the geometric mean 4. Results
size; this was computed after averaging the log-trans-
formed values from the mean antilogarithm. Datasets 4.1. General characteristics of study sites at Halea-
were tested for normality based on their kurtosis and kala
skewness (Jones, 1969). All particle-size data popu-
lations showed extreme negative skewness; seven The numbering of sites investigated in the crater
datasets were platykurtic, and the three others (sites follows Kobayashi (1973); study sites were located
2, 3, 57) were markedly leptokurtic (i.e., had many between 2175 and 2755 m altitude. Nine of the sites
particles both near the mean and the tails) (Sokal and included the upper elevation distribution range of the
Rohlf, 1969). All datasets were subjected to a log10 silversword and were located along or close to the Ke-
transformation; about half of all value-frequency dis- one-he’ehe’e (Sliding Sands) trail. One additional site
tributions tested as normal ( p < 0.05) (Jones, 1969), was studied at a lower elevation at the Silversword
but because of their inordinate skewness, the others Loop, near the Ko’olau Gap (Fig. 1c). Study sites
remained non-normal. Thus, textural differences were selected on Macdonald’s (1978) map to examine
among datasets were compared with the nonparamet- diverse geologic substrates and a variety of landscape
ric Kolmogorov – Smirnov (KS) test, where the con- positions (Table 1). Sites 1 and 5 (hb – ta) were
dition of data normality is not critical (Miller and mantled with ash, cinder, and blocks. Site 5 was on
Kahn, 1962). Chi-square (v2) tests were used to a nearly level area; site 1 was on a steep talus, below
compare clast data obtained by photosieving to similar dark gray alkali-olivine basalt lava outcrops of pre-
clast data acquired along plant surveys. dominantly aa but with some pahoehoe material. Site
After sieving, fine volcaniclastic samples were 2 (hco-ta) was on a gentle slope blanketed by cinder
grouped into three size fractions, following U.S. and spatter, mainly reddish- to yellowish-brown and
ASTM standards (Gardiner and Dackombe, 1983; locally red, below weathering aa lava outcrops. Sites 3
Poesen and Lavee, 1994b; Pérez, 2000): fine gravel and 4 (hc-ta) occupied steep talus slopes mantled with
( < 4.75 mm), medium gravel (4.75 – 19 mm), and some cinder and spatter and with many blocks, pre-
coarse gravel (>19 mm). Cumulative particle-size dominantly black to gray but reddened locally, below
182 F.L. Pérez / Geomorphology 55 (2003) 173–202

undifferentiated lava outcrops exposed along crater late eruptive stages and coalesces into ruff-like agglu-
walls (Fig. 4). Site 6 (ha-hb) was located on the gently tinate layers around crater rims (Fig. 6). As aggluti-
sloping edge of the crater floor, covered by largely nate deposits gradually weather, they provide clasts
windblown, but also water-washed, primarily gray ash that roll and slide downhill to accumulate on cinder
and by scattered reddish-gray volcanic blocks. Site 7 cone slopes.
(hb) consisted of a gentle slope with a continuous
layer of reddish-brown to dark gray alkali-olivine 4.2. Clast size in silversword areas, as determined
basalt blocks, largely derived from weathering out- along photosieving transects
crops of lava, mostly aa but with some pahoehoe. Site
57 (hm) had a gently undulating topography and The combined percentage of the ground surface
numerous porphyritic alkali-olivine basalt aa lava occupied by all volcaniclastic fragments was high on
flows from Pu’u o Maui, with dark reddish-brown the 10 sites. The modal (i.e., most common) clastic
to dark gray clastic fragments and rocky crags; the texture in 7 of the 10 sites consisted of cobbles (Table
troughs between weathering lava ridges were heavily 2). Overall, cobbles comprised 36.3% of the ground
mantled with fine ash and cinder, also from Pu’u o surface at all sites, followed by pebbles (26.5%) and
Maui’s eruptions (Fig. 5). Sites 10 and 11 (hc) were blocks (19.5%); finer material (gravel and/or sand)
located on the Pu’u-o-Pele cinder cone: 10 was on the made up only 17.7% of the surface along photo-
inner crater slope; 11 on the outer, N-facing flank. graphic transects (Fig. 7a). The overall shape of the
Both were covered with mainly reddish-brown to dark textural curves shows a broad granulometric variation
gray cinder and spatter volcanic bombs—including among sites (Fig. 8a). The coarsest volcaniclastic
some with ropy ‘‘cow-dung’’ appearance (Macdonald, covers (e.g., sites 3, 4, 7) exhibit steep particle-size
1967)—and abundant fragments of welded-spatter distribution curves; clasts here are mainly within the
agglutinate. This welded spatter, called Pohaku-o- cobble and block classes and show a narrow size
Hanalei, or ‘‘wreath of stones’’ (Wentworth and Mac- variation (Table 2). Most remaining samples display
donald, 1953; Macdonald, 1967), is ejected during more gradual curves, have a comparatively greater

Fig. 4. Highest-altitude silversword population investigated (site 3), 2755 m altitude. This site occupies a steep slope on the south wall of
Haleakala’s crater, completely covered by clastic fragments (mostly large cobbles and blocks) derived from weathering of extensive block- and
aa-lava outcrops exposed along the crater walls. Note two outcrop segments, protruding above the mantle of weathered rubble, on the upper left
and upper right of the photograph. July 25, 2001 (Hk—01.30d).
 F.L. Pérez / Geomorphology 55 (2003) 173–202 183

Fig. 5. View, looking south, of site 57, near Silversword Loop trail; 2180 m altitude. Note how silversword rosettes and shrubs (mainly kupaoa,
D. menziesii) are confined to elongated rocky ridges (weathered aa lava flows) crossing the photograph from left to right, but are largely absent
from the intervening troughs, heavily covered with fine ash and cinder. August 5, 2001 (Hk—01.241d).

Fig. 6. Pohaku-o-Hanalei (‘‘wreath of stones’’) deposit on the northern flank rim of Pu’u-o-Pele cinder cone; 2440 m altitude. The f 2.8-m-tall
rockwall was produced by accumulation of welded-spatter agglutinate, presumably produced during late eruptive stages. Site 11 lies directly
downslope from this agglutinate rampart (on the left side of the photograph); site 10 is also located below this area (on the inner crater area, right
side of the photograph). July 25, 2001 (Hk—01.83d).
184 F.L. Pérez / Geomorphology 55 (2003) 173–202

Table 2
Percentage of textural surface types on 10 study sites with large populations of Hawaiian silversword, Haleakala’s cratera
Texture types (%) Sampling sitesb Grand
1 2 3 4 5 6 7 10 11 57 averagec

Gravel 11.7 37.5 0 4.2 6.7 28.3 5.8 23.3 42.5 16.7 17.7
Pebbles 26.7 28.4 5.8 23.3 35.8 30.0 30.0 20.8 21.7 42.5 26.5
Cobbles 36.6 18.3 56.7 38.3 45.0 31.7 47.5 37.5 23.3 28.3 36.3
Blocks 25.0 15.8 37.5 34.2 12.5 10.0 16.7 18.4 12.5 12.5 19.5
Geometric mean size (cm)d 5.5 3.4 8.8 7.4 5.4 3.6 5.6 4.4 2.6 4.3 5.1
Median texturee C P C C C P C C P P C
a
Data obtained from photosieving transects.
b
Site numbers correspond to those on Fig. 1.
c
Grand average of texture types shows mean arithmetic values for the 10 sites sampled.
d
Geometric mean size is the anti-logarithm average clast size after a log10 transformation of all datasets (a axis length).
e
Key for median texture types: P = pebbles, C = cobbles. See text for additional sedimentological data.

fraction of sand + gravel and an intermediate clast in Haleakala are largely covered by coarse volcani-
size. A third, distinctive type of sediment distribution clastic debris, and (ii) the coarsest volcaniclastic
is found at site 11, which has the lowest percentage of layers are better sorted than those with finer sediment,
cobbles + stones (just 35.8%) and the smallest average reflecting the wider range of particle sizes—both
size (2.6 cm). Its curve shows a strongly bimodal coarse and fine—in the latter.
distribution of both fine and coarse grains. These data The analysis above deals with clasts on sites where
indicate that (i) the areas associated with silverswords silverswords grow; the following data describe surface

Fig. 7. Pie charts showing (a) percentage of the ground surface occupied by different types of volcaniclastic particles on 10 study sites covered
by coarse fragments in Haleakala’s crater. Particle size (a axis) was determined from photographic transects; 120 clasts were measured at each
site. (b) Percentage distribution of the largest volcaniclastic particles and organic litter next to 1000 silversword rosettes on 10 study sites in
Haleakala’s crater. Particle data from plant surveys; 100 silverswords and adjacent particles were examined at each site. Key: G = gravel,
P = pebbles, C = cobbles, B = blocks, O = outcrops, L = litter (dead silverswords). Letters inside boxes show probability levels for statistical tests
(Kolmogorov – Smirnov) comparing percentages of particles on photographic transects with those obtained from plant surveys at different sites.
Only circled particle sizes show significant differences. Key (site number in parenthesis): a = p < 0.001; b = p < 0.005; c = p < 0.01; d = p < 0.025;
e = p < 0.05. Consult text for further statistical information.
 F.L. Pérez / Geomorphology 55 (2003) 173–202 185

Fig. 8. (a) Particle-size distribution (cm) for surficial volcaniclastic sediments on 10 study sites with abundant coarse fragments in Haleakala’s
crater; numerous silversword rosettes were present at all these sites. (b) Particle-size distribution (cm) on four study plots—adjacent to study
sites above—covered by primarily fine debris. Silverswords were absent from plots 1f, 57f, and 6f, and scarce on plot 4f. Particle size on all
areas, except for 1f, was measured from photographic transects; 120 clasts were measured at each location. Grain size for plot 1f was determined
by mechanical sifting in the laboratory. Numbers designate individual sites investigated (Fig. 1).

sediment on comparable plots adjacent to four of the present; blocks were absent from the three unvege-
study sites not occupied by plants (Fig. 8b). Mean tated areas, and even on plot 4f—the only one with
size was 1.3 cm. Modal texture on all plots was some silverswords—made up just 1.7% of its surface
gravel, which covered an average of 82% of the (Table 3). Fig. 8a and b show little, if any, overlap
ground surface. A few pebbles and cobbles were between the two sets of granulometric curves, except
186 F.L. Pérez / Geomorphology 55 (2003) 173–202

Table 3 4.3. Texture types and clast size near silversword

Percentage of textural surface types on four study plots devoid of rosettes
silverswords (1f, 6f, 57f) or with very few rosettes present (plot 4f)
Texture types (%) Study plotsa Grand
The previous section indicates that locations with
1f 57f 6f 4f averageb
widespread stone mantles are conducive to rosette
Gravel 95.8 90.8 80.0 60.8 81.9 growth. However, as phototransects only examined
Pebbles 4.2 7.5 16.7 28.3 14.2 the general areas within plant populations, a more
Cobbles 0 1.7 3.3 9.2 3.5
specific study was needed to determine the actual size
Blocks 0 0 0 1.7 0.4
Geometric mean size (cm)c 0.6 1.1 1.3 2.2 1.3 of stones immediately adjacent to the (base of) silver-
Median textured G G G G G swords. This could show (i) whether plants are truly
a
Data for all plots except 1f obtained from photosieving rooted next to stones and (ii) if they show any inherent
transects. Data for 1f were obtained by direct sediment sifting and preference for (a) particular substrate(s). Clast data
represent percentage by (dry) weight of the sediment sample. Plot collected along the WQ plant surveys (‘‘observed’’
numbers correspond to those on Fig. 1. distribution) were compared to data obtained by pho-
b
Grand average of texture types shows mean arithmetic values
for the four plots sampled.
tosieving (‘‘expected’’ distribution) with v2 tests. The
c
Geometric mean size is the anti-logarithm average clast size high v2 value (45.4) obtained after contrasting all data
after a log10 transformation of all datasets (a axis length); plots points on the 10 sites indicated significant differences
arranged by increasing geometric mean size. ( p < 0.001) between datasets. Independent v2 tests
d
Median texture: G = gravel. See text for additional sedimento- revealed comparable statistical differences ( p < 0.001
logical data.
or < 0.005) at the site level in eight sampling loca-
tions; while data from two sites showed no significant
variance (Table 4).
for site 11, with sharply bimodal surface sediments. Although useful in detecting overall differences, the
Instead, a gradual shift in decreasing particle size is v2 tests cannot determine which texture type or size
apparent from the locations with abundant silver- class(es) actually differ(s). Thus, the percentages of the
swords (Fig. 8a), to plot 4f, which had some rosettes, ground surface occupied by each texture were com-
and then to the three remaining plots devoid of plant pared, both at the individual site and all-sites level,
cover (Fig. 8b). with KS tests. The mean fraction of the ground surface

Table 4
Percentage of surface types and largest clast type found next to Hawaiian silversword plants on 10 study sites in Haleakala’s crater
Texture types (%) Sampling sitesa Grand
1 2 3 4 5 6 7 10 11 57 averageb

Sand 0 0 1 1 2 4 0 0 0 3 1.1
Gravel 6 22 0 8 2 19 7 1 22 2 8.9
Pebbles 24 56 3 28 35 24 33 41 28 27 29.9
Cobbles 19 16 52 38 39 21 59 44 31 30 34.9
Blocks 44 6 43 23 20 19 1 14 15 32 21.7
Outcrops 5 0 0 2 0 5 0 0 1 2 1.5
Organic litter 2 0 1 0 2 8 0 0 3 4 2.0
Median texturec C P C C C P C C P C C
Geometric mean (cm)d 9.5 3.5 9.7 6.7 5.9 4.6 5.4 5.9 4.8 7.0 6.3
v2 valuese 34.6 39.8 2.6 22.6 6.4 25.4 18.1 43.1 15.2 50.9 45.4
p < levele 0.001 0.001 n.d. 0.001 n.d. 0.001 0.001 0.001 0.005 0.001 0.001
a
Data obtained from plant surveys (WQ transects); site numbers correspond to those on Fig. 1.
b
Grand average of texture types shows mean arithmetic values for the 10 sites sampled.
c
Key for median texture types: P = pebbles, C = cobbles.
d
Geometric mean size is the anti-logarithm average clast size after a log10 transformation of all datasets (a axis length).
e 2
v values and probability ( p) levels refer to a series of statistical tests comparing clast data acquired during plant surveys with the clast
data obtained by photosieving (cf. Table 2); n.d. = no statistical difference.
 F.L. Pérez / Geomorphology 55 (2003) 173–202 187

Fig. 9. Pie charts showing the percentage distribution of the largest volcaniclastic particles and organic litter found upslope and downslope from
600 silversword rosettes on six study sites (1, 2, 4, 6, 10, 11) in Haleakala’s crater. Particle data from plant surveys; 100 silverswords and
adjacent particles were examined at each site. (a) Substrates upslope from rosettes. (b) Substrates downslope from rosettes. Key: S = sand,
G = gravel, P = pebbles, C = cobbles, B = blocks, O = outcrops, L = litter (dead silverswords). Letters inside boxes show probability levels for
statistical tests (Kolmogorov – Smirnov) comparing the percentages of particles upslope and downslope from rosettes. Only circled particle sizes
show significant differences. Key (site number in parenthesis): a = p < 0.001; b = p < 0.005; c = p < 0.025. Consult text for further statistical
information.

Fig. 10. Small silversword rosette located directly below an isolated lava block ( f 29 cm wide) on a steep slope ( f 27j) largely covered by a
thin layer of fine black-gravel fragments. Downslope is toward the bottom of the photograph. Site 1, 2675 m altitude. July 26, 2001 (Hk—
01.98d).
188 F.L. Pérez / Geomorphology 55 (2003) 173–202

covered by sand + gravel in all 10 sites was lower plants grew directly on the organic litter left behind by
(10.0%) next to plants than in their general vicinity dead silverswords (Pérez, 2001) (Table 4).
(17.7%) (Tables 2 and 4). This difference is statisti-
cally significant ( p < 0.005) and suggests that rosettes 4.4. Clast size upslope and downslope from silver-
are primarily excluded from fine-debris substrates and/ sword rosettes
or that their survival rate—assuming a homogeneous
seed dispersal—may be lower in these areas. The Ground texture and maximum clast size above and
percentages of the substrate occupied by other ground below rosettes were compared with a series of KS tests
textures next to plants were similar to—if slightly for all sites. No appreciable variance between these two
higher than—those along phototransects, and no sig- slope positions turned up at sites 3, 5, 7, or 57 but was
nificant variance was detected between grand averages found at the other six sites. KS tests comparing all 600
(Tables 2 and 4); however, statistical differences were data points (6 sites  100 plants) showed that the
found on several individual sites. KS tests confirmed percentages of all substrate categories, except outcrops
that sand + gravel areas were specifically underrepre- and organic litter—which had very few data points—
sented near rosettes at sites 4, 10, and 11. In addition, were significantly different uphill and downhill from
pebbles were overrepresented (i.e., more common) silverswords (Fig. 9). The widest differences were
next to plants at sites 2 and 10, while blocks were found within finer size classes. The slope below 22%
more frequently found adjacent to rosette bases at sites and 44.5% of plants was occupied by sand and gravel,
1, 11, and 57 (Fig. 7b). Sites 3, 5, 6, and 7 showed no respectively, but only 2.2% and 19.2% of the rosettes
statistical differences between datasets. Besides these were associated with such textures on their upslope side
volcaniclastic textures above, a few plants along WQ (both differences significant at p < 0.001); clearly, the
surveys occupied two marginal substrates: 1.5% of the ground surface was, on average, finer downhill from
rosettes were rooted at the base of (and usually down- plants. In contrast—as expected—the trends in larger
slope from) massive lava outcrops, while 2.0% of the size classes were the opposite: 34.4%, 24.3%, and

Table 5
Textural parameters and modal color of fine volcaniclastic sediment (mainly gravel with some pebbles) collected on 10 study sites with large
populations of Hawaiian silversword, Haleakala’s crater
Sediment Sampling sites
characteristics 1 2 3 4 5 6 7 10 11 57
Clast colora 10YR 4/2 5YR 4/3 10YR 3/1 10YR 3/1 10YR 3/2 5YR 4/2 5YR 4/4 5YR 3/4 5YR 4/3 5YR 3/2
(dark (reddish (very dark (very dark (very dark (dark (reddish (dark (reddish (dark
grayish brown) gray) gray) grayish reddish brown) reddish brown) reddish
brown) brown) gray) brown) brown)
Fine gravel 28.2 29.2 0.2 4.0 2.6 17.5 0.5 12.8 42.0 3.4
( < 4.75 mm) (%)
Medium gravel 40.8 56.3 57.3 65.3 22.0 49.6 42.3 27.4 5.5 67.8
(4.75 – 19 mm) (%)
Coarse gravel 31.0 14.5 42.5 30.7 75.4 32.9 57.2 59.8 52.5 28.8
(>19 mm) (%)
Median sizeb 9.7 7.1 17.7 13.6 24.9 14.3 21.6 22.0 20.4 14.7
(D50), mm
Graphic meanc 14.6 9.7 18.1 15.9 25.3 14.0 21.6 20.8 19.9 15.3
(Mz) (mm)
r Sorting, / units 1.49 (2.8) 1.27 (2.4) 0.56 (1.5) 0.93 (1.9) 0.65 (1.6) 1.29 (2.4) 0.71 (1.6) 1.22 (2.3) 1.88 (3.7) 0.80 (1.7)
(and mm)d
a
Colors indicated are for dry samples (Munsell Soil Color Charts, 1992).
b
D50 is the grain diameter of the 50th percentile.
c
Mz is the graphic mean (Folk, 1966).
d
r sorting is a standard deviation measure of size (Inman, 1952) given in / units and in millimeters. Consult text for additional
sedimentological information.
 F.L. Pérez / Geomorphology 55 (2003) 173–202 189

Fig. 11. Particle-size distribution (mm) for fine surficial volcaniclastic sediments—mainly gravel and pebbles—on 10 study sites in Haleakala’s
crater; silversword rosettes were abundant at these sites. Sediment samples were gathered from 10 15  15 cm sampling miniplots; particle size
was determined by mechanical sifting in the laboratory. Numbers designate individual sites investigated (Fig. 1).

Fig. 12. Percentage frequency-distribution histogram for 100 measurements of the depth of a volcaniclastic layer (cm) found on the soil surface
at site 3, 2755 m. Class interval is 2 cm. Mean depth: 98.6 F 43.7 mm.
190 F.L. Pérez / Geomorphology 55 (2003) 173–202

16.1% of the silverswords grew beneath pebbles, downward from large stones, but areas below them
cobbles, or blocks, respectively; but only 14.3%, consisted mainly of bare sand or fine gravel (Fig. 10).
11.4%, and 5.4% of the rosettes were rooted uphill KS tests for the six individual sites identified relevant
from those stone classes (differences for P and C are local differences: pebbles, cobbles, or blocks were
significant at p < 0.001 and for B at p < 0.005). In other overrepresented above plants at sites 1, 2, 6, and 10
words, most silverswords had become established (Fig. 9a). By the same token, exposed sand or gravel

Table 6
Representative soil profile from clast areas with dense populations of Hawaiian silversword, Haleakala’s cratera
Horizon Depth Description
(cm)
Andic Humitropept
C + 5–0 Openwork layer of slightly weathered gravel and pebbles: small (D50: 20.7 mm), loose, irregular, oblate to
subspherical, angular to subangular, volcaniclastic scoriaceous lapilli, with many small (2 – 3 mm) regular
spherical to medium-size (6 – 8 mm) irregular vesicles. Lapilli fragments were light (density: 1.97 – 2.06 mg m 3).
Layer had 95.5% gravel (D50 of gravel fraction: 24.2 mm); soil fraction: 79.6% sand, 20.4% fines. Color: 5YR
4/4 (reddish brown). This layer had a pronounced sieve effect. Smaller grains (fine and medium gravel) were
concentrated at the layer bottom within voids between larger particles (coarse gravel and pebbles); voids remain
empty at layer top. Abrupt smooth boundary to:
A11 0 – 15 Sand with 9.3% gravel; soil fraction: 91.8% sand, 8.2% fines. Single grain structureless. Consistence: loose (0) dry
and moist; very slightly sticky (0.5) and nonplastic (0) wet. Bulk density: 1.47 mg m 3. Color: 10YR 3/3.5 (dark
yellowish brown). Organic matter: 2.6%; pH: 7.25; C/N ratio: 9.8; (%) C: 0.18; total N: 184 ppm. Available P:
43 ppm; Ca: 235 ppm; Mg: 25 ppm; K: 38 ppm. CEC: 1.5 cmol+ kg 1. Clear smooth boundary to:
A12 15 – 18 Sand with 7.9% gravel; soil fraction: 92.1% sand, 7.9% fines. Single grain structureless. Consistence: loose (0)
dry; very friable (1) moist; slightly sticky (1) and nonplastic (0) wet. Bulk density: 1.40 mg m 3. Color: 10YR
3.5/3 (brown). Organic matter: 2.9%; pH: 7.3; C/N ratio: 8.5; (%) C: 0.095; total N: 111 ppm. Available P:
92 ppm; Ca: 594 ppm; Mg: 63 ppm; K: 56 ppm. CEC: 3.6 cmol+ kg 1. Abrupt smooth boundary to:
A13 m 18 – 18.5 (See description for Aj13 m, a similar horizon, below). Abrupt smooth boundary to:
AV12 18.5 – 21 (See description for A12, a similar horizon, above). Abrupt smooth boundary to:
AV13 m 21 – 21.5 (See description for Aj13 m, a similar horizon, below). Abrupt smooth boundary to:
A14 21.5 – 23.5 Sand with 3.7% gravel; soil fraction: 89.5% sand, 10.5% fines. Slightly coherent to massive structureless.
Consistence: soft (1) dry; very friable to friable (1.5) moist; sticky (2) and slightly plastic (1) wet. Bulk density:
1.37 mg m 3. Color: 10YR 3/4 (dark yellowish brown). Organic matter: 3.2%; pH: 7.25. Abrupt smooth
boundary to:
AW13 m 23.5 – 24 (See description for Aj13 m, a similar horizon, below). Abrupt smooth boundary to:
AV14 24 – 27 (See description for A14, a similar horizon, above). Abrupt smooth boundary to:
Aj13 m 27 – 28 Sandy loam, with 3.3% fine gravel: fine volcanic ash, consolidated, weakly cemented, but not welded, into tuff.
Soil fraction: 72.4% sand, 27.6% fines (25.5% silt, 2.1% clay). Massive structureless, with tendency to break into
small (8 – 15 mm) angular blocky peds. Consistence: slightly hard to hard (2.5 – 3) dry; firm to very firm (3.5)
moist; sticky to very sticky (2.5) and plastic (2) wet. Bulk density (after crushing): 1.29 mg m 3. Color: 10YR
4/4 (dark yellowish brown). Organic matter: 4.9%; pH: 7.25. Abrupt smooth boundary to:
AW14 28 – 34 (See description for A14, a similar horizon, above). Clear, smooth to wavy boundary to:
2Cb 34 – 65 + Fine gravelly sand (D50: 2.84 mm) with 65.3% gravel (D50 of gravel fraction: 3.75 mm): unweathered,
little-altered, very small, fragile, elongated to subspherical, vitric volcanic ash. Soil fraction: 97.8% sand, 2.2%
fines. Single grain, structureless. Consistence: very loose (0) dry (particles separate and samples collapse easily);
loose (0) moist; nonsticky (0) and nonplastic (0) wet. Bulk density (soil fraction): 0.98 mg m 3 (whole sample):
0.81 mg m 3. Color: 10YR 2/1 (black). Organic matter: 0.4%; pH: 7.1. Horizon or profile bottom were not
reached during excavation.
a
Profile examined next to Sliding Sands trail, between sites 6 and 7 (Kobayashi, 1973) at a location devoid of plant cover. Altitude: 2505 m;
aspect: Nj19NE; slope angle: 7j. Soil fraction includes particles < 2 mm; fines include silt and clay ( < 0.063 mm). D50 is the grain diameter of
the 50th percentile. Colors are for dry soils (Munsell Soil Color Charts, 1992). CEC: cation-exchange capacity. Soil terminology follows Soil
Survey Staff (1975, 1994). See American Geological Institute (1976), Macdonald et al. (1983), and Pérez (2000) for terminology on
volcaniclastic ejecta and see Gardiner and Dackombe (1983) for particle description and shape. Consult Fig. 13 for further details on the horizon
sequence.
 F.L. Pérez / Geomorphology 55 (2003) 173–202 191

were underrepresented above silverswords on all sites among two hue groups: dark gray to grayish-brown
(Fig. 9b). colors, included in the Munsell 10YR sheet; and
Small piles of accumulated (fine) debris upslope reddish-brown to reddish-gray colors, included in the
from rosettes of all sizes were found at six sites. The 5YR hue sheet (Table 5). Lastly, the fine-debris sam-
convex piles bulged noticeably, forming miniature ples described here were representative not only of the
steps above the plants (Pérez, 1994; Pérez et al., smaller clasts exposed at the ground surface throughout
2002). The number of rosettes with perceptible debris the study sites, but also exemplify well the interstitial
piles varied among sites, from 12 to 19 plants (sites 6, debris infilling pore spaces between the larger stones of
10, 1, 11) to 78 (site 2). Step height ranged from 1.5 to volcaniclastic covers (see below) (Pérez, 2000).
31.5 cm; a brief analysis for sites 2 and 4, with the
greatest number of steps, follows: a total of 78 plants at 4.6. Vertical structure of volcaniclastic layers and soil
site 2, and 65 at site 4 showed appreciable steps above profile description
them. Average step height at gently sloping site 2 (12j)
was 9.9 F 4.2 cm; while the steeper site 4 (25j) was In addition to the size of surficially exposed volca-
13.8 F 7.5 cm. At both sites, the thickness of debris niclastic sediments, the vertical variation of these
piles was positively correlated with plant stature, but substrates was also examined by shallow excavation
this relationship was only significant—as determined at seven sites (1, 2, 3, 6, 7, 10, 11). All clastic layers
by polynomial regressions—at site 4 (r2 = 0.679, showed a well-developed stratification with depth;
p < 0.001). Most of the debris piles at site 2 consisted even in areas where the ground surface was fully
of pebbles ( f 59%) or sand/gravel ( f 35%), and the
remaining 6% were made up of small cobbles. Corre-
sponding percentages at site 4 were f 38% (P),f 17%
(S/G), f 38% (C), and only f 7% blocks.

4.5. Particle-size distribution of fine volcaniclastic

fragments

All 10 fine volcaniclastic samples were lapilli (6.4 –

38 mm; Macdonald et al., 1983). Based on median
(D50) and graphic mean (Mz) particle size, the sedi-
ments would be designated as gravel at all sampling
locations, except perhaps at site 5 where they contained
many small pebbles (Table 5). As in previous analyses
of Haleakala’s lapilli (Pérez, 2000), all the specimens
had little fine gravel, and most of their mass fell within
the medium- or coarse-gravel size classes. Thus, most
granulometric curves for fine debris were steep, with
the glaring exception of site 11, which had a sharply
platykurtic (bimodal) distribution (Fig. 11). Grain-size
dispersion (/r) fluctuated widely (Table 5). As in the
coarse (photosieving) clast data, fine debris with bigger
grains (i.e., sites 5, 7, 10; Mz = 22.6 mm) were much
better sorted (mean /r = 0.86) than those with the Fig. 13. Sketch of soil profile investigated at 2505 m, near sites 6
smallest grains (sites 1, 2, 6; Mz = 12.8 mm, mean and 7. Horizon designations are to the right of the profile, depths in
/r = 1.35). Thus, comparable trends in sorting are centimeters to the left. Horizon terminology follows Soil Survey
Staff (1975, 1994). Key: C = reddish volcaniclastic, scoriaceous
apparent in both the coarse and fine components of lapilli and blocks; A11, A12 = sandy soil; A13 = fine volcanic,
volcaniclastic layers at Haleakala (Figs. 8a and 11). The weakly cemented, ash tuff; A14 = sandy soil; 2Cb = black vitric
colors of lapilli samples were nearly equally divided volcanic ash. Consult Table 6 for additional horizon details.
192 F.L. Pérez / Geomorphology 55 (2003) 173–202

covered by blocks or cobbles, subsurface layers largely unconformably over sandy, very dark gray (10YR 3/1)
consisted of fine sandy soil with few stones included soil with f 7% fines. Above this horizon lay a sharply
within it (see profile below). In many places, stones separated, 2- to 3-cm-thick layer of black (10YR 2/1)
formed only a thin openwork, continuous, shingle-like gravel and pebbles (8– 30 mm) overlain by a f 6-cm-
pavement just one to two stones (3– 6 cm) thick over thick band of very dark gray to reddish-brown (5YR 4/
finer soil; this segregation was particularly striking on 4) pebbles and cobbles 2 – 8 cm (mean f 3-4 cm).
level surfaces around the base of Pu’u-o-Pele and other The thickness of volcaniclastic covers varied
cinder cones where extensive soil areas were covered broadly among sites, from barely 15 to 20 mm in
by such pavements. Typically, average clast size the finer gravels to >20 cm in areas with large blocks;
within the openwork layer decreased gradually with however, within-site variance was often pronounced.
depth. In most cases, the largest stones had a similar Clastic pavement depth next to rosettes at site 3
size at the bottom and top of the openwork layer, but fluctuated between 30 and 220 mm; but mean thick-
the spaces between them were infilled at depth by ness was 98.6 F 43.7 mm, and more than 75% of all
smaller particles; this is the so-called sieve effect measurements were < 140 mm (Fig. 12).
(Carniel and Scheidegger, 1974), to be discussed later. The soil profile examined near site 6 was complex
In other places, pavements exhibited an exceptionally and was classified (Soil Survey Staff, 1994) as an
well-developed separation of layers; an example from Inceptisol with some andic properties (andic humitro-
site 7 will serve to describe this stratification; colors pept), a mesic temperature regime, and an aridic to
are reported, as they readily allowed for separation of ustic moisture regime. The overall nature of the
layers in the field. A 9-cm-thick layer of clasts rested profile corresponds closely to that for the subgroup

Fig. 14. Cross-sections showing the root morphology of four silversword (dry) seedlings in relation to the underlying slope; all scales shown are
15 cm long. Approximate position and inclination of slope surface is indicated by a thick dashed line. Seedling (a) was examined at site 2 on
August 9, 1998; seedlings (b), (c), and (d) were inspected at site 11 on August 1, 2001. Diagram for plant (a) was directly traced in the field; the
other seedlings were photographed with a digital camera. Position in the field of a surficial volcaniclastic layer of pebbles and gravel is shown
for seedling (a).
 F.L. Pérez / Geomorphology 55 (2003) 173–202 193

in Soil Survey Staff (1975, p. 262) ‘‘soils. . . derived In the ORSTOM system (Segalen, 1984), the soil is
mainly from pyroclastic materials and lava (that) have an organic primarosol with a dark, pachic (>18 cm),
many but not all of the properties of Dystrandepts fine gravelly, arenic to slightly vitroarenic—but with-
(now within Andisols).’’ Table 6 provides a thorough out Andic properties—humon (a normal Sombron).
description of the profile, which had a 5-cm-thick This upper humon was eutric-neutral, calco-magnesic,
openwork C layer of lapilli fragments overlying a 34- and had a noncoherent particulate (psammoclodic)
cm-thick sandy A1 horizon resting over a thick, uni- structure. The main humon layer was underlain by a
form, buried C horizon of fine volcanic ash (Fig. 13). melanic, thick (pachic), lithic gravelly, neutral, miner-
The complex A1 horizon consisted of three layers alon (vitropsammic Andon) composed of very friable
(A11, A12, A13). The upper A11 was homogeneous, cinder aggregates with a mealy (fluffy) structure, very
but the A12 and A13 layers were interbedded with weak cohesion, and high porosity. The humon surface
several thin (5– 12 mm) bands of weakly cemented was, in turn, covered by a nearly continuous, leptic
silt-sized grains of fine volcanic ash, not welded but (thin), lithic gravelly, epimineralon (vitropsammic
consolidated into a tuff. Emplacement of these thin Andon) of red dusky pyroclastic openwork fragments.
continuous layers—traced laterally for several feet The soil did not qualify in the ORSTOM system as an
along the trenched trailside—may have been caused Andisol either, mainly because of its high bulk-density
by pyroclastic (ash) explosions, pellicular mudflow values (>0.9 mg m 3) (Segalen, 1984, p. 71).
events, or post-depositional runoff, including that
caused by snow melting. The intricate profile mor- 4.7. Root patterns of silversword seedlings
phology attests to the complexity of volcanic and
other geomorphic processes in Haleakala’s crater dur- Several dead silverswords, particularly small seed-
ing soil development. lings, were found on steep, fine-debris areas. These

Fig. 15. Adult silversword (f 54 cm tall) on a 28j slope; site 1, 2670 m altitude. This plant, which may be several decades old, has a basal
caulescent segment (dark gray area) sharply tilted downhill, but its live foliage (bright rosette) has become phototropically reoriented. July 26,
2001 (Hk—01.105d).
194 F.L. Pérez / Geomorphology 55 (2003) 173–202

plants were invariably desiccated and shriveled, and vertically reoriented by phototropism, and the rosettes’
either lay partially exposed on the ground surface or foliage showed a strong asymmetry when viewed from
had still-buried roots. Many such plants were ins- above.
pected; all were similarly deformed: their roots had
developed a prominent S-shaped bend or ‘‘elbow’’ in 4.8. Population structure and regeneration of the
their upper portion nearest to the slope surface (cf. Hawaiian silversword
Schumm, 1964; Kobayashi, 1973) and were irregu-
larly curved underground, trailing uphill from the Botanical data collected along the WQ transects
rosette’s base in all cases (Young and León, 1990) will be published elsewhere, but some relevant geo-
(Fig. 14). Excavation suggests that some root irregu- ecological findings are briefly summarized here.
larity was promoted by growth around clasts (Pérez, Apparently, a brisk silversword recovery has taken
1991), and upslope elongation was due to burial by place at Haleakala in areas covered by volcaniclastic
shifting debris (Kershaw and Gardner, 1986). Sixteen materials. Comparing my population estimates (Table
dead seedlings examined on steep sites had trailing 1) with Kobayashi’s (1973, p. 83)—taken during
segments that extended 20 – 130 mm (mean: 1969 – 1972—some rosette groups have apparently
72.0 F 34.9 mm) uphill from the plant’s base—mostly expanded dramatically over the past 40 years, expe-
along the bottom of a thin clast cover—before riencing up to a fivefold increase in plant numbers.
descending into finer soil (Fig. 14a,b). In addition to This is most conspicuous in smaller plant populations
the root patterns, many tall silverswords on steep (1, 3, 5, 7) that only had 26– 43 rosettes each in the
gradients were noticeably tilted downhill at their base early 1970s, but where I counted at least 100 indi-
(Fig. 15); in these plants, leaf crowns had become viduals in 2001 and for which my estimates are most

Fig. 16. Extensive volcaniclastic pavement (5 – 8 cm thick) of reddish aa lava cobble- and pebble-sized fragments with high silversword
regeneration; some 21 seedlings and young rosettes ( f 3 – 25 cm tall) appear on this 2.7-m-wide view. Site 7, 2510 m altitude. July 25, 2001
(Hk—01.48d).
 F.L. Pérez / Geomorphology 55 (2003) 173–202 195

reliable. The clearest indicator of successful silver- p < 0.001) from an expected uniform seedling distri-
sword regeneration is shown by the exceedingly high bution and, therefore, clast color may have influenced
numbers of small seedlings growing at these sites rosette propagation.
(Fig. 16). The remaining larger plant populations have
also increased in size but probably only within the
50 –75% range, which still is impressive and shows a 5. Discussion and interpretation of results
promising ecological trend of recovery for the Hawai-
ian silversword in Haleakala. The different analyses of ground texture and par-
The population structure of all sites combined ticle size amply confirm the initial impressions
shows a generally healthy rosette regeneration. The obtained in the field: silverswords are, indeed, closely
size-class distribution (Fig. 17) evinced a smooth associated with stones and clastic pavements but are
pattern of gradually decreasing silversword numbers scarce—at best—in fine-debris areas. The general
with increasing plant stature (a general proxy for age). granulometric correspondence between datasets (Figs.
This robust structure suggests that silversword repro- 7, 8 and 11) generated by three different methods
duction during the past three or four decades has been (photosieving, direct clast measurement near plants,
exceptionally successful: 32.5% of all rosettes sur- and mechanical sifting) bolsters the confidence in the
veyed were seedlings (plants V 75 mm tall), 57.0% observed spatial correlations between volcaniclastic
were V 15 cm in height, and 74.2% did not exceed debris and silverswords at Haleakala.
22.5 cm in height; maximum plant height was 76 cm. Only the two finest textures (sand and gravel) were
Overall, plant regeneration seemed better on the six underrepresented next to plants, but all coarser clasts
sites with reddish clasts (Table 5), where an average (P, C, and B) were common near them (Fig. 7). The
36.3% of the individuals were seedlings; but only percentages of clasts obtained from photosieving data
26.8% of the rosettes on four sites covered by gray- and WQ transects were comparable: the fraction of the
ish-brown clasts were comparable juveniles. A v2 test ground surface occupied by either P, C, or B’s varied
showed these data varied significantly (v2 = 83.2, just 4 – 11% between datasets, but the total area

Fig. 17. Population structure (size – class distribution) of 1000 Hawaiian silverswords censused on 10 study sites within Haleakala’s crater; 100
plants were examined at each site. Plant-size classes based on plant height—excluding inflorescence, if any present. Class interval is 7.5 cm.
196 F.L. Pérez / Geomorphology 55 (2003) 173–202

represented by sand + gravel at the study sites was vertical structure within clastic layers indicate abun-
77% higher than near plants (Tables 2 and 4). The dant water and nutrients may be stored in the finer soil
above trends strongly suggest that silverswords are beneath stones and utilized by the rosettes growing on
not dependent on any particular clast type, as long as clasts.
the sandy-gravelly soil is well insulated by stones. Stones modify soil moisture in other ways. Hen-
Dunkerley (1996) asserted that photosieving may be drix (1981) found plant colonization in volcanic
inaccurate unless data are corrected to account for bias substrates of the Galápagos Islands occurred at the
toward large stones (Leopold, 1970), but his criticism base of blocks, where fog interception on the cool
seems valid only for fine sediments with a much rock face added some crucial fog-drip water to the soil
smaller size, such as those he tested (mean axis: below. I have often seen this in Haleakala, where soil
1.4– 1.6 cm). In contrast, the particles I sampled along surfaces may be moist—down to f 2 – 4 cm depth—
phototransects had a grand average = 5.1 cm, and site just below the N-facing side of blocks, outcrops, and
means ranged as high as 8.8 cm (Table 2). In fact, that even tall silverswords, as these receive the fog drip
is why I felt it incumbent to sample finer sediments intercepted from moist air masses that periodically
directly, to avoid some of the limitations of photo- invade the crater through the Ko’olau Gap (Fig. 2).
sieving. In addition, surface-based methods like pho- Kobayashi (1973) also noted the occurrence of silver-
tosieving do not allow true volumetric sampling of swords on slopes intercepting moist NE trade winds.
sediment when only a stone veneer lies over finer- Rapidly falling temperatures at sunset may also result
textured material (Dunkerley, 1996). Since my basic in condensation in the interstices between stones in
hypothesis is that it is precisely the surficial exposed clastic layers (Evenari et al., 1975), a process exceed-
volcaniclastic debris that are significant for the estab- ingly common in high tropical mountains (Coe, 1969;
lishment of the silversword, a surface-based method Pérez, 1991).
such as photosieving is most adequate for this study’s A cover of stones will also affect substrate temper-
purposes. atures (Othieno and Ahn, 1980; Pérez, 1991, 1998a:
I will briefly discuss the geomorphic and ecolog- Pérez et al., 2002; del Moral and Bliss, 1993). In
ical factors that could be altered by the presence of a Haleakala (F.L. Pérez, unpublished data) soils below
volcaniclastic layer and review the possible effects stones remain considerably cooler during the day than
that might be operative at Haleakala. By far the most adjacent bare ground, which may reach noon maxima
oft-cited influence of a cover of volcanic scoria is over 45 jC. This cooling effect stems from the
increased soil moisture availability caused by a drastic efficient insulation of the soil surface from solar
reduction in evaporation from the soil below stones radiation and results in further reduction of water
(Kobayashi, 1973, p. 48; Othieno, 1980; Swedberg, losses from the soil beneath clasts. Silversword seeds
1986; Groenevelt et al., 1989; del Moral and Bliss, are highly heat-sensitive, as no germination occurs if
1993; Pérez, 1998a; Titus and del Moral, 1998) and soil temperature exceeds 35 jC (Siegel et al., 1970).
by concentration of water in the interstitial spaces Indeed, silversword germination is substantially great-
between stones, often filled with soil or smaller chips er under shade (Walker and Powell, 1995). Silver-
of rock (Yair and Lavee, 1976; Pérez, 1991; Poesen sword reproduction may be better on reddish clast
and Lavee, 1994b). My own data at Haleakala (Pérez, areas because this coloration, with a greater albedo
2000) indicated that surface clasts reduced evapora- than comparable brown/black stones, allows soils
tion from the soil beneath mainly by disrupting beneath to reach noon maxima that are f 9.9 jC
capillary flow to the ground surface. Poorly sorted lower than in soils below darker stones (F.L. Pérez,
clasts with abundant fine (10 –20 mm) gravel are most unpublished data). This difference may result in
effective in this respect, as larger pore spaces between greater water conservation under reddish-brown clasts.
blocks and cobbles are efficiently sealed. This is one On exposed surfaces, stones may also be more likely to
reason why the presence of gravel or small pebbles, trap seeds (del Moral and Bliss, 1993; Titus and del
mixed with larger stones by the sieve effect, is crucial Moral, 1998); if propitious conditions are found there,
for water conservation in volcanic pavements. Both seedlings will preferentially grow in such protected
the soil profile (Table 6, Fig. 13) and variation of microsites (Kobayashi, 1973, p. 46).
 F.L. Pérez / Geomorphology 55 (2003) 173–202 197

Pyroclastic substrates are too loose and unstable for tions (Pérez, 1993). The slow rates of movement—a
some plants, and the instability of the surface soil layer few centimeters/year—reported by Kobayashi (1973)
may inhibit plant establishment and cause quick death in clast areas are consistent with theoretical rates of
to seedlings germinating on them (Tsuyuzaki, 1991). creep solely induced by thermal variation (Scheideg-
Stone areas may provide safe sites (del Moral and Bliss, ger, 1970). In essence, while thermal creep and rock-
1993) that offer protection from various geomorphic falls occur at Haleakala, they are probably only minor
agents of ground disturbance; several surface processes agents of slope disturbance.
can be mitigated by volcaniclastic covers, especially on In contrast, fine-debris areas may be recurrently
sloping areas. Steep debris slopes like taluses and cinder affected by several processes. Surprisingly, slope dis-
cone flanks are affected by many geomorphic pro- turbance by runoff may be significant at Haleakala.
cesses; different slope substrates are modified by agents Medeiros and Loope (1994) indicated that heavy rains
that may operate with widely dissimilar intensities and caused much shifting of cinders on slopes, which
recurrence intervals, thus mobilizing sediment at vari- buried or washed away many rosette seedlings.
able rates (Iwata, 1983). Studies on alpine talus of the Kobayashi (1973, p. 27) also noted how winter storms
Andes and the Cascades Mountains have showed that removed clasts at the foot of cinder cones, when
rates of descent are inversely correlated with size of copious runoff was generated. A powerful storm
surface particles. Markers on sand shifted 22.5 cm/year; brought 15.5 inches of snow, rain and hail to Haleakala
those on pebbles and cobbles moved 15.2 cm/year, but on February 1 – 3, 1936 (Stearns and Macdonald (1942,
block areas only experienced a mean downslope dis- p. 32), when ‘‘. . .the slope was white with streams and
placement of 3.7 cm/year (Pérez, 1993). Embedded waterfalls. Sheetflood erosion was extremely fast, the
blocks or outcrops were also very stable, and allowed loose clinker and pumice moving rapidly down the
an annual shift of only f 5 cm, while markers placed at steep slopes.’’ Kobayashi’s (1973, p. 56) records of
the margins of rocks moved 23 –45 cm/year (Pérez, slope movement on Ka-Moa-o-Pele show pronounced
1990). Similar data from a steep cinder cone (Ka-Moa- activity during the rainy season, when 87 – 90% of the
o-Pele) in Haleakala show that clasts moved 1.8 – 6.3 movement on areas with both cinder and clasts took
cm in 1 year, while adjacent fine cinder shifted 25.3 cm place. This strong seasonal variation raises the possi-
(Kobayashi, 1973, p. 56). The reason for this inordinate bility that displacement in both coarse and fine areas
spatial disparity in movement rates is simple: finer- may be triggered by rain and/or presence of moisture.
grained substrates are disturbed more frequently and by Debris flows appeared to occur only infrequently at
more geomorphic processes than coarse clast sites. Haleakala. I observed some fresh, 50 –75-m-long deb-
Blocky areas are mainly affected by sporadic rockfall ris-flow channels near the Sliding Sands trail, but they
and thermal creep, but steep sand/gravel slopes erode had not affected sites with silverswords.
by a combination of miniature dry-debris slides, debris Frost creep in tropical mountains is frequently
flows, runoff, and frost creep, including that caused by caused by needle-ice growth. This process affects a
needle-ice activity (Pérez, 1993, 1998b). shallow soil layer that moves downhill after nightly
Field evidence for some of these processes at heaving by ice crystals. Miniature sorted stripes,
Haleakala may be only circumstantial, but it is not ubiquitous on the upper slopes of Haleakala, are
scarce. Small rockfalls could affect sites that lie among the most common features produced by frost
directly below lava outcrops or Pohaku-o-Hanalei activity. Noguchi et al. (1987) noted needle ice and
deposits (Fig. 6), which might disturb the slopes hoarfrost near Haleakala’s summit in December, when
below while adding fragments to them. Other than the climate is cold and wet, but found only ‘‘indis-
occasional boulders rolling downhill, I have not tinct’’ stripes at 2774 m. This suggests that Haleakala
witnessed any sizable rockfalls and have seen only occupies a marginal periglacial environment, due both
sparse signs of such activity (i.e., impact craterlets, to climatic (insufficient freezing) and pedological
aligned bump holes; Pérez, 1998b). Thermal creep (dry, coarse soils) reasons. Nonetheless, I did find
would leave no obvious traces behind, as it occurs many—albeit faintly developed—sorted stripes at site
gradually due to alternating contraction/expansion of 2 (2725 m) both in 1998 and 2001. Downhill-oriented
blocks following broad diurnal temperature fluctua- stripes had formed on areas of fine gravel and scoria
198 F.L. Pérez / Geomorphology 55 (2003) 173–202

and were closely related to fine-earth flags and pebble Pérez et al., 2002). This is amply confirmed by the
streaming at the margins of blocks (see below). analysis of small piles above plants on sites 2 and 4.
Miniature dry-debris slides are common on steep Because the former location is steeper, accumulated
( z 24j) fine-debris (sand and gravel) slopes. Such debris wedges are thicker (13.8 cm) than on gentler site
shallow ( z 7 cm) mass wasting events normally 2 (9.9. cm). The fact that steps are substantially larger
affect small segments of just a few square meters above taller silverswords indicates these plants actively
and travel a short distance downhill, but are easily intercept moving sediment through their lives, which
triggered. Agents include falling rocks or dislodged may span 63 to 90 years (Rundel and Witter, 1994, p.
boulders which cause ‘‘impact avalanches’’ (Pérez, 304). Thus, debris piles may well represent the cumu-
1985), trampling by people (Kobayashi, 1973) or lative effect of several decades of slope creep and
animals—particularly feral goats in the past (Yocom, debris sliding; the same logic applies to the tilting
1967)—and any other process that causes local debris observed in adult rosettes (Fig. 15). The prevalent
accumulation and oversteepening (Pérez, 1998b). distorted patterns of silversword roots (Fig. 14) are
Intermittent downhill streaming of fine debris has also suggestive of periodic burial by shifting debris.
some important effects on slope morphology. The Only inclined, high-altitude (>2610 m) sites had a
sieve effect is largely responsible for debris stratifica- significant fraction of rosettes with steps upslope of
tion and the stone packing observed with depth in them; this suggests frost disturbance may be respon-
volcaniclastic layers: as fine debris move downslope, sible for much debris mobilization at these sites.
they fall into the openwork voids between larger Although site 3 was the steepest (27j) and also had
stones, gradually infilling them. This causes not only the highest elevation (2755 m), silverswords there had
heterogeneous, poorly sorted deposits—more efficient no pronounced debris steps. The reason is simple: this
in preserving moisture—but also has the effect of slope was practically 100% covered by coarse clasts
‘‘coarsening’’ the slope surface, as fine debris disap- (Table 2), and no fine debris were exposed at this site.
pear within stone interstices. The lack of steps on this thick, coarse clastic cover
Whether caused by frost creep and/or shallow further suggests that cobbles and blocks move downhill
debris sliding, several features associated with rapid slowly, in agreement with expected rates effected
debris descent are common on sites over 2610 m mainly by thermal creep, as discussed above. I tested
(Table 1). On steep areas, isolated blocks obstruct the the subsidiary hypothesis that stone size might be
descent of fine debris (boulder dams; Pérez, 1994); greater upslope from larger rosettes; this would indicate
these accumulate upslope from stones, but are pro- that at least some large blocks/cobbles were not present
gressively deflected laterally toward the block mar- when the plant germinated but were instead immobi-
gins. A sheltered area below the stone does not lized at a later life stage. A linear regression between
receive any debris input and thus stands as a distinc- plant height and clast type upslope from 65 rosettes
tive, elongated, clast-free ribbon or fine-earth flag with steps at site 4 revealed that clast size increased
(Fig. 18). Such features are common on tropical with rosette age, but the correlation was not highly
mountains in South America (Francou, 1983; Pérez, significant ( p < 0.025) and could ‘‘explain’’ only
1993, 1994), Africa (Hastenrath, 1973) and Mexico f 10% (r2) of the observed variance. In essence, these
(Heine, 1977). The extensive stone pavements near data suggest that (i) plants on steep, fine-debris slopes
cinder cones have probably collected by gradual gradually arrest modest amounts of shifting sediment,
downslope migration of stones, caused by frost heav- which accumulates upslope from them; but (ii) most of
ing and creep (cf. Schubert, 1975) and/or by runoff this intercepted debris consists of small clasts, although
transport (Stearns and Macdonald, 1942). an occasional block or cobble may come to rest above a
The differences in clast size and ground texture plant; therefore (iii) it would seem that most large
observed above and below silverswords (Fig. 9) may blocks immediately upslope from adult plants were
be at least partially ascribed to creep processes. Some already there when the plant(s) colonized the slope;
rosettes growing below clasts must have initially ger- and (iv) this probably occurs because blocks offer
minated there, but other plants, once established, may eminently safe sites from excessive plant burial by
have become debris-damming agents (Pérez, 1987; the constant streaming of debris.
 F.L. Pérez / Geomorphology 55 (2003) 173–202 199

Fig. 18. Fine-earth flag exposed below a 13-cm-wide block. The 32-cm-long flag is the only area where the underlying sand is exposed, as
surficial black-gravel fragments forming a continuous f 3-cm-thick layer over the ground are deflected sideways by the stable block during
their descent. White ruler is 15 cm long. Site 1, 2680 m altitude; slope angle is f 26j. July 26, 2001 (Hk—01.70d).

In summary, volcaniclastic covers at Haleakala nation of runoff, frost creep, and dry-debris slides and
provide shelter to silverswords because they increase thus offer an inhospitable substrate for plant coloni-
soil moisture availability and can collect water by fog zation and growth.
interception and possibly also by condensation. In
addition, maximum substrate temperatures are ame-
liorated below clasts, and stones may help in trapping Acknowledgements
plant seeds. Clastic covers are disrupted less fre-
quently and by fewer geomorphic processes than fine I am grateful to Ronald J. Nagata (Chief, Re-
cinder. Stone-covered slopes are sporadically affected sources Management), Elizabeth Gordon (Manager,
by rockfalls and by slow thermal creep. In contrast, Cultural Resources Program), and Dr. Lloyd L. Loope
fine sand/gravel flanks are easily eroded by a combi- (Research Scientist) for kindly providing the research
200 F.L. Pérez / Geomorphology 55 (2003) 173–202

permits for Haleakala National Park during 1996, Corey, A.T., Kemper, W.D., 1968. Conservation of Soil Water by
1998, and 2001. I sincerely thank my wife, Inés, and Gravel Mulches. Hydrology Papers, vol. 30. Colorado State
University, Ft. Collins. 23 pp.
our two sons, Andrés and Alejandro, for hiking with del Moral, R., Bliss, L.C., 1993. Mechanisms of primary succes-
me into Haleakala’s crater and for their enthusiastic sion: insights resulting from the eruption of Mount St. Helens.
help during field work. Critical comments and Advances in Ecological Research 24, 1 – 66.
suggestions by Dr. I.L. Bergquist (University of Dinkins, S., 1969. Lanzarote, the strangest Canary. National Geo-
graphic Magazine 135 (1), 116 – 139.
Texas, Austin) and various anonymous reviewers
Doolittle, W.E., 1998. Innovation and diffusion of sand- and gravel-
were very helpful in focusing and improving the mulch agriculture in the American Southwest: a product of the
original manuscript. My deepest appreciation to all. eruption of Sunset Crater. Quaternaire 9, 61 – 69.
Doolittle, W.E., 2000. Cultivated Landscapes of Native North
America. Oxford Univ. Press, Oxford, UK.
References Dunkerley, D.L., 1996. Stone cover on desert hillslopes: extent of
bias in diameters estimated from grid samples and procedures
Abrahams, A.D., Parsons, A.J., Luk, S.-H., 1986. Resistance to for bias correction. Earth Surface Processes and Landforms 21,
overland flow on desert hillslopes. Journal of Hydrology 88, 573 – 580.
343 – 363. Evenari, M., Schulze, E.D., Kappen, L., Buschbom, U., Lange,
Alexander, W.D., 1870. On the crater of Haleakala, island of Maui, O.L., 1975. Adaptive mechanisms in desert plants. In: Vernberg,
Hawaiian group. American Journal of Science and Arts 49 (145), F.J. (Ed.), Physiological Adaptation to the Environment. Intext,
43 – 48. New York, pp. 111 – 129.
American Geological Institute, 1976. Dictionary of Geological Folk, R.L., 1966. A review of grain-size parameters. Sedimentology
Terms, 2nd ed. (Rev.). Anchor-Doubleday Press, Garden City, 6, 73 – 93.
NY. Folk, R.L., 1980. Petrology of Sedimentary Rocks, 3rd ed. Hemp-
Araña, V., López, J., 1974. Volcanismo. Dinámica y Petrologı́a de hill, Austin, TX.
sus Productos. Ed. Itsmo, Madrid. Francou, B., 1983. Dynamiques périglaciaires et Quaternaire dans
Ball, D.F., 1964. Loss-on-ignition as an estimate of organic matter les Andes centrales. Rapports Scientifiques et Techniques-
and organic carbon in non-calcareous soils. Journal of Soil Sci- Centre National de la Recherche Scientifique, Centre de Géo-
ence 15, 84 – 92. morphologie 2 (Caen, France, 63 pp.).
Blake, G.R., Hartge, K.H., 1982. Bulk density. Methods of Soil Gardiner, V., Dackombe, R.V., 1983. Geomorphological Field Man-
Analysis: Part 1. Physical and Mineralogical Methods, 2nd ed. ual. Allen & Unwin, London.
Monograph, vol. 9. American Society of Agronomy, Madison, Giambelluca, T.W., Nullet, D., 1991. Influence of the trade-wind
WI, pp. 363 – 375. inversion on the climate of a leeward mountain slope in Hawaii.
Blumenstock, D.I., Price, S., 1967. Climates of the States: Hawaii. Climate Research 1, 207 – 216.
Climatography of the United States. U.S. Department of Com- Groenevelt, P.H., van Straaten, P., Rasiah, V., Simpson, J., 1989.
merce, Washington, DC, 60-51. 27 pp. Modifications in evaporation parameters by rock mulches. Soil
Brückl, E., Brunner, F.K., Gerber, E., Scheidegger, A.E., 1974. Technology 2, 279 – 285.
Morphometrie einer Schutthalde. Mitteilungen Oesterreichische Hastenrath, S., 1973. Observations on the periglacial morphology
Geographische Gesellschaft 116, 79 – 96. of Mts. Kenya and Kilimanjaro, east Africa. Zeitschrift fur Geo-
Bruegmann, M.M., 1995. Protecting habitat for silversword recov- morphologie. Supplementband 16, 161 – 179.
ery. Endangered Species Bulletin 20 (4), 6 – 7. Heine, K., 1977. Zur morphologischen Bedeutung des Kammeises in
Caine, T.N., 1969. The analysis of surface fabrics by means of der subnivalen Zone randtropischer semihumider Hochgebirge.
ground photography. Arctic and Alpine Research 1, 127 – 134. Zeitschrift fur Geomorphologie. Supplementband 21, 57 – 78.
Carniel, P., Scheidegger, A.E., 1974. Morphometry of an alpine Hendrix, L.B., 1981. Post-eruption succession on Isla Fernandina,
scree cone. Rivista Italiana di Geofisica 23, 95 – 100. Galápagos. Madrono 28, 242 – 254.
Catana, A.J., 1963. The wandering quarter method of estimating Inman, D.L., 1952. Measures for describing the size distribution of
population density. Ecology 44, 349 – 360. sediments. Journal of Sedimentary Petrology 22, 125 – 145.
Church, M.A., Stock, R.F., Ryder, J.M., 1979. Contemporary sedi- Iwata, S., 1983. Physiographic conditions for the rubble slope for-
mentary environments on Baffin Island, N.W.T., Canada: debris mation on Mt. Shirouma-dake, the Japan Alps. Geographical
slope accumulations. Arctic and Alpine Research 11, 371 – 401. Reports of Tokyo Metropolitan University 18, 1 – 51.
Cline, M.G., 1955. Soil Survey of the Territory of Hawaii. Soil Jones, T.A., 1969. Skewness and kurtosis as criteria of normality in
Survey Series 25. USDA, Soil Conservation Service, Washing- observed frequency distributions. Journal of Sedimentary Pet-
ton, DC. 644 pp. rology 39, 1622 – 1627.
Coe, M.J., 1969. Microclimate and animal life in equatorial moun- Kershaw, L.J., Gardner, J.S., 1986. Vascular plants on mountain
tains. Zoologica Africana 4, 101 – 128. talus slopes, Mt. Rae area, Alberta, Canada. Physical Geography
Colton, H.S., 1965. Experiments in raising corn in the Sunset Crater 7, 218 – 230.
ashfall area east of Flagstaff, Arizona. Plateau 37, 77 – 79. Kobayashi, H.K., 1973. Ecology of the silversword, Argyroxiphium
 F.L. Pérez / Geomorphology 55 (2003) 173–202 201

sandwicense DC. (Compositae), Haleakala Crater, Hawaii. PhD Pérez, F.L., 1985. Surficial talus movement in an Andean paramo of
Dissertation, Univ. of Hawai’i, Manoa. Venezuela. Geografiska Annaler 67A, 221 – 237.
Leopold, L.B., 1949. The interaction of tradewind and sea breeze, Pérez, F.L., 1987. Downslope stone transport by needle ice in a high
Hawaii. Journal of Meteorology 6, 312 – 320. Andean area (Venezuela). Revue de Geomorphologie Dynami-
Leopold, L.B., 1970. An improved method for size distribution of que 36, 33 – 51.
stream bed gravel. Water Resources Research 6, 1357 – 1366. Pérez, F.L., 1989. Talus fabric and particle morphology on Lassen
Leuschner, C., Schulte, M., 1991. Microclimatological investiga- Peak, California. Geografiska Annaler 71A, 43 – 57.
tions in the tropical alpine scrub of Maui, Hawaii: evidence Pérez, F.L., 1990. Talus shift near rock outcrops in the California
for a drought-induced alpine timberline. Pacific Science 45, Cascades (Abstract). Association of American Geographers,
152 – 168. Program and Abstracts, AAG Annual Meeting, Toronto, Cana-
Liegel, E.A., Simson, C.R., Schulte, E.E., 1980. Wisconsin Proce- da, p. 193.
dure for Soil Testing, Plant Analysis and Feed and Forage Anal- Pérez, F.L., 1991. Soil moisture and the distribution of giant An-
ysis. Soil Fertility Series, vol. 6. University of Wisconsin dean rosettes on talus slopes of a desert paramo. Climate Re-
Extension, Madison. search 1, 217 – 231.
Loope, L.L., Medeiros, A.C., 1994. Biotic interaction in Hawaiian Pérez, F.L., 1993. Talus movement in the high equatorial Andes: a
high elevation ecosystems. In: Rundel, P.W., Smith, A.P., synthesis of ten years of data. Permafrost and Periglacial Pro-
Meinzer, F.C. (Eds.), Tropical Alpine Environments. Plant cesses 4, 199 – 215.
Form and Function. Cambridge Univ. Press, Cambridge, UK, Pérez, F.L., 1994. Geobotanical influence of talus movement on the
pp. 337 – 354. distribution of caulescent Andean rosettes. Flora 189, 353 – 371.
Macdonald, G.A., 1967. Forms and structures of extrusive basaltic Pérez, F.L., 1998a. Conservation of soil moisture by different stone
rocks. In: Hess, H.H., Poldervaart, A. (Eds.), Basalts: The Pol- covers on alpine talus slopes (Lassen, California). Catena 33,
dervaart Treatise on Rocks of Basaltic Composition, vol. 1. 155 – 177.
Interscience-Wiley, New York, pp. 1 – 62. Pérez, F.L., 1998b. Talus fabric, clast morphology, and botanical
Macdonald, G.A., 1978. Geologic map of the crater section of indicators of slope processes on the Chaos Crags (California
Haleakala National Park, Maui, Hawaii. U.S. Geological Sur- Cascades) USA. Geographie Physique et Quaternaire 52,
vey, Miscellaneous Investigation Series Map I-1088, 8 pp. 47 – 68.
+ map (1:24,000). Pérez, F.L., 2000. The influence of surface volcaniclastic layers
Macdonald, G.A., Abbott, A.T., Peterson, F.L., 1983. Volcanoes in from Haleakala (Maui, Hawaii) on soil water conservation. Cat-
the Sea. The Geology of Hawaii. University of Hawai’i Press, ena 38, 301 – 332.
Honolulu. Pérez, F.L., 2001. Geoecological alteration of surface soils by the
Mack, J., 1984. Haleakala: The Story Behind the Scenery. KC Hawaiian silversword (Argyroxiphium sandwicense DC.) in Ha-
Publications, Las Vegas, NV. leakala’s crater, Maui. Plant Ecology 157, 215 – 233.
Maule, S.H., 1963. Corn growing at Wupatki. Plateau 36, 29 – 32. Pérez, F.L., 2002. Geobotanical relationship of Draba chionophila
Medeiros, A.C., Loope, L.L., 1994. Rare Animals and Plants of (Brassicaceae) rosettes and miniature frost-sorted stripes in the
Haleakala National Park. Hawai’i Natural History Association, high equatorial Andes. Flora 197, 24 – 36.
Hawai’i National Park, HI. Poesen, J., Lavee, H. (Eds.), 1994a. Rock Fragments in Soils: Sur-
Miller, R.L., Kahn, J.S., 1962. Statistical Analysis in the Geological face Dynamics. Catena, vol. 23. 198 pp.
Sciences. Wiley, New York. Poesen, J., Lavee, H., 1994b. Rock fragments in top soils: signifi-
Missionary Herald, 1829. Ascent of an extinguished volcano. The cance and processes. Catena 23, 1 – 28.
Missionary Herald (Boston), August 25 (8), 247 – 248. Pukui, M.K., Elbert, S.H., Mookini, E.T., 1974. Place Names of
Mueller-Dombois, D., Ellenberg, H., 1974. Aims and Methods of Hawaii. University of Hawai’i Press, Honolulu.
Vegetation Ecology. Wiley, New York. Pukui, M.K., Elbert, S.H., Mookini, E.T., Nishizawa, Y.M., 1992.
Munsell Soil Color Chart, 1992. Macbeth, Kollmorgen Instruments New Pocket Hawaiian Dictionary. University of Hawai’i Press,
Corp., Newburgh, NY. Honolulu.
Noguchi, Y., Tabuchi, H., Hasegawa, H., 1987. Physical factors Ruhle, G.C., 1959. A Guide for the Haleakala Section, Island of
controlling the formation of patterned ground on Haleakala, Maui, Hawaii. Hawaii Nature Notes. Hawaii Natural History
Maui. Geografiska Annaler 69A, 329 – 342. Association, Honolulu.
Olsen, S.R., Sommers, L.E., 1982. Phosphorus. Methods of Soil Rundel, P.W., Witter, M.S., 1994. Population dynamics and flower-
Analysis: Part 2. Chemical and Microbiological Properties, 2nd ing in a Hawaiian alpine rosette plant, Argyroxiphium sandwi-
ed. Monograph, vol. 9. American Society of Agronomy, Mad- cense. In: Rundel, P.W., Smith, A.P., Meinzer, F.C. (Eds.),
ison, WI, pp. 403 – 430. Tropical Alpine Environments. Plant Form and Function. Cam-
Othieno, C.O., 1980. Effects of mulches on soil water content and bridge Univ. Press, Cambridge, UK, pp. 295 – 306.
water status of tea plants in Kenya. Experimental Agriculture Scheidegger, A.E., 1970. Theoretical Geomorphology, 2nd ed.
16, 295 – 302. Springer Verlag, Berlin.
Othieno, C.O., Ahn, P.M., 1980. Effects of mulches on soil temper- Schubert, C., 1975. Glaciation and periglacial morphology in the
ature and growth of tea plants in Kenya. Experimental Agricul- northwestern Venezuelan Andes. Eiszeitalter und Gegenwart 26,
ture 16, 287 – 294. 196 – 211.
202 F.L. Pérez / Geomorphology 55 (2003) 173–202

Schumm, S.A., 1964. Seasonal variation of erosion rates and pro- U.S. Fish and Wildlife Service, 1992. Final listing rules for 53
cesses on hillslopes in western Colorado. Zeitschrift fur Geo- species. Endangered Species Technical Bulletin 17 (3 – 8), 15.
morphologie. Supplementband 5, 215 – 238. U.S. Geological Survey, 1971. State of Hawai’i, Principal Islands.
Segalen, P., 1984. Project of Soil Classification (Projet de Classi- U.S. Geological Survey Map (1:500,000).
fication des Sols). International Soil Reference and Information U.S. Geological Survey, 1983. Kilohana and Luala’i-lua Hills,
Centre, Wageningen, The Netherlands. Maui, Hawai’i. U.S. Geological Survey Maps (1:24,000).
Siegel, S.M., Carroll, P., Corn, C., Speitel, T., 1970. Experimental Wagner, W.L., Herbst, D.R., Sohmer, S.H., 1990. Manual of the
studies on the Hawaiian silverswords (Argyroxiphium spp.): Flowering Plants of Hawai’i. Bishop Museum Special Publi-
some preliminary notes on germination. Botanical Gazette 131, cation, vol. 83. University of Hawai’i Press, Honolulu. Two
277 – 280. volumes.
Soil Survey Staff, 1975. Soil Taxonomy. Agriculture Handbook, Walker, L.R., Powell, E.A., 1995. Factors affecting seed germina-
vol. 436. U.S. Department of Agriculture, Soil Conservation tion of the Mauna Kea silversword in Hawai’i. Pacific Science
Service, Washington, DC. 49, 205 – 211.
Soil Survey Staff, 1994. Keys to Soil Taxonomy, 6th ed. U.S. Walker, L.R., Powell, E.A., 1999. Regeneration of the Mauna Kea
Department of Agriculture, Soil Conservation Service, Washing- silversword Argyroxiphium sandwicense (Asteraceae) in Ha-
ton, DC. waii. Biological Conservation 89, 61 – 70.
Sokal, R.R., Rohlf, F.J., 1969. Biometry: The Principles and Practice Weaver, J.E., 1919. The Ecological Relations of Roots. Publica-
of Statistics in Biological Research. Freeman, San Francisco. tion 286. Carnegie Institution of Washington, Washington,
Stearns, H.T., 1942. Origin of the Haleakala crater, island of Maui, DC. 128 pp.
Hawaii. Geological Society of America Bulletin 53, 1 – 14. Wentworth, C.K., Macdonald, G.A., 1953. Structures and Forms of
Stearns, H.T., Macdonald, G.A., 1942. Geology and ground-water Basaltic rocks in Hawaii. U.S. Geological Survey Bulletin 994
resources of the island of Maui, Hawaii. Bulletin, Division of (Washington, DC, 98 pp.).
Hydrography, Hawaii. 7, 344 pp. + map. Whiteaker, L.D., 1983. The vegetation and environment in the
Stone, C.P., Loope, L.L., 1987. Reducing negative effects of intro- Crater District of Haleakala National Park. Pacific Science 37,
duced animals on native biotas in Hawaii: what is being done, 1 – 24.
what needs doing, and the role of National Parks. Environmental Wilkes, C., U.S.N., 1845. Narrative of the United States Exploring
Conservation 14, 245 – 258. Expedition During the Years 1838, 1839, 1840, 1841, 1842,
Swedberg, K.C., 1986. Effects of Mount St. Helens ash on plant vol. 4. Lea & Blanchard, Philadelphia.
populations of shallow stony soils. In: Keller, S.A.C. (Ed.), Yair, A., Lavee, H., 1976. Runoff generative process and runoff
Mount St. Helens: Five Years Later. Eastern Washington Univ. yield from arid talus mantled slopes. Earth Surface Processes
Press, Cheney, pp. 256 – 260. 1, 235 – 247.
Titus, J.H., del Moral, R., 1998. Seedling establishment in different Yocom, C.F., 1967. Ecology of feral goats in Haleakala National
microsites on Mount St. Helens, Washington, USA. Plant Ecol- Park, Maui, Hawaii. American Midland Naturalist 77, 418 – 451.
ogy 134, 13 – 26. Young, K.R., León, B., 1990. Curvature of woody plants on slopes
Tsuyuzaki, S., 1991. Species turnover and diversity during early of a timberline montane forest in Peru. Physical Geography 11,
stages of vegetation recovery on the volcano Usu, northern 66 – 74.
Japan. Journal of Vegetation Science 2, 301 – 306.
 Geomorphology 55 (2003) 203 – 217
www.elsevier.com/locate/geomorph

Lithologic, structural, and geomorphic controls on ribbon forest

patterns in a glaciated mountain environment
David R. Butler a,*, George P. Malanson b, Matthew F. Bekker c, Lynn M. Resler a
a
The James and Marilyn Lovell Center, Department of Geography, Southwest Texas State University, 601 University Drive, ELA 120-B,
San Marcos, TX 78666-4616, USA
b
Department of Geography, University of Iowa, Iowa City, IA 52242, USA
c
Department of Geography, Brigham Young University, Provo, UT 84602, USA
Received 1 June 2002; received in revised form 30 August 2002; accepted 10 March 2003

Abstract

So-called ‘‘ribbon forests’’ have been attributed to snowdrift patterns and fire history without reference to geomorphology
[Vegetatio 19 (1969) 192.]. This paper illustrates how site conditions of geomorphology and geology explain the origin of
ribbon forests. In Glacier National Park, MT (USA), regional tectonic uplift associated with the Laramide Orogeny produced
structural features that amplify lithologic differences. Pleistocene glaciation scoured deeply along the strike of bedding planes,
highlighting this pattern and in some cases producing fine-scale parallel finger lakes between forested ribbon strips.
Twelve ribbon forest sites on both sides of the Continental Divide were closely studied on stereoscopic aerial photographs,
and several of these sites were examined in the field or from helicopter overflights. In all cases, geologic and geomorphic
conditions explain the location and distribution of the ribbon forests. Change-detection of the distribution of trees versus
nontree-covered surfaces in an area of ribbon forest on Flattop Mountain, a complex uplifted synclinal structure, was
undertaken using panchromatic, low-altitude aerial photographs from 1966 to 1991. Areas changed from forest to meadow and
from meadow to forest in roughly equal amounts in a generally random spatial pattern. No evidence was seen to suggest that the
creation of one ribbon eventually created another downwind, as suggested by Billings. Aerial photograph interpretation, field
examination and soils analyses of forest ribbons and adjacent unforested meadows clearly illustrated that trees occupy higher,
parallel to subparallel, well-drained sites where the spatial pattern is in turn a distinct reflection of the spatial pattern of structure
and stratigraphy. Meadows occupy topographically lower positions between ridges where erosion along bedding plane strike
was concentrated. Topography sets conditions that allow tree growth in certain locations while precluding it in immediately
adjacent areas. Ribbon forests there are thus a spatial manifestation of the interaction between structure, lithology, and
topography.
D 2003 Elsevier Science B.V. All rights reserved.

Keywords: Ribbon forest; Stratigraphy; Lithology; Montana; Glacier National Park; Aerial photography change detection

1. Introduction

In locations around the world, local variations in

* Corresponding author. soils, moisture regimes, geomorphic history, and/or
E-mail address: [email protected] (D.R. Butler). structural and lithological controls have produced

0169-555X/03/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0169-555X(03)00140-5
204 D.R. Butler et al. / Geomorphology 55 (2003) 203–217

relatively linear vegetation features on the landscape. aligned parallel to the direction of prevailing winds are
Terms such as ‘‘banded vegetation,’’ ‘‘tiger-striped sometimes referred to as ‘‘hedges’’ (Holtmeier, 1982).
vegetation,’’ ‘‘brousse tigrée,’’ ‘‘hedges,’’ ‘‘vegetation Intimately associated with ribbon forests are ‘‘snow
ripples,’’ ‘‘vegetation arcs,’’ ‘‘vegetation stripes,’’ and glades’’ that are hypothesized to result from eolian
‘‘ribbon forest’’ describe relatively linear vegetation deposition of snow in large drifts in the lee of each
patterns found from the savanna region of the African forested ribbon (Billings, 1969). Ecologists have been
Sahel, to Australia, to the Rocky Mountains of North at great pains to understand the mechanism by which
America (Billings, 1969; White, 1969, 1970, 1971; ribbon forests actually develop, postulating compli-
Buckner, 1977; Vale, 1978; Holtmeier, 1982; Doering cated origins that typically involve some combination
and Reider, 1992; Bryan and Brun, 1999; Dunkerley of wind and snow (as described by Billings) and/or
and Brown, 1999; Zonneveld, 1999). In many cases, occasionally the addition of fire as a primary causal
these banded forms of vegetation have been attributed factor (Buckner, 1977; Holtmeier, 1982).
to underlying variations in soils and soil-moisture Billings’ (1969) primary hypothesis of ribbon
characteristics reflective of differing geomorphology forest development may be summarized as follows:
or lithology or to processes of surface wash or eolian a group of seedlings becomes established on shallow
deposition interacting with the pre-existing landform soil toward the windward edge of an open area (the
surface. Notably, however, works examining the so- origin of the open area is not specified by Billings).
called ribbon forests of the Rocky Mountains have The forested patch expands laterally (in a north –south
largely ignored the influence of geomorphology, land- direction, perpendicular to an assumed prevailing
form history, and underlying geological structure and westerly wind) through layering and establishment
lithology. Instead, those studies have attempted to of seedlings; however, Billings gives no explanation
describe the cause(s) of ribbon forests as an interaction as to why the established patches would expand
between snowdrift and prevailing winds, in some cases laterally. A snowdrift builds up on the lee side of
interacting also with forest fire history of a given area. the patch. The snow drift melts in the summer,
These causes imply positive feedback and potential providing ‘‘ideal moisture conditions for tree seedling
connections to complexity theory (Malanson, 1999). establishment wherever its far edge happens to be late
In the course of our research in Glacier National in June or early July’’ (Billings, 1969, p. 198). These
Park (GNP), MT, USA, we have encountered numer- seedlings will be protected by the larger drift that
ous examples of ribbon forest in the field, as viewed forms during winter. Where the drift persists beyond
from a helicopter, and on stereoscopic aerial photo- July, moist or wet meadow vegetation will be favored,
graphs. In this paper, we demonstrate that ribbon and thus ribbon forests will be spaced apart roughly
forests of GNP are the result of interactions between one-half the maximum width of the winter drift. When
the area’s geological structure and lithology, Pleisto- mature trees die, they create ‘‘blowouts’’ in the lee-
cene glacial history, current geomorphic processes and ward drift, which allow seedlings to establish in the
landforms, soils, and/or slope. In no case do we deem snow glade, thus causing bends or angles in the
it necessary to invoke complex snowdrift/wind/fire ribbons. The ribbons thus break and heal and migrate
interactions as causal agents of ribbon forest in this downwind over time. Although no firm time frame for
environment, although we recognize that snow may this sequence to occur was given, it would doubtless
assist in maintenance of ribbon forest patterns under take decades to centuries in the harsh subalpine
current climatic conditions. environment of the Rocky Mountains.
A related hypothesis posited by Billings (1969)
attempted to incorporate possible effects of fire. In
2. Background that scenario, crown fires may occur and clear large
areas of timber (thus accounting for the creation of the
Ribbon forests are roughly linear strips of forest that needed bare area mentioned above), resulting in
are theoretically aligned perpendicular to the direction increased wind speeds. When the accelerated wind
of prevailing winds, as described by Billings (1969) in eventually hits nonburned forest, it ‘‘dumps’’ large
a now-classic paper in ecology. Forested strips that are amounts of snow 80– 100 m in from the new forest
 D.R. Butler et al. / Geomorphology 55 (2003) 203–217 205

edge. This snow kills mature trees, creating a new forest could develop in the slow fashion Billings
snow glade. The encroaching snow eventually breaks described.
the forest up into ribbon-like patches. Other hypotheses for ribbon forest development
The bulk of Billings’ evidence for his theory of the have briefly alluded to, but not specifically addressed,
development of ribbon forests came from work in the relationships between ribbon forests and geological
Medicine Bow Mountains of SE Wyoming, USA. conditions. Buckner (1977) gave four hypotheses for
There, as few as one, and no more than three, ribbons the development of ribbon forest: (i) standing waves
had developed at any of his study sites. In such in air flow over the surface; (ii) subtle variations in
limited cases, his postulated mechanisms for ribbon surface relief; (iii) a priori difference in substrates; and
creation could conceivably function as he described. (iv) areas opened by fire and now filling in. Although
His discussion was obfuscated, however, by inclusion two of his four hypotheses suggest a geomorphic or
of an oblique aerial photograph over West Flattop geological origin, he nevertheless believed that differ-
Mountain in GNP, Montana, for which he offered no ences in soil characteristics between forested and
field evidence, aerial photograph interpretation, or nonforested strips were more attributable to effects
analysis. Its inclusion was meant to illustrate the of differing vegetational cover than to pre-existing
concept of ribbon forests but, in doing so, brought differences in underlying geological properties. Holt-
into question how such an extensive area of ribbon meier (1982) examined ribbon forests in the Colorado

Fig. 1. Map of location of Glacier National Park in NW Montana. Numbers refer to ribbon forest sites described in Table 1.
206 D.R. Butler et al. / Geomorphology 55 (2003) 203–217

Front Range and described ribbon-topped bedrock carpa), five-needle pines (Pinus flexilis and Pinus
ridges with snow glades occupying the surface albicaulis), and Engelmann spruce (Picea engelman-
between the ridges. Doering and Reider (1992) exam- nii). In spite of the differing climatic and vegetative
ined one of Billings’ study sites, at Cinnabar Park, communities on either side of the Continental Divide
WY, and attributed the development of the single in GNP, however, ribbon forest may be found on both
ribbon and adjacent snow glade (Cinnabar Park itself) sides of the Park (Fig. 1). Although local winds may
there to differences in edaphic factors. Their analysis blow from any direction, the prevailing wind direction
of soils morphology and stratigraphy and radiocarbon is from the W– SW (Finklin, 1986).
dating showed no evidence of migration of the snow
glade comprising Cinnabar Park.
4. Methods

3. The study area In our examination of ribbon forests in GNP, we

employed a variety of techniques to sample a broad
Glacier National Park is located in the Rocky range of ribbon forest sites. We examined several sites
Mountains of NW Montana, USA, along the USA/ during low-altitude helicopter overflights (Sue Lake,
Canadian border (Fig. 1). It is bisected by the Con- Preston Park, Granite Park, West Flattop Mountain,
tinental Divide, which runs roughly from the NW to Flattop Mountain, and the peak west of Camas Lake).
the SE in the park. Two mountain ranges comprise During field reconnaissance, we studied the ribbon
portions of the divide—the Livingston Range occu- forests of Preston Park, Logan Pass, and Flattop
pies the NW quarter of the park, and the Lewis Range Mountain; and at Flattop Mountain, we took soil
runs the length of the park from the north-central to samples and detailed field notes from ribbon forests
southern tip. The Park is comprised primarily of and the adjacent meadow. We studied all our sites on
widely distributed, thinly bedded, Precambrian sedi- stereoscopic aerial photographs with a mirror stereo-
mentary and metasedimentary limestones and argil- scope. At Flattop Mountain, we performed a photo-
lites of the Belt Series that were thrust eastward (the graphic change detection to determine if directional
Lewis Overthrust) during the Laramide Orogeny changes in ribbon and glade positions (sensu Billings,
(Ross, 1959; Whipple, 1992). The thrust sheet has 1969) can be discerned. Each of these techniques is
numerous secondary faults and folds associated with described more fully in the following sections.
it; chief among the folds is the Akamina Syncline,
which runs NW – SE through the center of the park 4.1. Aerial photo interpretation of ribbon forest sites
(Whipple, 1992), cutting partly through West Flattop
Mountain (Fig. 1) and bisecting Flattop Mountain into On high-quality, 9  9-in. panchromatic vertical
two approximately equal portions. The sedimentary stereoscopic aerial photographs of Glacier National
and metasedimentary rocks of the Belt Series were no Park (from late July or August 1966, 1991, and 1995),
match for the extensive Pleistocene glaciation in the we identified 12 ribbon forest sites on either side of
area (Carrara, 1990), resulting in both widespread the Continental Divide in GNP (Fig. 1). Of these 12
plucking and scouring along lithologic contacts and sites, two are astride the Continental Divide (West
bedding planes and the area’s striking glacial scenery. Flattop Mountain, #12, and Logan Pass, #1), five are
The presence of the Continental Divide establishes west of the Divide (Flattop Mountain, #11; Granite
two distinct climatic regimes in the park: the western Park, #7; the peak west of Camas Creek, #8; the
slopes experience a modified Pacific maritime climate cirque floor SE of Avalanche Lake, #9; and Thomp-
with some continental tendencies, and the eastern son Creek headwaters, #10), and the remaining five
slopes have a strongly continental regime (Finklin, are east of the Divide (#2 – 6, Table 1). Each site was
1986). The differing moisture regimes result in differ- examined in detail under the stereoscope; and notes
ent forest communities west and east of the Divide, were taken on the nature of glacial scouring, stream
although upper treeline on both sides of the Divide is incisions where present, strike and dip directions
typically dominated by subalpine fir (Abies lasio- (unfortunately, most ribbon forest sites in GNP have
 D.R. Butler et al. / Geomorphology 55 (2003) 203–217 207

Table 1 geomorphic and geologic conditions associated with

Areas of ribbon forest in Glacier National Parka the presence of the ribbons. At Flattop Mountain, we
Location Interpreted geomorphic/geologic cause took detailed notes on conditions at the site, and
Logan Pass, #1 Interaction of glacial plucking with collected two soil samples each from one forested
stratigraphy/dip; see Fig. 5, Klasner ribbon and its adjacent ‘‘snow glade.’’ These soil
and Fagre (2002)
samples were dried and transported to the laboratory
Sue Lake bench, #2 Interaction of glacial scouring and
plucking with stratigraphy/dip for analysis. Dry Munsell colors were recorded. Par-
Unnamed Hanging Interaction of glacial scouring and ticle size analysis was performed using standard
Valley S of Heavy plucking, stratigraphy/dip, and fluvial hydrometer settling, and organic matter content was
Runner Mountain, incision determined using the loss-on-ignition technique.
#3
Preston Park, #4 Interaction of glacial scouring,
stratigraphy/dip, and fluvial incision; 4.3. Change detection analysis of ribbon forests on
see Fig. 2b Flattop Mountain
Cirque floor SW of Interaction of glacial scouring and
Citadel Mountain, plucking, stratigraphy/dip, minor Two panchromatic, vertical, aerial photographs of
#5 fluvial incision
Flattop Mountain (from 1966 and 1991) served as
Unnamed point NE of Interaction of geologic structure,
Red Eagle Glacier, stratigraphy/dip, and glacial scouring; data in this study. The 1966 photograph (nominal
#6 see Fig. 2c scale of 1:34,000) was taken by the U.S. Geological
Granite Park, #7 Interaction of geologic structure, Survey (USGS) on August 24 and the 1991 photo-
stratigraphy/dip, and glacial scouring/ graph (nominal scale of 1:48,000) was taken by the
plucking
National Aerial Photography Program (NAPP) on
Peak west of Camas Interaction of glacial scouring with
Lake, #8 stratigraphy/dip; see Fig. 2a August 22. These photographs were chosen because
Cirque floor SE of Interaction of glacial plucking with of their temporal similarity, cloud-free conditions, and
Avalanche Lake, #9 stratigraphy/dip; see Fig. 3 availability. These images were scanned using a
Thompson Creek Interaction of glacial scouring and flatbed scanner.
headwaters, #10 plucking with stratigraphy/dip
The 1991 image was georeferenced to a 1996
Flattop Mountain, #11 Interaction of geologic structure,
stratigraphy/dip, and glacial scouring image of the same area that was previously geocoded
and plucking; see Figs. 3, 4, 5, 6 and 7 to a U.S. Geological Survey basemap and acquired
West Flattop Interaction of geologic structure, from the TerraServer website. Once the 1991 image
Mountain, #12 stratigraphy/dip, slope, glacial was geocoded, image-to-image registration was used
scouring and plucking, and fluvial
to register the 1966 image to the 1991 image (the
incision; see Figs. 4 and 5
a
1996 image from TerraServer was not used in this
Number following site name refers to location on Fig. 1.
study because of its poor image quality). An error of
F 12 m, an error value accepted for aerial photogra-
little or no strike and dip data available for them from phy (Baker et al., 1995), was achieved using nearest
the geologic maps of the Park), presence of folding neighbor interpolation and several ground control
structures, and any other notable features (such as points on distinguishable landmark features. Some
prominent ridges and low-lying points). From this distortion in the areas with complex topography was
information, we determined which combination of evident; however, distortion in the relatively small
geomorphic and geologic conditions accounted for area of interest was quite low. Both images were
the presence of ribbon forests. In every case, the resampled to a 2-m per pixel resolution.
position of the ribbon forests could be explained as Following geometric corrections of each aerial
resulting from the geomorphic and geologic factors. image, a low pass filter with a 3  3 kernel was used
to reduce the pixilated, systematic error caused in the
4.2. Fieldwork scanning process. Applying a 3  3 window size
produces the best overall effect to remove systematic
We observed ribbon forests in situ at Preston Park, biases and improve classification performance (Muel-
Logan Pass, and Flattop Mountain and noted the ler and Hoffer, 1989).
208 D.R. Butler et al. / Geomorphology 55 (2003) 203–217

An unsupervised approach using ERDAS Imagine eral to improve the accuracy of the classification
ISODATA Initial Arbitrary Cluster Allocation algo- (Jensen, 1996). This clustering algorithm employs
rithm independently classified each image. This user-specific parameters (such as the desired number
method has been used successfully by others in of clusters, minimum number of pixels per cluster,
studies of the forest-alpine tundra ecotone to classify and cluster iteration counter) to generate statistical
vegetation and snow cover (Allen and Walsh, 1993; clusters from the input (Allen and Walsh, 1993).
McGregor, 1998) and is a common algorithm to use in The goal of classifying the two images was to
post-classification analysis methodologies (Jensen, break the landscape surface into four categories. The
1996; Sohl, 1999). Unlike other unsupervised classi- classification scheme in this study was modest
fication algorithms that make only two passes through because the images were to be used for change
the dataset to achieve results, ISODATA makes sev- analysis of vegetation type rather than change detec-

Fig. 2. Ribbon forest types in Glacier National Park. (a) On flank of peak west of Camas Lake, glacial scouring has accentuated stratigraphic
bedding planes along strike. Particularly deep glacial scouring along strike of dipping bedding planes has produced ribbon forests as well as
subparallel ‘‘finger lakes’’ above Avalanche Lake (b) and on the western side of Flattop Mountain (c).
 D.R. Butler et al. / Geomorphology 55 (2003) 203–217 209

Fig. 2 (continued).

tion of species. Furthermore, the use of panchromatic tion. Each image was classified independently so that
imagery limits analysis to visible wavelength reflec- they could be used in a post-classification change
tance. Using a more complex classification scheme detection (Jensen, 1996) using the ISODATA algo-
could cause unnecessary errors in the change detec- rithm to generate 50 initial spectrally similar clusters.

Fig. 3. In Preston Park, glacial plucking interacts with stratigraphic bedding planes to produce ribbons that are dissected by fluvial incision.
210 D.R. Butler et al. / Geomorphology 55 (2003) 203–217

Fig. 4. Ribbons follow bedding planes in complex folded structure east of Red Eagle Glacier.

More clusters were generated than needed to allow for photography for comparison. The resulting image
the recoding of confused clusters (Sohl, 1999). Recod- consisted of four land-cover classes.
ing was done with the aid of field observations, Post-classification change detection (Sohl, 1999)
historical maps and graphics, field notes, and aerial was used to show change within and between land-

Fig. 5. U.S. Geological Survey topographic map of Flattop Mountain and West Flattop Mountain (note 1 square mile sections for scale). The
dashed line cutting across the top of both West Flattop and Flattop Mountains is the Continental Divide. Elevations in meters.
 D.R. Butler et al. / Geomorphology 55 (2003) 203–217 211

cover classes for the two different aerial photograph tion of geomorphic and geologic control. We present
dates. The 1966 and 1991 classified and recoded several examples here and in Table 1 to illustrate these
images served as input into a change detection matrix interpretations.
algorithm. The algorithm generated two change detec- A common interaction resulting in ribbon forests
tion images and two 4  4 matrices of changes that was uneven depth of glacial scouring parallel to the
occurred within classified pixel values between 1966 strike of dipping bedding planes (Fig. 2a). In some
and 1991. The derived values described the nature of cases, such as those shown in Fig. 2b and c, the
change each pixel underwent between the two dates. scouring depth was sufficient to produce small sub-
parallel ‘‘finger lakes.’’ In these cases, as also revealed
by our field inspection on Flattop Mountain, the
5. Results ribbon forests occupy the topographically well-
drained bedrock ridges, whereas ‘‘snow glades’’ (or
5.1. Aerial photograph and field reconnaissance ‘‘finger lakes’’) occupy the more deeply scoured
zones along stratigraphic and bedding contacts.
The aerial photograph interpretation and field The converse of parallel scouring along dipping
reconnaissance revealed that every case of ribbon bedding planes occurred at Logan Pass and Sue Lake.
forest encountered was attributable to some combina- At those sites, Pleistocene-aged glacial plucking

Fig. 6. Ribbon forests on West Flattop Mountain (a) and Flattop Mountain (b). The latter photograph was taken in 1935 from Swiftcurrent
Lookout and is used courtesy of the U.S. Geological Survey’s Glacier Field Station.
212 D.R. Butler et al. / Geomorphology 55 (2003) 203–217

moved roughly perpendicular to the strike of bedding of the Logan Pass site). The lee (plucked) sides of the
planes, resulting in widespread roche moutonée (see roche moutonée provide protection from wind and the
Klasner and Fagre, 2002, for an in-depth description sites occupied by ribbon forests.

Fig. 7. Ground photographs taken while conducting fieldwork and soils collection on Flattop Mountain, 6 August 2001. (a) Person stands in
lower center in meadow, flanked by parallel ribbons on other side. (b) Close-up of meadow (left)/ribbon (right) contact, note the topographically
elevated position of the ribbon forest. (c) Soil sampling in meadow between two subparallel ribbon forests. Note the topographically elevated
positions of the ribbon forests in comparison to the meadow.
 D.R. Butler et al. / Geomorphology 55 (2003) 203–217 213

A somewhat similar situation developed in Preston axis of the syncline in a N – NE to S –SW direction.
Park (Fig. 3). There, Pleistocene glaciation moved out This interaction of geologic structure with topogra-
of the cirque to the upper right of Fig. 3, and flowed to phy, in turn interacting with glacial scouring along
the SW across strata dipping 10j in the same direction the strike of bedding planes, is further complicated
(Whipple, 1992). This coincidence of flow direction by fluvial incision across the surface of the syncli-
and dip produced deeply plucked roche moutonée. nal peak. The net result is the pattern illustrated in
Ribbon forests run along the base of these features, Fig. 6a, the same area illustrated as in the oblique
but their spatial pattern is disrupted by spring snow- aerial photo shown in Billings (1969). On Fig. 6a,
melt and permanent stream channels that have incised one can trace the direction of the bedding planes,
deeply across the floor of the park. observe the glacial scouring along them (note a
East of Red Eagle Glacier (Fig. 4), Pleistocene small ‘‘finger lake’’ near the bottom center of the
scouring and plucking occurred roughly perpendicular photo), and observe the disruption to ribbons by
to the strike of bedding planes, but a nearby dip wandering streams.
measurement (Whipple, 1992) suggests that the dip On nearby Flattop Mountain, similar conditions
here was up-ice rather than down-ice as at Preston exist except that this synclinal peak is more sym-
Park. This site is further complicated by complex metrical than West Flattop, with Akamina Syncline
folding of the bedrock into a chevron pattern visible forming the long axis directly down the middle of the
on Fig. 4 and by stream incision that disrupts the mountain (Figs. 5 and 6b). Ribbon forests occupy
continuity of the ribbon forests. bedrock ridges (Fig. 6b) that are a result of glacial
Because of the significance of the photograph of scouring along the strike of inwardly dipping (toward
West Flattop Mountain in Billings’ (1969) paper the center of the synclinal structure) bedding planes
and the clear continuation of the patterns of ribbon and lithologic contacts.
forest there on the adjacent Flattop Mountain to the In the field, the topographically elevated posi-
SE, we focused our attention on these two sites tions of the ribbon forests are striking (Fig. 7).
(Figs. 5 and 6). West Flattop Mountain is no longer Minor conifer encroachment into the adjacent, lower
accessible by trail, so our interpretations there are meadows is apparently hindered by the presence of
based on airphoto interpretation and viewing from a late-lying snow; we observed black ‘‘snow fungus’’
helicopter overflight (approximately 300 m above on the tips of nearly every seedling and sapling
the ground). Although we did not perform a along the margins of the meadow shown in Fig. 7.
detailed change detection of ribbon forest patterns In this sense, Billings (1969) was correct in stating
there, a comparison of forest positions on 1966 and that the areas between the ribbon forests are snow
1995 panchromatic and 1975 color-infrared aerial glades, and the presence of the late-lying snow that
photographs shows no movement of ribbons such as accumulates in these topographic depressions hin-
would occur according to the theory of development ders successful conifer invasion under current
postulated by Billings (1969). We observed topo-
graphically elevated, parallel to subparallel ridges
occupied by ribbon forests, separated by intervening Table 2
low-lying meadows (Billings’ snow glades). The Soil characteristics, Flattop Mountain meadows and ridgesa
topographically elevated ridges follow the strike of Organic matter Particle size Munsell color
bedding planes there. The fact that the prevailing Meadows
wind here is westerly is sheer coincidence; the 20 – 22% LOI 37 – 73% Sand, 7.5 or 10 YR 3/2
ribbon forests follow the strike of the bedding 13 – 23% Silt,
planes and the direction of glacial scouring in the 14 – 40% Clay
Ridges
same direction. The situation is made more complex 30 – 48% LOI 13 – 20% Sand, 5 YR 2.5/2
by the asymmetry of West Flattop Mountain, where 7 – 17% Silt,
the Akamina Syncline passes across its NE quad- 63 – 80% Clay
rant, in combination with the location of the Con- a
Ranges for meadows and ridges based on two soil samples
tinental Divide running roughly perpendicular to the each; soil samples taken from depth of 0 – 20 cm.
214 D.R. Butler et al. / Geomorphology 55 (2003) 203–217

Fig. 8. (A) Aerial photograph of the southern one-half of Flattop Mountain, in vicinity where soil sampling and field reconnaissance occurred.
The eastern one-half of this view is recognizable in Fig. 7b. (B) Vegetation change detection matrix of the eastern portion of the southern one-
half of Flattop Mountain, for period from 1966 to 1991. Approximately one-half of the study site experienced no change during this period, and
the other one-half changed in approximately equal proportions (about one-half increasing and one-half decreasing). No directionality in change
was discerned.
 D.R. Butler et al. / Geomorphology 55 (2003) 203–217 215

climatic conditions. However, although the presence 6. Discussion

of the snow glades is spatially related to the
presence of windward ribbon forests, glade origin Similar to our change detection results on Flattop
is due to the topographic differences between ridges Mountain, Klasner and Fagre (2002) observed no
and intervening hollows; and those topographic detectable change in the position of ribbon forests
differences resulted from the interaction of Pleisto- along the fronts of the roche moutonée at Logan Pass
cene glaciation with the structural geology and during the 70 years between 1927 and 1997. They did
stratigraphy of Flattop Mountain. note thickening of the forest and upward extension of
Soil conditions at our field site on Flattop Mountain leader growth. At other treeline sites in GNP, where
(Fig. 7), although based on limited sampling, never- early twentieth century ground photographs were
theless emphasize the different edaphic conditions on replicated in the early 1990s, similar patterns of
ridge and in glade (Table 2). The ribbon forest soils non-movement of trees in association with in situ
were higher in organic matter content, and their thickening of tree patches and upward growth were
particle size distribution reflects the argillitic nature revealed (Butler et al., 1994). Visual, nonquantitative
of the underlying Mount Shields Formation (Whipple, comparisons of aerial photographs from 1966 to 1991
1992). Coarser materials in the snow glade (Table 2) or 1995 from other ribbon forest sites similarly
appear similar to what Carrara (1990) mapped as revealed no observable change in the positions of
‘‘alluvial and colluvial deposit (Holocene and late ribbon forests or adjacent meadows/snow glades.
Pleistocene)’’ and are locally derived from the adjacent Under changing climates, we can conceptualize
bedrock ridges washing down into the adjacent glades. how ribbon forests become established and how they
These coarser glade soils would be, in turn, more may displace meadows and how meadows may dis-
susceptible to summer drying due to their coarser place ribbon forests. These conceptualizations are
nature as well as the absence of shade in comparison predicated on the lithologic and geomorphic condi-
to the adjacent ribbon forests. The net result is a system tions prevailing in GNP at sites such as West Flattop
that enhances the unlikeliness of tree invasion from Mountain, Flattop Mountain, the peak west of Camas
bedrock ridges into the adjacent meadow depressions Lake, and Granite Park, where bedrock ridges support
under prevailing climatic conditions. ribbon forests and meadows occupy adjacent topo-
graphic depressions (Fig. 9). Glaciation along the
5.2. Change detection on Flattop Mountain strike of bedding planes produces the ridge-and-
trough morphology in the uppermost section of Fig.
The change detection performed on the SE quad- 9. Topography thus defines establishment site con-
rant of Flattop Mountain (Fig. 8) showed no evi- ditions (ESC) for trees, and a treeless surface looks the
dence for ribbon movement and sequential change as same as the topography (uppermost Fig. 9). As
described by Billings (1969); rather, it revealed climate change occurs in a direction that improves
widespread areas of stability of forest ribbons and ESC for trees, the ESC surface lowers and intersects
meadows. Some areas of tree-to-meadow change the ridges, at which point trees begin to occupy the
were encountered, but this was interpreted as result- topographic ridges (Fig. 9, second row). A feedback
ing from mortality due to pine bark beetle blight. The develops, in which glades then do collect more snow,
change detection also revealed some change from increasing the difference between the ESC for ridges
meadow to tree seedlings/saplings, perhaps a reflec- and glades, which tends to stabilize the pattern. If
tion of a diminishing impact of late-lying snow in climate change moves toward a drier state, decreased
some meadows. Snowpack in GNP was below aver- snow retention in the glades allow trees to migrate
age many years in the 1990s, so it is possible that down into the glades (rows 3 and 4, Fig. 9). The
future changes will occur as trees are more success- existence of an ESC maximum is theoretically possi-
ful in invading the topographic depressions currently ble beyond which ridgetops become too dry for trees,
occupied by meadows (see Discussion). We see no resulting in a ribbon-and-meadow pattern that is topo-
suggestion, however, that meadows are attempting to graphically reversed from current conditions, i.e.
displace trees on the bedrock ridges. meadows would occupy the drier ridgetops. Ribbon
216 D.R. Butler et al. / Geomorphology 55 (2003) 203–217

forests would occupy the former glades, sustained by 7. Conclusions

runoff from the surrounding ridges (bottom row, Fig.
9). Notably, however, these conditions currently do Ribbon forests in GNP are stable, well-established
not prevail anywhere in our study area in GNP nor has features that are roughly linear because of the effects of
movement toward such a condition been identified. current topography, past and present geomorphic pro-

Fig. 9. Hypothesized establishment site conditions relationship with topography and ribbon forest development. See text for further explanation.
 D.R. Butler et al. / Geomorphology 55 (2003) 203–217 217

cesses, and conditions of bedrock lithology and struc- Bryan, R.B., Brun, S.E., 1999. Laboratory experiments on sequen-
ture. No evidence exists to suggest that ribbon forests tial scour/deposition and their application to the development of
banded vegetation. Catena 37, 147 – 163.
there are migrating in the fashion Billings (1969) Buckner, D.L., 1977. Ribbon forest development and maintenance
described, nor is there a need to invoke positive feed- in the central Rocky Mountains of Colorado. PhD Dissertation,
back or complexity. The association of ribbon forest University of Colorado, Boulder.
with adjacent topographically low positions where Butler, D.R., Malanson, G.P., Cairns, D.M., 1994. Stability of al-
pine treeline in Glacier National Park, Montana, U.S.A. Phyto-
snow accumulates and lingers, thus inhibiting tree
coenologia 22, 485 – 500.
growth, does present an interesting spatial pattern that Carrara, P.C., 1990. Surficial geologic map of Glacier National
may be sensitive to climatic change. Ecological re- Park, Montana. U.S. Geological Survey Miscellaneous Investi-
sponses to climatic change in this location are best gations Series, Map I-1508-D, Washington, DC.
considered in light of their biogeomorphic connections. Doering, W.R., Reider, R.G., 1992. Soils of Cinnabar Park, Medicine
A need for further research clearly exists, especially a Bow Mountains, Wyoming, U.S.A.: indicators of park origin and
persistence. Arctic, Antarctic, and Alpine Research 24, 27 – 39.
more detailed, comprehensive survey of the character- Dunkerley, D.L., Brown, K.J., 1999. Banded vegetation near Bro-
istics and causes of ribbon forests throughout western ken Hill, Australia: significance of surface roughness and soil
North America, because previous studies have focused physical properties. Catena 37, 75 – 88.
on ribbon forests in only one location and then have Finklin, A.I., 1986. A climatic handbook for Glacier National
extrapolated a proposed origin from that site to all Park—with Data for Waterton Lakes National Park. U.S. Forest
Service General Technical Report INT-204, Intermountain Re-
ribbon forests. Finally, any such future studies should search Station, Ogden, UT.
first examine the on-site geomorphic and geologic con- Holtmeier, F.-K., 1982. ‘‘Ribbon Forest’’ und ‘‘Hecken’’. Erdkunde
ditions before postulating complex ecological scenar- 36, 142 – 153.
ios for explaining the spatial distribution of ribbon Jensen, J.R., 1996. Introductory Image Processing: A Remote Sens-
ing Perspective, 2nd ed. Prentice Hall, Upper Saddle River, NJ.
forest.
Klasner, F.L., Fagre, D.B., 2002. A half century of change in alpine
treeline patterns at Glacier National Park, Montana, U.S.A. Arc-
tic, Antarctic, and Alpine Research 34, 49 – 56.
Acknowledgements
Malanson, G.P., 1999. Considering complexity. Annals Association
of American Geographers 89, 746 – 753.
This research was funded by a grant to DRB and McGregor, S.J., 1998. An integrated geographic information system
GPM from the U.S. Geological Survey, Biological approach for modeling the suitability of conifer habitat in an
Resources Division, in cooperation with its Glacier alpine environment. Geomorphology 21, 265 – 280.
Mueller, P.W., Hoffer, R.N., 1989. Low-pass spatial filtering of
Field Station and Global Change Research Coordina-
satellite radar data. Photogrammetric Engineering and Remote
tor, Dr. Dan Fagre. Chelsea Summer Kelley provided Sensing 55, 887 – 895.
field assistance. We thank Stephen J. Walsh, Richard A. Ross, C.P., 1959. Geology of Glacier National Park and the Flat-
Marston, and an anonymous reviewer for their very head Region, Northwestern Montana. U.S. Geological Survey
helpful comments on an earlier draft of the paper. This Professional Paper, vol. 296. Washington, DC.
Sohl, T.L., 1999. Change analysis in the United Arab Emirates, an
paper is a contribution of the Mountain GeoDynamics
investigation of techniques. Photogrammetric Engineering and
Research Group. Remote Sensing 65, 475 – 484.
Vale, T.R., 1978. Tree invasion of Cinnabar Park in Wyoming.
American Midland Naturalist 100, 277 – 284.
References Whipple, J.W., 1992. Geologic map of Glacier National Park, Mon-
tana. U.S. Geological Survey Miscellaneous Investigations
Allen, T.R., Walsh, S.J., 1993. Characterizing multitemporal alpine Series Map I-1508-F, Washington, DC.
snowmelt pattern for ecological inferences. Photogrammetric White, L.P., 1969. Vegetation arcs in Jordan. Journal of Ecology 57,
Engineering and Remote Sensing 59, 1512 – 1529. 5461 – 5464.
Baker, W.L., Honaker, J.J., Weisberg, P.J., 1995. Using aerial White, L.P., 1970. ‘Brousse tigré’ patterns in southern Niger. Jour-
photography and GIS to map the forest-tundra ecotone in nal of Ecology 58, 549 – 553.
Rocky Mountain National Park, Colorado, for global change White, L.P., 1971. Vegetation stripes on sheet wash surfaces. Jour-
research. Photogrammetric Engineering and Remote Sensing nal of Ecology 59, 615 – 622.
61, 313 – 320. Zonneveld, I.S., 1999. A geomorphological based banded (‘tiger’)
Billings, W.D., 1969. Vegetational pattern near alpine timberline as vegetation pattern related to former dune fields in Sokoto (north-
affected by fire – snowdrift interactions. Vegetatio 19, 192 – 207. ern Nigeria). Catena 37, 45 – 56.
 Geomorphology 55 (2003) 219 – 234
www.elsevier.com/locate/geomorph

Zoogeomorphology in the Alpine: some observations on

abiotic–biotic interactions
Kevin Hall *, Neil Lamont
Geography Program, University of Northern British Columbia, 3333 University Way, Prince George, BC, Canada V2N 4Z9
Received 9 January 2002; received in revised form 31 May 2002; accepted 10 March 2003

Abstract

Cryonival processes typify the alpine environment. These cold-based processes operate synergistically with wind,
rainsplash, surface wash, and mass movement to create the local morphology. These traditionally accepted abiotic processes are
rarely, if ever, identified as operating in conjunction with biotic processes. Paradoxically, the existence and activity of animals
within the alpine zone is widely reported. Here the argument will be made for the interoperation of the biotic with the abiotic in
the development of the alpine terrain.
Animals frequently exert a strong geomorphic influence in alpine environments through such activities as grazing,
trampling, digging, burrowing, and direct erosion of even bedrock. The impact of the animals, in a geomorphic context, can be
both direct and indirect. Direct effects are soil compaction, removal of sediments, loading causing slope failure, and the
introduction or removal of chemicals (nutrients). As significant as these impacts can be, so too are the indirect effects:
preparation of sediments for removal by abiotic processes, decreasing of slope stability by burrowing, increasing surface wash
and/or concentration of overland flow as a result of compaction, pedoturbation, and changing of soil-water chemistry because of
facilitating water penetration as a result of burrowing. The abiotic and biotic impacts occur both in parallel (i.e., at the same
time) and/or in series (i.e., one subsequent to another), frequently as a function of seasonal climatic influences on both forces.
Climate exerts a significant impact on both biotic and abiotic processes, and aspect can greatly influence these. Solar
radiation, ground temperature, wind, and precipitation affect biotic and abiotic attributes and their interaction. Slope angle and
parent material(s) will also affect the degree of influence. Biotic – abiotic interactions are complex outcomes of parent material,
aspect, slope, and season, and all of the factors that these influence (i.e., vegetation).
Here a series of flow diagrams are used to identify both abiotic and biotic elements and to show the manner in which they
interoperate to create the local alpine terrain. Examples from a variety of alpine locations (Canada and China) will be used to
exemplify geomorphic outcomes.
D 2003 Elsevier Science B.V. All rights reserved.

Keywords: Alpine geomorphology; Zoogeomorphology; Biotic – abiotic interactions; Burrowing; Climate; Aspect

1. Introduction

The alpine zone is characterized by cold, snowy

* Corresponding author. winters and warm, sometimes wet, summers with
E-mail address: [email protected] (K. Hall). abundant flora and fauna. The alpine is a dynamic

0169-555X/03/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0169-555X(03)00141-7
220 K. Hall, N. Lamont / Geomorphology 55 (2003) 219–234

environment affected by a seasonally varying range of number of common elements are characteristic of
geomorphic processes. Simplistically, land-forming any alpine environment. With increasing demands of
processes are identified as being associated with sustainable agriculture and of tourism, it is becoming
freezing conditions and the impact of snow during imperative that an understanding of these interactions
the winter coupled with mass movement and fluvial be gained in order to better recognize the human-
activity during the summer. Although alpine meadows influenced impacts. In an attempt to describe the
abound, the broader scene encompasses rock expo- interactions between the abiotic and biotic factors,
sures, cliffs and steep slopes, vegetated slopes, and and their positive and negative feedbacks, a series of
other slopes consisting of active scree. The alpine simplified flow diagrams (readily modifiable for any
zone is also recognized for its abundance and diversity local situation) is used. Where appropriate, specific
of wildlife: from various species of ungulates, includ- examples will be presented from zoogeomorphic
ing mountain goats, sheep and deer; various species of studies by the authors in the alpine regions of Canada
bear; ground burrowing mammals such as marmots, and China.
pika, pocket gophers, and ground squirrels, and large
quantities of insects and birds. Despite the wide
diversity of animals associated with alpine regions, 2. Study area and methods
little recognition has been given to the interaction of
the biotic with the abiotic in these dynamic and Data provided are from a preliminary study on the
relatively sensitive zones. effects of burrowing and digging activities in the
Studies related to zoogeomorphic impacts in the MacGregor Ranges of the Canadian Rockies
alpine zone have focused on the role of burrowing (120j50VW, 54j14VN) and is situated above the local
animals (Grinnel, 1923; Ellison, 1946; Price, 1971; treeline in the alpine zone. The regional geology is
Thorn, 1978, 1982; Smith and Gardner, 1985; Marti- one of folded and faulted sedimentary and metasedi-
nez Rica and Pardo Ara, 1990; Hall et al., 1999), mentary rocks, mostly Paleozoic in age, composed
digging activities by grizzly bears (Holcroft and primarily of limestone, quartzite, schist, and shales
Herrero, 1984; Butler, 1992; Hall et al., 1999), and (Valentine et al., 1978). Data were collected on the
the effects of trampling and grazing by hoofed ani- east-facing and west-facing slopes in early August
mals on trails and slopes (Summer, 1980, 1986; Perez, 2000 and on the north-facing and south-facing slopes
1993; Watanabe, 1994; Cole, 1995; Whinam and in early August 2001. During a previous study on
Comfort, 1995; Deluca et al., 1998). Generally, the burrowing and digging impacts in this area, Hall et al.
focus of research is to demonstrate the effectiveness of (1999) provided estimates of sediment removal for
zoogeomorphic activity by providing information on burrowing and digging activities for the east-facing
the specific impact (i.e., the amount of sediment aspect only.
removed by burrowing or the degree of trail degrada-
tion by trampling) and by comparing this information 2.1. North-facing aspect
with other geomorphic processes such as mass move-
ments and fluvial erosion. Few papers consider in any The majority of the north-facing aspect is a steep,
detail the combined influence of climate (precipitation bedrock headwall rising above a cirque lake. The
and wind), topography (aspect and slope), and eco- slope consists of exposed rock faces and scree and
logical factors (vegetation and soil properties) on boulder fields. The slope ranges in elevation from
animal activity in the alpine zone. 1650 to 1826 m. The lower extent of the boulder field
The aim of this discussion is to consider the is covered with a thin colluvial soil and a blanket of
interaction between vegetation, aspect, slope, climate, pink mountain heather (Phyllodoce empetriformis)
soil properties, and the geomorphic role of animals in and the upper ridge is exposed to the prevailing wind,
the development of the alpine terrain. Abiotic and with a thin soil over bedrock. The average slope angle
biotic interactions are highly complex and commonly is 15.5j and ranges from 7j to 30j (but also includes
vary as a function of geographic location. Although some sheer faces), with gentler slopes dominating the
abiotic – biotic interactions must be site-specific, a upper ridge and lower boulder fields.
 K. Hall, N. Lamont / Geomorphology 55 (2003) 219–234 221

2.2. East-facing aspect bottom-right of each 25-m2 quadrat. The volume of

each burrow mound is calculated based on the cross
The east-facing aspect ranges in elevation from section of a wedge (for details, see Thorp, 1949, p.
1706 to 1875 m, with the slope consisting of concave 205) and for bear diggings the volume of sediment is
and convex sections with short, straight, slope breaks calculated based on the cross section of ellipsoid.
in between. The varied topography results in slope Particle size analyses were done using the pipet
angles ranging from 11j to 41j, with an average slope method (see Gee and Bauder, 1986 for details).
angle of 28.4j. This aspect is on the leeward side to Standard statistical procedures were used throughout
the prevailing winds and experiences late-lying snow- the analyses. Analysis of variance (ANOVA) was
patches that provide moisture long into the growing applied to measures of gradient and particle size
season. analysis, and Fisher’s least significant difference test
(Fisher’s LSD) to determine if differences occurred
2.3. South-facing aspect between aspects at the 5% confidence level.

The south-facing aspect can be divided into an

upper and lower section; the upper section is com- 3. Climatic considerations
posed of steep, exposed bedrock ridges with scree
fans at their base. The lower section consists of well- Climatic factors play an extremely important role
drained diamict deposits that are initially steep and in influencing both animal distributions and geo-
concave in form, which transform into straight, deeply morphic processes within the alpine (Fig. 1). Al-
dissected, gentle slopes. The elevation ranges from though consideration will first be given to the
1750 to 1816 m, with an average slope angle of 12.8j climatic factors, Fig. 1 indicates the framework
and a range from 1j to 39j. Steeper slope angles are within which alpine attributes interact to influence
associated with the well-drained moraine deposits and zoogeomorphic activity. Climatic factors cannot be
the stream banks associated with the deeply incised considered in any singular direction because they
channels. directly influence vegetation distributions both spa-
tially and temporally, which in turn affects animal
2.4. West-facing aspect species diversity and distribution. Further, those same
climatic attributes will also be important in determining
The west-facing aspect is the most homogeneous in geomorphic processes; but here, consideration must
form of all the slopes. It ranges in elevation from 1694 also be given to parent material, moisture content, and
to 1800 m, with a mean slope angle of 17.4j and slope angle in determining outcomes. The key is that
range from 9j at the bottom of the slope, to 27j near the climatic parameters are interacting with both the
the crest of the slope. The slope is straight to slightly biotic and the abiotic, which are, in turn, affecting each
concave in form and is dissected by troughs from other.
inactive and active ploughing blocks. Climatic attributes considered important in the
alpine, and how these attributes interact with the
2.5. Methods abiotic/biotic components of the landscape, are shown
in Fig. 2. Aspect, slope angle, and altitude play
Each aspect was sampled using randomly distrib- significant roles in determining the local climate,
uted 5  5 m quadrats. Aspect, gradient, percent which influences animals and their geomorphic
ground cover of vegetation and bare ground/rock, impact. A slope’s orientation determines the degree
and estimates of the volume of burrowing mounds of influence solar radiation and wind has in modifying
and/or bear digs were recorded for each quadrat. Soil the air temperature. Furthermore, air temperature
samples were taken at a 0- to 10-cm depth and later decreases with increasing altitude, thus promoting
analyzed to determine the particle size distribution. precipitation in the form of snow and its longevity
A 1-m2 plot was used to estimate percent ground in the landscape. Slope angle further influences the
cover, which was sampled in the top-left, center, and intensity of solar radiation received on a given
222 K. Hall, N. Lamont / Geomorphology 55 (2003) 219–234

Fig. 1. Generalized climatic considerations for abiotic – biotic interactions.

aspect—the lower the slope angle, the higher the will have an effect on the abiotic and biotic compo-
angle of incidence—resulting in a greater reception nents of the landscape.
of solar radiation. For example, north-facing aspects The impacts of ground temperatures on abiotic and
with gentle slope angles can, during the growing biotic processes are presented in more detail in Fig. 3,
season, receive comparable amounts of solar radiation where the relationship of the key climatic parameters
to a south-facing aspect (Davidson, 1992). Although to ground freezing is shown. Ground freezing affects
air temperatures are lower at higher elevations, on the plant distributions by reducing moisture availability
microscale, the absorption of intense solar radiation and by discouraging seedling establishment as a result
by rocks, soils, and plants creates warmer near-ground of frost processes, such as needle ice (Bennet, 1976;
air temperatures, which provide better growing con- Campbell, 1997). This, in turn, influences animal
ditions for alpine plants (Bliss, 1988). Thus, at any activity—the richer the flora the better the food
specific site, altitude, slope angle, aspect, and albedo resource plus, for some smaller animals, it also
will influence the ground temperature, which in turn provides protection from predators (Reichel, 1986).
 K. Hall, N. Lamont / Geomorphology 55 (2003) 219–234 223

Fig. 2. Expanded version of climatic considerations to illustrate the complexity of interactions between the abiotic and biotic.

Thus, where plant-food resources are low, animal altitudinal/latitudinal impact upon precipitation occur-
activity is minimal, resulting in a negligible zoogeo- rence can, in conjunction with the impacts of wind
morphic impact. Conversely, the typical species-rich and temperature, significantly influence soil moisture
alpine meadow offers an ideal location for extensive distribution and availability (Fig. 5). Soil moisture
animal activity and, ultimately, a significant zoogeo- components are the controlling factors for growth and
morphic impact. production of alpine plant species, as well as being
In the alpine, wind (Fig. 4) and precipitation (Fig. determinants of community patterns (Campbell,
5) can both have a significant impact on plant and 1997). Too much snow (unavailable moisture and
animal distributions, as well as on subsequent geo- covered ground) in the upper part of the alpine can
morphic activity. Wind can play a major role in be as detrimental to plant growth as situations where
determining plant species distribution (Bliss, 1988), evaporation exceeds precipitation in some wind-
thus having an influence on the food resources exposed lower parts of the alpine. The role of snow,
available to the animals (Fig. 4). Equally, precipita- a recognized element in the alpine, is critical. While
tion may play a role in several ways (Fig. 5). The snowbanks can be highly beneficial in releasing water
224 K. Hall, N. Lamont / Geomorphology 55 (2003) 219–234

Fig. 3. The impact of freezing on animal action.

during otherwise dry periods, too extensive a snow thermal blanket under which much biological activity
cover will be detrimental to plant growth despite the takes place during the winter (Marchand, 1996). For
protection it may offer from ground freezing. Thorn many small mammals, the presence of an adequate
(1978) suggests that areas of prolonged snow cover, snow cover is critically important to their overwinter-
that melt out late in the summer, preclude vegetation ing success. Young (1990), in southwestern Alberta,
and therefore are unattractive to burrowing animals. compared the location of Columbian ground squirrel
Thus, plant species diversity (and the consequent hibernacula based on soil temperature, soil moisture,
animal activity) is intimately linked with the type and ground cover, and suggested that the position and
and temporal distribution of precipitation at any given depth of hibernacula may reflect a preference for a
site (Cox, 1933; Billings and Bliss, 1959; Thorn, temperature regime during hibernation. Slopes with
1978). colder soil temperatures regimes related to grass
The presence of snow can also influence the cover or little snow had deeper hibernacula when
location of hibernacula for both burrowing animals compared to slopes with deep snow covers, and
and grizzly bear. Snow acts as both a reflective cover steep, snow free slopes were avoided (Young,
against the warming effects of solar radiation and as a 1990). Grizzly bears also demonstrate an aspect
 K. Hall, N. Lamont / Geomorphology 55 (2003) 219–234 225

Fig. 4. The influence of wind on animal action.

preference when constructing winter dens, preferring burrows that go to the lower substrate) plus offers
sites located on steep slopes with deep snow cover; some thermal protection for hibernating species. Thus,
Vroom et al. (1980) suggests that den sites are perhaps paradoxically, the areas of more ground-
located on leeward slopes away from the prevailing protective vegetation are sites of more zoogeomorphic
winds. disturbance (Thorn, 1978) than are the less protected,
vegetation-deficient sites. Within all of this, parent
material must also exert a significant influence (Fig.
4. Vegetation and parent material 7). Clearly sites of steep bedrock, not easily weath-
ered, will be plant-deficient whatever the positive
The vegetation that results from the above factors attributes of the other climatic factors; however,
(Figs. 2 –5) is clearly very important in influencing impacts can occur even on bedrock or scree slopes
animal activity (Fig. 6). As stated before, plants during animal movement across them. Field experi-
provide both a food resource and some measure of ments on the transport of rock fragments on scree
protection from predators. Further, where it is abun- slopes by animal trampling by Govers and Poesen
dant, plant growth tends to provide a deeper organic (1998) suggest that animal trampling may have a
layer more suitable for burrowing (or at least initiating significant impact on the evolution of scree slopes
226 K. Hall, N. Lamont / Geomorphology 55 (2003) 219–234

Fig. 5. The influence of precipitation on animal action.

in areas of high grazing pressure and that trampling is parent material can have a strong influence on soil
most efficient in causing vertical sorting of fine rock formation during the initial stages of development and
fragments compared to downslope sorting. Parent upon the resulting soil texture, if other soil-forming
material, through its influence on soil properties, plays factors are considered equal. Soil texture influences the
a role in helping determine plant species and abun- rate and depth of leaching and the thermal properties of
dance, the ability of animals to burrow (Hansen and the soil; in general, coarse-textured soils retain less
Morris, 1968), plus the subsequent geomorphic water, are drained to greater depths, and warm more
impacts. Soil formation in alpine areas can be consid- quickly than fine-textured soils (Foth, 1990; Birkeland,
ered as modest due to the relatively young age of 1999). The occurrence of coarse-textured soils may
landforms, the extreme climate, the reduction of bio- influence burrow location because burrows function as
logical activity, and by the frequent interruption of favorable micro-sites against the harsh alpine environ-
geomorphic processes (Valentine et al., 1978; Benis- ment (Woods, 1980; Young, 1990), therefore sites that
ton, 2000). Although factors such as climate, topog- are easily drained and warm quicker are likely to be
raphy, and vegetation will exert a strong local effect preferred over areas of poor drainage and cold soils.
upon soil development, Birkeland (1999) states that The cohesiveness of the soil may be another important
 K. Hall, N. Lamont / Geomorphology 55 (2003) 219–234 227

Fig. 6. The influence of vegetation on animal action.

factor with respect to burrow location; for example, of alpine vegetation. Grizzly bears utilize a variety of
Onda and Itakura (1997) in an experimental study on food sources, such as starchy plants, small mammals,
the burrowing activity of sand crabs observed that the and insects (Hamer and Herrero, 1987). Grizzly bears
sand crabs in fine sands would cease burrow construc- usually dig for cow parsnip (Hedysarum sulpheres-
tion due to the noncohesive nature of the sand causing cens) and glacier lily (Erythronium grandiflorum) on
the burrow to collapse. south aspects, at high elevations, open and dry mead-
Many animals burrow for breeding/hibernation as ows, and on inclined slopes (Hamer and Herrero,
well as for predator evasion; conversely, some pred- 1983). Other studies have reported diggings for Hedy-
ators (e.g., grizzly bears) dig to acquire food and to sarum on northern exposures. McCrory and Herrero
construct winter dens. Food palatability and the par- (1981) and Hamer and Herrero (1983) observed dig-
ticle size distribution of a soil may both influence the gings in late spring when Hedysarum was in the
occurrence of grizzly bear diggings for various types earlier growth stage at which time grizzly bears prefer
228 K. Hall, N. Lamont / Geomorphology 55 (2003) 219–234

Fig. 7. The influence of particle size distribution, as inherited from parent material, on animal action.

to eat the roots, whereas on the south aspect, the abundance of plant species. The disturbance caused
Hedysarum was already beginning to flower, thus by grizzly bear digging for plants may also influence
being less palatable. plant distribution. Tardiff and Stanford (1998) observe
Relationships other than aspect and time of year that when digging for glacial lily bulbs, grizzly bears
have been made with respect to foraging preference removed the sod to a fairly consistent depth of 10 cm,
by grizzly bears. Holcroft and Herrero (1984) ob- which is related to the rooting depth, thereby exposing
served that digging for Hedysarum roots was often the bulbs that are then eaten. After digging, the pieces
concentrated in certain areas even though the plants of sod are left intact with the rest of the plant, which
were widely distributed. Bear diggings for bulbs were can then resprout in the following year. Disturbance of
usually on sandy loam soils, loose gravel, and on the sod influences the regrowth of glacial lilies, thus
scree. Factors such as loose soil structure, poorly increasing soil mineral nitrogen by exposing the soil
developed sod, the presence of scree and moist to to more wetting and drying events, improving soil
wet soils led Holcroft and Herrero (1984) to conclude structure and the removal of plants and litter (Tardiff
that ease of digging was more important than the and Stanford, 1998).
 K. Hall, N. Lamont / Geomorphology 55 (2003) 219–234 229

The preceding discussion has suggested that the Table 2

greatest zoogeomorphic impacts in an alpine area are Average percent ground cover stratified by aspect, August 2000/
2001
constrained to areas of optimal living conditions and/
North East South West
or feeding habitat. In the Canadian Rockies research
site, the two most commonly observed burrowing Ground cover (%): herbs/ 10 66 36 34
grasses/palatable vegetation
animals were the Columbian ground squirrel (Sper-
Ground cover (%): 39 12 25 36
rmophilus columbianus) and the hoary marmot (Mar- dwarf-shrubs/cushion plant/
rmota caligata). Both species can be considered as unpalatable vegetation
herbaceous grazers feeding on a variety of grasses and Ground cover (%): bare ground 51 22 39 30
forbs, feeding on the roots and stalks of plants in the and rock
Number of observed plant species 8 23 19 10
spring, and then switching to leaves and fresh shoots;
late in the season, they concentrate on flowers and
berries (Banfield, 1974; Woods, 1980). Research on the west) and pink mountain heather (P. empetrifor-
burrowing animals from the Canadian Rockies (Hall mis; 39% on the north and 30% on the west), which is
et al., 1999) suggests that burrowing is significant on a low-lying shrub well adapted to wind-exposed sites
slopes oriented northeast to southeast, which is similar with dry, cold, or frozen soils (Kershaw et al., 1998).
to the preference of burrowing activity on a southeast The low diversity of plant species and the dominance
orientation suggested by Price (1971). However, both of bare ground may reflect the limited amount of
of these studies only made observations on slopes burrowing observed on the north and west aspects.
where burrowing activity was noticeable and did not Dyni and Yensen (1996) suggest that the Columbian
report any findings from other aspects in their study ground squirrel inhabit areas with more luxuriant
areas. Data collected during the 2000/20001 field vegetation and avoid areas with dry and shallow soils
seasons in the Canadian Rockies suggest that burrow- that may not provide enough protection from digging
ing activity is strongly influenced by aspect character- grizzly bears. Areas of preferred vegetation in high
istics. In this area, the north- and west-facing slopes densities provide the Columbian ground squirrel with
have the lowest volumes of sediment removed by a high-energy food source that can be foraged in an
burrowing animals (Table 1). The north aspect has efficient and safe manner (Dyni and Yensen, 1996).
shallow soils, late-lying snow patches, and an upper In contrast, the east- and south-facing aspects
ridge that is often windswept. The latter characteristic experience a significantly greater impact by burrow-
is shared with the west aspect, which faces the strong ing animals (Table 1). The east-facing aspect and, for
prevailing westerly winds. Due to the corresponding the most part, the south-facing aspect are in the lee of
climatic influences on these two aspects, similar plant the prevailing winds, allowing deeper snow packs and
species were observed. Both slopes have relatively a wind-protected environment that promotes the
low plant species diversity (Table 2) and are domi- growth of tall herbaceous plants (Bennet, 1976).
nated by bare ground (51% for the north and 30% for Due to earlier melting snow and higher evaporation
rates in the summer, the south-facing aspect has a
Table 1 varied ground cover (Table 2), with herb and grasses
Estimate of total volume of sediment displaced stratified by aspect, (36% of the mean ground cover) on wetter microsites,
August 2000/2001 and pink and white mountain heather (25% of the
North East South West mean ground cover) related to dry, well-drained areas.
Sediment displaced by 7 186 95 11 As previously stated, burrowing animals in alpine
burrowing animals areas graze on a wide variety of herbs and grasses;
(m3/km2/year) the higher level of burrowing activity on the east-
Sediment displaced by n.d. 400 532 n.d.
Grizzly bear (m3/km2/year)
facing aspect maybe related to its dominant cover of
Total sediment displaced 7 586 627 11 herbaceous plants (66% of the mean ground cover).
(m3/km2/year) Results from the particle size analysis (Table 3)
Plots sampled 19 43 46 20 show a sandy loam texture for all four of the aspects
n.d. = no data. sampled. The coarse fraction (>2 mm) was found to
230 K. Hall, N. Lamont / Geomorphology 55 (2003) 219–234

Table 3 digging, and trampling. Burrowing and digging as

Particle size analysis (0 – 10 cm depth) stratified by aspect; mean animal actions are differentiated, although the
values are percent of total weight, August 2000/2001
impacts may have significant similarities. Animals
North East South West
burrow to create living spaces (for breeding, hiber-
Fraction greater 34.3 (6.7) 44.7 (16.8) 31.4 (9.5) 28.9a (12.4)
a
nation, prey avoidance, etc.), while digging is often
than 2 mm
to excavate these small mammals as a food source
Fraction less 65.7 (6.7) 55.3 (16.8) 68.6a (9.5) 71.1a (12.4)
than 2 mm for larger mammals; digging can also be for plant
% Sand 60.6 (3.9) 59.2 (7.8) 63.4 (8.8) 63.4 (10.9) roots or tubers for food as well. Both burrowing and
% Silt 31.1 (4.2) 32.3 (6.6) 29.5 (7.7) 27.5 (8.6) digging involve direct removal of sediment; vegeta-
% Clay 8.3 (2.1) 8.4 (2.2) 7.2 (1.4) 9.1 (2.7) tion disruption occurs as a direct result of the
Values in parenthesis denote two standard deviations. digging as well as by the removed sediments cover-
a
Significantly different from the east-facing aspect; p < 0.05. ing plants downslope from the activity. The very
action of sediment removal can lead (particularly in
vary as a function of aspect. For example, the east- steeper and wetter areas) to a decrease in slope
facing aspect has the highest coarse fraction (44.7% stability and, ultimately, failure. Further, the removed
by weight), which is significantly different from the material is exposed to subaerial processes of rain-
south-facing (31.4% by weight) and the west-facing splash, wind action, needle ice activity, segregation
(28.9% by weight) aspects. Woods (1980) suggests ice, and/or wash. In other words, abiotic geomorphic
that both ground squirrels and marmots prefer to activity occurs as a direct result of animal activity
construct burrows in porous, well-aerated soils and that would not, in the absence of that animal activity,
will avoid areas of hardpan and shallow bedrock. Soil have otherwise occurred. As much as there are
textures in the Canadian Rockies site (Table 3) are surface responses and slope responses to the dig-
also consistent with Thorn (1978), who reported that ging/burrowing, so too can this have a localized
areas with high pocket gopher burrowing activity had impact upon pedogenesis (Thorp, 1949; Hole,
soils that contained 72% sand and less than 10% clay. 1981). While little impact is noted in terms of
Due to the similarity of soil properties between regional pedogenic processes, indirect results can
aspects, other parameters such as gradient, micro- be due to changes in resultant water chemistry. By
climate, and vegetation may exert a stronger influence increasing inflow of water into the ground, the
on where burrowing activity is concentrated in this subsequent downslope movement of solutes may be
area of the Canadian Rockies. significantly changed. In areas near to streams or
The observed grizzly bear diggings (Table 1) were rivers, fine-grained sediment transfers may increase
mostly excavations of burrows; therefore, intensity of as well. Although at first appearance seemingly
digging is strongly related to the aspects with high small, the actual result of thousands of burrows/
burrowing activity. The south-facing aspect exhibits diggings within a small area of a catchment can
the greatest amount of sediment displacement by have a major impact upon resultant stream chemistry
grizzly bear digging (532 m3/km2/year), of which and/or fine-grained debris transport.
approximately half is attributed to a winter den site. As much as the removal of sediments can cause
Grizzly bear diggings were not observed on either the slope failure, so too can loading of slopes by animals
north- or west-facing aspect in this study area. (Fig. 8). Steep, wet slopes can fail as a result of
movement across them by large mammals (e.g., bear,
moose, elk, caribou, sheep, goats), especially where
5. The role of animals some of these species may move in large herds (e.g.,
caribou). Examples from Tibet (Hall et al., 1999) have
All of the preceding diagrams, easily adjustable to shown extensive terracette development as a result of
any specific alpine site, lead to the main theme of loading of steep slopes by yaks. As an aside, burrowing
this discussion, namely the impact of animals (Fig. animals were observed frequently exploiting the veg-
8). Simplistically, animal geomorphic effects can be etation-free terracette front (the face along which fail-
broken down into the impacts of burrowing, loading, ure occurred) because of the ease of access to the rich,
 K. Hall, N. Lamont / Geomorphology 55 (2003) 219–234 231

Fig. 8. The influence of zoogeomorphic activity upon the alpine landscape. Activities considered are burrowing, digging, trampling, and
loading.

soft soil (Hall et al., 1999). The loading can also be In essence, the animals are working, geomorpho-
exacerbated by the action of trampling (Fig. 8). The logically, at two levels. First, at the level of their direct
movement of the animals leads to destruction of the influence (e.g., burrowing, loading, etc.). The second
soft-foliage alpine plants and to subsequent compac- level involves the operation of abiotic alpine geo-
tion of the underlying sediments, especially under wet morphic processes that take place as a result of the
soil conditions (Whinam and Comfort, 1995). As exposure/preparation of material by animals (i.e., the
loading can cause failure on steep, unstable slopes, making available of sediments from burrowing, etc.).
loading can be especially important on nonvegetated In some situations, the joining together of burrows
scree slopes as well as in vegetated but particularly wet down a slope has led to canalization and the facili-
situations. The action of trampling, as well as aiding tation of fluvial entrenchment combined with
failure by loading, can also operate to initiate fluvial enhanced downslope transport of sediments (Hall et
erosion (Fig. 8) by concentrating overland flow (itself al., 1999). Thus, while animals can be a significant
increased as a result of compaction) along animal factor in the landscape in their own right, the greater
tracks. impact on the landscape results from the synergistic
232 K. Hall, N. Lamont / Geomorphology 55 (2003) 219–234

interaction between animals and nonbiotic geomor- Animals clearly play an important role in alpine slope
phic processes. If and where permafrost exists, the processes because their contribution to sediment
effects can become even greater (Hall, 1997) as the removal can be substantial. Activity on scree or
removal of the protective active layer leads to perma- boulder fields was not considered nor was any other
frost melt and increased fluvial/solifluction activity animal activity, such as trail creation by caribou and
that both greatly increase mass movement. The impact moose.
on permafrost can be particularly significant; once
melting has started it tends to be self-sustaining and
potentially even worse under a warming climate. 6. Conclusion
Frequently, the sheer scale of animal activity is not
always appreciated. As Butler (1992) has shown, the The human exploitation of some alpine areas
digging impact of grizzly bears in Glacier National where domestic animal numbers are increased in
Park is said to exceed all other debris removal/trans- attempts to achieve sustainable development can lead
port processes combined. Debris exposed by digging to significant landscape impacts (Hall et al., 1999). A
is transported away, leaving the exposed substrate to comparable situation may arise where animal protec-
be affected by subsequent geomorphic processes. In tion is introduced within a national park or wilderness
other areas, the number of holes produced by burrow- area—the increase in animal numbers, which are
ing animals can be so substantial (Fig. 9) that it can beneficial for tourism, may have a detrimental impact
destroy the majority of the vegetation and accelerate on the landscape. The latter may be more significant
erosion (Thorp, 1949), which affects both slope in permafrost areas, especially if additionally im-
stability and debris transport. In addition, the com- pacted by global warming. A number of alpine areas,
bined effects of animals in a given area are often especially in developing countries, can be affected
overlooked. With reference to Table 1, the contrast simultaneously by both indigenous and increased
between aspects with respect to zoogeomorphic activ- numbers of domesticated animals, the latter some-
ity becomes even greater when both burrowing animal times in response to the improved grazing as a result
and grizzly bear sediment displacement are combined. of climatic warming.

Fig. 9. Example from China of burrowing activity; holes in close proximity can have a substantial impact upon the landscape by affecting both
slope stability and debris transport.
 K. Hall, N. Lamont / Geomorphology 55 (2003) 219–234 233

Although animals play a role in modifying the Cox, C.F., 1933. Alpine plant succession on James Peak, Colorado.
landscape and sediment production, they receive little Ecological Monographs 3, 301 – 372.
Davidson, M.A., 1992. Influence of direct beam solar radiation on
recognition as geomorphic agents in the alpine zone. hillslope morphology. PhD dissertation, University of Denver,
As a group, animals can have a significant impact upon Denver, CO. 239 pp.
the landscape through such activities as burrowing, Deluca, T.H., Patterson IV, W.A., Freimund, W.A., Cole, D.N.,
digging, trampling, and surface loading of slopes. To 1998. Influence of llamas, horses, and hikers on soil erosion
from established recreation trails in western Montana, USA.
fully appreciate the role of animals in alpine landscape
Environmental Management 22, 255 – 262.
evolution, an understanding of the climatic –biotic – Dyni, E.J., Yensen, E., 1996. Dietary similarity in sympatric Idaho
abiotic interactions is required. Climate, topography, and Columbian ground squirrels (Spermophilus brunneus and S.
and ecological factors vary between and within aspects, columbianus). Northwest Science 70, 99 – 108.
therefore influencing the spatial and seasonal distribu- Ellison, L., 1946. The pocket gopher in relation to soil erosion on
tion of zoogeomorphic activity. It is important to mountain range. Ecology 27, 101 – 114.
Foth, H.D., 1990. Fundamentals of Soil Science. Wiley, New York.
understand the abiotic – biotic interactions because it 360 pp.
is their synergy that contributes to the development of Gee, G.W., Bauder, J.W., 1986. Particle-size analysis. In: Klute, A.
the alpine landscape. (Ed.), Methods of Soil Analysis: Part 1. Physical and Minera-
logical Methods. American Society of Agronomy and Soil Sci-
ence Society of America, Madison, WI, pp. 363 – 375.
Govers, G., Poesen, J., 1998. Field experiments on the transport of
Acknowledgements
rock fragments by animal trampling on scree slopes. Geomor-
phology 23, 193 – 203.
This work was funded by NSERC grant 185756 Grinnel, J., 1923. The burrowing rodents of California as agents in
and the University of Northern British Columbia. soil formation. Journal of Mammalogy 4, 137 – 149.
Hall, K., 1997. Zoological erosion in permafrost environments: a
possible origin of dells? Polar Geography 21, 1 – 9.
Hall, K., Boelhouwers, J., Driscoll, K., 1999. Animals as erosion
References
agents in the alpine zone: some data and observations from
Canada, Lesotho, and Tibet. Arctic, Antarctic, and Alpine Re-
Banfield, A.W.F., 1974. The Mammals of Canada. University of search 31, 436 – 446.
Toronto Press, Toronto. 438 pp. Hamer, D., Herrero, S. (Eds.), 1983. Ecological studies of the griz-
Beniston, M., 2000. Characterization in mountain environments. In: zly bear in Banff National Park. Parks Canada Contract WR4-
Bradley, R.S., Roberts, N., Williams, M.A.J. (Eds.), Key Issues In 80. Final Report.
Environmental Change: Environmental Change in Mountains Hamer, D., Herrero, S., 1987. Grizzly bear food and habitat in the
and Uplands. Arnold, Great Britain, pp. 17 – 38. Front Ranges of Banff National Park, Alberta. International
Bennet, R.C., 1976. Notes on alpine climate. Proceedings of the Conference on Bear Research and Management 7, 199 – 213.
Workshop on Alpine and Subalpine Environments. BC Ministry Hansen, R.M., Morris, M.J., 1968. Movement of rocks by northern
of the Environment Resource Analysis Branch, Victoria, BC, pocket gophers. Journal of Mammalogy 49, 391 – 398.
pp. 13 – 20. Holcroft, A.C., Herrero, S., 1984. Grizzly bear digging sites for
Billings, W.D., Bliss, L.C., 1959. An alpine snowbank environment Hedysarum sulpurescens roots in southwestern Alberta. Cana-
and its effects on vegetation, plant development, and productiv- dian Journal of Zoology 62, 2571 – 2575.
ity. Ecology 40, 388 – 397. Hole, F.D., 1981. Effects of animals on soil. Geoderma 25, 75 – 112.
Birkeland, P.W., 1999. Soils and Geomorphology. Oxford Univ. Kershaw, L., MacKinnon, A., Pojar, J., 1998. Plants of the Rocky
Press, New York. 430 pp. Mountains. Lone Pine Publishing, Edmonton, Canada. 384 pp.
Bliss, W.D., 1988. Alpine vegetation. In: Barbour, M.G., Billings, Marchand, P., 1996. Life in the Cold: An Introduction to Winter
W.D. (Eds.), North American Terrestrial Vegetation. Cambridge Ecology. University Press of New England, Hanover, NH.
Univ. Press, New York, pp. 573 – 592. 304 pp.
Butler, D.R., 1992. The grizzly bear as an erosional agent in Martinez Rica, J.P., Pardo Ara, M.P., 1990. Initial data on erosion
mountainous terrain. Zeitschrift für Geomorphologie, N.F. 36, caused by small mammals in the central Pyrenees (Spain). Pyr-
179 – 189. enees Institute of Ecology, Jaca, Spain (Translated from). Éko-
Campbell, J.S., 1997. North American alpine ecosystems. In: Wiel- logiya 1, 27 – 36.
golaski, F.E. (Ed.), Ecosystems of the World: 3. Polar and Al- McCrory, W., Herrero, S., 1981. An Evaluation of Grizzly Bear
pine Tundra. Elsevier, Amsterdam, pp. 211 – 261. Autumn Feeding Sign and Habitat in the Upper Highwood Riv-
Cole, D.N., 1995. Experimental trampling of vegetation: I. Rela- er Valley, Alberta. Alberta Fish and Wildlife Division, Calgary,
tionship between trampling intensity and vegetation response. Alberta.
Journal of Applied Ecology 32, 203 – 214. Onda, Y., Itakura, N., 1997. An experimental study on the burrow-
234 K. Hall, N. Lamont / Geomorphology 55 (2003) 219–234

ing activity of river crabs on subsurface water movement and Blanquet vegetation units in the Niwot Ridge alpine tundra
piping erosion. Geomorphology 20, 279 – 288. zone, Colorado Front Range, USA. Arctic and Alpine Research
Perez, F.L., 1993. Alpine turf destruction by cattle in the high 14, 45 – 51.
equatorial Andes. Mountain Research and Development 13, Thorp, J., 1949. Effects of certain animals that live in soil. The
107 – 110. Scientific Monthly 68, 180 – 191.
Price, L.W., 1971. Geomorphic effect of the arctic ground squirrel Valentine, K.W.G., Sprout, P.N., Baker, T.E., Lavkulich, J. (Eds.),
in an alpine environment. Geografiska Annaler 53A, 100 – 106. 1978. The Soil Landscapes of British Columbia. The Resource
Reichel, J.D., 1986. Habitat use by alpine animals in the Pacific Analysis Branch, Ministry of the Environment, Victoria, BC.
Northwest, USA. Arctic and Alpine Research 18, 111 – 119. 197 pp.
Smith, D.J., Gardner, J.S., 1985. Geomorphic effects of ground Vroom, G.W., Herrero, S., Ogilvie, R.T., 1980. The ecology of
squirrels in the Mount Rae area, Canadian Rocky Mountains. winter den sites of grizzly bears in Banff National Park, Alberta.
Arctic and Alpine Research 17, 205 – 210. International Conference on Bear Research and Management 4,
Summer, R.M., 1980. Impacts of horse traffic on trails in Rocky 321 – 330.
Mountain National Park. Journal of Soil and Water Conserva- Watanabe, T., 1994. Soil erosion on yak-grazing steps in the Lang-
tion 35, 85 – 87. tang Himal, Nepal. Mountain Research and Development 14,
Summer, R.M., 1986. Geomorphic impacts of horse traffic on mon- 171 – 179.
tane landforms. Journal of Soil and Water Conservation 41, Whinam, J., Comfort, J., 1995. The impact of commercial horse
126 – 128. riding on sub-alpine environments at Cradle Mountain, Tasmania,
Tardiff, S.E., Stanford, J.A., 1998. Grizzly bear digging: effects on Australia. Journal of Environmental Management 47, 61 – 70.
subalpine meadow plants in relation to mineral nitrogen avail- Woods Jr., S.E., 1980. The Squirrels of Canada. Natural Museums
ability. Ecology 79, 2219 – 2228. of Canada, Ottawa. 199 pp.
Thorn, C.E., 1978. A preliminary assessment of the geomorphic Young, P.J., 1990. Structure, location and availability of hibernacula
role of pocket gophers in the alpine zone of the Colorado Front of Columbian Ground Squirrels (Spermophilus columbianus).
Range. Geografiska Annaler 60A, 181 – 187. The American Midland Naturalist 123, 357 – 364.
Thorn, C.E., 1982. Gopher disturbance: its variability by Braun –
 Geomorphology 55 (2003) 235 – 247
www.elsevier.com/locate/geomorph

Post-Oligocene river incision, southern Sierra Madre

Occidental, Mexico
David R. Montgomery a,*, Jorge López-Blanco b
a
Department of Earth and Space Sciences, University of Washington, Seattle, WA 98195, USA
b
Instituto de Geografia, UNAM, Circuito Ext., Cd. Universitaria, Mexico, 04510 D.F., Mexico
Received 2 November 2001; received in revised form 13 May 2002; accepted 10 March 2003

Abstract

The Sierra Madre Occidental of western Mexico consists of a granitic basement covered by Oligocene ignimbrites that
define a reference surface from which to estimate late Cenozoic river incision. A 90-m-grid digital elevation model was used to
characterize contemporary topography and interpolate the Late Oligocene surface of the ignimbrite plateau from a surface fit to
the highest points in the relatively undissected uplands between major river valleys. Long-term river incision rates calculated
from the difference between this reference surface and longitudinal profiles of 11 rivers that flow toward the Tepic-Zacoalco rift
zone range from about 0.01 to 0.2 mm year
1. River profiles of this region also show evidence of river capture driven by
flexural uplift along the flank of the rift zone. River profile concavity values (h) in the Sierra Madre Occidental range from 0.22
to 0.63, a range similar to that reported previously for a wide range of environments. In contrast, the empirically constrained
ratio of exponents in the stream power model of river incision (m/n) ranges from 0.44 to 0.52, close to the expected theoretical
value of 0.5. The wider range of observed h values may illustrate how h can differ from the driving values of m/n in non-steady-
state bedrock river systems.
D 2003 Elsevier Science B.V. All rights reserved.

Keywords: Tectonic geomorphology; Ignimbrite plateau; River incision; Mexico

1. Introduction ery and Gran, 2001; Tinkler and Wohl, 1998; Mas-
song and Montgomery, 2000; Snyder et al., 2000). In
In the last decade, renewed appreciation of the addition, strong interest in feedback between erosion
importance of bedrock river incision in landscape and rock uplift has focused attention on processes and
evolution has spurred research on the morphology rates of bedrock river incision (Seidl and Dietrich,
and distribution of bedrock channels (e.g., Miller, 1992; Wohl, 1992a; Wohl et al., 1994; Seidl et al.,
1991; Wohl, 1992b, 1998, 1999; Montgomery et al., 1994; Hancock et al., 1998; Pazzaglia et al., 1998;
1996; Montgomery and Buffington, 1997; Montgom- Pazzaglia and Brandon, 2001; Stock and Montgom-
ery, 1999; Whipple et al., 1999, 2000; Kirby and
Whipple, 2001). Isostatic response to river incision
* Corresponding author. can focus rock uplift in areas of concentrated erosion
E-mail address: [email protected] (D.R. Montgomery). (Beaumont et al., 1992; Willett et al., 1993; Zeitler et

0169-555X/03/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0169-555X(03)00142-9
236 D.R. Montgomery, J. López-Blanco / Geomorphology 55 (2003) 235–247

al., 1993, 2001; Finlayson et al., 2002), and landscape 2. Theory

evolution models increasingly incorporate assump-
tions about rates of bedrock channel incision into Many workers report that channel slope varies as
studies linking fluvial geomorphology and tectonic an inverse power law function of drainage area where
processes (Tucker and Slingerland, 1994; Whipple
S ¼ bA
h ð1Þ
and Tucker, 1999).
Rates of bedrock incision reflect both geophysical in which h ranges from 0.2 to 1.0 (Hack, 1957; Flint,
forcing and spatial variations in erosional processes, 1974; Tarboton et al., 1989; Moglen and Bras, 1995;
but little is known about how to calibrate quantita- Slingerland et al., 1998; Hurtrez et al., 1999; Snyder
tively the characterization of the resistance of bedrock et al., 2000; Kirby and Whipple, 2001). For detach-
to these processes and, therefore, how to predict rates ment-limited channel incision, modeling the local
of bedrock river incision. Stock and Montgomery erosion rate (E) as a function of drainage area (A)
(1999) found a more than five-order-of-magnitude and local slope (S) has become standard with
range in erosion resistance by modeling reported
E ¼ KAm S n ð2Þ
examples of long-term bedrock river incision with
initial profiles of known age for a wide range of where K is a constant that incorporates climatic factors
lithologies. Although the controlling variables are and bedrock erodibility, and m and n are thought to
not easily distinguished, let alone quantified, erosion vary with different erosional processes (Kirkby,
resistance scales the response time for evolution of 1971). Eq. (2) can be rearranged to cast channel slope
bedrock river profiles (Whipple and Tucker, 1999; as a function of erosion rate, drainage area and bed-
Whipple, 2001). Evaluations of bedrock erodibility rock resistance:
typically back-calculate erosion resistance from sin-
gle rivers and hence within region variability is less S ¼ ðE=KÞð1=nÞ A
ðm=nÞ ð3Þ
constrained than variability between regions.
In the case of steady-state topography, in which E
Previous work on bedrock river incision has
everywhere equals the local rock uplift rate U, Eq. (3)
focused on tectonically active regions (typically
can be recast in terms of the local uplift rate such that
convergent margin settings) with relatively little
attention being given to rift zones and passive S ¼ ðU =KÞð1=nÞ A
ðm=nÞ , ð4Þ
margins, although Young and McDougall (1993),
Seidl et al. (1994), and Weissel and Seidl (1998) allowing analysis of river profiles to yield information
provide notable exceptions. But investigations of about the ratio of uplift (or bedrock incision) rates to
passive and extensional margins reveal that even bedrock erodibility (e.g., Hurtrez et al., 1999; Kirby
in non-collisional settings erosion can strongly and Whipple, 2001). Although Eqs. (1) and (4) imply
influence large-scale patterns of isostatic uplift (Gil- that h = m/n for steady-state topography, the relation
christ and Summerfield, 1990, 1991; Pazzaglia and of h to m/n is not clear for nonequilibrium landscapes
Brandon, 1996). Of particular interest in the evolu- because of the potential effect of spatial and temporal
tion of rifted margins is the fact that simulation variability in E and S along an evolving river profile.
modeling of flexural uplift along rift zones reveals In addition, Roe et al. (2002) showed that feedback
the potential for drainage capture induced by margin between orographically variable precipitation and dis-
upwarping (Tucker and Slingerland, 1994). Here we charge-driven river incision implies that h p m/n for
report results of a study of long-term river incision steady-state landscapes with strong orographic precip-
into Oligocene ignimbrites in the southern Sierra itation regimes.
Madre Occidental, Mexico. In particular, we calcu-
late long-term river incision rates into Tertiary
ignimbrites in the western Mexican Volcanic Belt 3. Study area
and discuss evidence for river capture due to
flexural upwarping along the margin of the Tepic- The Sierra Madre Occidental of western Mexico,
Zacoalco rift zone. the world’s largest Tertiary silicic province, consists of
 D.R. Montgomery, J. López-Blanco / Geomorphology 55 (2003) 235–247 237

a Cretaceous granitic core covered by Oligocene digital elevation model (i.e., 90-m grid) is available
ignimbrites (Fig. 1) (McDowell and Keizer, 1977; for the region.
McDowell and Clabaugh, 1979; Albrecht and Gold- The Sierra Madre Occidental lies immediately north
stein, 2000; Henry and Aranda-Gomez, 2000). Ignim- of the western Mexican Volcanic Belt, an active
brite deposition from 30 to 20 Ma predates onset of volcanic arc with substantial Pliocene to Holocene
E – NE extension ca. 13 Ma (Fig. 2). The Sierra Madre volcanism focused into a series of grabens and fault-
Occidental is an excellent area to study long-term bounded valleys (Righter, 1997). The southern end of
river incision because it is covered by an ignimbrite the Sierra Madre Occidental is one of three large rift
sheet that provides a reasonably well-constrained zones—the Tepic-Zacoalco, Colima, and Chapala rift
initial condition of known age, and a 3-arc-second zones—that intersect in western Mexico to form a

Fig. 1. Location of the study area southern Sierra Madre Occidental, Mexico. Note the presence of the Tepic-Zacoalco rift to the south.
238 D.R. Montgomery, J. López-Blanco / Geomorphology 55 (2003) 235–247

structural triple junction that records the initiation of Table 1

continental rifting (Luhr et al., 1985; Luhr, 1997; Allan Morphometric characteristics of 11 rivers studied in the southern
Sierra Madre Occidental
et al., 1991). Onset of rifting is thought to be related to
River Name Length Watershed Maximun Minimun
changes in offshore plate boundaries, in particular the
(km) area elevation elevation
migration of the East Pacific Rise onto the continent (km2) (m.a.s.l.) (m.a.s.l.)
(Luhr et al., 1985; Allan et al., 1991) and possibly
River 1 Bolaños 317 14,943 2379 426
oblique subduction of the Rivera Plate (DeMets and River 2 Mango 241 8828 2900 390
Stein, 1990). The rifts are related to jumping of the River 3 Jesús Marı́a 176 4867 2724 217
Pacific – Rivera spreading ridge to the site of the River 4 Huajimic 55 573 1582 376
Colima rift; four similar eastward ridge jumps have River 5 La Palmilla 35 400 2018 246
River 6 Camotlán 85 2053 2225 865
progressed northward in sequence along the East
River 7 Joraviejo 56 629 2205 378
Pacific Rise during the last 12 Ma (Luhr, 1997). River 8 Chico 50 827 2232 557
Righter (1997) reported post-Pliocene bedrock River 9 Tlaltenango 167 3136 2568 1461
incision rates of 0.23 – 0.25 mm year
1 in the Aten- River 10 San Pedro 72 1706 2450 1627
guillo River valley on the Jalisco block south of the River 11 Juchipila 289 7986 2387 949
Tepic-Zacoalco rift zone. We studied the longitudinal
profiles of 11 rivers draining the southern Sierra
Madre Occidental on the north side of the Tepic- relatively undissected uplands of the Sierra Madre
Zacoalco rift (Table 1). These rivers rise in the Occidental and flow south through valleys incised into
the ignimbrites and underlying rocks. Upon crossing
into the Tepic-Zacoalco rift zone, the rivers join the
Rio Santiago to flow westward and empty into the
Pacific Ocean.

4. Methods

We integrated a digital mosaic of 25 1j by 1j

digital elevation model (DEM) files to create a single
composite DEM of the southern Sierra Madre Occi-
dental. The original source of this DEM, the GEMA
DEM database (INEGI, 1994), covers all of Mexico at
a grid resolution of 3-arc seconds (i.e., 90 m). The
composite DEM forms an array extending from 101j
to 106j W longitude and from 19j to 24j N latitude.
We analyzed 11 of the primary rivers in the study
area (Fig. 3). Through on-screen digitizing using the
afore mentioned DEM, we traced each river trajectory
from its drainage divide to its confluence with the Rio
Santiago, which bounds the study area to the south.
We also traced the bounding drainage divides of both
sides of each river basin. River trajectories were used
to guide extraction of river profiles using a custom-
Fig. 2. Ignimbrite age distribution. Data compiled from Damon et ized ARC/INFO macro (proffix2.aml) designed to
al. (1979), Clark et al. (1981), Nieto-Obregón et al. (1981, 1985),
address two major difficulties that plague efforts to
Labarthe-Hernández et al. (1982), Lang et al. (1988), Castillo-
Hernández and Romero-Rios (1991), Pasquaré et al. (1991), Moore extract river profiles from DEMs: (i) geographic
et al. (1994), Webber et al. (1994), and Nieto-Samaniego et al. projections are not equidistant, and (ii) river profiles
(1996, 1999). generated from DEMs typically yield a bilinear inter-
 D.R. Montgomery, J. López-Blanco / Geomorphology 55 (2003) 235–247 239

Fig. 3. Shaded relief digital elevation model of the southern Sierra Madre Occidental showing the river profile locations (R1 to R11) and high
points used to estimate late-Oligocene topograpahy.

polation of the four lattice points nearest to the sample depending on the algorithm used in DEM creation,
points and consequently can actually represent points the initial lower-elevation values that get raised
partway up valley walls, especially along narrow through pit filling routines may better record the
bedrock valleys lacking substantial flood plains. In actual river elevation.
part due to this latter problem, DEMs frequently The custom macro that we used accomodates
contain artificial ‘‘sinks’’ from which water cannot different distance values for vertical, horizontal, and
flow. Conventional river profile extraction algorithms diagonal lines and preserves the lowest elevations
treat these pits as artifacts that are filled to the along the represented stream bed rather than filling
elevation of their lowest outlet, or pour point. The pits. Instead, when proffix2.aml encounters a filled
usual solution is to flag the real sinks, such as lakes, sink, the program interrogates the raw DEM and
and then fill the rest, resulting in a profile that has traces the path of steepest descent. When it reaches
artificially flat zones composed of filled pits. But a closed depression, it assigns weights to paths of
240 D.R. Montgomery, J. López-Blanco / Geomorphology 55 (2003) 235–247

shallowest ascent and to the direct path to the sink’s A representation of the ‘‘original’’ topography of
pour point, from where it works its way uphill. When the ignimbrite plateau was estimated from a surface fit
proffix2.aml reaches a part of the profile upstream of a to the 42 highest points in the study area using a
filled section, it assigns a user-designated minimum smoothed quadratic interpolation. High points along
slope to the formerly filled section of river. This the flank of the Tepic-Zacoalco rift were not consid-
results in a profile with neither flat spots nor closed ered in creating this reference surface to avoid bias
depressions and yields a profile with a continuous because of the influence of local flexural uplift along
downstream slope. the rift flank. The resulting surface has a slope of
Using the modified river profiles, the contributing roughly 1j to the south and provides a reasonable
drainage area was calculated for each pixel along the interpretation of the depositional surface of the ignim-
river profiles. To reduce the effect of ‘‘noise’’ in the brite plateau.
original DEM data as well as interpolation algorithms Using the final delineation of the 11 river trajecto-
used to produce DEMs, we applied a smoothing ries, we developed a raster map for each river trajec-
function that averaged the river slope over 20 neigh- tory and assigned each pixel an ordinal number in a
boring grid cells (10 up and 10 down the river profile) downstream sequence from the drainage divide. Each
to obtain profiles that consistently dropped in eleva- river raster map was overlaid with both the original
tion in the downstream direction without major dis- DEM and the estimated topography of the original
continuities. ignimbrite plateau. We then generated a tabular data-

Fig. 4. Longitudinal profiles of several of the study rivers (R1, R2, R5, and R10).
 D.R. Montgomery, J. López-Blanco / Geomorphology 55 (2003) 235–247 241

base for each river trajectory containing the ordinal ence between the inferred ‘‘original’’ elevation of the
number for each pixel, its elevation value in the ignimbrite and the modern river elevation by the
present topography DEM, and the elevation value in average ignimbrite age of 26 Ma to yield an average
the ignimbrite plateau DEM. Total river incision was river incision rate. Estimates of the long-term incision
calculated for each grid cell by subtracting the present rate under the assumption that all incision post-dates
elevation from the interpolated original plateau ele- the 13 Ma onset of rifting would be double those
vation at each point along the river profile. estimated from the mean ignimbrite age.
The age of the ignimbrites provides a limiting age
on river incision. Although a river valley system
carved into the Cretaceous basement likely existed 5. Results
at the time of ignimbrite eruption, any such valleys
were most likely filled in by the massive deposition The form of river profiles in the study area ranges
associated with emplacement of the ignimbrite sheets. from smoothly concave to relatively linear and even
Incision of the present valleys therefore post-dates convex in the lower reaches of some profiles (Fig. 4).
ignimbrite deposition. Local long-term incision rates The convex reaches presumably reflect either active
were estimated by dividing the net elevation differ- subsidence of the Tepic-Zacoalco rift zone, or uplift

Fig. 5. Examples of river and ridge profiles illustrating the minimum estimates of post-Oligocene river incision versus distance along the river.
242 D.R. Montgomery, J. López-Blanco / Geomorphology 55 (2003) 235–247

of the Sierra Madre Occidental. In addition, a number rifting (13 Ma). Under either scenario, lower values
of rivers have distinct knickpoints midway along in basin headwaters yield to progressively greater
their profile. The knickpoints on rivers 1 and 2, values farther down the river profile.
however, are located just upstream of confluences Plots of drainage area versus channel slope show
with rivers 9 and 6, respectively, that initially drain the typical general inverse trend and wide scatter
away from the rift flank and flow north before (Fig. 6). Power law regressions of drainage area
turning east to join the rivers flowing in the neigh- versus slope for each river profile indicate h values
boring valleys. of 0.22 to 0.63 (Table 2). The wide range of R2
Long-term rates of river incision calculated by values (0.19 to 0.85) indicates substantial variability
dividing the elevation difference between the present to these relationships. Nonetheless, the rivers drain-
channel and the reconstructed ignimbrite plateau by ing the southern Sierra Madre Occidental exhibit
26 Ma (the mean ignimbrite age) increase downstream distinct relations between drainage area and slope in
along each of the river profiles (Fig. 5). Hence, these spite of the systematic variability in long-term
rivers are not in steady-state. Long-term rates of erosion rates along their profiles.
bedrock incision along these profiles vary from about Stratifying the individual drainage area –slope data
0.01 to 0.1 mm year
1 assuming onset of incision at points for each 20-grid-cell long reach from the
the mean ignimbrite age (26 Ma), or 0.02 to 0.2 mm profiles of all the study rivers into three ranges of
year
1 assuming incision initiated coincident with long-term erosion rates of < 0.03, 0.03 –0.05, and

Fig. 6. Examples of slope versus area plots for several of the study rivers (R1, R2, R5, and R10).
 D.R. Montgomery, J. López-Blanco / Geomorphology 55 (2003) 235–247 243

Table 2 Table 3
Power law relationships between drainage area and slopes Power law relationships between drainage area and slopes
(S = cA
h) for the study rivers stratified by erosion rate for the points from the study rivers
River Name c h R2 [S = (E/K)(1/n)A
(m/n)]

River 1 Bolaños 0.05 0.30 0.24 Erosion rate (E/K)(1/n) m/n R2 Ka

(mm year
1) (m0.5 year
1)
River 2 Mango 0.12 0.37 0.52
River 3 Jesús Marı́a 0.29 0.55 0.63 < 0.03 0.11 0.52 0.77 2.7  10
7
River 4 Huajimic 0.06 0.22 0.35 0.03 – 0.05 0.15 0.48 0.62 2.7  10
7
River 5 La Palmilla 0.18 0.36 0.74 > 0.05 0.20 0.44 0.48 2.5  10
7
River 6 Camotlán 0.33 0.63 0.53 a
Assuming n = 1.
River 7 Joraviejo 0.10 0.26 0.42
River 8 Chico 0.10 0.25 0.53
River 9 Tlaltenango 0.04 0.32 0.35
River 10 San Pedro 0.18 0.59 0.85 implying that K itself does not vary much in the
River 11 Juchipila 0.04 0.32 0.19 relatively homogenous rocks of the study area. Assum-
ing that n = 1 yields an estimated K of 2.5  10
7 m0.5
year
1.
>0.05 mm year
1 allows assessment of m/n and (E/
K)1/n values based on the empirical relation between
drainage area and slope for data from each erosion 6. Discussion
rate class. The exponents in power law regressions of
these stratified data (Fig. 7) yield m/n values of 0.44 The association of distinct knickpoints just up-
to 0.52, with R2 values of 0.48 to 0.77 (Table 3). The stream of the confluences of small, initially north-
variation in the coefficient, (E/K)1/n, follows the draining rivers with south-draining rivers in neighbor-
variation in E between the three groups of data, ing valleys implies a history of river capture along the

Fig. 7. Slope versus area data from all 11 rivers stratified by erosion rate determined as illustrated in Fig. 5. Power law regressions are for data
within each grouping (Table 4).
244 D.R. Montgomery, J. López-Blanco / Geomorphology 55 (2003) 235–247

northern flank of the Tepic-Zacoalco rift zone. Each of Block on the opposite side of the Tepic-Zacoalco rift.
these north-flowing rivers shares a N – S oriented As Righter’s (1997) rates were based on post-Pliocene
valley with a larger south-flowing river. In each case, incision, this difference could be due to either accel-
N –S flowing rivers turn abruptly downstream of their eration of bedrock incision in the Pleistocene, differ-
confluence to drop through a narrow E – W trending ences in erodibility in the two areas on either side of
gorge before merging with the river in the neighboring the rift zone, or because the ignimbrite ages only
valley. We suspect that these short north-flowing rivers provide limiting ages on the onset of river incision in
reflect flexural uplift of the margin of the Tepic- the Sierra Madre Occidental and, therefore, minimum
Zacoalco rift zone that was sufficient to reverse the incision rates for the period of active river incision. If
course of these rivers, and trigger their diversion and the incision of the Sierra Madre Occidental river
capture by overtopping their lateral drainage divide. valleys commenced at 13 Ma coincident with the
The apparent reversal, diversion, and capture of these onset of rifting, then the incision rates reported here
rivers is in accord with the predictions of Tucker and would double, bringing the upper end of the range of
Slingerland’s (1994) surface process model simula- incision rates into general agreement with those
tions of river-system evolution on rifted margins. reported by Righter (1997). While the K values
Although substantial variability in the combined inferred from Fig. 7 are also sensitive to the time
datasets exists, the K values are lower than those scale over which river incision is assumed to have
reported by Stock and Montgomery (1999) for other been effective, the range of 13– 26 Ma yields calcu-
areas with volcaniclastic rocks, and are more similar lated river incision rates of 0.01 – 0.2 mm year
1.
to values they report for more lithified granitoids and The m/n values of 0.42 to 0.52 determined by
meta-sediments. The rates of long-term river incision analysis of data grouped into locations with similar
indicated by our analysis are comparable to rates post-Oligocene erosion rates are similar to those
reported by other researchers for long-term river expected based on models for river incision driven
incision in extensional settings or volcanic islands by unit stream power and are not dependent on the
(Table 4). However, our rates are lower than those time scale over which incision is assumed to have
reported by Righter (1997) for channels on the Jalisco occurred. Neither are they dependent on the steady-
state assumption, as local values of E are known
independently from the long-term incision of the
Table 4 ignimbrite plateau. The empirically derived h values
Comparison of incision rates between the southern Sierra Madre
Occidental and other areas
of 0.22 to 0.63 based on slope – area relations for
individual rivers span much of the range of values
Locality Tectonic Incision rate Reference
characteristics (mm year
1)a
reported in previous studies of rivers in other regions,
suggesting that much of the range in previously
Southern Oligocene 0.01 – 0.2 This study
Sierra Madre ignimbritic
observed h values may simply reflect that some of
Occidental plateau the rivers in question do not have steady state profiles
Atenguillo Plio-Quaternary 0.23 – 0.25 Righter et al. and that h p m/n for these non-steady-state rivers.
River extension (1995) and
Righter (1997)
Grand Canyon Plio-Quaternary 0.09 – 0.11 Damon et al.
extension (1974)
7. Conclusions
Utah Plio-Quaternary 0.30 Hamblin et al.
extension (1981) Ignimbrite sheets, such as the southern Sierra
Rio Grande Plio-Quaternary < 0.01 – 0.08 Grimm (1982) Madre Occidental, provide ideal locations for study-
Rift extension ing long-term river incision because of their relatively
Israel Extension 0.10 Wohl et al.
(1994)
simple initial condition as a depositional surface and
Hawaii Ocean islands < 0.01 – 0.08 Seidl et al. their amenability to conventional radiometric dating.
(1994) The long-term bedrock river incision rates of 0.01–
a
Assuming incision commenced 13 – 26 Ma, as constrained by 0.2 mm year
1 determined for the southern Sierra
onset of rifting (13 Ma) and mean ignimbrite age (26 Ma). Madre Occidental are comparable to rates reported for
 D.R. Montgomery, J. López-Blanco / Geomorphology 55 (2003) 235–247 245

a variety of extensional and volcanic landscapes G.A., Eastwood, R.L. (Eds.), Geology of Northern Arizona with
suggesting that such rates may be typical for rifted Notes on Archeology and Paleoclimate: Part I. Regional Studies.
Geological Society of America 27th Annual Meeting, Rocky
margin settings. Although the wide range of river Mountain Section, Flagstaff, AZ., pp. 221 – 235.
concavity (i.e., h) values for the southern Sierra Madre Damon, P.E., Nieto, O.J., Delgado, A.L., 1979. Un plegamiento
Occidental spans the range of values reported previ- neogénico en Nayarit y Jalisco y evolución geomórfica del
ously, much of the variability in our data likely is due Rı́o Grande de Santiago. Asoc. Ingenieros Mineros, Metalúrgi-
cos y Geólogos de México. Memoria Técnica XXIII, 156 – 191.
to the non-steady-state nature of the river profiles.
DeMets, C., Stein, S., 1990. Present day kinematics of the Rivera
This supposition is supported by the finding that the Plate and implications for tectonics in southwestern Mexico.
m/n values determined by normalizing for differences Journal of Geophysical Research 95, 21931 – 21948.
in bedrock lowering rates are consistent with values Finlayson, D.P., Montgomery, D.R., Hallet, B., 2002. Spatial co-
predicted by conventional models of river incision. incidence of rapid inferred erosion with young metamorphic
The non-steady-state nature of these river profiles massifs in the Himalayas. Geology 30, 219 – 222.
Flint, J.J., 1974. Stream gradient as a function of order, magnitude,
appears to impart a wider range in h than is apparent and discharge. Water Resources Research 10, 969 – 973.
in empirically estimated m/n values that describe the Gilchrist, A.R., Summerfield, M.A., 1990. Differential denudation
integrated effect of erosional processes. and flexural isostasy in formation of rifted-margin upwarps.
Nature 346, 739 – 742.
Gilchrist, A.R., Summerfield, M.A., 1991. Denudation, isostasy and
Acknowledgements landscape evolution. Earth Surface Processes and Landforms 16,
555 – 562.
Grimm, J.P., 1982. Base level changes and incision rates for can-
We thank Harvey Greenberg for developing the yons draining the Mount Taylor Volcanic field, New Mexico. In:
ARC/INFO macro proffix2.aml and for assistance Wells, S.G., Grambling, J.A., Callender, J.F. (Eds.), Albuquer-
with GIS analyses. This research was funded in part que Country II. Guidebook, vol. 33. New Mexico Geological
by CONACYT and DGAPA-UNAM, México. Society, Albuquerque, NM, pp. 60 – 61.
Hack, J.T., 1957. Studies of longitudinal stream profiles in Virginia
and Maryland. United States Geological Survey Professional
Paper 294-B. Washington, DC, 97 pp.
References Hamblin, W.K., Damon, P.E., Bull, W.B., 1981. Estimates of ver-
tical crustal strain rates along the western margins of the Colo-
Albrecht, A., Goldstein, S., 2000. Effects of basement composition rado Plateau. Geology 9, 293 – 298.
and age on silicic magmas across an accreted terrane – Precam- Hancock, G.S., Anderson, R.S., Whipple, K.X., 1998. Beyond
brian crust boundary, Sierra Madre Occidental, Mexico. Journal power: bedrock river incision process and form. In: Tinkler,
of South American Earth Sciences 13, 255 – 273. K., Wohl, E. (Eds.), Rivers Over Rock: Fluvial Processes in
Allan, J.F., Nelson, S.A., Luhr, J.F., Carmichael, I.S.E., Wopat, M., Bedrock Channels. Geophysical Monograph, vol. 107. Ameri-
Wallace, P.J., 1991. Pliocene – recent rifting in SW Mexico and can Geophysical Union, Washington, DC, pp. 35 – 60.
associated volcanism: an exotic terrane in the making. In: Dau- Henry, C.D., Aranda-Gomez, J.J., 2000. Plate interactions control
phin, J.P., Simoneit, B.R.T. (Eds.), The Gulf and Peninsular middle – late Miocene, proto-Gulf and Basin and Range exten-
Province of the Californias. American Association of Petroleum sion in the southern basin and Range. Tectonophysics 318, 1 – 26.
Geology Memoir, vol. 47, pp. 425 – 445. Tulsa, OK. Hurtrez, J.-E., Lucazeau, F., Lavé, J., Avouac, J.-P., 1999. Inves-
Beaumont, C., Fullsack, P., Hamilton, J., 1992. Erosional control of tigation of the relationships between basin morphology, tectonic
active compressional orogens. In: McClay, K.R. (Ed.), Thrust uplift, and denudation from the study of an active fold belt in the
Tectonics. Chapman & Hall, New York, pp. 1 – 18. Siwalik Hills, central Nepal. Journal of Geophysical Research
Castillo-Hernández, D., Romero-Rios, F., 1991. Estudio geológico- 104, 12796 – 12799.
regional de Los Altos, Jalisco y El Bajı́o. Comisión Federal de INEGI, 1994. GEMA (Geomodelos de Altimetrı́a del Territorio
Electricidad, Gerencia de Proyectos Geotermoeléctricos. Depto. Nacional). Instituto Nacional de Estadı́stica Geografı́a e Infor-
Exploración, Open File Report 02-91, Mexico City. 35 pp. mática. Compact Disc with 255 1j by 1j DEM files from topo-
Clark, K.F., Damon, P.E., Shafiquillah, M., Ponce, B.F., Cárdenas, graphic mapas at 1,250,000 scale. ISBN: 970-13-0474-8.
D., 1981. Sección geológica-estructural a través de la parte Sur Kirby, E., Whipple, K., 2001. Quantifying differential rock-uplift
de la Sierra Madre Occidental, entre Fresnillo y la costa de rates via stream profile analysis. Geology 29, 415 – 418.
Nayarit. Asoc. Ingenieros Mineros, Metalúrgicos y Geólogos Kirkby, M.J., 1971. Hillslope process-response models based on the
de México. Memoria Técnica XIV, 69 – 99. continuity equation. In: Brunsden, D. (Ed.), Slopes: Form and
Damon, P.E., Shafiqullah, M., Leventhal, J.S., 1974. K – Ar chro- Process, vol. 3. Institute of British Geographers Special Publi-
nology for the San Francisco Volcanic Field and rate of erosion cation, London, pp. 15 – 30.
of the Little Colorado River. In: Karlstrom, T.N.V., Swann, Labarthe-Hernández, G., Tristán-González, M., Aranda-Gómez,
246 D.R. Montgomery, J. López-Blanco / Geomorphology 55 (2003) 235–247

J.J., 1982. Revisión Estratigráfica del Cenozoico de la Parte zoica de la parte meridional de la Mesa Central, México. Revista
Central del Estado de San Luis Potosı́. Universidad Autónoma Mexicana de Ciencias Geológicas 13, 117 – 122.
de San Luis Postosı́, Instituto Geologı́a, Folleto técnico #85, Nieto-Samaniego, A.F., Ferrari, L., Alaniz-Alvarez, S.A., Labarthe-
Mexico City. 208 pp. Hernández, G., Rosas-Elguera, J., 1999. Variation of Cenozoic
Lang, B., Steinitz, G., Sawkins, F.J., Simmons, S.F., 1988. K – Ar extension and volcanism across the southern Sierra Madre
age studies in the Fresnillo Silver District, Zacatecas, Mexico. Occidental volcanic province, Mexico. Geological Society of
Economic Geology 83, 1642 – 1646. America Bulletin 111, 347 – 363.
Luhr, J.F., 1997. Extensional tectonics and the diverse primitive Pasquaré, G., Ferrari, L., Garduño, V.H., Tibaldi, A., Vezzoli, L.,
volcanic rocks in the Western Mexican Volcanic Belt. The Cana- 1991. Geologic Map of the Central Sector of the Mexican Vol-
dian Mineralogist 35, 473 – 500. canic Belt, States of Guanajuato and Michoacan, Mexico. Geo-
Luhr, J.F., Nelson, S.A., Allan, J.F., Carmichael, I.S.E., 1985. Ac- logical Society of America, Map and Chart Series MCH 072, 1
tive rifting in southwestern Mexico: manifestations of an incip- sheet. 21 pp.
ient eastward spreading ridge jump. Geology 13, 54 – 57. Pazzaglia, F.J., Brandon, M.T., 1996. Macrogeomorphic evolution
Massong, T.M., Montgomery, D.R., 2000. Influence of lithology, of the post-Triassic Appalachian mountains determined by de-
sediment supply, and wood debris on the distribution of bedrock convolution of the offshore basin sedimentary record. Basin
and alluvial channels. Geological Society of America Bulletin Research 8, 255 – 278.
112, 591 – 599. Pazzaglia, F.J., Brandon, M.T., 2001. A fluvial record of long-term
McDowell, F.W., Clabaugh, S.E., 1979. Ignimbrites of the Sierra steady-state uplift and erosion across the Cascadia forearc high,
Madre Occidental and their relation to the tectonic history of western Washington State. American Journal of Science 300,
western Mexico. In: Chapin, C.E., Elston, W.E. (Eds.), Ash- 385 – 431.
Flow Tuffs, vol. 180. Geological Society of America Special Pazzaglia, F.J., Gardner, T.W., Merritts, D.J., 1998. Bedrock fluvial
Paper, Boulder, CO, pp. 113 – 124. incision and longitudinal profile development over geologic
McDowell, F.W., Keizer, R.P., 1977. Timing of mid-Tertiary vol- time scales determined by fluvial terraces. In: Tinkler, T.J.,
canism in the Sierra Madre Occidental between Durango City Wohl, E.E. (Eds.), Rivers Over Rock: Fluvial Processes in Bed-
and Mazatlan, Mexico. Geological Society of America Bulletin rock Channels. American Geophysical Union Geophysical
88, 1479 – 1487. Mongraph, vol. 107, pp. 207 – 236.
Miller, J.R., 1991. Controls on channel form along bedrock-influ- Righter, K., 1997. High bedrock incision rates in the Atenguillo
enced alluvial streams in south-central Indiana. Physical Geog- River Valley, Jalico, western Mexico. Earth Surface Processes
raphy 12, 167 – 186. and Landforms 22, 337 – 343.
Moglen, G.E., Bras, R.L., 1995. The effect of spatial heterogene- Righter, K., Carmichael, I.S.E., Becker, T.A., Renne, R.P., 1995.
ities on geomorphic expression in a model of basin evolution. Pliocene to Quaternary volcanism and faulting at the intersec-
Water Resources Research 31, 2613 – 2623. tion of the Gulf of California and the Mexican Volcanic Belt.
Montgomery, D.R., Buffington, J.M., 1997. Channel reach mor- Geological Society of America Bulletin 107, 612 – 626.
phology in mountain drainage basins. Geological Society of Roe, G., Montgomery, D.R., Hallet, B., 2002. Effects of orographic
America Bulletin 109, 596 – 611. precipitation variations on steady-state river profiles. Geology
Montgomery, D.R., Gran, K.B., 2001. Downstream variations in the 30, 143 – 146.
width of bedrock channels. Water Resources Research 37, Seidl, M.A., Dietrich, W.E., 1992. The problem of channel erosion
1841 – 1846. into bedrock. In: Schmidt, K.-H., dePloey, J. (Eds.), Functional
Montgomery, D.R., Abbe, T.B., Peterson, N.P., Buffington, J.M., Geomorphology: Landform Analysis and Modelling. Catena
Schmidt, K.M., Stock, J.D., 1996. Distribution of bedrock and Supplement, vol. 23, pp. 101 – 124.
alluvial channels in forested mountain drainage basins. Nature Seidl, M.A., Dietrich, W.E., Kirschner, J.W., 1994. Longitudinal
381, 587 – 589. profile development into bedrock: an analysis of Hawaiian chan-
Moore, G., Marone, C., Carmichael, I.S.E., Renne, P., 1994. Basal- nels. Journal of Geology 102, 457 – 474.
tic volcanism and extension near the intersection of the Sierra Slingerland, R., Willett, S.D., Hovius, N., 1998. Slope – area scaling
Madre volcanic province and the Mexican Volcanic Belt. Geo- as a test of fluvial bedrock erosion laws. EOS, Transactions of the
logical Society of America Bulletin 106, 383 – 394. American Geophysical Union 79, F358 (Fall Meet. Suppl.).
Nieto-Obregón, J., Delgado-Argote, I., Damon, P.E., 1981. Rela- Snyder, N.P., Whipple, K.X., Tucker, G.E., Merritts, D.J., 2000.
ciones petrológicas y geocronológicas del magmatismo de la Landscape response to tectonic forcing: digital elevation model
Sierra Madre Occidental y el Eje Neovolcánico en Nayarit, analysis of stream profiles in the Mendocino triple junction
Jalisco y Zacatecas Asoc. Ingenieros Mineros, Metalúrgicos y region, northern California. Geological Society of America Bul-
Geólogos de México. Memoria Técnica XIV, 327 – 361. letin 112, 1250 – 1263.
Nieto-Obregón, J., Delgado-Argote, I., Damon, P.E., 1985. Geo- Stock, J.D., Montgomery, D.R., 1999. Geologic constraints on bed-
chronologic, petrologic and structural data related to large mor- rock river incision using the stream power law. Journal of Geo-
phologic features between the Sierra Madre Occidental and the physical Research 104, 4983 – 4993.
Mexican Volcanic Belt. Geofı́sica Internacional 24, 623 – 663. Tarboton, D.G., Bras, R.L., Rodriguez-Iturbe, I., 1989. Scaling and
Nieto-Samaniego, A.F., Macias-Romo, C., Alaniz-Alvarez, S.A., elevation in river networks. Water Resources Research 25,
1996. Nuevas edades isotópicas de la cubierta volcánica ceno- 2037 – 2051.
 D.R. Montgomery, J. López-Blanco / Geomorphology 55 (2003) 235–247 247

Tinkler, K., Wohl, E., 1998. A primer on bedrock channels. In: Willett, S., Beaumont, C., Fullsack, P., 1993. Mechanical model for
Tinkler, K.J., Wohl, E.E. (Eds.), Rivers Over Rock: Fluvial Pro- the tectonics of doubly vergent compressional orogens. Geology
cesses in Bedrock Channels. Geophysical Monograph, vol. 107. 21, 371 – 374.
American Geophysical Union, Washington, DC, pp. 1 – 18. Wohl, E.E., 1992a. Gradient irregularity in the Herbert Gorge of
Tucker, G.E., Slingerland, R.L., 1994. Erosional dynamics, flexural northeastern Australia. Earth Surface Processes and Landforms
isostasy, and long-lived escarpments: a numerical modeling 17, 69 – 84.
study. Journal of Geophysical Research 99, 12229 – 12243. Wohl, E.E., 1992b. Bedrock benches and boulder bars: floods in the
Webber, K.L., Fernández, L.A., Simmons, B., 1994. Geochemistry Burdekin Gorge of Australia. Geological Society of America
and mineralogy of the Eocene – Oligocene volcanic sequence, Bulletin 104, 770 – 778.
southern Sierra Madre Occidental, Juchipila, Zacatecas, Mexico. Wohl, E.E., 1998. Bedrock channel morphology in relation to ero-
Geofı́sica Internacional 33, 77 – 89. sional processes. In: Tinkler, K.J., Wohl, E.E. (Eds.), Rivers
Weissel, J.K., Seidl, M.A., 1998. Inland propagation of erosional Over Rock: Fluvial Processes in Bedrock Channels. Geophys-
escarpments and river profile evolution across the southeast ical Monograph, vol. 107. American Geophysical Union, Wash-
Australian passive continental margin. In: Tinkler, K.J., Wohl, ington, DC, pp. 133 – 151.
E.E. (Eds.), Rivers Over Rock: Fluvial Processes in Bedrock Wohl, E.E., 1999. Incised bedrock channels. In: Darby, S.E.,
Channels. Geophysical Monograph, vol. 107. American Geo- Simon, A. (Eds.), Incised River Channels. Wiley, New York,
physical Union, Washington, DC, pp. 189 – 206. pp. 187 – 217.
Whipple, K.X., 2001. Fluvial landsacpe response time: how plau- Wohl, E.E., Greenbaum, N., Schick, A.P., Baker, V.R., 1994. Con-
sible is steady-state denudation? American Journal of Science trols on bedrock channel incision along Nahal Paran, Israel.
301, 313 – 325. Earth Surface Processes and Landforms 19, 1 – 13.
Whipple, K.X., Tucker, G.E., 1999. Dynamics of the stream-power Young, R., McDougall, I., 1993. Long-term landscape evolution:
river incision model: implications for height limits of mountain early Miocene and modern rivers in southern New South Wales,
ranges, landscape response time scales, and research needs. Australia. Journal of Geology 101, 35 – 49.
Journal of Geophysical Research 104, 17661 – 17674. Zeitler, P.K., Chamberlain, C.P., Smith, H.A., 1993. Synchronous
Whipple, K.X., Kirby, E., Brocklehurst, S.H., 1999. Geomorphic anatexis, metamorphism, and rapid denudation at Nanga Parbat
limits to climate-induced increases in topographic relief. Nature (Pakistan Himalaya). Geology 21, 347 – 350.
401, 39 – 43. Zeitler, P.K., et al., 2001. Erosion, Himalayan geodynamics, and the
Whipple, K.X., Hancock, G.S., Anderson, R.S., 2000. River inci- geomorphology of metamorphism. GSA Today 11, 4 – 9.
sion into bedrock: mechanics and relative efficacy of plucking,
abrasion, and cavitation. Geological Society of America Bulletin
112, 490 – 503.
 Geomorphology 55 (2003) 249 – 261
www.elsevier.com/locate/geomorph

The shapes of cold, high mountains in sedimentary rocks

D.M. Cruden *
Department of Civil and Environmental Engineering, University of Alberta, Edmonton, Alberta, Canada T6G 2G7
Received 20 October 2001; received in revised form 16 February 2002; accepted 10 March 2003

Abstract

Terzaghi (Geotechnique 12 (1962) 251) and Young (Young, A., 1972. Slopes. Oliver and Boyd, Edinburgh, 288 pp.)
described the stable forms of slopes in sedimentary rock masses, assuming penetrative discontinuities, which are parallel to
bedding and joints which are perpendicular to bedding. The only movements considered were slides along bedding. Experience
in the Canadian Rockies indicates that the cohesionless rock masses that exist at or above tree line may also move by toppling,
buckling and sliding along joints. These processes also act to limit the inclinations of stable slopes. Rock strength is a factor in
the critical height of a slope that buckles. The processes can be represented as fields on a process diagram, a plot of slope
inclination against bedding dip, using the basic friction angles of the rocks present.
The process diagram also separates five common mountain peak shapes, which form on homoclinal sequences of beds.
Castellate and Matterhorn mountains occur in sub-horizontal beds, cuestas develop in gently to moderately dipping beds.
Hogbacks formed in moderately to steeply dipping beds have similar slope angles on both cataclinal and anaclinal slopes.
Dogtooth mountains occur in steeply dipping sub-vertical beds.
D 2003 Elsevier Science B.V. All rights reserved.

Keywords: Cuesta; Hogback; Matterhorn; Slope; Slide; Topple

1. Introduction hundreds of deaths. As Terzaghi (1962, p. 251) pointed

out ‘‘The failure of the Malpasset Dam and various
Modern rock mechanics began with the publication catastrophic slides on slopes above the portals of
of Terzaghi’s (1962) paper on the stability of steep pressure tunnels have aroused the concern of public
slopes on hard, unweathered rock. That paper was a building authorities. Consequently they have become
response to a catastrophic failure. On December 2, more and more reluctant to issue permits to construct
1959, a 60-m high, thin, concrete arch dam built on the on rock unless the applicants demonstrate by means of
LeReyran River to provide irrigation and water supply, stability computations that the proposed structure will
collapsed. The failure of the Malpasset Dam sent a not cause a rock failure.’’ Terzaghi, the father of soil
flood wave through the French town of Frejus on the mechanics, had the repose to write about rock mechan-
Cote d’Azur between Nice and St. Tropez. It caused ics forced on him. In November 1960, aged 78, he had
undergone a 9 1/2 h operation to repair heart damage.
Subsequent blood clots forced the amputation of his left
* Fax: +1-780-492-8198. leg above the knee (Goodman, 1999, p. 282). Terzaghi
E-mail address: [email protected] (D.M. Cruden). died on 25 October, 1963.

0169-555X/03/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0169-555X(03)00143-0
250 D.M. Cruden / Geomorphology 55 (2003) 249–261

The 1962 paper was enthusiastically received by The paper referred specifically to three common
some geomorphologists. Carson and Kirkby (1972, p. associations of mechanical defects of rocks. First,
123) wrote in their Chapter on Instability Processes in Terzaghi described ‘‘unstratified, jointed rock. . ...mas-
Rock Masses, ‘‘it is refreshing to come across a ..massive rocks with a random joint pattern. . ..crystal-
description (of the geomorphological evolution of a .crystalline rocks such as granite or marble,’’ which
landscape) based on the underlying principles of are not my subject. Second, ‘‘stratified sedimentary
mechanics in contrast to the speculative, non-quanti- rock’’ which is my focus. The third association,
tative and confused thinking of early geomorpholo- faulted rock, can be illuminated by the principles
gists dealing with the matter.’’ Young (1972) developed to describe the second association or fabric
elaborated and clarified Terzaghi’s ideas on the impor- in the terminology of Terzaghi’s Vienna colleague,
tance of cross-joint spacing and his figures have been Sander (1970). This terminology underlies modern
reproduced in subsequent texts (Trenhaile, 1987; structural geology (Turner and Weiss, 1963). The
Gerrard, 1990). Most recently, they have appeared in fabric of sedimentary rocks produced a particular
the second edition of Selby’s (1993, Fig. 15.6) well- mechanical model according to Terzaghi (1962, p.
established memoir and might be considered to rep- 258).
resent a consensus of geomorphological thought on
the topic. Stratified sedimentary rocks consist of layers with
The discipline of rock mechanics has grown since a thickness averaging between a few inches and
1962 and Terzaghi’s ideas should now be updated. many feet. These are commonly separated from
Some of the problems he identified have been each other by thin films of material with a
solved, others still require attention. They are the composition different from that of the rest of the
subjects of this paper. I first describe and extend the rock. The bedding planes are almost invariably
application of Terzaghi’s mechanical model of a surfaces of minimum shearing resistance. They
sedimentary rock mass to sliding. Then, the same are likely to be continuous over large areas.
mechanical model is used in descriptions of toppling The cross-joints, generally nearly perpendicular
and buckling. These two processes can be added to to the bedding joints, are commonly staggered at
the kinematic limitations on sliding to form a these joints. The cohesive bond along the walls of
process diagram. The addition of some new results the cross-joints is equal to zero. The intersections
in the analysis of assisted sliding in gently dipping between the cross-joints and the bedding planes
beds to the diagram explains how some rock falls may be more or less parallel to one of two or
occur. more directions, or less commonly, the intersec-
tions may have a nearly random orientation.
Because of the almost universal presence of
2. Terzaghi’s conditions for the stability of slopes bedding and cross-joints, stratified sedimentary
on stratified sedimentary rock rock with no effective cohesion (ci = 0) has the
mechanical properties of a body of dry masonry
Terzaghi’s (1962, p. 252) consideration of the composed of layers of more or less prismatic
stability of steep slopes on hard, unweathered rock blocks which fit each other. The boundaries
began with a review of the properties of unweathered between the individual layers of blocks constitut-
rock. He demonstrated that ‘‘the critical height, Hc, of ing the masonry correspond to the bedding planes
a vertical slope on the weakest rock. . ..would of the rock. The cohesion across the joints
be. . ...4200 feet. . .. . . Yet no vertical slopes with such between all the blocks of each layer is zero, and
height exist, and many gentler slopes with a smaller most of the joints between the blocks of two
height than Hc have failed. This fact indicates that the adjacent layers are staggered at the boundaries
critical height of slopes on unweathered rock is between layers. The stability of a slope on a rock
determined by the mechanical defects of the rock with the mechanical properties of such a body of
such as joints and faults and not by the strength of masonry depends primarily on the orientation of
the rock itself.’’ the bedding planes with reference to the slope.
 D.M. Cruden / Geomorphology 55 (2003) 249–261 251

The model gave rise to critical slope angles, bc, different geometric arrangements of the ‘‘cross-
which if steepened, caused sliding of the super- joints’’, as Terzaghi (1962) termed the joints formed
incumbent rock mass to occur. To discuss the critical perpendicular to bedding. A more precise term for
slope angles fully, the terminology of Powell (1875) these joints is kathetal, Hancock’s (1964) term for
has been supplemented with some new terms shown joints ‘‘normal to a bedding surface where the ori-
in Fig. 1. Terzaghi (1962, Fig. 2) showed three entation of the joint is a function of the orientation of
figures, one of a cataclinal slope parallel to the bed- the bedding.’’
ding and at the angle of friction. The other two figures Young’s diagrams indicated that the critical slope
explored anaclinal slopes with different arrangements angle is dependent on ‘‘the relative spacing and offset,
of cross-joints. Young (1972, Fig. 45) expanded these c, of bedding and cross joints’’. Pursuing Terzaghi’s
figures to a total of seven. analogy with the stacking of bricks, the slopes may
Fig. 2 shows the seven critical slopes, bc, on a show different bonds. The critical slope bc, is then:
single diagram with the dip of the bedding as the
other axis. The heavy line separates those slopes bc ¼ ð90  wÞ þ tan1 C=D ð1Þ
above on which sliding is possible from slopes below
on which sliding is not possible. So the diagram is a where w is the dip of the bedding, C the offset
process diagram. The letters A to G refer to Young’s between kathetal joints in successive beds of thick-
(1972, Fig. 45) seven figures and indicate where the ness, D. Eq. (1) shows that for values of C which were
slopes illustrated in these figures plot on the diagram. large compared to D the critical slope may be vertical.
Fig. 2 is supplemented by Fig. 3, an overlay to show Eq. (1) corrects an error in the last line of Young
the orientation of bedding and slope at points within (1972, p. 120) which was unfortunately reproduced
Fig. 2. (with different symbols) by Selby (1993, Fig. 15.4).
Fig. 2 can be simplified by a more detailed con- The text clarified Young’s intent. ‘‘For a given w, bc
sideration of cases E, F and G which represent increases with increasing values of C/D ’’ (p. 121).

Fig. 1. Classification of anaclinal and cataclinal slopes after Powell (1875) and Cruden and Hu (1996). The thick lines are the surfaces of slopes,
thin lines on the sections are the traces of bedding. The six slopes diagrammed are plotted in Fig. 8.
252 D.M. Cruden / Geomorphology 55 (2003) 249–261

Fig. 2. Young’s (1972 Fig. 45, A – G) seven figures arranged as a process diagram for sedimentary rock masses with / = 30j, cV= 0. Heavy
dashed lines indicate critical angles for slopes with; (A) horizontal bedding, (B) bedding dip less than friction angle, (C) bedding dip greater than
friction angle, (D) dip of cross joints less than friction angle, (E) cross joints not offset, (F) cross joints offset, h/b z 1, (G) cross joints offset, h/
b < 1.

Studies of joint spacing (Priest, 1993, Chapter 5; sample of spacings contains a substantial proportion
Giani, 1992, Chapter 3; Price and Cosgrove, 1990, of spacings close to zero. Typical mean spacings are
Chapters 2 and 9) have shown that joints are dis- of the same order as the thickness of the bed, about a
tributed along beds in a random way, that the spacing metre. Other kathetal joint sets divide the rock mass
is proportional to the thickness of the bed and that into rectangular prisms about a cubic metre in vol-
spacing increases with depth from the slope surface. ume. These blocks are familiar to geomorphologists
While the most likely value of the offset, C, is a little who work in cold mountains in screes and block
less than the mean spacing along a bed, a large fields.

Fig. 3. An overlay on the process diagram to show slope and bedding orientation within the diagram. Heavy solid lines are slopes and dashed
lines are traces of bedding on cross-sections of the slopes.
 D.M. Cruden / Geomorphology 55 (2003) 249–261 253

Fig. 4. Slopes which may slide in a bedded rock mass with kathetal joints. Stable slopes are shaded for / = 30j, c = 0.

With random variations in spacing, offset and bed existed in the rock masses with low values of i (Bruce
thickness, there are localities on any extensive rock and Cruden, 1980). Alternatively, cleft-water pres-
slope where tan 1 (C/D) is insignificantly small sures (Terzaghi, 1962, p. 262) may be sufficient to
compared to (90  w). These localities may expand dilate the rock mass and overcome the asperities’
into couloirs or chutes extending down the slope as contribution to resistance to movement. Experiments
moisture and freeze and thaw processes induce further with tilting tables (Bruce et al., 1989) demonstrated
joints and reduce C values. On a wide slope, the that /b could be conveniently estimated on carefully
minimum value of bc should approach (90  w). chosen loose blocks of sedimentary rocks, specifically
Cases G and F, which rely on large values of C can sandstones and limestones, in the mountains. Typi-
be at least temporarily ignored and a symmetrical cally, values ranges from 21.5j to 41.4j for carbo-
diagram produced (Fig. 4). nates (Cruden and Hu, 1988) and from 22.5j to 35.8j
Fig. 4, in which stable slopes are shaded, demon- for clastic rocks (Hu and Cruden, 1992, Table 1).
strates that Terzaghi (1962) reduced the problem of Thirty degrees, therefore, was selected as an appro-
apparently stable, vertical slopes by about 2/3. Gently priate value of / (and /b) to use in the construction of
dipping cataclinal slopes and steeply dipping anaclinal Fig. 4.
slopes were still without modes of movement even
when vertical, as Selby (1993, p. 326) noted. The only
material parameters necessary for the construction of 3. Toppling
the diagram are the friction angles, /, along the bed-
ding and the kathetal joints. Patton (1966) suggested: Both geologists (Lugeon and Oulianoff, 1922)
and geomorphologists (Cotton, 1958; Sharpe, 1938)
/ ¼ /b þ i ð2Þ had observed movements on rock slopes caused by
the rotation of rock blocks with steeply dipping
where the basic friction angles /b was measured on an margins. Lugeon and Oulianoff (1922) called these
artificially prepared specimen of the rock material and movements ‘‘balancement superficial des tétes de
i was the average angle of the asperities on the natural couches’’, Cotton (1958, p. 29) referred to them as
rock surfaces, bedding or joints. causing ‘‘outcrop curvature’’ and Sharpe (1938, Fig.
Analyses of rock slides in the Canadian Rockies 2) illustrated them, including them in the phenomena
(Cruden, 1985) indicated that / was reasonably he called ‘‘rock creep’’. Sharpe (1938, p. 47) pro-
estimated by /b, suggesting that surfaces of rupture vided a comprehensive list of the active causes of
254 D.M. Cruden / Geomorphology 55 (2003) 249–261

Fig. 5. Toppling on anaclinal slopes.

creep. Two likely to be particularly effective in Fig. 7. Toppling on cataclinal slopes.

mountain environments are wedging and prying by
expansion of water freezing in joints or by hydro- involving overturning of interacting columns. In rock,
static pressure of water. such columns are formed by regular bedding planes,
The introduction of the term ‘‘toppling’’ to describe cleavage or joints which strike parallel to the slope crest
these movements is usually attributed to de Freitas and and dip into the rock mass; this contrasts with the
Watters (1973). Goodman and Bray (1976, p. 201) structure of slides in which the controlling discontinu-
commented ‘‘Toppling is a failure mode of slopes ities dip into the open space’’. A more precise definition

Fig. 6. Toppling added to Fig. 4.

 D.M. Cruden / Geomorphology 55 (2003) 249–261 255

of a topple is ‘‘the forward rotation, out of the slope, of ping anaclinal and cataclinal slopes (Cruden, 1988).
a mass of soil or rock about a point or axis below the These seem implausible.
centre of gravity of the displaced mass’’ (WP/WLI, Cruden (1989) explored toppling on steep, under-
1993; Dikau et al., 1996, Appendix 1). dip, cataclinal slopes in which the bedding dips under
Goodman and Bray’s (1976) definition suggested the slope in the same direction as the slope (Fig. 7).
that toppling was possible away from steeply dipping Shear was possible along the bedding if
bedding on anaclinal slopes (or steeply dipping joints
on gently dipping cataclinal slopes). Then, if the slope bz/  ð90  wÞ ð4Þ
was steep, the largest principal stress acted down
slope and slip on bedding could occur if As the subsequent discussion clarified (Bovis, 1991;
Cruden, 1991), additional forces are required to rotate
bz/ þ ð90  wÞ ð3Þ individual blocks to the vertical. Further rotation can
then proceed under the blocks’ own weight. If, how-
Fig. 5 shows that the blocks formed with bedding and ever, the rock mass responds as a closely jointed
the kathetal joints should have suitable geometries. continuum as in a block-flexure (Cruden and Varnes,
Because Eq. (3) includes only variables already 1996, Fig. 3 – 20c), gravity may still be sufficient to
used in the construction of the process diagram (Fig. rotate the blocks of the continuum. Sharpe (1938, Fig.
4), it can be added to produce the fields seen in Fig. 6. 2) sketched the results.
This addition provides a movement mode for the Analyses of the mechanics of any of these pro-
vertical slopes unaffected by sliding the slopes that cesses on underdip cataclinal slopes remain to be
concerned Selby (1993). Abrupt transitions remain on carried out. One limit on their effectiveness can be
the process diagram between both very steeply dip- sketched from field studies (Cruden et al., 1993;
ping anaclinal and cataclinal slopes and gently dip- Cruden and Hu, 1994; McAffee and Cruden, 1996).

Fig. 8. Processes on anaclinal and cataclinal slopes, slope angle b. The bedding dip is w and /b is the basic friction angle. The six slopes in Fig.
1 are plotted as triangles. Heavy dashed lines represent hypothesized lower bounds to processes.
256 D.M. Cruden / Geomorphology 55 (2003) 249–261

Underdip topples where the initial dip of the bedding Hw has increased by less than a third when the bed is
is below 60j have yet to be observed. dipping at 40j and, then, increases very rapidly as the
The effect of these processes, toppling from both dip of the bed approaches the friction angle (as an
steep bedding and steep kathetal joints, is to blunt the inspection of Eq. (8) shows). The buckles mapped in
sharp boundaries on the process diagram (Fig. 8) at the Highwood Pass (Hu and Cruden, 1993, Table 2)
90j and 0j by reducing the stable slope angles of were apparently limited to beds dipping at over 80j. So
cataclinal slopes with steep bedding dips and anaclinal there is little field evidence, as yet to support Eq. (8).
slopes with shallow bedding dips. Schmidt and Montgomery (1995) have proposed a
limit to relief based on a simpler mechanical model of
cohesive rock masses. Buckling is the only slope
4. Buckling process observed in the Rocky Mountains, which
depends on the strength of the rock mass through its
Where extensive slabs of rock dip steeply and Young’s modulus. As Young’s modulus is usually
parallel to a mountain slope, movement can occur correlated with the uniaxial compressive strength of
by buckling. Theory requires definition of the the rock material (Selby, 1993, p. 71), buckling might
mechanical conditions at each end of the column provide limits to the relief of dip slopes that appear to
whose deformation by buckling under its own weight be strength controlled. The buckling movements we
is being considered. Hu and Cruden (1993) assumed have described suggest these limits might be hundreds
the upper end of the column was free and the lower of metres in steeply dipping, strong (Barton et al.,
end restrained, perhaps by talus. These assumptions 1978) rock masses and tens of metres in poor rock
lead to the minimum critical length of the column. masses. As dips approach friction angles, these limits
Euler’s model of buckling then gave a critical length may increase by an order of magnitude. So the region
Lcr of a buckled bed as on the process diagram (Fig. 8) in which buckling may
take place has been truncated at 5j above the friction
Lcr ¼ ½1:39 Et 3 =ðtcsinw  tccoswtan/  cÞ1=3 ð5Þ angle. While buckling along kathetal joints have yet to
be reported that possibility is included in Fig. 8.
where E is the Young’s modulus of the bed, thickness Because bedding planes and slopes may be rough
t and cohesion, c, (Hu and Cruden, 1993, Eq. (6)). enough to show measurable departures from their
With zero cohesion, Eq. (5) becomes mean orientations (Bruce and Cruden, 1980; Cruden
and Charlesworth, 1976), the buckling field on the
Lcr ¼ ½1:39Et 2 cos/=csinðw  /Þ1=3 ð6Þ process diagram extends for 3j on either side of the
slope dip.
When the bedding is vertical, w = 90j, and / is 30j,
then the critical length, Lcr, and height, Hcr, of the slab
are the same and are minima,
5. Falls
2 1=3
Hcr ¼ ð1:39Et =cÞ ð7Þ
Falls are usually preceded by topples or slides of
Because the critical height or relief on a slab dipping rock blocks. Besides the kinematic restrictions on slip,
at w, Hw, is Lcr cosec w Eqs. (3) and (4), which should be satisfied for top-
pling to occur, there are geometric considerations
Hw ¼ Hcr ðcos/sin3 w=sinðw  /ÞÞ1=3 ð8Þ which can be summarized by limits on the block
ratios of rock blocks which may topple. The block
The relationship between the critical height, Hw, and ratio is the ratio of the bedding thickness to the
the dip angle has been graphed by Cruden and Hu spacing between successive kathetal joints. These
(1996, Figs. 5 and 6) for bed thicknesses ranging 0.01 – limits are described in Selby (1993, pp. 338– 339)
5 m and by Cruden and Hu (1998, Figs. 5 and 6) for and derived from de Freitas and Watters (1973).
Young’s moduli ranging from 2 to 300 Gpa. The critical In flat lying sedimentary rocks, kathetal joints are
height is clearly a minimum when the bed is vertical; commonly sufficiently widely spaced that the rock
 D.M. Cruden / Geomorphology 55 (2003) 249–261 257

Fig. 9. A wedge of rock sliding on horizontal bedding under cleft water pressures, V.

mass does not topple. This situation has been exten- c and cw are the relative densities of the rock wedge
sively explored in rock mechanic texts (Hoek and and the water in the crack. So the factor of safety, F, of
Bray, 1981; Norrish and Wyllie, 1996) which demon- the rock wedge is
strate how sliding may then take place and lead to
F ¼ ðW  U Þtan/=V
rock falls. Movement requires that the joints be full of
water and that cleft water pressures (Terzaghi, 1962,
p. 262) be exerted. ¼ ðc  cw Þtan/=cw tanb ð9Þ
Norrish and Wyllie’s (1996, Fig. 15.6) description
of a planar failure stability analysis can be consider- For typical values of c, about 2.5,
ably simplified. Consider a uniform slope inclined at
an angle b with a vertical crack, height H, filled with F ¼ 1:5tan/=tanb
water (Fig. 9). The crack follows kathetal joints and
For F = 1, / = 30j, b = 41j. So the new simplification
the bedding is horizontal with no cohesion and a
in Eq. (9) suggests a lower bound for sliding along
friction angle, /. Then the force, V, that disturbs the
horizontal bedding (in rock masses whose block ratios
equilibrium of the wedge of rock results from the
do not permit toppling) is about 40j.
pressure on the water on the vertical face of the wedge.
When the bedding is gently inclined at w out of the
V ¼ 0:5cw H 2 slope to produce an overdip slope, Eq. (9) becomes

The force resisting motion is (W  U) tan /, where W tanb ¼ tanð/  wÞðc  cw Þ=cw ð10Þ
is the weight of the wedge and U is the uplift force
for a factor of safety F of 1. So in Fig. 8, b declines
exerted by the water pressure on the wedge,
almost linearly with increasing w to intersect the
W ¼ 0:5cH 2 cotb lower bound of overdip slopes at about 18j. This
limit is approximately the lower limit for sliding of
U ¼ 0:5cw H 2 cotb bedding-parallel blocks down the bedding surface.
258 D.M. Cruden / Geomorphology 55 (2003) 249–261

Again, these blocks are driven by cleft water pres- the classification and rating of the strength of rock
sures. masses (Selby, 1993, Table 6.5) so, presumably, these
If the bedding is gently inclined into the slope, Eq. processes leading to falls, represent structural controls
(10) becomes of the rock mass slopes rather than the adjustment of
the slopes to strength – equilibrium, the strength –
tanb ¼ tanð/ þ wÞðc  cw Þ=cw ð11Þ equilibrium slopes of Selby (1993, pp. 103 – 105).
The distinction between structurally controlled slopes
So, on anaclinal slopes, the relationship between b and strength – equilibrium slopes is, however,
and w is, again, approximately linear with b increas- obscured by Selby’s inclusion of ‘‘Joint orientation’’
ing to intersect the field of sliding on kathetal joints at in his rating scheme, contributing 20% or more to the
about 65j (Fig. 8). total strength rating (Selby, 1993, Table 6.5).
As the cleft water pressures dilate the kathetal
joints, water pressures fall (unless a substantial inflow
of water occurs into the joint). The joint may even 6. Peak shapes in homoclinal sedimentary rocks
drain. So the wedge is stabilized by movement except,
perhaps, for the small volume of rock now cantilevered The processes, sliding, toppling, buckling and fall-
over the slope (Fig. 9) This exposed portion of the ing, that have been described might leave traces of
mass displaced by sliding is prone to falling (or rolling, their activity in the shapes of mountain peaks in the
depending on the steepness of the slope beneath the Canadian Rockies. Cruden and Hu (1999) report
displaced material). Rocks from above fall, topple or anaclinal and cataclinal slopes of the first 200 m or
roll into the gaping cracks, propping them open and so down from 34 well-described peaks. They are all
reducing the volume of snow, ice and run off necessary above treeline and formed from uniformly dipping
to fill them before subsequent wedge movements sedimentary rocks, most frequently Palaeozoic carbo-
(Cruden, 1989; Cruden et al., 1993). nates. All these mountain slopes have been plotted on
Stable slope angles are functions of c, / and w the process diagram after the mountains themselves
(Eqs. 9 –11). Rock density, c, and friction angles on had been placed in four shape classes; matterhorns and
bedding surfaces, /, do not enter Selby’s scheme for castles, cuestas, hogbacks and dogteeth (Fig. 10).

Fig. 10. The sample of Rocky Mountain rock slopes on the process diagram.
 D.M. Cruden / Geomorphology 55 (2003) 249–261 259

When bedding is gently dipping, slope angles may mountain chains and important in risk assessment
vary from 37j to 65j reflecting the full range of (Abbott et al., 1998a,b).
friction angles of bedding in sedimentary rocks. The mountains beside the transcontinental railways
Steep-sided castles and matterhorns form. When bed- through the Rockies were surveyed by terrestrial
ding is moderately dipping, from 15j to 60j, from photogrammetry as the railways were built (Thomson,
0.5/b to (90  /b), rock masses slide along bedding 1967, Chapter 8). The 735 photographs by W.P.
surfaces and kathetal joints to form dip slopes and Bridgland from the survey through Jasper National
normal escarpments. Where the bedding is steeper, Park in 1915 have recently become easily accessible
anaclinal and cataclinal slopes are similar and the on the INTERNET through http://.bridgland.sunsi-
mountain is a hogback. Pronounced asymmetry is te.ualberta.ca. They may be compared there with
apparent at lower dips to produce cuestas. If the modern repeat photography. A reconnaissance of this
bedding dips at over 60j, slopes may be subject to archive suggests that the railway belt of Jasper
toppling and buckling which seem to have limited National Park encompasses fine examples of the
both cataclinal and anaclinal slopes to about 60j. The shape classes identified, principally in the southern
symmetric steep slope forms of such mountains are Rockies by Cruden and Hu (1999). Fig. 11 shows an
described as dogteeth in the Rockies. easily recognizable symmetric dogtooth.
These extensions of Terzaghi’s model of the cohe- Though the heights and relief of individual peaks
sionless rock mass thus describe and explain several are matters of interest to an increasing population of
characteristic mountain forms in the Rocky Moun- recreational mountain climbers (Gadd, 1995, p. 736),
tains. However, more remains to be done. Kinematic these features appear in models as mere relicts of
models should be extended into the third dimension, fluvial and ice erosion (Hovius, 2000) of the mountain
for instance. While plagioclinal and orthoclinal slopes chain. The dominance of flat-lying rocks among the
are not as common in the Rockies as the anaclinal and peaks over 3500 m in the Rockies (Gadd, 1995, p.
cataclinal slopes considered in this paper, they are 157) is unlikely to be a series of coincidences. An
frequent along transportation routes traversing the understanding of the history of the peaks over longer

Fig. 11. View southeastwards from Mt. Cinquefoil to an unnamed dogtooth peak in the Jacques Range, Jasper National Park, Bridgland Station
48, photograph 386.
260 D.M. Cruden / Geomorphology 55 (2003) 249–261

time periods is needed to distinguish the processes al Association for Engineering Geology and the Environment,
controlling their development. vol. 2. Balkema, Rotterdam, pp. 1105 – 1200.
Abbott, B., Bruce, I., Keegan, T., Oboni, F., Savigny, W., 1998b.
Application of a new methodology for the management of rock-
fall risk along a railway. Proceedings, 8th Congress Internation-
7. Conclusions al Association for Engineering Geology and the Environment,
vol. 2. Balkema, Rotterdam, pp. 1201 – 1208.
Shapes seen in the Rockies are consistent with a Barton, N., et al., 1978. Suggested methods for the quantitative
model of the rock mass which has no cohesion along description of discontinuities in rock masses. International Jour-
nal of Rock Mechanics and Mining Sciences 15, 319 – 368.
bedding planes and at least two sets of kathetal joints. Bovis, M.J., 1991. Limits to common toppling: discussion. Cana-
Friction angles along the bedding and joints control dian Geotechnical Journal 28, 463 – 464.
rock mass response. Bruce, I., Cruden, D.M., 1980. Simple rockslides at Jonas Ridge,
Processes which act on this rock mass model Alberta, Canada. Proceedings, 3rd International Symposium on
include toppling, buckling, sliding and sliding Landslides, New Delhi, vol. 1. Sarita Prakashan, Meerut, India,
pp. 185 – 190.
driven by cleft water pressures. Toppling and sliding Bruce, I.G., Cruden, D.M., Eaton, T.M., 1989. The use of tilting
lead to rock falls. Only buckling explicitly considers table to determine the basic friction angle of hard rock surfaces.
the strength-related properties of the rock material. Canadian Geotechnical Journal 26, 472 – 479.
So, it is unlikely that slopes other than dip slopes Carson, M.A., Kirkby, M.J., 1972. Hillslope Form and Process.
will be found to show any strength-related limits to Cambridge Univ. Press, Cambridge, UK. 475 pp.
Cotton, C.A., 1958. Geomorphology: An Introduction to the Study
relief. It is also unlikely that steep slopes in the of Landforms. Whitcombe and Tombs, Christchurch, New Zea-
Rockies are in strength – equilibrium (in Selby’s use land. 502 pp.
of the term). Cruden, D.M., 1985. Rock slope movements in the Canadian Cor-
Natural slopes frequently exceed the theoretical dillera. Canadian Geotechnical Journal 22, 528 – 540.
lower bounds suggested by the process models. These Cruden, D.M., 1988. Thresholds for catastrophic instabilities in
sedimentary rock slopes, some examples from the Canadian
excesses indicate the volumes of rock, which might Rockies. Zeitschrift fur Geomorphologie Supplementum 76,
move prior to a relatively stable slope being formed. In 67 – 76.
particular, vertical slopes are unstable in areas where Cruden, D.M., 1989. The limits to common toppling. Canadian
cleft water pressures may develop. In such areas, the Geotechnical Journal 26, 737 – 742.
maximum stable slope is about 65j and occurs on Cruden, D.M., 1991. Reply to discussion by M. Bovis of Cruden
(1989). Canadian Geotechnical Journal 28, 465.
gently dipping anaclinal slopes. Cruden, D.M., Charlesworth, H.A.K., 1976. Errors in strike and dip
measurements. Bulletin Geological Society of America 87,
977 – 980.
Acknowledgements Cruden, D.M., Hu, X.-Q., 1988. Basic friction angles of carbonate
rocks from Kananaskis Country. Bulletin International Associ-
ation of Engineering Geology 38, 55 – 59.
This work in the Rockies has been supported by Cruden, D.M., Hu, X.-Q., 1994. Topples on underdip slopes in the
Alberta Environment, Alberta Transportation, the Highwood Pass, Alberta, Canada. Quarterly Journal of Engi-
Geological Surveys of Alberta and of Canada and neering Geology 27, 57 – 68.
the Natural Sciences and Engineering Research Cruden, D.M., Hu, X.-Q., 1996. Hazardous modes of rock slope
movement in the Canadian Rockies. Environment and Engineer-
Council. I am grateful to colleagues at the University
ing Geoscience 2, 507 – 516.
of Alberta and elsewhere for stimulating discussions. Cruden, D.M., Hu, X.-Q., 1998. Landslides in the Rocky Moun-
Karen Hincks selected photographs for Fig. 11 from tains of Canada. In: Kalvoda, J., Rosenfeld, C.L. (Eds.), Geo-
the Bridgland collection. morphological Hazards in High Mountain Areas, Geojournal.
Kluwer, Dordrecht, pp. 133 – 148.
Cruden, D.M., Hu, X.-Q., 1999. The Shapes of Some Mountain
Peaks in the Canadian Rockies. Earth Surface Processes and
References Landforms 24, 1 – 13.
Cruden, D.M., Varnes, D.J., 1996. Landslide types and processes.
Abbott, B., Bruce, I., Oboni, F., Savigny, W., 1998a. A method- In: Turner, A.K., Schuster, R.L. (Eds.), Landslides: Investigation
ology for assessment of rockfall hazard and risk along linear and Mitigation, Transportation Research Board. Special Report,
transportation corridors. Proceedings, 8th Congress Internation- Serial, published in Washington D.C., 247, pp. 36 – 75.
 D.M. Cruden / Geomorphology 55 (2003) 249–261 261

Cruden, D.M., Hu, X.-Q., Lu, Z.Y., 1993. Rock topples in the McAffee, R., Cruden, D.M., 1996. Landslides at Rock Glacier Site,
highway cut west of Clairvaux Creek, Jasper, Alberta. Canadian Highwood Pass, Alberta. Canadian Geotechnical Journal 33,
Geotechnical Journal 30, 1016 – 1023. 685 – 695.
De Freitas, M.H., Watters, R.J., 1973. Some field examples of top- Norrish, N.I., Wyllie, D.C., 1996. Rock slope stability analysis.
pling failure. Geotechnique 23, 485 – 514. Transportation Research Board. Special Report, Serial, pub-
Dikau, R., Brunsden, D., Schrott, L., Ibsen, M.-L., 1996. Landslide lished in Washington D.C., 247, pp. 391 – 425.
Recognition. Wiley, Chichester, UK. 251 pp. Patton, F.D., 1966. Multiple modes of shear failure in rock. Pro-
Gadd, B., 1995. Handbook of the Canadian Rockies. Corax Press, ceedings, 1st International Congress, vol. 1. International Soci-
Jasper, Canada. 831 pp. ety for Rock Mechanics, Lisbon, Portugal, pp. 509 – 513.
Gerrard, A.J., 1990. Mountain Environments: An Examination of Powell, J.W., 1875. Exploration of the Colorado River of the West
the Physical Geography of Mountains. Bellhaven Press, Lon- and its Tributaries. Government Printing Office, Washington,
don, UK. 317 pp. USA. 291 pp.
Giani, G.P., 1992. Rock slope Stability Analysis. Balkema, Rotter- Price, N.J., Cosgrove, J.W., 1990. Analysis of Geological Struc-
dam. 361 pp. tures. Cambridge Univ. Press, Cambridge. 502 pp.
Goodman, R.E., 1999. Karl Terzaghi, the Engineer as Artist. ASCE Priest, S.D., 1993. Discontinuity Analysis for Rock Engineering.
Press, Reston, VA, USA. 340 pp. Chapman & Hall, London. 473 pp.
Goodman, R.E., Bray, J.W., 1976. Toppling of rock slopes. Pro- Sander, B., 1970. An Introduction to the Study of the Fabric of
ceedings, Specialty Conference on Rock Engineering for Foun- Geological Bodies. Pergammon, Oxford. 641 pp.
dations and Slopes. ASCE, Boulder, CO, pp. 201 – 234. Schmidt, K.M., Montgomery, D.R., 1995. Limits to relief. Science
Hancock, P.L., 1964. The relations between folds and late-formed 270, 617 – 620.
joints in South Pembrokeshire. Geological Magazine 101, Selby, M.J., 1993. Hillslope Materials and Processes. Oxford Univ.
174 – 184. Press, Oxford. 451 pp.
Hoek, E., Bray, J., 1981. Rock Slope Engineering. The Institution Sharpe, C.F.S., 1938. Landslides and Related Phenomena; a Study
of Mining and Metallurgy, London. 358 pp. of Mass-Movements of Soil and Rock. Columbia Univ. Press,
Hovius, N., 2000. Macroscale process systems of mountain belt New York. 137 pp.
erosion. In: Summerfield, M.A. (Ed.), Geomorphology and Terzaghi, K., 1962. Stability of steep slopes on hard unweathered
Global Tectonics. Wiley, Chichester, UK, pp. 77 – 105. rock. Geotechnique 12, 251 – 270.
Hu, X.-Q., Cruden, D.M., 1992. A portable tilting table for on site Thomson, D.W., 1967. Men and Meridians: The History of Survey-
tests of the friction angles of discontinuities in rock masses. ing and Mapping in Canada, Volume 2, 1867 – 1917. Queen’s
Bulletin International Association of Engineering Geology 46, Printer, Ottawa, pp. 131 – 146.
59 – 62. Trenhaile, A.S., 1987. The Geomorphology of Rock Coasts. Oxford
Hu, X.-Q., Cruden, D.M., 1993. Buckling deformation in the Univ. Press, Oxford. 384 pp.
Highwood Pass, Alberta. Canadian Geotechnical Journal 30, Turner, F.J., Weiss, L.E., 1963. Structural Analysis of Metamorphic
276 – 286. Tectonites. McGraw-Hill, New York. 545 pp.
Lugeon, M., Oulianoff, N., 1922. Sur le balancement superficial des WP/WLI (International Geotechnical Societies’ UNESCO Working
couches et sur les erreurs que ce phénomène peut faire com- Party on World Landslide Inventry), 1993. Multilingual Land-
mettre. Bulletin de la Société Vaudoise des Sciences Naturelles slide Glossary. Bitech, Richmond, British Columbia. 59 pp.
54, 383 – 390. Young, A., 1972. Slopes. Oliver and Boyd, Edinburgh. 288 pp.
 Geomorphology 55 (2003) 263 – 281
www.elsevier.com/locate/geomorph

Inferring erosional resistance of bedrock units in the east

Tennessee mountains from digital elevation data
Hugh H. Mills *
Department of Earth Science, Tennessee Technological University, Cookeville, TN 38505, USA

Received 26 November 2001; received in revised form 24 May 2002; accepted 10 March 2003

Abstract

Measures of local relief, regional relief, and slope were calculated from digital elevation models (DEMs) for 50 bedrock
units in the Ridge and Valley and Blue Ridge provinces of Tennessee. Each of these measures was normalized and the three
were then averaged to produce the erosional resistance index (ERI). Bedrock units with higher ERI values include coarse
clastics, intermediate clastics, and metaplutonics. Units with lower values include shales, limestones, limestones plus
dolostones, and carbonates plus fine clastics. Dolostones tend to have intermediate values. The calculated ERI values were
compared with subjective ratings by a geologist with decades of field experience in east Tennessee. Generally, the agreement
between the two ratings was good, the most glaring exception being several shales with improbably high ERI values. These
turned out to be thin units cropping out beneath very hard sandstones, allowing them to stand higher and steeper than would
otherwise be possible. A systematic method for detecting such erroneously high ERI values is suggested. Inspection of a
drainage map superimposed on the geology map shows that in a given area, streams tend to flow on rock units with the lowest
ERI values. In addition, statistical analysis shows that bedrock units with the lowest ERI values are, on average, almost three
times closer to the nearest stream and six times as likely to have streams flowing on them than are units with highest values.
Further, the effect of ERI on stream location is strongest for streams with drainage areas between 1 and 30 km2. Thus, small
streams appear to be subject to greater lithologic control than are larger streams.
D 2003 Elsevier Science B.V. All rights reserved.

Keywords: Tennessee; Appalachians; Rock resistance; Lithology; Relative relief; Lithologic control

1. Introduction Valley is a particularly good location in which to

study these relationships. Besides the regularity of
The effect of geology on topography has attracted landforms here, numerous rock units that vary greatly
the interest of geologists and geographers since the in erosional resistance crop out as narrow bands in
earliest geologic maps were compiled. In the Appa- close proximity, allowing the effects of differential
lachians, as early as 1856, Lesley (1856) described erosion to be readily perceived. The rocks of the Blue
many relationships between bedrock and topography Ridge province show more subtle differences in rock
in the Ridge and Valley province. The Ridge and resistance and crop out in broader belts; but with by
far the highest altitudes in the southern Appalachians,
* Fax: +1-931-372-3363. this province also invites research on erosional resist-
E-mail address: [email protected] (H.H. Mills). ance.

0169-555X/03/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0169-555X(03)00144-2
 264
H.H. Mills / Geomorphology 55 (2003) 263–281
Fig. 1. Index map and orthogonal plot of study area and adjacent areas in Tennessee and North Carolina. K = Knoxville, Tennessee; A = Asheville, North Carolina. Coordinates are
Tennessee State Plane, in meters. Vertical exaggeration is approximately 14.
 H.H. Mills / Geomorphology 55 (2003) 263–281 265

With the advent of high-resolution digital elevation regression equation was calculated between the river
data and digitized geologic maps, the relation of distance from each grid square to tidewater and the
bedrock to topography can now be studied in a more mean elevation (highest – lowest points) of the square.
systematic fashion than previously possible. Unfortu- They found that the pattern of mean altitude predicted
nately, the grid approach appropriate for computer from river distance values using regressions for each
analysis does not lend itself to the analysis of the rock resistance class simulates very closely the overall
effect of structure on topography, as creating a dense pattern of British relief. Interestingly, they also re-
grid of dip and strike values is very difficult. There- ported that rank-order age is a stronger influence on
fore, the analyses herein will deal solely with the relief than is lithology.
effect of lithology. Geographically, the study is con- Slope is another topographic attribute that reflects
fined to the Tennessee parts of the Ridge and Valley erosional resistance, with more resistant rocks gener-
and Blue Ridge provinces (Fig. 1). ally maintaining steeper slopes (e.g., Selby, 1980).
Specifically, the thrust of the present study is an Apparently, the first effort in making slope measure-
attempt to infer the erosional resistance of bedrock ments from gridded elevation data in the Appala-
units from topographic evidence consisting of two chians was by Grender (1973) for an area near New
indices of relief and an index of slope and to see how Castle, SW Virginia. Using a grid with a resolution of
these indices correlate with stream location. 158 m, he was able to successfully distinguish the
slope distribution of several rock units. Very low
slopes ( < 0.05) were strongly associated with Devon-
2. Previous work ian shales, whereas very high slopes were strongly
associated with sandstones.
Efforts to link relative relief to lithology began in The present research took advantage of standard
the 19th century. Hayes (1899), for example, pub- digital elevation models (DEMs) for topographic
lished a diagram showing the relation of topographic analysis. A DEM is a regularly spaced and georefer-
relief to rock type in the Chattanooga district. Cooper enced grid of point elevations. These models have
(1944) published a histogram showing the ‘‘relative been utilized for geomorphic and other research for
topographic potency’’ of bedrock units in the Burkes more than a decade (e.g., Moore et al., 1991), but
Garden 15’ quadrangle, SW Virginia. Neither of these apparently have not been used to study the effect of
efforts were based on actual quantitative data. Thomp- geology on topography. The best source of large-scale
son (1941) tabulated summit elevations for various DEMs in the United States is the U.S. Geological
bedrock units in several quadrangles in western Vir- Survey National Elevation Dataset (NED), which
ginia, clearly showing harder rocks to have higher provides seamless coverage at a resolution of 1 arc
summits. Apparently the first effort in the Appala- sec (approximately 30 m). NED incorporates the
chians to describe topography with gridded elevation highest quality and highest resolution data. Older
data was by Flint (1963), who measured elevation DEMs have been filtered during the NED assembly
points at 0.25- or 0.50-mile spacings to show differ- process to minimize artifacts commonly found in
ences in altitude distribution on various formations in these data, thereby greatly increasing the quality of
the Fall Zone of Connecticut. A modern effort of this the synthetic drainage networks derived from the
nature, utilizing digital data and computers, was by elevation data (Anonymous, 1999).
Clayton and Shamoon (1998), who made a study of
the topography and geology of Great Britain in terms
of the kilometer squares of the National Grid. They 3. Geologic setting
developed a system to classify the most common
geologic units of Great Britain according to the extent The Ridge and Valley province of Tennessee con-
to which they form high ground. They used a number sist of sedimentary rock ranging in age from Cam-
of discriminating variables to arrange the units into brian to Mississippian. The rocks are extensively
rank order, eventually developing a sixfold classifica- folded and thrust-faulted. The province is lower than
tion of rock resistance. For each of the six classes, a flanking provinces, with altitude ranging from about
266 H.H. Mills / Geomorphology 55 (2003) 263–281

200 to about 950 m. The Blue Ridge province (also digitized by the U.S. Geological Survey for a hydro-
known in Tennessee as the Unaka province, after the geologic study of this area (Hollyday and Hileman,
Unaka Mountains) consists of Precambrian rocks, 1996).) The map was transformed into a grid with
ranging from low-grade metasedimentary rocks to resolution of 60 m in ARC/INFO. Bedrock units with
gneiss and metaplutonic rocks. The southern Blue outcrop areas of at least 25 km2 were included in the
Ridge is the highest province in the Appalachians, study for a total of 50 units. The stratigraphic termi-
with altitude in the Tennessee part ranging from about nology of the 1966 map is used herein to avoid
250 to 2024 m. Fig. 2 is a simplified geologic map of confusion, although some of it is dated. In Idrisi,
the study area, with rock units divided into eight DEM-derived grids were windowed to the same size
categories. In terms of these generalized categories, as the geology map grid to allow overlays. When
the Ridge and Valley consists mainly of limestone determining elevations or topographic indices from
plus dolostone, carbonate with fine clastics, shale, and grids, larger reservoirs were masked to avoid counting
dolostone. There is little coarse or intermediate clastic data points from the water surface.
rock. The Blue Ridge consists mainly of coarse and The main indicator of erosional resistance for a
intermediate clastic rock (including low-grade meta- given rock outcrop is the height of the outcrop. Most
sedimentary rock) and metaplutonic rock. previous studies have simply used altitude as the
measure of height. However, this approach does not
take into account the position of the outcrop in the
4. Methods regional drainage system. Upstream outcrops tend to
stand higher than downstream ones in a given drain-
The first step was to construct three grids of topo- age system, independently of differences in resistance.
graphic indices (regional relief, local relief, and slope) Clayton and Shamoon (1998) attempted to get around
for the study area, and then overlay a geology map this problem by using statistics derived from the rate
grid in order to determine the mean values of the at which elevation increases with distance inland from
indices for the various geologic units The topographic the sea. As my study area does not border the sea, and
index grids were produced from an 800-quadrangle in fact is connected to the sea via a very circuitous
DEM from the U.S. Geological Survey National route, this approach did not seem appropriate. Instead,
Elevation Dataset. The block included not only the a measure of relief was devised that is based on the
study area but most of the area that drains into eastern vertical distance of the land surface above base level,
Tennessee as well. where the latter is the surface defined by the altitude
The original DEM, obtained as an Arc Grid, was of streams having drainage areas of 100 km2 or
projected into the Tennessee State Plane projection in greater. To create this surface, elevation points along
ARC/INFO and given a resolution of 60 m. (This the streams were interpolated to produce a grid for the
decreased resolution was necessary to avoid excessive study area, using the Kriging option in Surfer. This
run times in processing the large grids. An exception grid was then smoothed by a 33  33-cell mean filter,
was made for the slope map, which was calculated then subtracted from the altitude grid to produce the
using a 30-m resolution and then transformed to a 60- ‘‘regional relief’’ grid (Fig. 3). Note that whereas the
m grid by thinning.) The grid was exported to Riv- altitude grid (Fig. 3A) shows a rise in altitude to the
erTools, which was used to generate a flow-accumu- NE, the regional relief grid (Fig. 3C) eliminates this
lation grid as well as a slope map. The grid was also trend. The mean regional relief was then calculated for
exported to Idrisi, which was used for all the analysis. each geologic unit.
Surfer was used for generating orthogonal plots from Although the regional relief measure is an im-
grids and for interpolating the base-level surface from provement, a problem remains. Note that although
stream elevations, as explained below. For the bed- mean regional relief values generally are much lower
rock geology, an unpublished digitized version of the than mean altitudes (Table 1), the regional relief plot
Ridge and Valley and Blue Ridge parts of the (Fig. 3C) is grossly similar to the altitude plot (Fig.
1:250,000-scale Geologic Map of Tennessee (Harde- 3A). Thus, despite correction for the effect of
man, 1966) was obtained. (This part of the map was regional drainage, the Blue Ridge still stands well
 H.H. Mills / Geomorphology 55 (2003) 263–281 267

Fig. 2. Map of rock types in study area. Low-grade metasedimentary rocks of Blue Ridge province have been included with sedimentary rocks.
Coordinates are in Tennessee State Plane (m).
Fig. 6. Map of bedrock units classified by erosional resistance index. Note that low-grade metasedimentary rocks have been classified together
with sedimentary rocks. Coordinates are in Tennessee State Plane (m).
268 H.H. Mills / Geomorphology 55 (2003) 263–281

Fig. 3. Orthogonal plots of study area. Coordinates are Tennessee State Plane, in meters. (A) Plot of altitude. Vertical exaggeration is
approximately 26. (B) Plot of base level derived from stream elevations. Vertical exaggeration is approximately 26. (C) Plot of regional relief
(altitude minus base level). Vertical exaggeration is approximately 14.
 H.H. Mills / Geomorphology 55 (2003) 263–281 269

Table 1
Mean altitude, regional relief, local relief, and slope of geologic units
Geologic unit Age Mean altitude S.D. Mean regional S.D. Mean local S.D. Mean slope S.D.
(m) relief (m) relief (m) (degrees)
Athens Shale O 291.8 31.4 39.9 28.9 4.89 29.6 8.74 6.55
Bays Formation O 418.2 128.6 110.4 92.3 37.1 69.9 13.07 7.43
Beech Granite pC 1001.0 146.1 297.6 157.7 23.0 106.3 16.60 7.61
Cades Sandstone pC 636.6 115.2 230.7 127.0 20.1 92.6 17.80 8.40
Chattanooga Shale M 404.5 58.3 67.8 44.7 52.9 43.6 12.02 7.07
Chepultepec Dolomite O 351.1 105.6 65.9 57.7 11.1 29.2 7.03 4.54
Chickamauga Group O 325.9 83.9 43.4 44.7 28.4 32.0 5.63 5.12
Clinch Sandstone S 573.4 89.8 245.7 85.9 122.8 83.5 20.55 6.57
Cochran Conglomerate C 638.5 121.0 263.0 109.2 93.3 94.6 19.30 8.10
Conasauga Group C 307.8 83.6 32.7 36.5 17.7 27.3 5.65 5.33
Copper Ridge Dolomite C 558.0 68.8 65.8 43.7 8.4 33.8 8.93 6.15
Cranberry Granite pC 865.9 142.3 207.6 144.3 42.8 98.8 16.44 7.75
Erwin Formation C 348.5 173.5 286.3 143.0 51.8 116.1 19.21 8.15
Fort Payne Chert M 294.1 84.3 55.6 58.0 30.4 56.9 10.34 6.92
Grainger Formation M 419.3 80.9 94.3 58.1 13.4 51.0 15.53 7.59
Great Smoky Group pC 848.8 335.8 402.9 292.7 24.2 132.3 20.14 9.26
Hampton Formation C 859.1 206.6 318.7 160.6 81.7 125.1 20.37 8.20
Hesse Sandstone C 837.9 124.2 307.7 116.9 154.3 93.5 14.47 7.86
Holston Formation O 296.4 31.6 42.7 27.6 5.2 26.0 8.09 5.26
Honaker Dolomite C 590.4 59.4 82.3 43.7 6.2 37.5 7.34 5.05
Juniata Formation O 539.3 68.8 206.0 68.2 90.4 61.3 21.80 5.82
Knox Group O–C 354.7 101.7 50.7 37.9 4.7 31.0 5.83 4.21
Lenoir Limestone O 299.4 44.0 30.1 26.8 7.1 19.8 4.93 3.53
Longview Dolomite and O 426.6 58.1 114.7 47.5 25.8 35.1 10.53 6.06
Chepultepec Dolomite
undivided
Longview Dolomite O 348.9 81.1 57.6 38.7 0.4 33.3 6.54 4.21
Martinsburg Shale O 417.0 77.2 97.7 65.9 0.6 45.8 10.59 6.42
Mascot Dolomite O 349.5 32.0 50.4 24.4 9.0 19.8 4.27 2.77
Maynardville Limestone C 490.3 75.0 47.9 42.8 1.7 27.3 8.99 6.76
Murray Shale C 319.0 137.8 327.4 124.9 167.2 100.7 16.63 8.38
Nebo Sandstone C 342.3 142.9 369.5 141.9 199.4 119.2 18.00 8.64
Newala Formation O 363.3 76.8 64.4 46.8 3.8 27.8 6.49 4.74
Newman Limestone M 369.4 136.4 86.9 84.4 23.5 65.6 10.94 8.21
Nichols Shale C 640.0 142.4 327.1 134.3 152.8 107.4 20.44 8.92
Nolichucky Shale C 593.3 60.4 50.8 38.1 14.3 23.7 7.93 6.06
Ocoee Supergroup pC 640.4 145.0 228.3 124.0 17.1 89.1 15.99 7.82
Ottosee Shale O 283.5 24.5 30.2 20.8 9.0 19.1 5.56 3.98
Pennington Formation M 426.9 152.3 142.2 114.0 18.7 92.8 15.06 8.00
Pumpkin Valley Shale C 602.7 48.4 54.8 38.4 19.3 39.9 10.61 6.71
Rich Butt Sandstone pC 757.6 198.5 368.1 180.9 106.5 110.0 22.29 7.86
Roan Gneiss pC 1082.0 204.1 343.7 177.1 40.1 124.4 19.61 7.61
Rockwood Formation S 539.6 89.1 195.3 82.2 77.6 74.9 17.67 5.88
and Clinch Formation
undivided
Rockwood Formation S 313.8 67.4 70.6 54.8 0.4 59.6 11.31 6.87
Rome Formation C 426.8 178.2 72.4 53.5 15.7 55.1 11.59 7.77
Sandsuck Formation pC 403.3 159.5 119.9 99.2 22.4 62.3 13.66 8.08
Sequatchie Formation O 293.9 73.1 43.7 46.2 39.6 51.3 9.05 6.56
Sevier Shale O 390.3 82.6 45.1 44.3 15.4 35.9 8.18 6.55
Shady Dolomite C 656.7 171.1 109.7 77.9 61.4 69.0 11.27 7.50
(continued on next page)
270 H.H. Mills / Geomorphology 55 (2003) 263–281

Table 1 (continued )
Geologic unit Age Mean altitude S.D. Mean regional S.D. Mean local S.D. Mean slope S.D.
(m) relief (m) relief (m) (degrees)
Snowbird Group pC 653.8 202.6 228.4 166.8 31.2 105.0 17.49 8.46
Unicoi Formation C 832.2 222.1 296.8 164.4 55.6 116.0 20.19 8.52
Walden Creek Group pC 429.8 84.9 121.2 71.1 18.3 56.2 15.38 8.37
Abbreviations: pC = Precambrian, C = Cambrian, O = Ordovician, S = Silurian, M = Mississippian, S.D. = standard deviation. Note that Blue
Ridge province bedrock units are Precambrian, except for units in the foothills, which are largely Cambrian. Ridge and Valley province units
range from Cambrian to Mississippian.

above the Ridge and Valley. Does this mean that all The bedrock units with the highest mean local
the rocks in the Blue Ridge are more resistant than relief values occur in the Blue Ridge foothills and
all those in the Ridge and Valley? Not necessarily. Ridge and Valley province, whereas the values on the
Perhaps a preponderance of rocks in the one prov- high peaks of the Blue Ridge are relatively low. Such
ince is harder than the preponderance of rocks in the low values probably stem in part from the likelihood
other and that is sufficient to produce the observed that many bedrock units in this province are similar
difference in regional relief. Thus, nonresistant out- in erosional resistance, so differential erosion does
crops may appear resistant simply because of the not produce high local relief. A related factor is that
influence of their neighbors. To address this diffi- many outcrops of rocks in this province are quite
culty, a second measure of relief was devised: ‘‘local broad, so that in the relatively small area imposed by
relief,’’ which determines relief in a somewhat an 8-km window, the generalized surface used for
smaller area, thus allowing rock units to be com- measuring relief is computed mainly on the same unit
pared with other units in the immediate vicinity. The rather than on multiple units with varying resistance.
desired area was that occupied by a typical mountain In contrast, although it is true that rocks on the whole
and adjacent valleys, commonly contained within a are softer in the Ridge and Valley, it is also true that
7.5-min quadrangle. An area of 8  8 km (about there is much more variability in rock resistance than
40% of the area of a quadrangle) seemed appropri- in the Blue Ridge. A hard rock unit surrounded by
ate. To accomplish this, a generalized surface of the soft ones develops much more local relief in this
study area was first created by applying an 8  8-km province than would a comparable unit in the Blue
(133  133 cells) unweighted mean filter to the Ridge. For resistant units, the regional relief measure
altitude grid (Fig. 4A). This surface, when subtracted thus tends to produce higher values in the Blue Ridge
from the altitude grid, provides a measure of relief than in the Ridge and Valley, whereas the reverse is
within an 8  8-km square (Fig. 4B). The result of true for the local relief measure. It therefore seemed
this subtraction for a given cell is the vertical appropriate to use both measures to evaluate ero-
distance above or below the generalized surface. sional resistance.
The mean local relief is then calculated for each A third measure of rock resistance is slope. This
geologic unit, as was done for the regional relief. was determined by calculating a standard queen’s-
Note that the generalized surface made with the 8- case (eight-way adjacency) slope map. The steepest
km-wide mean filter (Fig. 4A) is much more irreg- slopes (Fig. 4C) occur in the Blue Ridge, the foothills
ular than the stream base-level surface (Fig. 3B), and of the Blue Ridge, and on some ridges in the NW part
unlike the regional relief plot, the local relief plot of the Ridge and Valley. Compare the plot of slope
(Fig. 4B) shows little resemblance to the altitude plot (Fig. 4C) with that of altitude (Fig. 3A). Note that the
(Fig. 3A). Note that whereas the foothills of the Blue foothills on this surface are much more prominent
Ridge (e.g., Chilhowee Mountain in Fig. 1) are much than they are in the actual topography, because
lower than the high peaks in altitude, they are much although low in altitude relative to the Blue Ridge,
closer in amplitude on the local relief plot because their slopes are somewhat steeper. Note also that in
they are steep and narrow and thus have high local the Blue Ridge, the slope surface has few peaks,
relief. because the slope angles are fairly uniform here.
 H.H. Mills / Geomorphology 55 (2003) 263–281 271

Fig. 4. Orthogonal plots of study area. Coordinates are Tennessee State Plane, in meters. (A) Plot of generalized surface (altitude smoothed with
8-km-wide mean filter). Vertical exaggeration is approximately 14. (B) Plot of local relief (altitude minus generalized surface). Vertical
exaggeration is approximately 43. (C) Plot of slope (degrees) derived from altitude grid with 30-m resolution.
272 H.H. Mills / Geomorphology 55 (2003) 263–281

Table 2
Topographic indices and other data for bedrock units
ERI Geologic unit Lithology ERI Regional Local Slope Hatcher’s Area Thickness
rank relief relief index rating (km2) (m)
index index
1 Nebo Sandstone CC 89.0 91.0 100.0 76.1 75.0 39.3 76.2
2 Rich Butt Sandstone CC-MS 85.0 90.7 64.4 100.0 70.0 25.4 457.2
3 Nichols Shale FC 83.8 79.7 82.1 89.7 25.0 44.6 213.4
4 Murray Shale FC 78.7 79.7 87.7 68.6 25.0 26.7 152.4
5 Hampton Formation IC 73.9 77.4 54.9 89.3 45.0 304.1 381.0
6 Great Smoky Group IC-MS 73.6 100.0 32.8 88.1 85.0 1238.0 5943.3
7 Clinch Sandstone CC 72.9 57.8 70.7 90.3 70.0 108.1 182.9
8 Hesse Sandstone CC 71.2 74.4 82.7 56.6 67.5 60.3 182.9
9 Cochran Conglomerate CC 68.4 62.5 59.3 83.4 67.5 96.2 365.7
10 Unicoi Formation IC 68.2 71.5 44.9 88.3 75.0 469.6 1066.7
11 Juniata Formation FC 67.6 47.2 58.2 97.3 25.0 35.0 91.4
12 Erwin Formation CC 65.0 68.7 43.4 82.9 60.0 444.0 381.0
13 Roan Gneiss MP 59.1 84.1 8.2 85.1 60.0 67.1 –
14 Beech Granite MP 57.5 71.8 32.3 68.4 60.0 259.4 –
15 Rockwood Fm and IC 57.3 44.3 53.3 74.4 55.0 62.1 213.4
Clinch Sandstone
undivided
16 Cades Sandstone IC-MS 53.4 53.8 31.3 75.1 85.0 101.4 457.2
17 Snowbird Group IC-MS 46.0 53.2 11.6 73.4 60.0 522.7 5029.0
18 Ocoee Supergroup CC-MS 45.1 53.2 17.0 65.0 62.5 175.8 15,239.0
19 Cranberry Granite MP 40.7 47.6 7.1 67.5 50.0 182.8 –
20 Pennington Formation FC 40.2 30.1 30.7 59.9 25.0 173.6 167.6
21 Bays Formation IC 36.0 21.5 37.8 48.8 40.0 233.1 304.8
22 Walden Creek Group IC-MS 34.2 24.4 16.5 61.7 40.0 1176.0 2438.3
23 Grainger Formation FC 32.7 17.2 18.4 62.5 40.0 95.9 365.7
24 Sandsuck Formation IC-MS 30.4 24.1 14.9 52.1 30.0 266.4 609.6
25 Longview Dolomite and LD 30.3 22.7 33.4 34.7 22.5 517.1 335.3
Chepultepec Dolomite
undivided
26 Martinsburg Shale CF 25.6 18.1 23.8 35.0 20.0 164.8 304.8
27 Rockwood Formation IC 24.4 10.9 23.4 39.1 40.0 128.0 137.2
28 Rome Formation CF 23.2 11.3 17.5 40.6 40.0 1086.0 609.6
29 Newman Limestone CF 22.3 15.2 14.5 37.0 10.0 190.1 213.4
30 Copper Ridge Dolomite DL 20.7 9.6 26.7 25.8 35.0 2068.0 304.8
31 Shady Dolomite LD 20.1 21.3 00.0 38.9 10.0 355.6 304.8
32 Pumpkin Valley Shale SH 19.3 6.6 16.1 35.2 20.0 47.7 106.7
33 Chattanooga Shale SH 18.8 10.1 3.3 43.0 15.0 112.6 152.4
34 Maynardville Limestone LS 18.0 4.8 22.9 26.2 15.0 317.1 83.8
35 Chepultepec Dolomite DL 17.6 9.6 27.8 15.3 15.0 228.4 243.8
36 Fort Payne Chert CH 17.5 6.8 11.9 33.7 35.0 112.1 91.4
37 Honaker Dolomite DL 17.4 14.0 21.2 17.0 12.0 372.5 457.2
38 Holston Formation CF 16.7 3.4 25.5 21.2 17.5 297.4 121.9
39 Athens Shale SH 16.4 2.6 21.7 24.8 10.0 334.6 457.2
40 Nolichucky Shale FC 14.6 5.6 18.1 20.3 25.0 78.9 152.4
41 Newala Formation LD 14.5 9.2 22.1 12.3 15.0 627.4 251.5
42 Longview Dolomite LD 14.5 7.4 23.4 12.6 30.0 109.9 91.4
43 Sevier Shale SH 14.5 4.0 17.6 21.7 15.0 2456.0 1371.5
44 Sequatchie Formation CF 12.8 3.7 8.4 26.5 27.5 68.0 91.4
45 Knox Group LD 12.0 5.5 21.7 8.7 20.0 5376.0 838.2
46 Ottosee Shale SH 9.09 0.0 20.1 7.2 12.5 483.0 304.8
 H.H. Mills / Geomorphology 55 (2003) 263–281 273

Table 2 (continued)
ERI Geologic unit Lithology ERI Regional Local Slope Hatcher’s Area Thickness
rank relief relief index rating (km2) (m)
index index
47 Mascot Dolomite LD 8.51 5.4 20.1 0.0 35.0 28.0 175.3
48 Conasauga Group CF 8.37 0.7 16.8 7.7 15.0 2624.0 609.6
49 Lenoir Limestone CF 8.16 0.0 20.8 3.7 10.0 206.4 80.0
50 Chickamauga Group CF 7.92 3.6 12.7 7.6 30.0 1127.0 2011.6
Abbreviations: ERI = erosional resistance index, CC = coarse clastics (sandstone and conglomerate), CF = carbonate and fine clastics,
CH = chert, DL = dolostone, FC = fine clastics (shale, siltstone, some thin-bedded sandstone), IC = intermediate clastics (sandstone with shale
and siltstone), LD = limestone and dolostone, LS = limestone, MP = metaplutonic rocks, -MS = low-grade metasedimentary rocks, SH = shale.

To allow the three measures discussed above to be regional relief is a fairly high 0.544, whereas that
combined, the values of each were normalized to a between altitude and local relief is virtually zero
value range of 0 –100; these products will be referred (0.043), in agreement with the visual impressions
to as the regional relief index, the local relief index, given by the orthogonal plots. Another interesting
and the slope index. The indices were weighted finding is that whereas the correlation between slope
equally and averaged to yield the erosional resistance and altitude is only 0.435, that between slope and
index (ERI). The mean values of all indices for all 50 regional relief is a very high 0.800. On the other hand,
geologic units are listed in Table 2. the correlation between slope and local relief is only
The final step in the analysis was to determine the 0.260.
effect of rock resistance on stream location. To accom- Fig. 5 illustrates variations and relationships of
plish this, a flow-accumulation grid was first calcu- erosional resistance indices in the two provinces.
lated. (In a flow-accumulation grid, the value in each Altitude has been normalized to a 0 –100 range
cell is the number of cells that drain to that cell. This in analogous to the indices. In Fig. 5A, Great Smoky
effect provides a stream map of the area.) This grid was Group and Roan Gneiss are in the Blue Ridge
much larger than the actual east Tennessee study area in province. Typically, resistant units in this province
order to allow an accurate determination of upstream show high mean altitudes and high regional relief
areas for streams that head outside of the study area and indices but relatively low local relief indices. Slope
then flow into it. Two measures of the effect of ERI on indices are high. Thus, two out of three indices are
stream location were determined. First, a distance map high, resulting in relatively high ERI values.
was constructed, giving the distance of each grid cell to The Cochran Conglomerate and Hesse Sandstone
the nearest stream. In this manner, the mean distance to are resistant units from the foothills. Note that values
the nearest stream was calculated for each bedrock unit of regional relief and local relief indices are similar,
on the geology map. Second, the relative abundance of unlike those of the two Blue Ridge units. Fig. 5B
streams on each rock unit was determined by creating a shows units from the Ridge and Valley province. The
boolean stream map and then using the Idrisi
EXTRACT command to determine the total number
of stream cells associated with each rock unit. The idea
Table 3
is that rock units with low ERI values should tend to be R2 values between topographic indices for 47 bedrock units
closer to streams and to have more streams overlying
Altitude Regional Local Slope
them than should units with high ERI values. relief relief index
index index
Altitude 1.000 0.544 0.043 0.435
5. Erosional resistance index and lithology Regional 0.544 1.000 0.398 0.800
relief
The correlation table for the topographic indices Local relief 0.043 0.398 1.000 0.260
(Table 3) shows that the R2 between altitude and Slope 0.435 0.800 0.260 1.000
274 H.H. Mills / Geomorphology 55 (2003) 263–281

Fig. 5. Comparison of dimensionless indices (values 0 – 100) of mean altitude, regional relief, local relief, and slope for nine bedrock units. (A)
Blue Ridge and foothills. (B) Ridge and Valley.

Chickamauga Group and Ottosee Shale are two of the (IC), and metaplutonic rocks (MP). Units showing the
weakest units of the province. They show very low lowest values generally are limestone (LS), limestone
values of all indices. Copper Ridge Dolomite and plus dolostone (LD), carbonate plus fine clastics (CF),
Rome Formation are slightly stronger, commonly and shale (SH). Dolostone (DL) values generally are
forming low ridges. Their indices are moderately higher than those of the low group, but much lower
higher than those of the weaker two units. The Clinch than those of the high group. Fig. 6 shows geologic
is a strong quartz sandstone and shows a higher local units mapped by ERI values; it may be compared to
than regional relief index. Its regional relief is still the geologic map (Fig. 2). In Table 2, note that among
relatively high, however, and with the aid of a high the units with the highest ERI values are two shales
slope index, the unit has a high ERI value. and the Juniata Formation, which is mainly shale.
Table 2 shows the 50 geologic units ranked in Their high ERI values appear as suspect and will be
order by ERI value. Also shown are the rock types so discussed further below.
that it is possible to see the association of ERI to One check on these findings is to compare them
lithology. Units showing the highest ERI values gen- with our subjective knowledge of the association
erally are coarse clastics (CC), intermediate clastics between lithology and topography. Of course, the
 H.H. Mills / Geomorphology 55 (2003) 263–281 275

hope is that the ERI will be more dependable than the

subjective knowledge, but nevertheless, a wide dis-
parity between the two would raise suspicions about
the ERI. In order to carry out this test, the subjective
ratings of bedrock resistance would best be made by
someone very well acquainted with rock units of the
Ridge and Valley and Blue Ridge province. Therefore,
Prof. Robert D. Hatcher Jr. of the University of
Tennessee-Knoxville, an authority on southern Appa-
lachian geology with decades of field experience in
east Tennessee, was requested to rate the erosional
resistance of the units in Tables 1 and 2. Hatcher said
he based his ratings on his knowledge of the compo-
sition of each unit, together with ‘‘. . .the kinds of
slopes or valleys or ridges that specific rock units
produce on a local (say part of a [7.5-min] quadran-
gle) scale’’ (Hatcher, personal communication, 2001).
Fig. 7 shows Hatcher’s ratings (expressed as 0–
100) plotted against ERI values. There is fairly good
agreement between the two. Of the disparities, the
lesser problem concerns units that have ERI values

Fig. 8. Plot of erosional resistance index values vs. distance to

nearest hard (ERI>50) bedrock outcrop for units with ERI values at
least 50% higher than Hatcher’s ratings.

that seem too low based on Hatcher’s ratings. The

reasons for low ERI values for Great Smoky Group
and Cades Sandstone, both Blue Ridge units, have
been discussed previously. The greater problem, how-
ever, is that of units which, based on their lithology,
would seem to be nonresistant but which have very
high ERI values. The most glaring examples are the
Juniata Formation, the Murray Shale, and the Nichols
Shale. How might such high values for shales be
explained? All three are relatively thin units immedi-
ately overlain by resistant sandstones. The Juniata un-
derlies the Clinch Sandstone (ERI = 72.9), the Murray
underlies the Hesse Sandstone (ERI = 71.2), and the
Nichols underlies the Nebo Sandstone (ERI = 89.0).
The relief and slope of these shale units are thus
determined by the overlying sandstones.
Although the cause of the high ERI values for
these units was found by map inspection, such erro-
Fig. 7. Plot of subjective ratings of erosional resistance by Robert D. neous values can be detected in a more systematic
Hatcher Jr. vs. computed erosional resistance index values. way. For each unit with an ERI more than 50% greater
276 H.H. Mills / Geomorphology 55 (2003) 263–281

than Hatcher’s rating, the mean distance to the nearest solved simply by looking at the mean altitudes of
hard rock unit was calculated, where ‘‘hard’’ was the two units—the higher unit being the controlling
defined by an ERI of at least 50. (The value of 50 one.).
was selected from an inspection of Table 2, which The Hampton Formation and the Shady Dolomite
shows that the cutoff between what we normally show a more distant association with resistant units,
consider soft rock and what we consider hard falls while the Pennington Formation, the Newman Lime-
in the 40 –60 interval, and so a median value of 50 stone, and the Athens Shale lie so far from the nearest
was used.) Fig. 8 confirms the map observations—the hard rock that their apparently excessive ERI values
three suspect shales are shown to lie very close to hard (relative to Hatcher’s ratings) cannot be attributed to
bedrock, on average, little more than 100 m. (Note that this cause. On the basis of these findings, the Juniata
for two closely associated units, the question might be Formation, Murray Shale, and Nichols Shale were
raised of which unit controls the ERI of the other. For eliminated from further consideration, as their ERI
example, the Murray Shale [ERI = 78.7] occurs very values are clearly misleading.
close to the Hesse Sandstone [ERI = 71.2]—might Note that the correlation of Hatcher’s ratings with
the Murray control the ERI of the Hesse, instead of the ERI is R2 = 0.764 after omission of the three above
vice-versa? Above, we simply assumed that the shale units. Looking at the R2 values between his
sandstone controlled the shale. However, in a less ratings and regional relief (0.762), local relief (0.345),
obvious case, this question could usually be re- slope (0.696), and altitude (0.360) is also of interest.

Fig. 9. Stream network superposed on enlarged portion of Ridge and Valley geological map, showing affinity of streams for units with lowest
ERI values. Coordinates are in Tennessee State Plane (m).
 H.H. Mills / Geomorphology 55 (2003) 263–281 277

Despite his statement that he considered topography outcrop were determined. Each of these variables was
over part of a quadrangle, the correlation between his then correlated with area. The idea was to see whether
ratings and local relief is much lower than for regional resistant units would produce significant positive
relief. His ratings also show a fairly high correlation correlations and nonresistant units, significant nega-
with slope, although this correlation may simply be tive correlations. This procedure was carried out for
due to the high correlation (0.800) that exists between 27 bedrock units. The highest positive correlation
regional relief and slope (Table 3). found was an R2 of 0.318 for the local relief index
of the Cochran Conglomerate, and the highest neg-
ative correlation was an R2 of 0.115 for the altitude of
6. Other factors that may affect ERI values the Knox Group. Most correlations, however, were
much lower, with the average positive correlation (R2)
Several other factors that might affect the ERI were for all variables being 0.020 and the average negative
investigated. correlation (R2) being 0.017.
(i) Thickness of bedrock units: The idea is that Thus, the above three variables apparently have a
thicker units would be better able to demonstrate their small-to-negligible influence on the ERI and related
erosional resistance. That is, thick resistant units topographic indices.
would tend to resist erosion more than thin ones and
so would tend to show higher relief and steeper
slopes. Thick nonresistant units, in contrast, would 7. Effect of ERI on stream location
erode more rapidly than thin ones (which might be
influenced by nearby resistant rocks), resulting in Streams should have more of an affinity for bed-
lower relief and slope. This hypothesis was tested rock units with low ERI values than for those with
by, first, correlating the ERIs of all the harder rocks high ERI units, as the former should tend to underlie
(ERI>50) with their thicknesses (obtained from
descriptions on the state geologic map), the expect-
ation being that a significant positive correlation
would result. Second, the ERIs of all the weaker
rocks (ERI < 20) were correlated, the expectation
being that a significant negative correlation would
result. However, the correlation (R2) for the harder
rocks was only 0.003 and that for the weaker was
0.002.
(ii) Age of bedrock units: Clayton and Shamoon
(1998) showed a fairly good correlation between rank
order of geologic periods and relative relief in Great
Britain. An R2 value of only 0.104 between this age
measure and ERI was found. However, unlike Clayton
and Shamoon, no Cenozoic or Mesozoic rocks are
present in this study area, so the range in ages may
have been too small to see a higher correlation.
(iii) Area of outcrop: The idea was that, for a given
unit, really larger outcrops, compared to smaller out-
crops, are more likely to stand higher and have steeper
slopes if the unit is resistant and to stand lower and
have lower slopes if the unit is nonresistant. This
hypothesis was tested as follows. For each bedrock
unit having at least 20 separate outcrops, the mean Fig. 10. Plot of mean distance to nearest stream vs. erosional
altitude, slope index, and local relief index for each resistance index values.
278 H.H. Mills / Geomorphology 55 (2003) 263–281

to the nearest stream (defined in this case as having a

drainage area of at least 3 km2, for reasons made clear
below) was first determined. In Fig. 10, the mean
distance to the nearest stream has been calculated for
each geologic unit and plotted against the ERI for that
unit. A fairly high correlation between distance and
ERI results (R2 = 0.685). As Fig. 10 shows, the mean
distance to the nearest stream is nearly three times as
great for units with the highest ERI values as it is for
units with the lowest values.
A follow-up question that might be asked about
this relationship concerns the size of the streams
involved. Do all streams show a relation like this or
only streams of a certain size? In Fig. 11, this question
is explored by plotting the R2 values between the ERI
and the mean distance to streams with the specified
drainage area against the drainage area. Note that the
highest correlations are found for a stream size
between 1 and 30 km2 (the very highest is for 3
km2, hence the use of this value in Fig. 10). Beyond
30 km2, the R2 value drops off dramatically, declining
to 0.1 for 100 km2.
Fig. 11. Plot of correlation (R2) between ERI and mean distance to
stream of specified drainage basin as a function of drainage basin
area.

valleys and the latter should tend to underlie ridges. In

Fig. 9, streams draining at least 1 km2 are super-
imposed on an enlarged section of the Ridge and
Valley geologic map. The ERI values are shown for
selected bedrock units. Note that strike-following
streams consistently flow on the units with the lowest
ERI values: the Chickamauga (7.9) and the Cona-
sauga (8.4). For units with higher, but still relatively
low, ERI values, such as the Rome (23.2) and Copper
Ridge Dolomite (20.7), streams arise on or cut across
the outcrops but do not follow their strike. For units
with high ERI values, such as the Clinch (72.9), not a
single stream occurs on the outcrops.
Inspection of other Ridge and Valley areas on the
state geologic map shows similar results. However,
demonstrating the relationship between ERI and
stream location systematically for the whole study
area is desirable. Observations such as those provided
by Fig. 9 suggest that, on average, bedrock units with
low ERI values should be closer to, and more likely to Fig. 12. Plot of proportion of grid cells with flow-accumulation
be overlain by, streams. To test the first hypothesis, for value of 1 km2 or greater on a given geologic unit vs. the ERI of
each grid cell of each geologic unit, the linear distance that unit.
 H.H. Mills / Geomorphology 55 (2003) 263–281 279

An alternative approach to the relationship of results show high values for most of the smaller
drainage to bedrock erosional resistance is to see if stream sizes, with a maximum again in the 1 –30
streams are more likely to overlie units with low ERI km2 range. A steep drop-off in correlation occurs
values than they are units with high ERI values. This beyond 30 km2. The decline is not as precipitous as
relationship was tested by determining, for each seen for stream distance, but the results still show
geologic unit, the proportion of cells that are overlain low correlations for the largest streams. Thus,
by grid cells with flow-accumulation values of 1 km2 whereas relatively small streams show a strong rela-
or greater. This proportion was then plotted against tionship to bedrock resistance, large ones appear to
the ERI for each geologic unit. Fig. 12 shows the be relatively independent of this factor.
expected relationship, although the correlation is
lower than that obtained for the stream distance case.
Note that the proportion of geology grid cells over- 8. Discussion
lain by stream grid cells is more than six times
greater for units with low ERI values than for units The correlations between topographic indices
with high ERI values. (Table 3) show wide differences that suggest the
In a manner analogous to that done for stream two relief indices are distinct. As might be expected,
distance, the effect of stream size on this relationship the R2 value between regional relief and altitude is
can be determined. In Fig. 13, the R2 value between fairly high (0.544). Less expected, however, is a
ERI and the proportion of cells of the specified flow- very strong 0.800 between regional relief and slope.
accumulation area overlying each geological unit is Part of this may be explained by the tendency of the
plotted against the flow-accumulation area. The high mountains to have steeper slopes, yet it is not
simply an effect of altitude, as the R2 between slope
and altitude is only 0.435. The local relief index is
very different from the regional relief index, show-
ing an R 2 with altitude of only 0.043, and a
relatively low R2 of 0.260 with slope. Because slope
has such a high R2 with regional relief (0.800), the
two relief variables alone can account for most of
the variance.
The two relief indices used in this study may be
criticized in that they are arbitrary: 100 km2 or
greater for the streams used to interpolate the base-
level surface for the regional relief index, and 8 km
for the width of the moving window used to create
the generalized surface for the local relief index.
However, although the details would change accord-
ing to the specific numbers, the same general results
would probably be obtained as long as one index was
based on large areas and the other based on small
areas. Also, two different methods were used to
generate the indices, but very likely, either method
could be used to generate both indices.
One problem with analysis of lithologic effects
based on small-scale geologic maps such as the
Geologic Map of Tennessee (Hardeman, 1966) is
Fig. 13. Plot of correlation (R2) between ERI of a geologic unit and that, of necessity, many mapping units are quite
proportion of cells with specified flow-accumulation area on that broad. Commonly, several disparate lithologic units
unit, as a function of flow-accumulation area. are combined to make larger map units. For exam-
280 H.H. Mills / Geomorphology 55 (2003) 263–281

ple, the Chickamauga Group includes (of the units context. Another problem is that such schemes com-
considered herein) the Lenoir Limestone, the Hol- monly deal with susceptibility to slope failure. How-
ston Formation, and the Ottosee Shale. Obviously, ever, if the concern is with long-term landform
greater accuracy would be achieved by measuring evolution, solubility becomes an important factor.
relief and slope of the individual units comprising Limestones in the Appalachians, for example, may
the Chickamauga Group than of the Chickamauga have high mechanical strength yet end up forming the
as a whole, but the area of the study area mapped valley floor.
simply as Chickamauga Group far exceeds the area Several authors have related topography in the
mapped as individual units of the Chickamauga. Appalachians to measurable characteristics of the bed-
This generalization obviously decreases the relation- rock. Significantly, in each case, the measured param-
ship of map units to rock type and therefore to eter reflected the susceptibility of bedrock to chemical
topography. To minimize this problem ideally, anal- weathering. Rahn (1971), for example, measured
yses such as the current one would be done using weathering rates on tombstones in New England and
1:24,000-scale geoquads. However, in areas where showed that the rates so determined can predict the
only a small fraction of quadrangles have been average altitude of rock types within that region. Flint
mapped, as in east Tennessee, this option is not yet (1963) noted that the metamorphic rocks underlying
available. the highest areas in the Connecticut Fall Zone were
The possible effect of neotectonics has been characterized by greater quartz content than metamor-
ignored in this paper. The assumption has been made phic rocks underlying the lowest areas. Costa and
that because of the thick crust in the Appalachians Cleaves (1984) reported similar findings for the Pied-
and the relatively small size of the study area, the mont region of Maryland. Therefore, it is likely that the
only likely vertical movement would be a relatively appropriate rock parameters to measure in the Appa-
uniform uplift due to isostatic adjustment to denuda- lachians will differ from those used in alpine or dry
tion. Such uplift should have no differential effect on regions.
relief and slope of bedrock units within the study The finding that the mean ERI value of a geologic
area. If other Cenozoic tectonic activity has taken unit helps predict the affinity of streams for that unit,
place in the study area, then interpretation of rock with units having low mean values being on average
resistance may be in error. If, for example, the Blue both closer to the nearest stream and more likely to
Ridge rises above the Ridge and Valley partly be overlain by a stream than are units having high
because of differential uplift, then the resistance of mean values, is not surprising. Likewise, the finding
rocks in that province has been overestimated. Hack that this effect becomes weaker for larger streams is
(1980) has observed, however, that much of the compatible with the accepted knowledge that the
topography in the Appalachian Highlands can be courses of master streams are less controlled by
explained by differential erosion alone. bedrock outcrop distribution than are their smaller
Although not attempted in this paper, ultimately tributaries, in part because their courses are thought
the goal of studies such as this should be to compare to be determined as much by inheritance as by rock
the topographic indicators of erosional resistance to control (e.g., Feldman et al., 1968). However, the
physical parameters of the rocks. Applying Selby’s finding that the effect of rock type on stream location
(1980) Rock Mass Strength Classification would be begins to decline at a drainage basin area of 30 km2
difficult in this environment because this method is perhaps surprising, as master streams generally
requires that outcrops must be on slopes that are have drainage areas much larger than this.
natural and not actively undercut. In the Appalachian
setting, natural outcrops of all but the hardest rocks
are rare. An alternative method might be one in which 9. Conclusions
measurements are made on fresh unweathered rock,
mainly from artificial cuts and cores. Goudie (1990), The ERI (composed of indices of regional relief,
following Bell (1983), suggested some additional local relief, and slope) is suggested as an index to
parameters that might be useful to measure in this characterize the erosional resistance of bedrock units
 H.H. Mills / Geomorphology 55 (2003) 263–281 281

on a geologic map. The ranking of 47 geologic units References

in the study area by this index produces reasonable
results based on their likely resistances inferred from Anonymous, 1999. National elevation dataset fact sheet. U.S. Geo-
lithology. Coarse clastic, intermediate clastic, and logical Survey (1 – 4 pp.).
Bell, F.G., 1983. Fundamentals of Engineering Geology. Butter-
metaplutonic rocks have the highest ERI values, with
worth, London. 648 pp.
limestone, limestone plus dolostone, and limestone Clayton, K., Shamoon, N., 1998. A new approach to the relief of
plus fine clastics having the lowest. Dolomite gener- Great Britain: II. A classification of rocks based on relative
ally has a value higher than the second group, resistance to denudation. Geomorphology 25, 155 – 171.
although much less than the first group. Unit thick- Cooper, B.N., 1944. Geology and mineral resources of the Burkes
Garden Quadrangle, Virginia. Virginia Geological Survey Bul-
ness, area of individual outcrops, and age of bedrock
letin, vol. 60. Virginia Geological Survey, Charlottesville, VA,
seem to have little effect on ERI. Mean ERI values 299 pp.
accurately predict stream location in the Ridge and Costa, J.E., Cleaves, E.T., 1984. The piedmont landscape of Mary-
Valley, with along-strike streams generally located on land: a new look at an old problem. Earth Surface Processes and
units with lowest mean values. Statistically, geologic Landforms 9, 59 – 74.
Feldman, S., Harris, S.A., Fairbridge, R.W., 1968. Drainage patterns.
units with the lowest ERI values tend to be three times In: Fairbridge, R.W. (Ed.), Encyclopedia of Geomorphology.
closer to the nearest stream and six times more likely Dowden, Hutchinson & Ross, Stroudsburg, PA, pp. 284 – 291.
to have a stream flowing on them. The correlation Flint, R.F., 1963. Altitude, lithology, and the Fall Zone in Connect-
between ERI value and mean distance of the unit to icut. Journal of Geology 17, 683 – 697.
the nearest stream and between ERI and the propor- Goudie, A., 1990. The Landforms of England and Wales. Black-
well, Oxford, UK.
tion of the unit outcrop occupied by streams are
Grender, G.C., 1973. Slope distribution and relationship to bedrock
highest for streams with drainage areas between 1 in Appalachian ridges near Roanoke, Virginia. American Jour-
and 30 km2, suggesting that the courses of larger nal of Science 273A, 391 – 395.
streams are less affected by bedrock character than are Hack, J.T., 1980. Rock control and tectonism—their importance in
the courses of smaller streams. The concept that the shaping the Appalachian Highlands. U.S. Geological Survey
1126B, B1 – B17.
courses of large master streams are less affected by
Hardeman, W.D., 1966. Geologic Map of Tennessee. Tennessee Di-
geology than are their tributaries is not new. What the vision of Geology Map, Nashville, TN, 4 sheets, scale 1:250,000.
analyses here do is to pin down the size of stream that Hayes, C.W., 1899. Physiography of the Chattanooga District in
is most affected by bedrock resistance. Tennessee, Georgia, and Alabama. U.S. Geological Survey An-
Future research would probably best be carried out nual Report, Pt. 2, Washington, DC, pp. 1 – 58.
Hollyday, E.F., Hileman, G.E., 1996. Hydrogeologic terranes and
on 7.5-min geoquads, where map units are less gener-
potential yield of water to wells in the Ridge and Valley prov-
alized and finer-resolution DEMs can be employed. For ince in the eastern and southeastern United States. U.S. Geo-
example, measures of hillslope convexity and concav- logical Survey Professional Paper, vol. 1422-C. U.S. Geological
ity based on 10-m resolution DEMs may well be Survey, U.S. Government Printing Office, Washington, DC,
capable of differentiating bedrock units. pp. C1 – C30.
Lesley, J.P., 1856. Manual of Coal and its Topography. J.B. Lip-
pincott, Philadelphia. 216 pp.
Moore, I.D., Grayson, R.B., Ladson, A.R., 1991. Digital terrain
modelling: a review of hydrological, geomorphological, and
Acknowledgements
biological applications. Hydrological Processes 5, 3 – 30.
Rahn, P.H., 1971. The weathering of tombstones and its relationship
Funds for NED data and some software were to the topography of New England. Journal of Geological Edu-
provided by a Tennessee Technological University cation 19, 112 – 118.
Faculty Research grant. Thanks to Prof. Robert D. Selby, M.J., 1980. A rock mass strength classification for geomor-
Hatcher Jr. for his rating of east Tennessee geologic phic purposes with tests from Antarctica and New Zealand.
Zeitschrift für Geomorphologie N.F. 24, 31 – 51.
units, to Scott Lucas for his assistance in an early Thompson, H.D., 1941. Topographic of the analysis of the Monte-
version of this project, and to Dr. Peter Li for his helpful rey, Staunton, and Harrisonburg Quadrangles. Journal of Geol-
advice. ogy 49, 521 – 549.
 Geomorphology 55 (2003) 283 – 303
www.elsevier.com/locate/geomorph

Glacial erosion and geomorphology in the northwest

Sierra Nevada, CA
L. Allan James *
Department of Geography, University South Carolina, Columbia, SC 29208, USA

Received 27 November 2001; received in revised form 17 April 2002; accepted 10 March 2003

Abstract

Pleistocene glacial erosion left a strong topographic imprint in the northwestern Sierra Nevada at many scales, yet the specific
landforms and the processes that created them have not been previously documented in the region. In contrast, glaciation in the
southern and central Sierra was extensively studied and by the end of the 19th century was among the best understood examples
of alpine glaciation outside of the European Alps. This study describes glacially eroded features in the northwest Sierra and
presents inferred linkages between erosional forms and Pleistocene glacial processes. Many relationships corroborate theoretical
geomorphic principles. These include the occurrence of whalebacks in deep ice positions, roches moutonnées under thin ice, and
occurrence of P-forms in low topographic positions where high subglacial meltwater pressures were likely. Some of the
landforms described here have not previously been noted in the Sierra, including a large crag and tail eroded by shallow ice and
erosional benches high on valley walls thought to be cut by ice-marginal channels.
D 2003 Elsevier Science B.V. All rights reserved.

Keywords: Quaternary landforms; Erosion; Alpine glaciation

1. Introduction: knowledge of Sierra glaciation Sierra Nevada. In some ways, however, the geologic
structure and history of the northern Sierra Nevada are
This paper describes Pleistocene glacial landforms very different and variations in the geomorphology
in the NW Sierra Nevada, CA and links them with reflect the differences. Much has been learned about
glacial processes. Erosional landforms dominate the glaciological processes over the last 50 years. This
landscape and are emphasized because of their ubiquity knowledge has not been applied to glacial features in
and persistence across the severely eroded landscape. the Sierra Nevada because paleoglacial research there
Considerable study of glaciation in the southern and has moved away from linking landforms and glacio-
central Sierra over the last 140 years provided an early logical processes to an emphasis on stratigraphic ques-
understanding of Sierra glacial geomorphology and tions. Early researchers (e.g., Gilbert, 1904) were
stratigraphy and, to some extent, basic principles drawn fascinated by glacial landforms to the south, but scien-
from those studies are directly applicable to the NW tific understanding of subglacial processes was limited.
Few modern studies have been concerned with gla-
cially eroded landforms in the Sierra Nevada.
* Tel.: +1-803-777-6117; fax: +1-803-777-4972. This study presents a survey of geomorphic features
E-mail address: [email protected] (L.A. James). ranging from small, local striae and gouges to large

0169-555X/03/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0169-555X(03)00145-4
284 L.A. James / Geomorphology 55 (2003) 283–303

glacial troughs, roches moutonnées, and a crag and the scientific community. He traveled through the
tail. While qualitative in nature, these geomorphic south Tahoe basin and down the South Fork American
descriptions provide indicators of important glacial River and provided the only nineteenth century scien-
processes and constraints on their former patterns tific descriptions of glacial evidence in the northern
and characteristics. The intent of this paper is to (i) Sierra (LeConte, 1873) until Lindgren (1897, 1900)
describe glacially eroded forms in the NW Sierra recognized glacial deposits in mapping the general
Nevada and (ii) link these forms to modern concepts geology of the area. John Muir’s accounts of the
of glacial processes in order to facilitate inferences Yosemite Valley spread glacial knowledge to a wide
about the nature of glaciation at the time the landforms audience because of the novelty of the glacial premise
were generated. The glaciological literature is far too and the grandeur and passion of his landscape descrip-
extensive and complex to be covered in this brief tions (e.g., Muir, 1873a,b).
treatment, so discussion is restricted to a few key
examples. Hopefully, future geomorphic studies will 1.2. The next generations of glacial geomorphology
advance the linking of form to process and will
integrate glaciological principles elsewhere across Fascination for alpine glacial geomorphology in the
the Sierra. Sierra continued through the turn of the century when
relationships between form and process at all scales
1.1. Early application of glacial theory in California received scrutiny that has not since been paralleled in
the region. G.K. Gilbert, who had written earlier of the
Introduction of the glacial theory to formal western pre-glacial Sierra topography (Gilbert, 1883), revisited
science was followed rapidly by the arrival of Euro- the central Sierra to study small- and intermediate-
pean culture in the California interior. The radical new scale glacial features and processes (Gilbert, 1904,
geologic ideas about glaciation were transported from 1906a,b). Johnson (1904) concluded that glacial ero-
Europe across the English Channel and Atlantic sion had created and deepened valleys by back-wear-
Ocean, then across North America to California with ing of cirques.
astonishing speed. Evidence of alpine glaciation in the As the concept of multiple glaciations arose, atten-
mountains of California was very quickly recognized tion began to shift from geomorphic process – form
and soon became the best-known example of alpine relationships to the complexities of glacial stratigra-
glaciation in the New World. Much as the settlement of phy. The work of Russell (1889) in the east – central
California leaped ahead of settlement on the western Sierra had clearly documented at least two major
frontier, so application of the glacial theory bounded glacial advances and several lesser ‘‘fluctuations.’’
across the western interior into California. He noted multiple lateral moraines and recognized a
The gathering of geologists into the California gravel deposit separating lacustrine beds in Mono
Geological Survey in the early 1860s brought Josiah Lake as evidence for a period of interglacial lake des-
Whitney and William Brewer fresh from geological sication. Ultimately, glacial studies in the southern and
training at Yale and travels in the European Alps, central Sierra Nevada by Matthes (1930) and Black-
together with Clarence King fresh from course work welder (1931) provided a basic stratigraphic frame-
under Louis Agassiz at Harvard (Guyton, 1998). work for Pleistocene glaciations. Matthes (1930) was
Although Whitney (1865) downplayed the geomor- deeply concerned with geomorphic questions around
phic importance of glacial erosion, the discovery and Yosemite. He extrapolated valley-side profiles across
documentation of extensive glacial evidence in the canyon inner gorges to illustrate depths of canyon
central Sierra Nevada in the early 1860s provided incision and defined three erosional surfaces in the
examples of the wide application of the new theory Yosemite area. Matthes (1930) also recognized three
(Whitney, 1865; King, 1872; Brewer, 1966). Joseph glacial advances in the Yosemite area and qualitatively
LeConte, another student of Agassiz, arrived in Cal- noted general characteristics of the moraine morphol-
ifornia from South Carolina in 1869, took a post at ogies, boulder frequencies, and weathering character-
Berkeley, and soon introduced several important gla- istics. Blackwelder (1931) recognized multiple glacial
cial concepts to the educated public, John Muir, and advances east of Yosemite Valley across the Sierra
 L.A. James / Geomorphology 55 (2003) 283–303 285

crest and correlated them with Matthes’ (1930) map- this information was not well known. Chamberlin’s
ping units. (1888) map of the extent of ice in the United States
showed no glacial ice in the Sierra Nevada north of
1.3. Glacial knowledge in the NW Sierra Carson Pass or the Stanislaus River (Fig. 1).
The dearth of glacial geomorphic knowledge of the
In spite of early recognition and on-going study of NW Sierra persisted through most of the 20th century.
glaciation in the southern and central Sierra Nevada, Several geologic maps of the region included Pleisto-
little work was done in the NW Sierra where the extent cene deposits (Lindgren, 1897, 1900; Hudson, 1951;
of ice remained unknown until the turn of the twentieth Harwood, 1980), but they were primarily interested in
century. LeConte (1872, 1873) briefly visited and older rocks and did not identify glacial landforms.
described glacial evidence in the area south of Lake Jones (1929) provided general descriptions of the
Tahoe and down the South Fork American River. Yet, geography of glaciated areas south of Lake Tahoe

Fig. 1. Detail of northern California from Chamberlain’s (1888) map of glaciation (shaded) in the United States. The northern glacial limit
terminates south of Lake Tahoe indicating no Pleistocene glaciation of NW Sierra. Apparently, Chamberlin was not aware of LeConte’s (1873)
surveys. Box (added) shows study area of this paper. Letters (added) represent American (A), Yuba (Y), and Feather (F) Rivers.
286 L.A. James / Geomorphology 55 (2003) 283–303

and Desolation Valley in the South Fork American of the Tioga glaciers at their utmost extent, i.e., the last
drainage, but did not consider processes. Blackwelder glacial maximum, although at least two older glacia-
(1931) did not map in the NW Sierra, but he described tions reached slightly higher elevations.
three stratigraphic units along the Southern Pacific
Railroad. South of Bear Valley near the Blue Canyon
Station he interpreted two old till outcrops as Sherwin 2. Physiography of the study area
and extensive upland till deposits as Tahoe. He
described Tahoe ice as 180 m thick at Cisco Station 2.1. Geography
and 3 km wide and more than 300 m thick where it
capped the ridge above Emigrant Gap. Blackwelder The NW Sierra Nevada includes the drainages of
also described small Tioga moraines near the Soda the three forks of each of the American, Yuba, and
Springs railroad station as having ‘‘weak cirques,... Feather Rivers. These rivers flow west from the Sierra
and some very low moraine loops at an altitude of crest (Fig. 1) with the South Fork American beginning
6500 feet [1980 m] a few miles farther west. The SW of Lake Tahoe and the northernmost North Fork
glacier was only about 200 feet [656 m] thick and 5 Feather River slightly north of the latitude of Pyramid
miles [8 km] long’’ (Blackwelder, 1931, p. 909). Lake in Nevada. The study area is primarily in the
Recent mapping and surface-exposure ages corrobo- South Yuba drainage, although ice flowed out into the
rate high pre-Tioga (younger Tahoe?) tills at Emigrant Middle Fork Yuba, the North Fork American, and Bear
Gap but show that Tioga ice was at least 320 m thick Rivers (Fig. 2). The geology of this region is similar to
above the valley bottom at Cisco and extended through the central and southern Sierra in that much of the
Bear Valley, much thicker and more extensive than higher elevations are underlain by Mesozoic granitic
Blackwelder’s estimate (James et al., 2002). batholiths that have been overridden by ice, and at
In the 1960s, reviews of the Quaternary geology of lower elevations deformed Paleozoic metamorphics
the Sierra range included small-scale maps showing are common with ridges capped by flat-lying Cenozoic
Tioga glaciers in the northern Sierra Nevada (Wahr- volcanics. However, rock types are more diverse at
haftig and Birman, 1965; Bateman and Wahrhaftig, high elevations of the NW Sierra than to the south
1966). In the NE Sierra, Birkeland (1964) mapped where granite predominates. Both metamorphic base-
deposits and developed a glacial stratigraphy around ment rocks and a Cenozoic cover of conglomerates
Donner Lake and the Truckee River. He applied and volcanics are more widespread here. The top-
Blackwelder’s (1931) Tahoe and Tioga units to the ography is also more subdued in this region, with the
dominant moraines around Truckee, interpreted the exception of the North Fork American River along the
Tahoe as Wisconsin in age, and recognized two pre- south margin of the study area.
Wisconsin tills he named Donner Lake and Hobart
Mills. He also mapped a post-Tioga Frog Lake till 2.2. The preglacial landscape
confined to high cirques such as on the east flank of
Castle Peak. No other large-scale glacial stratigraphic Preservation of an extensive Tertiary channel sys-
or geomorphic mapping had been done until recently tem in the northern Sierra provides important con-
(James and Davis, 1994; James, 1995; James et al., straints on landform evolution in the region (Lindgren,
2002). 1911). Eocene channels were incised into a landscape
At least three major Pleistocene glacial advances that has been interpreted as a broad eroded upland with
occurred in the region and each of them may have had rolling hills and tropical weathering (Yeend, 1974).
multiple stades (James et al., 2002). While small-scale For example, along the lower South Yuba River, the
geomorphic features are likely to be products of the main ancestral Yuba channel had cut f 300 m below a
latest glaciation to override them, larger landforms surface defined by the tops of San Juan, Washington,
may be polygenetic. They were not only eroded by and Harmony Ridges. The present elevation of modern
multiple glaciations but also experienced substantial channels below the Tertiary channels represents can-
weathering intervals during interstadials. The follow- yon deepening that was encouraged by uplift of the
ing discussions of glacial ice concentrate on the nature Sierra block (Lindgren, 1911).
 L.A. James / Geomorphology 55 (2003) 283–303 287

Fig. 2. Tioga glacial flows (arrows) and ice fields (dashed lines). Topography from Chico-E 1j DEM (1:250,000 base). C = Castle ice field,
F = Fordyce ice field, S = Spaulding ice field. 1 = Devils Peak, 2 = two roches moutonnées: Cisco Butte and Hill 6642, 3 = Old Man Mountain,
4 = Red Mountain, 5 = Castle Peak. BV = Bear Valley, CC = Canyon Crk., DP = Donner Pass, GR = Grouse Ridge, HP = Huysink Pass, LL = Loch
Leven Pass, LSY = lower So. Yuba Cn., LV = Lake Valley, NF = No. Fk. American Cn., RC = Rattlesnake Cn., USY = upper So. Yuba Cn.,
YG = So. Yuba Gorge.

During a late Cenozoic volcanic period, the pre- inspecting Alaskan fjords (letter to Wm. M. Davis in
existing drainage filled with sediment and a series of 1899; Davis, 1900). Although this connection between
new valleys developed flowing WSW. These valleys glacial erosion and hanging valleys is often assumed,
formed parallel drainage patterns as is common with hanging valleys are quite common in the Sierra
over-steepened channel systems (Howard, 1967), pre- Nevada foothills below the glacial limit. Thus, unfor-
sumably in response to uplift and steepening of the tunately, neither the degree of glacial erosion nor the
Sierra block. Deep channel incision ensued so that the occupation of valleys by a glacier can be established
modern drainage is much lower than the pre-volcanic simply from the presence of hanging valleys in this
Tertiary channel system. For example, the modern region. Discordant tributaries in the lower South Yuba
lower South Yuba Canyon is f 250 m below the level and North Fork American Canyons are the result of
of the Eocene channel bed near the town of Washing- steepening of the main E – W trending valleys by late
ton. Erosion below the ancestral channel probably was Cenozoic uplift that left tributaries relatively unaf-
not substantial until the voluminous production of fected (Lindgren, 1911; Matthes, 1930). The increased
volcaniclastic rock began to subside in the Pliocene. fluvial erosive power and incision rates of main
Thus, much incision in the Yuba and American Can- channels left tributaries hanging. Stream piracy of
yons apparently occurred during the Quaternary. the former upper Bear River also played a role in the
Hanging valleys (tributaries that are discordant to South Yuba by increasing stream power and incision
main valleys) were first named by G.K. Gilbert while rates (James, 1995).
288 L.A. James / Geomorphology 55 (2003) 283–303

At higher elevations in the glaciated valleys of the extensive transection glacier. This general description
South Yuba, hanging valleys are best expressed in the can be subdivided, however, into two high ice fields
South Yuba gorge below Bear Valley (Fig. 2) (e.g., feeding a lower ice field through two major valley
James, 1995, Fig. 8). Hanging tributaries here could not glaciers. The two upper ice fields extended across the
have been formed prior to the Pliocene because a large Sierra crest steeply down to the east (see Birkeland,
Tertiary channel passed through the ridge at Bear 1964), although descriptions here stop at the crest.
Valley only 1 km to the south of the gorge. This They also formed a more or less continuous band in the
paleochannel is filled with andesitic lahar material N – S direction along the crest, although they are
presumably deposited between the late Miocene and separated in the following discussion based on the
early Pliocene (Slemmons, 1960). Ironically, at higher canyons they fed to the west.
glaciated elevations hanging valleys are lacking along Three broad ice fields occupied large topographic
the South Yuba or Fordyce Canyons and modern depressions: the Fordyce, Spaulding, and Castle ice
streams tend to be graded to valley bottoms. fields (Fig. 2). The Fordyce ice field extended f 8 km
north to south and 13 km from the Sierra crest west to
2.3. Ice fields and valley glaciers Old Man Mountain. Some Fordyce ice spilled north
into the Middle Yuba Canyon, some flowed south to
Tioga (Late Wisconsin) glaciers formed a series of the South Yuba through Rattlesnake Canyon, but most
ice fields and valley glaciers that not only flowed down flow was to the west in a valley glacier f 600 m thick
into the lower study area but also spilled across drain- between Old Man and Red Mountains. The Fordyce
age divides in all directions. The following description valley glacier flowed a few kilometers down to a lower
of glaciers at Tioga maxima is based on topographic ice field around Lake Spaulding, which was f 9 km
relationships and, in some locations, field mapping of long by 5 km north to south. An outlet glacier entered
striae and other directional features. The glaciological the Spaulding ice field from the accumulation area
terminology used here (Table 1) is conventional for north of Grouse Ridge. It formed a lateral moraine
glaciers constrained by topography with one excep- along Granite Creek that descended steeply 130 m in
tion: small, steeply descending glaciers issuing from elevation ending abruptly at the level of the Spaulding
an upland ice field are referred to as outlet glaciers, a ice field surface. Ice from the Spaulding ice field
term usually applied to steep glaciers flowing out of discharged through three valley glaciers. The main
large ice sheets or ice caps. Most of the glaciers discharge flowed steeply down the deep South Yuba
described here were connected to some degree, so Gorge. A broad valley glacier more than 200 m thick
the glacier system could be broadly described as one extended SW through Bear Valley, although its low
large ice field at high elevations feeding down to an gradient and a constriction at the base of Bear Valley
moderated flow. A broad outlet glacier flowed west
across a high shallow trough NW of Lake Spaulding
Table 1 and spilled into the South Yuba Gorge as an ice fall.
Types of glaciers constrained by topographya To the east near the Sierra crest at Donner Pass, the
Icefield Large expansive ice masses free to flow so Castle ice field extended across a basin 8 km wide and
ice doming is absent. more than 10 km north to south. This ice field covered
Valley glacier Ice flowing in a deep bedrock valley; may
include branching system.
both Upper and Lower Castle Valleys that were con-
Transection glacier Interconnected valley glaciers in a web-like nected across an arête at Castle Gate, although little ice
pattern with minor flows between valleys. flowed between these upper accumulation areas. Some
Cirque glacier Ice emanating from small basins in ice flowed between Castle and Fordyce ice fields, but
valley heads. flows were primarily to the east, south, and west. To
Outlet glacier Ice flowing steeply out of a high
accumulation area; may be associated with
the south, ice spilled from the Castle ice field in a
rapid flow, crevasses, or ogives. broad thin sheet across a plateau into the deep North
Small glacier Niche glaciers, glacierets, ice aprons, Fork American Canyon and, to the west, ice fed a deep
ice fringes. valley glacier in South Yuba Canyon. Ice from Upper
a
Sugden and John (1976) and Benn and Evans (1998). Castle Valley bifurcated and flowed both east and
 L.A. James / Geomorphology 55 (2003) 283–303 289

west. Birkeland (1964) postulated correctly that the tonnées. Dissolution can be an important mode of
Tioga ice divide was west of the topographic divide, glacial erosion in calcareous rocks, but these are not
although the ice divide was not more than f 2 km common in the area and this mode of erosion is not
west of Donner Pass. discussed.
Most flow from the Castle ice field was west down
South Yuba Canyon where it fed a valley glacier more 3.1. Polish, striae, and grooves
than 300 m deep. Much of the valley glacier in this
part of the canyon topped a low plateau f 200 m Interactions between sediment in the basal layer
above the South Yuba valley bottom at several points of temperate glaciers and the bedrock surface may
along the southern divide and flowed into the Amer- create small-scale erosional landforms such as polish,
ican River drainage in relatively thin sheets on the striations, or grooves. These abrasion features are
order of 100 m thick. The ice-surface slope to the west common in the NW Sierra, although their spatial
was steeper than the ridge line of the southern divide, distribution varies with preservation and paleoenvir-
so the ridge emerged from the ice to the west. Con- onments. Preservation of Pleistocene polish and striae
sequently, ice flows south into the North Fork Amer- is favored by hard rock that resists weathering and by
ican Canyon decreased westward along the South surface conditions that have been protected from
Yuba valley glacier. Yet, even in the lower west end weathering and erosion. The ubiquitous granodiorites
of this canyon, some ice spilled south to the North in the area are prone to exfoliation and disintegration
Fork American Canyon through high passes in the and do not preserve surface features well, although
Loch Leven and Huysink Lake areas. Near the mouth polish and striae are common on Tioga-age grano-
of the upper South Yuba Canyon, ice topped Cisco diorite surfaces (Fig. 3). No known example of striae
Butte and Hill 6640, two roches moutonnées that on a pre-Tioga surface has been found in this area,
impeded flow. The dominant flow of the South Yuba although Matthes (1930, p. 70) describes one such
valley glacier passed NW through a deep trough be- location near artist’s Point in the Yosemite region as
tween Cisco Butte and Red Mountain to the Spaulding an anomaly. Abundant polish, striae, and grooves can
ice field. Substantial discharges of ice also spilled into be found on metamorphic rocks, particularly at the
Lake Valley and behind Cisco Butte. These glaciers north end of Bear Valley and the entrance to South
ultimately flowed SW to the North Fork American Yuba Gorge.
drainage. Patterns of basal erosion result from variations in
basal pressures, ice velocity, and sediment (Boulton,
1979) as well as rock structures. Processes that favor
3. Glaciation and glacially eroded landforms abrasion include basal melting and sliding with
abundant basal sediment that is hard relative to the
This discussion focuses on erosional landforms bed. Basal melting causes a downward ice velocity
that predominate in the region. Till deposits and that forces basal clasts against the bed (Iverson,
lateral moraines are common in the area (James, 1991a). Basal melting is common on stoss sides of
1995), but they are far less extensive than the ero- protruding rock masses and under thick ice. Con-
sional features. Glacial erosion is generally driven by ditions that promote rapid sliding include steep
four processes: abrasion, plucking, subglacial melt- gradients, a wet bed with ice near the pressure –
water, and dissolution (Benn and Evans, 1998). Abra- melting point, and high hydrostatic pressures,
sion is greatest on the stoss sides of protruding rock although the latter condition may reduce abrasion.
masses, is reliant on basal debris, and is discussed The abundance of basal sediment is affected by local
under the sections on polish, striae, grooves, and bed erosion and the introduction of sediment from
gouges. Plucking is greatest on the lee side of pro- glacier margins. Without sediment, the ability of the
truding rock masses and is discussed under the section ice to abrade its bed is severely limited. High shear
on roches moutonnées. Subglacial meltwater can be stresses are not necessary for abrasion, so polished
important where high pressures develop and is dis- and striated surfaces can occur under relatively thin
cussed in the sections on P-forms and roches mou- ice.
290 L.A. James / Geomorphology 55 (2003) 283–303

on fine-grained metamorphic rocks, quartz veins, and

on granodiorites where exfoliation has not yet
occurred. Extensive polished surfaces in the area sug-
gest the former presence of fine-textured basal till that
has largely been stripped away.
Glacial striae are linear scratches caused by drag-
ging coarse clasts along the basal contact of the glacier.
Three forms of striae, noted by Chamberlin (1888),
were related by Iverson (1991a) to striater point sharp-
ness and rotation. Iverson concluded that striation
depth was a function of the propensity for a clast to
rotate and that nonspheroidal angular fragments pro-
duce the deepest striae because they resist rotation.
Nailhead striae (Fig. 4), a variant of Iverson’s (1991a)
Type I striae, can serve as directional indicators
because they result from shifting of the clast at the
down-ice limit of the striation (Benn and Evans, 1998).
Striae are common throughout the region, most fre-
quently on stoss sides of rock protrusions but also on
flat surfaces and along valley walls. They can be found
in locations that were under thick and thin ice. They are
best preserved on hard metamorphic rocks but can be
found on some granodiorites where weathering has
been limited.
Glacial grooves are deep linear depressions repre-
senting abrasion concentrated along an ice flow path.
They are oriented parallel to ice-flow directions and
Fig. 3. Striae and polish on granodiorite in Loch Leven provide often occur in massive rocks, so they are not generally
evidence of Tioga ice flowing south across a high pass. Ice flowed the result of structural weaknesses in the rock such as
right to left. bedding planes or folia. Boulton (1979) described
debris-rich bands of ice above grooves under the
Glacial polish is a highly smoothed rock surface Breid̄amerkurjökull, Iceland. He postulated a positive
caused by abrasion. Polish represents the work of fine- feedback between bed irregularities that caused
grained sediment as opposed to striae that represent the streaming of debris and locally concentrated bed
focused translation of energy from large clasts to the erosion in grooves that, in turn, enhanced the bed
bedrock. Most glacially smoothed surfaces have a irregularities causing further streaming of debris.
much greater area of polished surface than area of Grooves are not common on granodiorites in the area
striae and grooves. This led Hindmarsh (1996) to but can be found in many distal areas on metamorphic
conclude that erosion by fine-grained sediment far rocks of the Shoo Fly Formation. This may represent
surpasses erosion by larger particles. This concept is the nature of basal debris carried over the two types of
supported by the fact that large clasts break down to rock. Granodiorite tends to erode by plucking of large
fine-grained sediment that continues to polish. True jointed blocks that are comminuted into rounded
polish with a high sheen is found only in isolated clasts, while metamorphic rocks of the Shoo Fly break
patches on the Pleistocene glacial surfaces in the study are entrained as small, hard angular fragments. Con-
area. Polished surfaces are common, however, with a versely, grooves in the Shoo Fly rocks may represent
smooth but granular surface due to post-glacial weath- their lower position in the glacial system, where
ering of selected surface mineral grains to a depth of meltwater and high hydrostatic pressures play a
less than 1 mm. Slightly weathered polish is common greater role.
 L.A. James / Geomorphology 55 (2003) 283–303 291

Fig. 4. Nailhead striae on slate bench eroded into north side of Monumental Ridge in Lake Valley. Ice flowed from bottom to top.

3.2. Crescentic gouges extremely large crescentic gouges on granodiorite

rocks in the central Sierra Nevada. He explained that
Small arcuate fractures on brittle glaciated rock they formed by two pressure fields: an oblique pres-
surfaces, collectively referred to as friction cracks, sure established a conoid fracture of percussion and a
have long been known in the geomorphic literature direct pressure set up a conchoidal fracture. A rock
(Chamberlin, 1888). They are typically small, ranging entrained in the base of the glacier exerts a great
from a few centimeters up to 50 cm (Fairbridge, 1968). forward and downward pressure focused at a point in
Two types of friction cracks are most common: cres- the bedrock. The underlying rock is elastically
centic fractures with horns pointing down-ice and deformed in front of the boulder and bulges until the
crescentic gouges with horns pointing up-ice. In both elastic limit of the rock is exceeded. A brittle fracture
cases, a low-angle plane dips (relative to the rock occurs and a sliver of rock is dislodged from the
surface) in the direction of ice flow (Gilbert, 1906a; intersection of the two fracture surfaces (Gilbert,
Harris, 1943; Benn and Evans, 1998). Crescentic 1906a). Harris (1943) performed laboratory experi-
gouges often have a steep arcuate face dipping about ments and found that the size of crescentic gouges
70j up-ice and a second fracture surface dipping 20j increased with the surface area of the contact. The
down-ice (Harris, 1943). largest gouges in the study area occur in the deepest
Crescentic gouges are by far the most common type position of the valley, however, suggesting that ice
of arcuate rock fracture in the area. Numerous large depth may somehow influence gouge size.
crescentic gouges near Big Bend in the South Yuba
Canyon extend from the valley bottom to f 100 m up 3.3. P-forms
the valley side. They range in size up to 2 m (linear
horn-to-horn) in the valley bottom where ice was more A suite of subglacially sculpted bedrock surface
than 300 m thick (Fig. 5). Although small crescentic features known collectively as P-forms have been
gouges are commonly described as occurring in lon- defined based on various shapes and orientations
gitudinal rows, these are rarely preserved on grano- including sichelwannen, flutes, and potholes (Benn
diorite in the study area. Large crescentic gouges may and Evans, 1998). The nature of subglacial meltwater
cluster in a particular locale and occasionally occur erosion and its importance to glacial erosion has been a
side-to-side as pairs in double crescents, but not in a subject of considerable debate, and P-forms are often
longitudinal array. Gilbert (1906a) described similar evoked as evidence of subglacial meltwater erosion.
292 L.A. James / Geomorphology 55 (2003) 283–303

Fig. 5. Large crescentic gouge in granodiorite at bottom of South Yuba Canyon. Radius f 134 cm, maximum depth f 9.6 cm. Vertical scale
bars in centimeters (left) and inches (right).

Some writers have likened P-forms to fluvially eroded shallow, smoothly scalloped, undulating surfaces are
bedrock forms and argue that subglacial erosion alone developed on fine-grained metasedimentary rocks of
is responsible for them (Shaw, 1994). Some of these the Shoo Fly Formation. These nondirectional features
features have been observed in association with till in may be incipient potholes but are so shallow that
subglacial environments, however, and combinations horizontal forces were clearly responsible for their
of fluvial and glacial processes may be involved in the sculpture. Similar features have been described in the
formation of some P-form features. Boulton (1974) central Sierra (Gilbert, 1906b, Fig. Y) and explained as
argued that sichelwannen result from streaming of the result of migrating moulins that trained meltwater
debris concentrations as was discussed under the from crevasses onto rocks beneath the glacier. Cre-
formation of grooves. In deep subglacial environments vasses may have been located in this extensional
where P-forms tend to be located, saturated till under environment where ice spilled through the constric-
high pressure may behave like a plastic. tion, but moulins are ephemeral features in such an
Temperate glaciers often have elaborate subglacial environment and not necessary for the generation of
drainage systems with hydraulic pressures that vary subglacial meltwater. Gilbert’s example included large
substantially at a variety of time scales (reviewed by potholes in close association with the shallow scal-
Willis, 1995). For example, subglacial hydrostatic loped surfaces; but no potholes occur at the Bear
pressures may increase in the early spring melt season Valley site and attempts to explain the two types of
due to closure of ice tunnels over the preceding winter features are best separated unless a link can be dem-
(Hooke et al., 1985). Glacial sliding velocities and onstrated. These forms are not striated in spite of their
erosion rates may be strongly related to subglacial basal position in a valley constriction and the lack of
hydrostatic pressures, so subglacial drainage condi- weathering as evidenced by fresh polish. This suggests
tions can exert considerable influence on local sub- that either the basal layer was starved of coarse sedi-
glacial processes that may be responsible for P-forms. ment or hydrostatic pressures were sufficient to lift the
Between a pair of bedrock benches in the lower basal layer above its bed. Abundant erratics in the area
Bear Valley where ice flow was constricted, a series of suggest that coarse sediment was not in short supple.
 L.A. James / Geomorphology 55 (2003) 283–303 293

Given the low topographic position within a major stepped valleys shown in standard geomorphology
valley constriction, the role of subglacial meltwater or references (Matthes, 1930, p. 90; Fairbridge, 1968, p.
high hydrostatic pressures are suspected to have 467; Flint, 1971, p. 128). Cotton (1942) argued that
played an important role in forming the streamlined individual riegel can be caused by truncation of valley
undulating features in the Bear Valley constriction, spurs, structurally induced differential glacial erosion,
either by direct meltwater erosion or by rapid basal or vertical glacial corrasion. While present in the NW
sliding. Numerous P-forms are also located at the north Sierra, riegel are not the dominant transverse erosional
end of Bear Valley and in the entrance to Yuba Gorge. features in the area.
Substantial hydraulic heads are likely to have devel- Extensive clusters of closely spaced transverse ribs
oped at both of these locations that were the lowest across the granodiorite floors of the Fordyce, Spauld-
points and served as outlets for subglacial meltwater ing, and Castle ice fields range in height from 5 to 20
from the Spaulding ice field. m (Fig. 6). The individual ridges of rock will be
referred to here as rock bars for lack of a better term.
3.4. Riegel, rock bars, and knock-and-lochain top- They are more closely spaced and lack the asymmetry
ography of riegel. They are structurally controlled and oriented
parallel to master joints approximately perpendicular
Riegel are bedrock ledges oriented transverse to the to the ice-flow direction. These rock bars are relevant
prevailing ice-flow direction. Conventionally, the term to glaciology and glacial geomorphology in at least
has been used to describe well-spaced, individual two regards. First, they represent a very high bed
ridges at the lower rims of rock basins or cirque floors. roughness that resisted basal sliding. Any attempt to
They are commonly associated with the breaks in stair- model basal sliding, shear stress, or erosion in this area

Fig. 6. Granodiorite rock bars in floor of Fordyce Canyon at south base of Old Man Mountain where valley glacier entered upper Spaulding ice
field. View to NNW across north-striking master joints (from lower left to upper right) exerting strong structural control on knock-and-lochain
topography. Ice flow from lower right to upper left.
294 L.A. James / Geomorphology 55 (2003) 283–303

must incorporate a high roughness for these surfaces. sure, freezing onto rock fragments, increased shear
Second, these rock bars appear to reflect the manner in stress, and reduced frictional resistance. Rapid reduc-
which the jointed granodiorite rocks in this area erode, tions in cavity water pressures increase stress gradients
that is, by plucking of large blocks between the in the bed and encourage joint propagation (Iverson,
remaining rock knobs. Analysis of the spacing of rock 1991b). Meltwater can be delivered to lee-side cavities
bars and joints may reveal a periodicity or pattern through transverse crevasses that form in the exten-
representing a stable bed form under the most erosive sional environments common in these positions
conditions of deep ice. Boulton (1974, p. 68) described (Hooke, 1991), or large hydraulic heads may develop
theoretical interrelationships between bedform wave- through elaborate subglacial conduits.
lengths, amplitudes, and ice dynamics. A thorough Asymmetry is enhanced by pressure differentials in
analysis of the mechanics of these processes is beyond the longitudinal direction. Asymmetry tends to be best
the scope of this paper but could elucidate fundamental expressed under thin ice where the ice overburden
processes and rates of glacial erosion over much of the pressure is low so the longitudinal pressure differential
Sierra granitic terrain. is maximized (Benn and Evans, 1998). Thin ice also
Large areas of rough bedrock terrain have been facilitates the introduction of meltwater to lee side
referred to as knock-and-lochain topography from cavities. In theory, therefore, roches moutonnées and
Gaelic words for knoll and small lake (Linton, 1963; similar asymmetrical features are best expressed under
Benn and Evans, 1998). While this term is not com- thin ice. Under deep ice, symmetrical forms such as
mon in the North American glacial geomorphology whale backs are more likely to develop. The environ-
literature, it describes the large areas of eroded gran- ment in which large roches moutonnées form has been
odiorite beneath the Fordyce, Castle, and Spaulding the subject of much debate (Sugden et al., 1992). Some
ice fields where the local relief of large areas is have argued that erosion is greatest during maximum
dominated by rock bars. glacial periods (Sugden and John, 1976), while others
have argued for the greatest erosion rates during
3.5. Roches moutonnées periods of growth or ablation (Boulton and Clark,
1990).
Roches moutonnées are asymmetric forms with A pair of large roches moutonnées eroded in mafic
streamlined stoss sides and steep, rugged lee sides. crystalline rock impeded flow of the valley glacier at
They are common throughout the Sierra Nevada at a the bottom of the upper South Yuba Canyon (Fig. 2).
variety of scales ranging from a few meters to more They are strongly asymmetric, with stoss-side ramps
than 100 m tall. Roches moutonnées have been of Hill 6642 and Cisco Butte extending 500 and 400
described in the literature as ranging up to 150 m high, m, respectively, and lee-side faces extending only 200
while larger asymmetrical hills up to 350 m high are and 240 m, respectively (Fig. 7). These hills rise to
sometimes referred to as flyggbergs (Benn and Evans, between 120 and 150 m above the bench on which
1998). The asymmetry that characterizes these land- they rest, f 300 m above the floor of South Yuba
forms is caused by abrasion on the gentle stoss slopes Canyon. Lee faces are extremely steep, jagged, and
and plucking on the lee sides. Jahns (1943) concluded over-deepened below the elevations of the plateau on
that roche-moutonnée formation is dominated by lee- the stoss sides. The tops of both hills have abundant
side erosion that is much more rapid than abrasion on fresh striae. Striae on Hill 6642 extend right up to the
the stoss side. edge of the lee-side cliff face attesting to the effective-
Plucking on the lee side of large roches moutonnées ness of abrasion on the hill top and a lack of flow
requires rock fracturing that may occur by any of three separation until the abrupt face. The flat hill crest
mechanisms: frost shattering, wedging of rock frag- extends about 50 m with very little lowering near the
ments, and subglacial water pressure variations (Sug- sheer lee-side face.
den et al., 1992). Water pressure fluctuations not only Two cosmogenic radionuclide surface-exposure
induce rock fracture but also encourage entrainment of ages averaging 13.4 F 740 obtained from an erratic
fractured rock material. High water pressure in lee-side on Hill 6642 (James et al., 2002) indicate that Tioga ice
cavities is associated with reduced overburden pres- overtopped this hill. Mapping of the Tioga maximum
 L.A. James / Geomorphology 55 (2003) 283–303 295

Fig. 7. Hill 6642 roche moutonnée at mouth of upper South Yuba Canyon. View to south from top of Cisco Butte, another roche moutonnée
with an abraided stoss side (foreground). Ice flow from left to right.

elevation on nearby valley sides indicates that the flowed across Rattlesnake Ridge north of Rattlesnake
maximum height of the Tioga glacier was not deep Peak into Rattlesnake Canyon. Numerous smaller
over the top of Hill 6642 or Cisco Butte, probably less roches moutonnées are scattered throughout the region
than 30 m thick above the hill tops. At least two other and form a continuum of landforms of varying asym-
glaciations as high but not much higher than the Tioga metry and girth grading to whalebacks and rock
ice occurred in this valley, so these large roches benches.
moutonnées were apparently created under thin ice
during multiple glacial maxima. The location of these 3.6. Whalebacks
roches moutonnées on a high surface above the main
valley and away from valley sides suggests that melt- Whalebacks are streamlined rock knobs with sym-
water was not introduced by deep subglacial conduits. metrical longitudinal profiles caused by abrasion of
Thin ice over the hill-tops may have resulted in trans- both their stoss and lee sides. In theory, whalebacks
verse crevasses and probably facilitated meltwater tend to form in areas of deep ice with rapid flow
introduction to the lee sides from above. velocities. Small whalebacks can form under only a
Old Man Mountain is a flybberg or very large few hundred meters of ice, but larger examples form
roche moutonnée with a steep lee side on an asym- under deep ice streams (Evans, 1996). Abrasion on
metrical peak rising more than 500 m above the floor the lee side of a whaleback indicates a lack of sepa-
of Fordyce Canyon. Tioga glaciers did not overtop ration of the ice from the bed at least some of the time.
Old Man Mountain (James et al., 2002), but swept Thick ice or a low-viscosity basal layer tend to
around both sides of the peak at a high elevation and suppress bed separation (Evans, 1996). If flow sepa-
may have supplied meltwater to the lee side. One or ration occurs, over-steepening or plucking on the lee
more pre-Tioga glaciers probably overtopped Old side may be suppressed by lack of water pressure
Man Mountain. variations within the lee-side cavity. This may result
A much smaller roche moutonnée east of Tuttle from cold, thick ice preventing meltwater from reach-
Lake indicates, along with striae and erratics that ice ing the bed.
296 L.A. James / Geomorphology 55 (2003) 283–303

Whalebacks in the study area are found on valley (Fig. 8) and are often graded at or near the paleo-ice
bottoms where valley glaciers were quite deep. For surface. They may extend a few hundred meters but
example, a large granodiorite whaleback on the floor rarely more than that. Bench widths range from more
of the upper South Yuba Canyon at the Big Bend than 30 m wide to zero where the benches pinch out
ranger station is in the deepest part of the canyon. The against valley walls or end abruptly. Bench tops are
whaleback is about 25 m high, 80 m long, and 50 m relatively flat in the cross-valley dimension, even
wide with large crescentic gouges on both the stoss and when developed on steep hill slopes, and their prox-
lee sides. It presumably formed under thick ice and imate slopes meet valley walls at an over-steepened cut
was not altered to an asymmetrical form by shallow ice slope.
during the period of ablation and ice thinning. Lack of Several eroded benches are located on S-facing
erosion by thin ice does not preclude recessional ice in exposures, but only a few have been found on N-
these locations, but it suggests that no prolonged facing slopes. They often occur on the up-ice termi-
shallow glacial flows were effective in altering these nations of ridges where ice bifurcated around the
features. ridge. These positions were presumably subject to
high basal shearing, as well as contributions of water
3.7. Lateral benches and ice-marginal channels and sediment from upslope. A boulder line is some-
times found 5 –15 m above the bench indicating ice
High on many valley walls in the NW Sierra, this thick deposited erratics above the bench. The
eroded bedrock benches are graded longitudinally in ubiquity, clear relation to glaciation, and lack of pre-
a manner very similar to lateral moraines. They were vious recognition of valley-side benches in the Sierra
initially enigmatic because erosional benches have not calls for a review of possible explanations for these
been previously described in the Sierra Nevada. In erosional features.
several locations, these benches are the dominant Cotton (1942, pp. 286 –299) devoted a chapter to
valley-side morphological features and they can easily multiple-benched valley-side profiles and reviewed
be mistaken for lateral moraines from a distance. They possible explanations including structural controls,
have little or no till cover and are high on valley walls, small troughs eroded into large troughs, lateral mor-
sometimes more than 300 m above the valley floor aine terraces, and epiglacial benches (a.k.a., ice-mar-

Fig. 8. Erosional bench high on steep flank of Rattlesnake Mountain (off photo to right). Interpreted as erosional remnant of ice-marginal
channel. Flat striated surface eroded into vertically dipping layered metamorphic rocks of Sailor Canyon Fm. Floor of South Yuba Cn. is straight
ahead more than 300 m down.
 L.A. James / Geomorphology 55 (2003) 283–303 297

ginal channels). Structural controls can be ruled out The eroded lateral benches in the study area are
because the Sierra benches develop gentle gradients interpreted as ice-contact channels. They indicate
parallel to ice surface slopes in many kinds of rock, minimum ice elevations because they may represent
including vertically dipping layered metamorphics submarginal channels with the ice surface above. This
(Fig. 8), granodiorites with varying joint orientations, ice surface can often be located by boulder lines above
and massive volcaniclastics. Explanations that involve the bench and in many cases this suggests that the
valley deepening into a previously broader valley (cf. channels were not deep below the ice surface. While
Garwood, 1910) do not apply here. The benches are basal sliding fails to explain the flat top surface, basal
flat surfaces with fresh lateral incision into valley walls erosion of a pre-existing marginal or submarginal
and are too young to represent an old, high, valley-side channel could explain the apparent lack of stratified
surface. Erosion by basal sliding along the glacial drift. A polygenetic explanation may be appropriate
margin has been well documented. On the Auster- given that several glacial advances occurred in these
dalsbreen, Glen and Lewis (1961) measured ice mar- valleys. The apparent preferred S-facing aspect of
gin slippage of 26 cm/day or 65% of the velocity of the these features may reflect seasonal differences in sur-
maximum centerline velocity. Yet, this process does face snow melt and supraglacial runoff. Runoff earlier
not explain the flat bench floors, nor has such a in the melt season would not penetrate the ice surface
process-form link been suggested in the literature. as easily as later in the season when subglacial
Only one plausible mechanism in the glacial geo- channels are enlarged and better connected. Thus,
morphology literature can account for the features early thaw snow-melt runoff from south-facing slopes
observed in the study area; that is, ice-marginal or may have been associated with larger ice-marginal
submarginal channels. Although little stratified drift flows than later runoff from north-facing slopes that
has been found in association with these surfaces that percolated down into the glacier.
may be due to a lack of exposures, subsequent erosion,
or burial by supraglacial till or colluvium. Tarr (1914) 3.8. Crag and tail
described ice-marginal and submarginal channels,
sketched a cross section of a flat bedrock bench eroded Crag and tail is a Scottish expression for a resistant
on a steep valley side by such a process, and cautioned knob that obstructs ice flow with a tapered tail in the
that they should only be used for mapping as minimum protected lee zone (Fairbridge, 1968). Most com-
ice-surface elevations due to the possibility that they monly, these are small features with tails of erodible
formed in submarginal positions. Ice-marginal and or deformable basal till, but the tail also may be
submarginal channels up to 2 km in length and graded streamlined weak bedrock. Small erosional crag-and-
along valley sides have been described by Price tail features were described by Chamberlin (1888, p.
(1973). Ice-marginal channels occur at the subaerial 193) as flow-direction indicators. The landform is
contact between ice and bedrock. Both the ice surface scale independent, however; and Castle Rock, Edin-
and the valley side slope toward the contact so water burgh, with the streamlined Royal Mile, has often been
that does not seep into the glacier tends to flow toward described as a large crag and tail (Fairbridge, 1968;
the channel. If the downvalley gradient of the contact Benn and Evans, 1998; Martini et al., 2001). The
is steep, the channel will tend to cut into the ice and Edinburgh crag is a volcanic plug with a bedrock tail
leave no evidence at the ice margin (Price, 1973). If the extending about 1.4 km (Fig. 9). Bedrock mapping
channel gradient is gentle, a bench can be cut into the below a thin but variable till sheet indicates that the
hill slope, especially where the slope is mantled by Edinburgh plug stands about 110 m above a horse-
weak rock. Flint (1971) argued that marginal channels shoe-shaped frontal trough in the bedrock on the up-
will only form on slopes of moderate gradient, and ice side (Sissons, 1971; Evans and Hansom, 1996).
Maag (1969) suggested that steep valley-side slopes This suggests severe scour and high ice velocities in
may be associated with the formation of benches rather front of the crag.
than channels. Sugden and John (1976) described ice- Devil’s Peak is a large crag and tail on the divide
marginal channels that eroded simple flat benches in separating the South Yuba and North Fork American
hillsides. Canyons (Fig. 10). The crag protruded above the ice
298 L.A. James / Geomorphology 55 (2003) 283–303

Fig. 9. Edinburgh Castle, Scotland, on crag and tail. Royal Mile extends along tail to left. Ice flow was upper right to lower left. Photograph by
Alex Shepherd, 1973 (used with permission).

surface and is a Tertiary basalt capping andesite basalt flows with strong columnar jointing. Hudson
(Lindgren, 1897; Harwood, 1980). Harwood described (1951) described nearby basalt plugs and flows in the
Devils Peak as 170 m thick and composed of two Castle Peak area and concluded from thin sections that

Fig. 10. Devils Peak crag and tail. View ESE across upper South Yuba Cn. from Rattlesnake Ridge. Main ice flow was down South Yuba Cn.
(middle left to lower right), but high, thin ice spilled across plateau (left to right around Devils Peak) into North Fork American Cn. beyond
center horizon. Central Pacific Railroad on valley wall; Interstate 80 out of view in bottom of South Yuba Cn.
 L.A. James / Geomorphology 55 (2003) 283–303 299

many of the extensive andesites mapped by Lindgren upland plateau where ice was much thinner than in the
(1897) were Pliocene basalts lying disconformably on main South Yuba Canyon. Effective erosion by shal-
surfaces of low relief. Dalrymple (1964) obtained a K/ low ice around Devils Peak was presumably enhanced
Ar age of 7.4 My from a basalt on nearby Boreal by steep flow gradients into the deep Royal Gorge of
Ridge. the North Fork American Canyon where local relief is
The layered basalt in Devils Peak was resistant to 1340 m at Snow Mountain. It also reflects the relative
glacial erosion while the tail, composed of weaker weakness of the andesite volcanics surrounding the
andesitic material, was streamlined. The front of Dev- crag and of the rhyolites forming the low plateau on
il’s Peak has no topographic trough. Streamlining of the north side.
the tail was not complete and an asymmetric spur
protrudes on the SW side (Fig. 11). Yet, this asym- 3.9. Valley morphology
metrical form should not be confused with horned
crags described by Jansson and Kleman (1999), that A reoccurring question in glacial geomorphology is
are more symmetrical and formed under deep ice. the extent to which Pleistocene glaciers created the
While the asymmetry and lack of a trough suggest deep troughs that they occupied. Whether the classic
early stages of crag-and-tail development, the glacial U-shaped valley represents minor valley-bottom alter-
streamlining of Devil’s Peak represents substantial ations or major valley deepening has been debated.
erosion. Glacial erosion is often assumed to be con- Many examples of glaciated V-shaped valleys have
centrated in valley bottoms but negligible on uplands been documented-particularly in lower reaches of
(Clayton, 1965; Price, 1973). Yet, Devils Peak is on an glaciated valleys-that indicate limited glacial erosion
of the fluvial form (Embleton and King, 1968). The
South Yuba Gorge is an example of a V-shaped
glaciated valley (James, 1996). On the other hand,
examples of the ability of glacial ice to erode deep
troughs are well known. For example, the bottoms of
deep fiords below sea level and Paleozoic glacial
troughs cut into areas of continental shield (Fairbridge,
1968, p. 459).
Evans (1997) advanced eight propositions regard-
ing the effectiveness of temperate alpine glacial ero-
sion. Two of the propositions are directly concerned
with the effectiveness of glacial valley erosion: (i)
glacial and related processes dominate the geomor-
phology of glaciated mountains and (ii) troughs are
glacial paleochannels calibrated to the discharge of ice
and the erodibility of bedrock. Harbor (1992) linked a
two-dimensional, finite-element model with an erosion
model and simulated the production of U-shaped
valleys from V-shaped valleys. If the model was run
for a long period of time, simulating repeated occupa-
tions of a valley by ice during the Quaternary, the U-
shape was propagated downward and deep narrow
troughs resulted.
Deepening of Sierra valleys by Pleistocene erosion
was postulated by Small and Anderson (1995) based on
Fig. 11. Topographic map of Devils Peak (center) showing crag with an analysis of cosmogenic radionuclides in upland
streamlined tail to south. Ice flow from South Yuba Cn. (north of rocks that showed relatively little late Quaternary
map), across plateau, into deep Royal Gorge (south of map). erosion. They suggested that deep glacial erosion con-
300 L.A. James / Geomorphology 55 (2003) 283–303

centrated in valley bottoms together with isostatic moraines remain in the area, the overwhelming dom-
rebound have substantially increased local relief. Geo- inance of eroded bedrock attests to the efficiency of
morphic field evidence qualitatively supports the con- sediment removal from the area and its transport
cept of glacial valley deepening and relative stability of downvalley. Outwash terraces are poorly preserved
surfaces above the glacial limit. in the mountainous gorges of the NW Sierra, although
The importance of glacial erosion to the present occasional remnants can be found, as in the South
morphology of valleys in the region is difficult to Yuba Gorge near the town of Washington and in the
prove, but evidence of severe glacial erosion is abun- North Fork American River at Green Valley and
dant. This evidence includes gouges, roches mouton- between Ponderosa and Long Point. Most glaciofluvial
nées, whalebacks, and other features described above, sediment has been eroded from the mountain canyons
the U-shaped cross-section morphologies of valleys to low-gradient reaches of the rivers in the Sacramento
such as Bear Valley (cf. James, 1996), and evidence of Valley and beyond. As with the glacial record, the
erosion from analysis of cosmogenic radionuclides alluvial record from the Sierra is better understood to
(Fabel et al., in review). Where large valley glaciers the south in the San Joaquin Valley than to the north in
flowed out of ice fields, they left characteristic trough the Sacramento Valley.
forms. Exceptions can be found where valleys do not Alluviation in the Central Valley has largely been in
display typical glacial form, and this may involve response to glaciation in the Sierra Nevada (Marchand
combinations of limited ice flow and structural weak- and Allwardt, 1981). For example, the upper member
nesses in the rock. of the Modesto Formation in the San Joaquin Valley
Rattlesnake Canyon provides an excellent example has at least four terraces interpreted as Tioga outwash
of sudden change in glaciated valley morphology deposits (Marchand and Allwardt, 1981). Central
across a bedrock contact. The middle reaches of Valley alluvial cycles are often assumed to have been
Rattlesnake Canyon are developed in granodiorite synchronous not only with Sierra Nevada glaciations,
and are characterized by wide cross sections with nu- but also with global continental ice (Marchand and
merous rock bars. Glacial ice widened the base of the Allwardt, 1981; Dupré et al., 1991). Strict synchrone-
granodiorite portion of the valley and subsequent ity between alpine glacial events and sea level changes
fluvial erosion has not substantially lowered the valley that integrated global changes in continental ice should
floor. In contrast, the lower valley is developed in slate not be assumed, however (Gillespie and Molnar,
and is steep walled with a deep V-shaped bottom. 1995). Alpine glaciers respond relatively rapidly to
Rattlesnake Creek has cut a series of waterfalls in the minor climate fluctuations while continental glaciers
slate beginning immediately downstream of the gran- have a large thermal inertia that requires an extended
odiorite contact. The cataracts of lower Rattlesnake period of climate change to overcome. Subtle differ-
Canyon are the result of fluvial incision that has mo- ences in timing between alpine glaciers and sea-level
dified the shape of the valley cross section by incising changes may have strongly influenced alluvial sequen-
a V-shaped notch in the valley bottom. Rapid fluvial ces downvalley in basins influenced by coastal base
incision may have been encouraged by a lowered base levels.
level at the mouth of the canyon. South Yuba Canyon During Sierra glacial advances aggradation by out-
at the mouth of Rattlesnake Canyon appears to have wash left deposits now in alluvial terraces grading out
been deepened by late Pleistocene glaciers into a nar- from foothill fan areas. In distal zones, however, lower
row trough more than 300 m deep. sea levels during global full-glacial periods caused
degradation. During Sierra glacial recessions and
interglacial periods, sediment production was reduced
4. Sedimentary products of glacial erosion and outwash near the mountain front was incised.
Lower in the system, however, rising sea levels during
The widespread glacial erosion in this area clearly global interglacial periods encouraged aggradation.
generated a tremendous volume of sediment, yet the These complex spatial and temporal interactions point
landscape at high elevations is dominated by bare rock to the need for higher resolution glacial and fluvial
surfaces. While till deposits and occasional lateral chronologies to develop process – response linkages
 L.A. James / Geomorphology 55 (2003) 283–303 301

between climate change and glacial and fluvial land- implications of rapid and erosive thin ice, drawn from
forms. the Devils Peak crag and tail, are preliminary and
Atwater and Belknap (1980) attributed thorough should be tested with thorough field mapping to deter-
erosion of Pleistocene transgressive estuarine deposits mine the number of glaciations and thickness of ice in
to erosion during low eustatic cycles, implying sedi- that area and to the south. While extensive lateral
ment production was substantially reduced before sea- moraines support these interpretations, only reconnais-
level rise was complete. Much of the sediment from sance mapping has been done there.
the severe glacial erosion documented in this paper The fundamental relationships between glacial
was probably delivered downstream rapidly while sea landforms and inferred glacial processes bear few
levels were still relatively low. Retreat of late Tioga ice surprises. Whalebacks and crescentic gouges are larg-
from Fordyce Canyon was complete by f 14,100 est and best expressed in deep ice positions. Roches
26
Al YBP (James et al., 2002), about the time that moutonnées of various scales appear to have formed in
global sea levels began accelerating around 14,000 positions under thin ice. Large former ice fields are
YBP (Bard et al., 1996). Thus, relatively efficient now characterized by rugged granodiorite knock-and-
flushing of glacial sediment may have been well lochaine topography with numerous structurally con-
underway before marine transgression was completed. trolled rock bars. Zones where ice discharge was the
This may have reduced alluvial deposition in flood- greatest are characterized by deep troughs. These
plains and estuaries and facilitated degradation of generally display the classic U-shape with the excep-
alluvium in the Sacramento Valley. tion of a few locations where dominance by fluvial or
glaciofluvial incision is apparent, such as in lower
Rattlesnake Canyon and the South Yuba gorge.
5. Conclusion

The NW Sierra Nevada provide a diverse set of Acknowledgements

landscapes varying from high alpine granodiorite ter-
rain of moderate local relief to extremely rugged, Dave Mickelson’s zeal for glacial geomorphology
deeply incised gorges of the North Fork American at Wisconsin and during trips to Scandinavia and
River. Great contrasts exist between the inaccessible Alaska was contagious. I will always be inspired by
but widely admired Wild and Scenic North Fork memories of Blåisen, Norway, eskers, patterned
American River Canyon area, the heavily traveled ground explained by Jan Lundqvist, and Richard
but geomorphically ignored Interstate 80 corridor Goldthwait hiking up to the Burroughs and Muir
down the South Yuba Canyon, and the relatively glaciers at age 80. Steve James and Russell Towle were
unstudied isolated roadless area to the north. Much good sounding boards for ideas in the field. Phil
can be learned from glacial geomorphology, and these Sexton and other staff of the Tahoe National Forest,
landscapes exemplify processes dominated by the U.S. Forest Service provided information and facili-
sculpture of glacial ice. Evidence of glacial valley tated the logistics of extended camping. Dennis Dahms
deepening and relative stability of surfaces above the provided several valuable suggestions based on an
glacial limit is abundant. In contrast, some upland early draft that led to substantial improvements. This
areas covered by relatively thin ice show indications research benefitted from funding by the Geography
of severe erosion. and Regional Science Program, National Science
Some of the glacial landforms described in this Foundation (Grant SBR 9631437).
report have not been previously identified in the Sierra
Nevada. The interpretation of high, eroded, lateral
benches as ice-marginal channels should be treated as
References
a hypothesis for thorough field testing. They are
potentially important indicators of former minimum Atwater, B.F., Belknap, D.F., 1980. Tidal-wetland deposits of the
ice elevations that are otherwise difficult to map be- Sacramento-San Joaquin Delta, California. In: Field, M.E.,
cause of poor preservation of moraines. Similarly, Douglas, R.G., Bouma, A.H., Ingle, J.C., Colburn, I.P. (Eds.),
302 L.A. James / Geomorphology 55 (2003) 283–303

Quaternary Depositional Environments of the Pacific Coast. sion at a valley scale derived from terrestrial cosmogenic 10Be
Proc. Pacific Coast Paleogeography Symp. 4, Los Angeles, and 26Al concentrations in rock. Ann. Assoc. Am. Geogr.
CA, Pacific Section, Soc. Econ. Petr. Mineralogists. Fairbridge, R.W., 1968. The Encyclopedia of Geomorphology.
Bard, E., Hamelin, B., Arnold, M., Montaggioni, L., Cabioch, G., Reinhold Bk., NY. 1295 pp.
Faure, G., Rougerie, F., 1996. Deglacial sea-level record from Flint, R.F., 1971. Glacial and Quaternary Geology. Wiley, New
Tahiti corals and the timing of global meltwater discharge. Na- York. 892 pp.
ture 382, 241 – 244. Garwood, E.J., 1910. Features of alpine scenery due to glacial
Bateman, P.C., Wahrhaftig, C., 1966. Geology of the Sierra Nevada. protection. Geogr. J. 36, 310 – 339.
Geology of Northern California. Calif. Div. Mines and Geol., Gilbert, G.K., 1883. Peneplain on western slope of Sierra. Science
Bull., vol. 190, pp. 107 – 172. 1, 194 – 195.
Benn, D.I., Evans, D.J.A., 1998. Glaciers and Glaciation. Arnold Gilbert, G.K., 1904. Systematic asymmetry of crest lines in the
Pub., London. 734 pp. High Sierra of California. J. Geol. 12, 579 – 588.
Birkeland, P.W., 1964. Pleistocene glaciation of the northern Sierra Gilbert, G.K., 1906a. Crescentic gouges on glaciated surfaces.
Nevada, north of Lake Tahoe, California. J. Geol. 72, 810 – 825. Geol. Soc. Am. Bull. 17, 303 – 314.
Blackwelder, E., 1931. Pleistocene glaciation in the Sierra Nevada Gilbert, G.K., 1906b. Moulin work under glaciers. Geol. Soc. Am.
and the Basin Ranges. Geol. Soc. Am. Bull. 42, 865 – 922. Bull. 17, 317 – 320.
Boulton, G.S., 1974. Processes and pattern of glacial erosion. In: Gillespie, A.R., Molnar, P., 1995. Asynchronous maximum advan-
Coates, D.R. (Ed.), Glacial Geomorphology. Proc. Bingham- ces of mountain and continental glaciers. Rev. Geophys. 33,
ton Geomorphology Symp., State Univ. N.Y., Binghamton, 311 – 364.
pp. 41 – 87. Glen, J.W., Lewis, W.V., 1961. Measurements of side-slip at Aus-
Boulton, G.S., 1979. Processes of glacier erosion on different sub- terdalsbreen, 1959. J. Glaciol. 3, 1109 – 1122.
strata. J. Glaciol. 23 (89), 15 – 38. Guyton, B., 1998. Glaciers of California: Modern Glaciers, Ice Age
Boulton, G.S., Clark, C.D., 1990. The Laurentide Ice Sheet through Glaciers, Origin of Yosemite Valley, and a Glacier Tour in the
the last glacial cycle: drift lineations as a key to the dynamic Sierra Nevada Univ. Calif. Press, Berkeley. 197 pp.
behaviour of former ice sheets. Trans. R. Soc. Edinb. Earth Sci. Harbor, J.M., 1992. Numerical modeling of the development of U-
81, 327 – 347. shaped valleys by glacial erosion. Geol. Soc. Am. Bull. 104,
Brewer, W.H., 1966. Up and down California in 1860 – 1864 1364 – 1375.
(3rd ed.). Reprinted in: F.P. Farquhar (Ed.). Univ. Calif. Press, Harris, S.E., 1943. Friction cracks and the direction of glacial
Berkeley. movement. J. Geol. 51, 244 – 258.
Chamberlin, T.C., 1888. The Rock-Scorings of the Great Ice Inva- Harwood, D.S., 1980. Geologic map of the North Fork of the
sions. USGS 7th An. Rept. 1885 – 1886, Wash., DC. American River Wilderness Study Area and adjacent parts of
Clayton, K.M., 1965. Glacial erosion in the Finger Lakes region. the Sierra Nevada, California. U.S. Geol. Surv. Misc. Field
Z. Geomorphol. 9, 50 – 62. Studies Map MF-1177-A, 1:62,500.
Cotton, C.A., 1942. Climatic Accidents in Landscape-Making. Hindmarsh, R.C.A., 1996. Sliding of till over bedrock: Scratching,
Whitcombe and Tombs, Christchurch. 354 pp. polishing, comminution and kinematic-wave theory. Ann. Gla-
Dalrymple, G.B., 1964. Cenozoic chronology of the Sierra Nevada, ciol. 22, 41 – 47.
California. Univ. Calif. Pubs. Geol. Sci., vol. 47. 41 pp. Hooke, R.L.B., 1991. Positive feedbacks associated with erosion of
Davis, W.M., 1900. Glacial erosion in France, Switzerland, and glacial cirques and overdeepenings. Geol. Soc. Am. Bull. 103,
Norway. Proc. Boston Soc. Nat. Hist. 29, 273 – 321. 1104 – 1108.
Dupré, W.R., Morrison, R.B., Clifton, H.E., Lajoie, K.R., Ponti, Hooke, R.L.B., Wold, B., Hagen, J.O., 1985. Subglacial hydrology
D.J., Powell II, C.L., Matheison, S.A., Sarna-Wojcicki, A.M., and sediment transport at Bondhusbreen, southeast Norway.
Leithold, E.L., Lettis, W.R., McDowell, P.F., Rockwell, T.K., Geol. Soc. Am. Bull. 96, 388 – 397.
Unruh, J.R., Yeats, R.S., 1991. Quaternary geology of the Pa- Howard, A.D., 1967. Drainage analysis in geologic interpretation: a
cific margin. In: Morrison, R.B. (Ed.), Quaternary Nonglacial summation. Am. Assoc. Pet. Geol. 51 (11), 2246 – 2259.
Geology; Conterminous U.S. Geol. Soc. Am., The Geology of Hudson, F.S., 1951. Mount Lincoln-Castle Peak area, Sierra Neva-
North America, vol. K-2. Boulder, CO. da, California. Geol. Soc. Am. Bull. 62, 931 – 952.
Embleton, C., King, C.A.M., 1968. Glacial and Periglacial Geo- Iverson, N.R., 1991a. Morphology of glacial striae: implications for
morphology. Edward Arnold, London. 608 pp. abrasion of glacier beds and fault surfaces. Geol. Soc. Am. Bull.
Evans, I.S., 1996. Abraded rock landforms (whalebacks) developed 103, 1308 – 1316.
under ice streams in mountain areas. Ann. Glaciol. 22, 9 – 16. Iverson, N.R., 1991b. Potential effects of subglacial water-pressure
Evans, I.S., 1997. Process and form in the erosion of glaciated fluctuations on quarrying. J. Glaciol. 37 (25), 27 – 36.
mountains. In: Stoddart, D.R. (Ed.), Process and Form in Geo- Jahns, R., 1943. Sheet structure in granites, its origin and use as a
morphology. Routledge, London, pp. 145 – 174. measure of glacial erosion in New England. J. Geol. 51, 71 – 98.
Evans, D.J.A., Hansom, J.D., 1996. The Edinburgh Castle crag- James, L.A., 1995. Diversion of the upper Bear River: glacial
and-tail. Scott. Geogr. Mag. 112 (2), 129 – 131. diffluence and Quaternary erosion, Sierra Nevada, CA. Geomor-
Fabel, D., Harbor, J., Dahms, D., James, L.A., Elmore, D., Horn, L., phology 14, 131 – 148.
Daley, K., Steele, C., in review. Spatial patterns of glacial ero- James, L.A., 1996. Polynomial and power functions for glacial
 L.A. James / Geomorphology 55 (2003) 283–303 303

valley cross-section morphology. Earth Surf. Processes Landf. Matthes, F., 1930. Geologic history of Yosemite Valley. U.S. Geol.
21, 413 – 432. Surv. Prof. Pap. 160 (Wash., DC, 137 pp.).
James, L.A., Davis, J.D., 1994. Glaciation and Hydraulic Gold-Min- Muir, J., 1873a. Discovery of glaciers in Sierra Nevada. Am. J. Sci.,
ing Sediment in the Bear and South Yuba Rivers, Sierra Nevada. 3rd Ser. 5, 69 – 71.
Unpub. Field Trip Guide for Assn. Amer. Geogs. 106 pp. Muir, J., 1873b. Explorations in the great Tuolumne Cañon, over-
James, L.A., Harbor, J., Fabel, D., Dahms, D., Elmore, D., 2002. land monthly. Reprinted in. In: Gilliam, A. (Ed.), Voices for the
Late Pleistocene glaciations in the Northwestern Sierra Nevada, Earth: A Treasury of the Sierra Club Bulletin. Sierra Club
California. Quat. Res. 57 (3), 409 – 419. Books, San Francisco, CA.
Jansson, K.N., Kleman, J., 1999. The horned crag-and-tails of the Price, R.J., 1973. Glacial and Fluvioglacial Landforms. Oliver and
Ungava Bay landform swarm, Quebec-Labrador, Canada. Ann. Boyd, Edinburgh.
Glaciol. 28, 168 – 174. Russell, I.C., 1889. Quaternary History of Mono Valley, California.
Johnson, W.D., 1904. The profile of maturity of alpine glacial U.S. Geol. Surv. 8th Ann. Rpt., Pt. 1, Wash., DC., pp. 261 – 394.
erosion. J. Geol. 12, 569 – 578. Shaw, J., 1994. Hairpin erosional marks, horseshoe vortices and
Jones, W.D., 1929. Glacial land forms in the Sierra Nevada south of subglacial erosion. Sediment. Geol. 91, 269 – 283.
Lake Tahoe. Univ. Calif. Publ. Geogr. 3 (2), 135 – 157. Sissons, J.B., 1971. The geomorphology of central Edinburgh.
King, C., 1872. Mountaineering in the Sierra Nevada. James Os- Scott. Geogr. Mag. 87, 185 – 196.
good and Co., Boston. Slemmons, D.B., 1960. Cenozoic volcanism of the Central Sierra
LeConte, J., 1872. On some ancient glaciers of the Sierras. Proc. Nevada, California. Geology of Northern California. Calif. Div.
Calif. Acad. Sci., pp. 1 – 4. Mines and Geol. Bull., vol. 190, pp. 199 – 208. San Francisco,
LeConte, J., 1873. On some of the ancient glaciers of the Sierra. CA.
Am. J. Sci., 3rd Ser. 5 (29), 325 – 342. Small, E.E., Anderson, R.S., 1995. Geomorphically driven late
Lindgren, W., 1897. Description of the gold belt: description of the Cenozoic rock uplift in the Sierra Nevada, California. Science
Truckee Quadrangle, California. U.S. Geol. Surv. Geol. Atlas, 270, 277 – 280.
Folio 39; 1:125,000. 8 pp. Sugden, D.E., Glasser, N., Clapperton, C.M., 1992. Evolution of
Lindgren, W., 1900. Description of the Colfax Quadrangle, Cali- large roches moutonnées. Geogr. Ann. 74A, 253 – 264.
fornia. U.S. Geol. Surv., Geologic Atlas, Folio 66; 1:125,000. Sugden, D.E., John, B.S., 1976. Glaciers and Landscape. Edward
10 pp. Arnold, London. 376 pp.
Lindgren, W., 1911. Tertiary gravels of the Sierra Nevada of Cal- Tarr, R.S., 1914. College Physiography. MacMillan, New York,
ifornia. U.S. Geol. Surv. Prof. Pap. 73 (226 pp.). NY. 837 pp.
Linton, D.L., 1963. The forms of glacial erosion. Trans. Inst. British Wahrhaftig, C., Birman, J.H., 1965. The Quaternary of the Pacific
Geogr. 33, 1 – 28. mountain system in California. In: Wright, H.E.J., Frey, D.G.
Maag, H., 1969. Ice dammed lakes and marginal glacial drainage on (Eds.), The Quaternary of the United States. Princeton Univ.
Axel Hieberg Island. Heiberg Island Research Rpt., McGill Press, Princeton, NJ, pp. 299 – 340.
Univ., Montreal. Whitney, J.D., 1865. Geology of the Sierra Nevada. Geologic Survey
Marchand, D.E., Allwardt, A., 1981. Late Cenozoic stratigraphic of California. Geology, vol. 1. Calif. Legislature, CA. 498 pp.
units, Northeastern San Joaquin Valley, California. U.S. Geol. Willis, I.C., 1995. Intra-annual variations in glacier motion: a re-
Surv. Bull. 1470 (Wash., DC, 70 pp.). view. Prog. Phys. Geog. 19 (1), 61 – 106.
Martini, I.P., Brookfield, M.E., Sadura, S., 2001. Principles of Gla- Yeend, W.E., 1974. Gold-bearing Gravel of the Ancestral Yuba
cial Geomorphology and Geology. Prentice Hall, Upper Saddle River, Sierra Nevada, California. U.S. Geol. Surv. Prof. Pap.
River, NJ. 381 pp. 772 (Wash., DC, 44 pp.).
 Geomorphology 55 (2003) 305 – 316
www.elsevier.com/locate/geomorph

Advances in Holocene mountain geomorphology

inspired by sediment budget methodology
Olav Slaymaker a,*, Catherine Souch b, Brian Menounos a, Gabriel Filippelli c
a
Department of Geography, The University of British Columbia, 1984 West Mall, Vancouver, British Columbia, Canada V6T 1Z2
b
Department of Geography, Indiana University Purdue University Indianapolis, Indianapolis, IN 46202, USA
c
Department of Geology, Indiana University Purdue University Indianapolis, Indianapolis, IN 46202, USA
Received 19 November 2001; received in revised form 14 June 2002; accepted 10 March 2003

Abstract

The sediment budget, which links sediment sources to sediment sinks with hydroclimatic and weathering processes
mediating the response, is applied to the analysis of sediments in three alpine lakes in British Columbia. We provide two ways
of using the sediment budget as an integrating device in the interpretation of mountain geomorphology. These approaches differ
in their resolution and ability to budget the major components of the fine-sediment cascade in glaciated environments. Taken
together, they provide an integrated index of landscape change over the Holocene. The first example compares the
hydroclimatic controls of lake sedimentation for the last 600 years (A.D. 1370 – 1998) preserved in varved sediments from two
of the lake basins. This hydroclimatological approach incorporates contemporary monitoring, air photo analysis, and detailed
stratigraphy of sedimentation events within a single varve to infer the timing, sources, and preferred pathways of fine-grained
sediments reaching the lake basins. The results indicate that glaciers, hillslope, and channel instability within the major
subbasins are the principal sediment sources to the lake basins. Transitory sediment storage of glacially derived sediments
within the channels is believed to modulate the episodic and more frequent delivery of sediments from adjacent hillslope and
fluvial storage sites and direct routing of glacial rock flour during years of prolonged glacial melt. The second example, relying
on the phosphorus geochemistry of sediments in an alpine lake basin, considers the evolution of phosphorus forms (from
mineral to occluded and organic fractions) as a function of the soil development, inherent slope instability, and repeated cycles
of glaciation and neoglaciation over the Holocene. This geochemical approach demonstrates that both neoglaciation and full
glaciation have essentially zeroed the system in such a way that a high proportion of mineral phosphorus remains in the present
lake sediments and the bioavailability of phosphorus (a key to ecosystem development) is low. Both examples illustrate the
importance of variable sediment sources; the seasonality, frequency, and magnitude of sediment transfers; and the profound
influence of ice cover over contemporary, neoglacial and Pleistocene time scales. They also signal the value of including both
clastic and dissolved components in the sediment budget.
D 2003 Elsevier Science B.V. All rights reserved.

Keywords: Sediment budget; Hydroclimate; Phosphorus geochemistry; Varves

1. Introduction
* Corresponding author. Tel.: +1-604-822-3246; fax: +1-604-
822-6150. The macro-scale geomorphology of the Pacific
E-mail address: [email protected] (O. Slaymaker). ranges of the Coast Mountains of British Columbia,

0169-555X/03/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0169-555X(03)00146-6
306 O. Slaymaker et al. / Geomorphology 55 (2003) 305–316

the product of tens of millions of years of tectonic lag times between process change and landscape
evolution, is well understood (Ryder, 1981). Glacial response are, in general, not known. Lake sediments
modifications of that macro-scale geomorphology provide an archive of information on past environ-
express themselves characteristically at meso-scales mental changes. In particular, their analysis provides
and result from the last 1 –2 million years of recurring insight into variations in process rates and the activa-
glaciation. These too are moderately well understood. tion of sediment sources. This paper describes results
By way of contrast, the fine-scale detail of geomor- of analysis of both clastic sediment archives and the
phic modification during the 10,000-year Holocene phosphorus geochemistry of those sediments pre-
Epoch is not so well understood. This is because the served within a number of British Columbia’s Coast
effects of changing climate on weathering, soil for- Mountain lake basins (Fig. 1).
mation and erosional processes, the thresholds of Two approaches that depend heavily on applications
resistance presented by varying lithologies, and the of sediment budget methodology are described here.

Fig. 1. Study sites: (1) Kwoiek, including Kokwaskey Lake; (2) Duffey Lake; and (3) Green Lake basins.
 O. Slaymaker et al. / Geomorphology 55 (2003) 305–316 307

The first addresses the question of magnitude and or, alternatively (as much of the literature does), to
frequency of geomorphic events during the Holocene ignore two of the three components of the budget.
(and can be labelled a hydroclimatological approach). Recently, increasing emphasis has been placed on the
The second considers soil development and slope quantification of storage elements, their size and
stability (a geochemical approach). We will first review longevity in the landscape, and their response to
the origins and importance of sediment budget method- environmental change as a surrogate for detailed sedi-
ology. An explanation will follow of the premises on ment budget calculations. We have followed this trend
which the hydroclimatological and the geochemical by examining sediments stored in lakes and on slopes
approaches are based. We will then demonstrate the and estimating their magnitude and frequency of
key role played by sediment budgeting in these appli- deposition as a basis for first estimates of components
cations and, finally, we will return to the landscape of the sediment budget. Sediment storage elements in
scale and suggest ways in which these findings affect the landscape, whether in lakes or on slopes, are the
our understanding of the regional geomorphology. medium through which transport processes act and,
therefore, their quantification is critical input to the
sediment budget.
2. Variable source hydrology, variable solute Common to all the watersheds discussed in this
sources, dynamic sediment sources, and sediment paper is the presence of glacial ice and a more ex-
budgets tensive ice cover in the recent past. The major sedi-
ment sources believed to be important for this study
The idea of a sediment budget is disarmingly simple include the following:
in that it is an accounting of the sources, movements,
and sinks of sediment in the landscape. The first formal (i) rock flour derived from subglacial erosion and
sediment budget was published by Jackli (1957) in the glacier forefields—these glacial sediments have
context of a study of the upper Rhine watershed and its been directly coupled with downstream storage
geomorphology. Rapp (1960) emphasised the impor- sites, such as lake basins;
tance of including solutes; and Oldfield (1977) linked (ii) fluvioglacial deposits derived from paraglacial
solute routing, nutrient cycling, and ecology in his lake valley fill and terraces from the early Holocene;
basin sediment budgets. Subsequently, the solute frac- (iii) fluvioglacial deposits derived from exposed
tion and the role of weathering vis-à-vis sediment forefields and moraines during neoglacial ad-
budgeting have been comparatively neglected [though vances; and
note Gallie and Slaymaker (1984) and, more recently, (iv) sediments originating from hillslope instabilities.
Hill and Brooks (1996) and Devito et al. (1996).
Swanson et al. (1982) discussed the range of applica- Any inferences about sediment sources drawn from
tion of sediment budgets and Foster et al. (1985) the sediments in lakes at the lower end of these
formulated the sediment budget for lake basins. Reid systems require careful checking against air photo
and Dunne (1996) provided illustrations of ways in analysis and field site visitation. In addition to this
which sediment budgets could resolve environmental more standard sourcing methodology, the present
management problems. Sources, pathways, and sinks study introduces a technique of examining the status
of clastic sediments within relatively small basins have of the phosphorus geochemistry of the sediments that
been widely reinterpreted as a result of quantitative allows a potentially greater accuracy in the identifi-
sediment budget calculations. The study of sources, cation of basin sediment sources.
pathways, and sinks of solutes and sediments relies, in
a quantitative sense, on the understanding of variable
runoff sources, pathways, and sinks in watershed 3. A hydroclimatological approach to sediment
hydrology (Slaymaker, 2000). Comprehensive studies budgeting
of runoff, solutes, and clastic sediments are extremely
time consuming and expensive; and a strong incentive Because the majority of sediment moving through
exists, therefore, to define and use surrogate measures a basin is transported by water, sources, pathways,
308 O. Slaymaker et al. / Geomorphology 55 (2003) 305–316

and sinks of water in a basin normally are determined 4. A geochemical approach to sediment budgeting
first and then sediment transport is estimated via a
sediment concentration versus discharge relation at a The basic premise of this approach is that the
number of sites in the basin. This method gives a good geochemistry and accumulation rates of the phospho-
first estimate of total sediment flux during ‘‘normal’’ rus fraction in lake sediments over time reflect chang-
discharge events. Under high flow conditions, rating ing conditions in the contributing watersheds,
curves tend to break down and samplers are rarely in specifically the nature of the soils on the slopes and
place at the peak of the discharge. This means that a other up-system sources. A simple conceptual model
purely hydrologic approach will normally underesti- of these changes, based on what we know about
mate the total sediment flux. This is a recurring phosphorus cycling in soils (Walker and Syers,
problem with hydrologic estimates of magnitude and 1976; Gardner, 1990; Crews et al., 1995; Cross and
frequency of occurrence of extreme hydrological Schlesinger, 1995; Vitousek et al., 1997; Schlesinger
events. et al., 1998), is presented in Fig. 2. On a newly
The availability of records of sedimentation in exposed lithic surface, phosphorus typically is present
lakes and deltas for periods of time greatly in excess in the form of the mineral apatite (termed here mineral
of the monitored record is a critical archive of past phosphorus). Over time, and with soil development,
events (e.g., Desloges and Gilbert, 1994; Menounos et phosphorus is increasingly transformed into fractions
al., 2000). Where lake sediments are varved, the co-precipitated with and/or absorbed onto iron and
sedimentation history can be learned by measuring manganese oxyhydroxides (occluded phosphorus,
variation of the mass properties of each sedimentation formed in soils), in soil organic matter (organic
unit with depth. Typically, studies of this kind engage phosphorus), and incorporated into more readily bio-
in a detailed statistical analysis of discharge and available (labile) forms in soil pore spaces and
sediment transport over a shorter monitored period adsorbed onto soil particle surfaces (nonoccluded
of record (of perhaps the last 50 – 100 years). At the phosphorus). Concurrently, the total amount of phos-
same time, they supplement this analysis with analysis phorus in the soil profile decreases as soil phosphorus
of the sediment properties over millennia. A central
premise in patching the monitored records and the
accumulated sediment records is that the thickest
varves correspond with the highest magnitude dis-
charge events. The censoring of medium to low
magnitude events in the varve record contrasts with
the comprehensive record of all hydrologic events
included in the monitored record (Menounos and
Slaymaker, 2000).
Secular trends and cycles are therefore analysed
separately and compared. Decomposing the sediment
records provides insight into sediment source changes
over seasonal to millennial time scales. Although the
low frequency component of clastic sediment deliv-
ery agrees with well-constrained periods of ice
advance in the Coast Mountains and Canadian Cor-
dillera more generally, the complexity and difficulty
of assessing sediment sources in this way becomes Fig. 2. Modeled changes in soil phosphorus geochemistry over time
greater as finer detail is examined. Phosphorus geo- showing transformation of mineral phosphorus into nonoccluded
chemistry, which relies on totally different premises and organic forms before eventual dominance of occluded (oxide-
bound) and organic forms. Note the continual loss of total
from those of the hydroclimatological approach, is a phosphorus from system. From Filippelli and Souch (1999) based
useful method for checking consistency of the inter- on Walker and Syers (1976). The vertical lines indicate the extent of
preted record. the development of the Kwoiek watershed system.
 O. Slaymaker et al. / Geomorphology 55 (2003) 305–316 309

is lost through surface and subsurface runoff. Even- The 4.2-m lake sediment core analysed is de-
tually, the soil reaches a steady state when soil scribed by Souch (1994); chronology is provided by
phosphorus lost through runoff is replaced by new two radiocarbon dates (11,485 + 185 S-2935; 4900 +
phosphorus weathered from apatites at the base of the 325 S-2988) and two tephra layers (Mazama and
soil column. Bridge Rivers) (Souch, 1994). The lake sediments
Headwater, oligotrophic alpine lake sediments that (mean grain size 7.2 A) are composed almost exclu-
are dominated by terrestrial clastic inputs are best sively of terrigenous material from the upstream
suited for recording the geochemical transformations Chochiwa glacier and surrounding slopes, with little
of phosphorus of watershed soils over time. The material contributed by lake productivity. Trap effi-
dissolved inputs of phosphorus are so low that in situ ciency is not 100%, but a significant fraction of
organic production is at a minimum and low amounts inflowing sediment has accumulated throughout the
of labile organic matter in the sediments limit the lake’s history. Because the main goal of this study is
degree of diagenetic overprinting of the original sedi- to examine relative phosphorus geochemical frac-
ment record. The advantages of looking at alpine lake tions, it does not matter if a significant amount of
sediments in this way, rather than relying on soil material is flowing through the system without being
chronosequences, relate to the facts that (i) the lake deposited, as long as the relative proportion of phos-
sediment records provide an integrated record of phorus geochemical fractions in the lake sediments
watershed scale processes and (ii) discrete temporal reflects that coming off the slopes. Eolian input is
resolution allows interpretation of landscape develop- limited and primarily derived from local sources
ment through time. The following complications must (Souch, 1994).
be acknowledged, however. Post-depositional diage- The Duffey Lake basin (250 km2) was heavily
netic overprinting of phosphorus inputs; the leakiness glaciated but, at present, is lightly glacierised (1.7%).
of the lake sediment record; some phosphorus is lost Cayoosh and Van Horlick Creeks are the two main
from the system; and the calibration of the time scale stream systems that drain the basin and 40% of the
with phosphorus form changes depends on the dated basin is above tree line. Duffey Lake itself, with a
varve chronology and the net sediment accumulation surface area of 3.8 km2, is a 90-m-deep basin that is
over time. oriented parallel to the main valley axis of Cayoosh
Creek. Van Horlick Creek is a sandbed river draining
118 km2. Cayoosh Creek is largely a gravel bed river
5. Description of study sites: Kokwaskey, Green, draining 98 km2. Lateral instability and river incision
and Duffey Lake basins are important sediment sources in Cayoosh Creek,
whereas large-scale channel aggradation characterises
Geochemical analyses of phosphorus were con- the Van Horlick system. A large sandur extends 3 km
ducted on the sediments of Kokwaskey Lake, located downvalley from an active glacier in the east fork of
within the Kwoiek Creek watershed, Coast Moun- Van Horlick Creek. Forestry activity and general land
tains of British Columbia (50jN, 122.8jW; elevation use began in 1970 in the Duffey basin when a logging
1050 m) (Filippelli and Souch, 1999). Approximately road was constructed to link the towns of Pemberton
40% of the contributing watershed (total area f 42 and Lillooet. Approximately 8.6% of the basin has
km2) currently is above tree line and 10% is glaci- been logged. Total forest road length is 121.7 km,
erised. The watershed has a high range of relief giving a density of 0.48 km/km2. The geology is
(150 –2944 m), with most summits and ridge crests dominated by mid-Cretaceous to mid-Jurassic grano-
above 2000 m. Valley slopes are steep (average slope diorites, diorites, and quartz diorite.
31j), with vertical rock faces common. The water- The Green Lake basin drains f 180 km2. Three
shed is underlain by coarse-grained granodiorite of creeks drain the lake basin, but only one (Fitzsim-
the Coast Plutonic complex, or low-grade metamor- mons Creek) delivers the majority of sediment to
phosed stratified rocks. Beyond recent extensive log- Green Lake. Fitzsimmons Creek is a 19.5-km-long
ging, the watershed is unaffected by anthropogenic mountain creek draining rugged glacierised terrain
activity. (9% glacierised). The creek is steep and is incised
310 O. Slaymaker et al. / Geomorphology 55 (2003) 305–316

into thick valley fill. Green Lake has a surface area of sands are characteristic of these lowermost units of
2 km2 and typically has a hummocky, glaciated varve couplets. Glacial melt produces finer sediments,
bottom with two main basins-one of which is 40 m which are delivered directly to lake basins, and,
and the other 30 m deep. Land use in the basin has because of late summer lake stratification, they are
increased dramatically since the formation of Black- effectively transported throughout the lake basin.
comb and Whistler ski resorts. The first expansion Autumn cyclonic events, by contrast, are more inter-
into the basin occurred at the beginning of the 20th mittent and spatially discontinuous. Sediment is
century with the construction of the Pacific Great entrained from all basin sediment sources and larger
Eastern Railway, but general development of the volumes of sediment are produced.
region did not commence until the 1960s. The geol-
ogy is complex with 50% granodiorite, diorite, and 6.2. Spatial variability of impact
quartz diorite and roughly equal amounts of volcani-
clastics and conglomerates. An exceptional rain storm in August 1991 illus-
trates the spatial variability of impact. Immediately
west of the Coast Mountains topographic divide,
6. Application of a hydroclimatological approach within the Green Lake basin, significant channel
change and hillslope and terrace material failure
Varve chronologies from two lake basins in the occurred. East of the divide, within the Duffey Lake
southern Coast Mountains were examined to interpret basin, no perceptible change was visible on the air
major flood events. Varve thickness is correlated with photos. The total mass (after correction for density
annual maximum discharge at the nearest stream and nonclastic component) of sediment delivered to
gaging station (Fig. 3). The interpretation of flood Green Lake as a result of the rainstorm exceeded
events, however, is more subtle than this and a sediment yields to the lake basin for the previous
number of issues must be taken into account: decade while the event delivered only minor volumes
of sediment to Duffey Lake. A large fraction of the
(i) the hydrologic seasons for preferred sediment sediments delivered to the lake basins occurred during
transfer; the annual flood. Varve thickness is correlated with
(ii) the spatial variations in impact, both internally to the magnitude of the annual flood. However, most of
the basin and also regionally in mountain regions the truly exceptional floods for the basins west of the
with highly variable meteorological conditions; divide occurred during autumn, while the discharge
(iii) variable climatological controls during the period maxima most important for sediment transfers east of
A.D. 1370 – 1998; the divide were snow melt and rain on snow events.
(iv) anomalously high sedimentation rates in the early Peak flows at Duffey Lake (1997 – 2000) occurred in
part of the 20th century; May and June, and these floods transported fine-
(v) post-1946 changes of sediment source. grained sediments stored within the channel originat-
ing from neoglacial and early Holocene paraglacial
6.1. The hydrologic seasons for preferred sediment sediments. By contrast, the autumn floods, which
transfer dominated the more westerly basins, generated larger
volumes of coarser sediments from all four sediment
Based on monitoring of water and suspended sedi- sources in the basin.
ments (1997 – 2000), sediment transfers to the lake
basins occurred primarily during three distinct sea- 6.3. Variable climatological controls during the
sons: nival runoff, glacial melt, and intermittent period A.D. 1370 –1998
autumn cyclonic disturbance (Menounos and Slay-
maker, 2000). The particle size and physical charac- Without access to detailed air photo interpretation
teristics of the sediments reflect this seasonality. or climatological data, the inferring of sediment sour-
During snowmelt, channel sediment sources are dom- ces and sedimentation processes is difficult. Informa-
inant and normally graded; coarse silts to very fine tion for this region prior to the turn of the century
 O. Slaymaker et al. / Geomorphology 55 (2003) 305–316 311

Fig. 3. Relation between varve thickness in Green Lake and (A) annual peak discharges (log transformed) in Lillooet River and (B) mean annual
temperature departures from the mean at Agassiz. Note the improvement in correlation achieved from 1900 to 1945 with mean annual
temperature departures and from 1946 to 1998 with annual peak discharge.

relies heavily on proxy data. We have reconstructed proxies (Mann et al., 1999). Decadally smoothed,
the varve chronology of Green Lake from A.D. 1370 standardised records reveal statistically significant
to 1998 and compared this with estimates of northern correlation between the two records (JISOA, 2001).
hemispheric growing season temperature from a spa- This indicates that enhanced sedimentation occurred
tially distributed network of temperature sensitive during years that were warmer than average. Such
312 O. Slaymaker et al. / Geomorphology 55 (2003) 305–316

conditions and those where winter snow packs were 7. Application of a geochemical approach
light [commonly during the warm phase of the El
Nino Southern Oscillation (ENSO)] produced pro- The sequential extraction technique of Filippelli
longed periods of glacial runoff during the summer. and Delaney (1996), similar to Tiessen and Moir
(1993), was used to geochemically distinguish the
6.4. Anomalously high sedimentation in the early 20th mineral, occluded, and organic phosphorus fractions
century in multiple lake sediment samples from Kokwaskey
Lake. The extraction technique involved (i) a citrate –
Varve couplets deposited during this time period dithionite – bicarbonate reducing agent and magne-
show low organic, fine-grained sediments uniformly sium chloride (occluded fraction), (ii) dissolutions
distributed throughout the lake basins. Individual with sodium acetate and acetic acid solution and
couplets deposited during this time have many sub- hydrochloric acid (mineral fraction), and (iii) ashing
laminae overlying the lowest, coarse-grained unit. followed by dissolution with hydrochloric acid
These sublaminae are interpreted to reflect periods (organic fraction). Given accumulation rates in the
of sustained glacial melt as they are well distributed lake and sample sizes used (depth increments f 1
throughout the lake basins and are comprised of cm), each sample provides integrated results for time
significant proportions of clay-sized material. This periods of decades to centuries.
interpretation is partly supported by air photo evi- Souch (1994) documented that variations in sed-
dence that documents dramatic glacial recession imentation rates in Kokwaskey Lake clearly reflected
between 1931 and 1946. Indeed, meteorological data the deglacial and neoglacial history of the region (Fig.
confirmed that this period of glacial retreat corre- 4). Sedimentation rates immediately following degla-
sponded to the warmest and driest climate of the ciation (the paraglacial period of Church and Ryder,
period 1900 –1990. Evidence from strong correlation 1972) were very high as large expanses of unvege-
of varve thickness in Green Lake with annual regional tated surficial deposits were exposed by retreating ice
air temperature (pre-1946) and stronger correlation and thus easily eroded. Greater meltwater discharges
with maximum daily discharge events (post-1946) is also increased the capacity and competence of pro-
consistent with the importance of subglacial sediment glacial rivers. Rates of sedimentation decreased and
sources during the earlier period (Fig. 3). stabilized 10,000– 7000 YBP as the landscape stabi-
lised and soil development was initiated. Climatic
6.5. Post-1946 changes of sediment source deterioration in the mid-Holocene and renewed neo-
glacial activity subsequently (Ryder and Thompson,
The secular trends within these records are also 1986) increased glacial sediment supply, resulting in
consistent with a sediment source change after 1946. greater sedimentation rates 6000 –5000, 3500– 2900,
Although northern hemispheric temperature estimates and post-750 YBP.
remained well above average following 1946, varve Temporal variations in the phosphorus geochemis-
thickness declined appreciably. The suggestion then is try of the lake sediments provide further insight into
that glaciers and glacial runoff played a larger role as a rates of erosion, soil development, and landscape
sediment source prior to 1946. stability over the postglacial period. Continual erosion
This hydroclimatological approach to sediment from steep slopes and glacial sources is evidenced by
source and sediment budget assessments is greatly the dominance of mineral forms of phosphorus in the
expanding our understanding of Holocene alpine lake sediments over the last 12,500 years (mineral
landscape evolution. Periods of time when glacial phosphorus constitutes >90– 50%). Occluded phos-
sediment sources are dominant alternate with periods phorus is of secondary importance with organic phos-
when basin-wide sources are dominant. Maritime and phorus present only in relatively low concentrations
continental basins separated by only a few tens of (Fig. 4).
kilometers, respond to different climatic signals. Such Kokwaskey Lake was completely glacier covered
temporal and spatial variability of process is subtly ca. 12 ka, thus the initial starting point for this system
reflected in varve thickness variation. was nearly completely mineral phosphorus (>90%)
 O. Slaymaker et al. / Geomorphology 55 (2003) 305–316 313

Fig. 4. (A) Environmental conditions in the basin; (B) phosphorus concentration for each P-bearing fraction in sediments from Kokwaskey
Lake; (C) percent of total phosphorus; (D) accumulation rate of phosphorus fractions; (E) relative rates of sediment influx based on results and
analyses in Souch (1994). Arrows on age axis are age control points (radiocarbon dates and Mazama and Bridge River tephras).

from rock flour. Mineral phosphorus does exhibit a trol points (indicated on Fig. 4) and are thus not just
decrease, however, to f 50% during the latter part of driven by rates of sedimentation. Rates of phosphorus
the mid-Holocene warmer, drier interval indicative of accumulation in the early deglacial period, driven
landscape stabilisation and soil development (ca. 9– 6 largely by paraglacial processes, exceed 1600 Amol
ka). The proportions of occluded P, and to a lesser P cm 2 ka 1. Most of the material accumulating is
extent organic P, increased during this interval. From mineral phosphorus (f 1600 Amol P cm 2 ka 1)
the mid- to late-Holocene (6 – 1 ka), a period of derived directly from glacial rock flour (sediments are
cooler/wetter conditions (Pellatt and Mathewes, blue-gray glacial clays). From f 9 to 6 ka, landscape
1997), each fraction varied slightly, with mineral stabilisation began, although the landscape was still
phosphorus 60– 70% and occluded phosphorus 20– marked by high relief and rapid rates of erosion. In
30%. The last 1000 years of this record, marked by this period, total phosphorus accumulation rates in
the most extensive Holocene neoglacial activity Kokwaskey Lake dropped by a factor of 3 to f 450
(Ryder and Thompson, 1986), is characterized by a Amol P cm 2 ka 1. Of this, most still remained
rapid return of phosphorus geochemistry to glacial/ mineral phosphorus (f 400 Amol P cm 2 ka 1).
deglacial conditions—mineral phosphorus rises rap- Total rates of accumulation rose in the mid-Holocene
idly (to >80%), while the occluded and organic (8000 – 6000 YBP) to f 800 Amol P cm 2 ka 1,
fractions fall. driven both by increasing sedimentation rates and
Large changes in the accumulation rate of phos- increased contributions of occluded phosphorus (up
phorus over time are also recorded. These accumu- to f 400 Amol P cm 2 ka 1). Phosphorus input rates
lation rate changes are driven partly by changes in rose again to f 1200 Amol P cm 2 ka 1 by 4 ka,
bulk sedimentation rate, but several of the rapid shifts decreasing slightly thereafter. Accumulation rates
in phosphorus accumulation occur between age con- dropped in the last 1000 years, but the mineral
314 O. Slaymaker et al. / Geomorphology 55 (2003) 305–316

fraction increased again as the occluded and organic 8. Implications of findings for geomorphic
fractions dropped. interpretation at the landscape scale
Throughout the Holocene, the high-relief Coast
Mountains watershed led to constant loss of surface 8.1. Magnitude and frequency of geomorphic events
sediments with poorly developed soils and relatively
little organic phosphorus. In this high-relief alpine The lake phosphorus record illustrates the impor-
setting, soil development is retarded by rapid denu- tance of glaciation in setting the initial stage of soil P
dation and thus the terrestrial phosphorus cycle is geochemistry, effectively zeroing the system. This can
stuck in an ‘‘initial development stage’’ with high be achieved by neoglaciation as well as by full
mineral phosphorus and high phosphorus release glaciation. The slow rate of evolution of the phospho-
rates from the landscape. Vertical lines on Fig. 2 rus forms is a function of the hydroclimatological re-
indicate the limited extent of the relative develop- gime and inherent slope instability, reinforced by fre-
ment of the soil P geochemistry of the Kwoiek quent glaciation/neoglaciation. This has implications
watershed system over the Holocene. Detailed varia- for the geochemical form and total loss of phosphorus
tion of lake sediment P geochemistry shows how the from these alpine systems. A high proportion of the
system is effectively reset (zeroed) back to initial phosphorus is mineral phosphorus and the bioavail-
conditions by neoglacial advances of the headwater ability of phosphorus, a key to ecosystem develop-
glaciers (Fig. 4). By contrast, the transformation of ment, is low.
mineral phosphorus is more complete and more rapid The varved lake sediment record permits greater
in relatively lower elevation and warmer sites (Jack- resolution of magnitude-frequency relations from
son Pond, TN) (Filippelli and Souch, 1999). A A.D. 1370 to present, but we are unable to say much
chronosequence of Hawaiian soils revealed signifi- about events earlier in the Holocene. The lesson that is
cant transitions in mineral phosphorus after about emphasised from these data is the necessity to com-
1000 years of soil development (Crews et al., 1995; bine a variety of proxy data to achieve understanding
Vitousek et al., 1997), while significant transforma- of the secular variations in varve thickness and the
tions occurred in phosphorus biogeochemistry during inadequacy of inferred temperature data to provide
soil development on a Krakatoa lava flow in just estimates of magnitude and frequency of sediment-
over 100 years (Schlesinger et al., 1998). In the producing events.
Coast Mountain example presented here, this de-
crease occurred over time scales of 3000 – 5000 8.2. Time scale considerations
years. The alpine British Columbia site has never
achieved the relative stabilisation observed in tem- The resetting of the geological clock following
perate and tropical settings elsewhere. Thus, phos- neoglaciation and full glaciation is an important con-
phorus release remains relatively high and the tribution provided by the phosphorus geochemistry.
phosphorus geochemistry of this system can be reset At this time, this provides a low resolution record for
quickly. Readvance of alpine glaciers during the the Holocene, but in principle, higher resolution anal-
Little Ice Age was sufficient to move the phosphorus yses can be anticipated.
geochemical cycle back to near-glacial conditions The varve chronology provides decadal-scale res-
(Fig. 2). olution over the past 630 years. Positive residuals
These results are preliminary and work is currently from the correlation between varve thickness and
underway to expand the geographic range of sites temperature indices are interpreted as times when
considered, increase temporal resolution to the deca- glaciers are retreating most rapidly and negative
dal scale, and to examine downvalley lake sediment residuals imply reduced glacier melt runoff.
sequences as they integrate several watersheds. How-
ever, the results suggest that in alpine settings lake 8.3. Spatial scale considerations
sediments allow us to document the effects of climate
on soil and landscape development and on biogeo- From the phosphorus geochemistry, site and water-
chemical cycles. shed scale interpretation is sought. An intermediate-
 O. Slaymaker et al. / Geomorphology 55 (2003) 305–316 315

size headwater system (102 –103 km2) with strong with hydroclimatology and weathering processes
glacial input, where dissolved nutrient inputs are mediating the response provides a fruitful avenue for
minimal so as to limit the degree of diagenetic over- exploration of the evolution of these alpine land-
printing on the original sediment record, is the basic scapes.
experimental unit. Sediment sources are relatively
simple, incorporating direct glacial and proximal
slope input. Acknowledgements
The varve chronology from two lake basins pro-
vides an integrated account of clastic sedimentation Much of this work was facilitated by an operating
for two intermediate-size headwater systems and grant to Olav Slaymaker (NSERC: A7073). The
makes inferences about the spatial variability of sedi- geochemical work was supported by an NSF grant to
ment sources within those watersheds, based on their Gabriel Filippelli and Catherine Souch (Geography
response to hydroclimatological events (continental and Regional Science Program). Brian Menounos
versus maritime location). held a UBC University Graduate Fellowship. All
funding sources are gratefully acknowledged.
8.4. The paraglacial concept

Much of the work on sediment sources and sedi- References

ment budgets in the Canadian Cordillera takes its
inspiration from the work by Church and Ryder Church, M., Ryder, J.M., 1972. Paraglacial sedimentation: a con-
sideration of fluvial processes conditioned by glaciation. Geo-
(1972), introducing the paraglacial concept. In the logical Society America Bulletin 83, 3059 – 3072.
case of the phosphorus geochemistry study, glaciation Crews, T.E., Kitayama, K., Fownes, J.H., Riley, R.H., Herbert,
(both neoglaciation and full glaciation) results in a D.A., Mueller-Dombois, D., Vitousek, P.M., 1995. Changes in
pulse of sediment moving through the system. This is soil phosphorus fractions and ecosystem dynamics across a long
reflected in sediment accumulation rates as well as in chronosequence in Hawaii. Ecology 76, 1407 – 1424.
Cross, A.F., Schlesinger, W.H., 1995. A literature review and eval-
the geochemical signature. uation of the Hedley fractionation: applications to the biogeo-
The varve chronology allows some increased res- chemical cycle of soil phosphorus in natural ecosystems. Geo-
olution of the paraglacial effect from the Little Ice derma 64, 197 – 214.
Age, namely the identification of a period of anom- Desloges, J.R., Gilbert, R., 1994. Sediment source and hydrocli-
matic inferences from glacial lake sediments: the post glacial
alously higher sedimentation rates during the early
sedimentary record of Lillooet Lake, British Columbia. Journal
part of the 20th century. Sedimentation in the two lake of Hydrology 159, 375 – 393.
basins is also recognized as greater at the base of the Devito, K.J., Hill, A.R., Roulet, N., 1996. Groundwater-surface
sediment cores retrieved, which correspond with the water interactions in headwater forested wetlands of the Cana-
early Holocene paraglacial episode. Unfortunately, the dian Shield. Journal of Hydrology 181, 127 – 141.
varve records are incomplete, such that quantitative Filippelli, G.M., Delaney, M.L., 1996. Phosphorus geochemistry of
equatorial Pacific sediments. Geochimica et Cosmochimica Ac-
estimates are difficult to assess. ta 60, 1479 – 1495.
Filippelli, G.M., Souch, C., 1999. Effects of climate and landscape
8.5. The sediment budget as an integrator of development on the terrestrial phosphorus cycle. Geology 27,
geomorphic information 171 – 174.
Foster, I.D.L., Dearing, J.A., Simpson, A., Carter, A.D., Appleby,
P.G., 1985. Lake catchment based studies of erosion and denu-
Comprehensive quantitative sediment budgets for dation in the Merevale catchment. Earth Surface Processes and
the three alpine lake basins studied for the whole of Landforms 10, 45 – 68.
the Holocene epoch are not available. The advantage Gallie, T.M., Slaymaker, O., 1984. Variable sources and hydrolog-
that the sediment budget approach affords, as illus- ical pathways. In: Burt, T., Walling, D.E. (Eds.), Catchment
trated by these two variants in three lake basins, is a Experiments in Fluvial Geomorphology. Geobooks, Norwich,
UK, pp. 347 – 358.
conceptual framework that encourages new insights. Gardner, L.R., 1990. The role of rock weathering in the phosphorus
The linking of sediment sources (glacial, paraglacial, budget of terrestrial watersheds. Biogeochemistry 11, 97 – 110.
and slope failures) to sediment sinks (the three lakes) Hill, A.R., Brooks, A.M., 1996. Coarse particulate organic matter
316 O. Slaymaker et al. / Geomorphology 55 (2003) 305–316

inputs to a headwater swamp stream. Archiv fur Hydrobiologia Coast Mountains of British Columbia. Zeitschrift fur Geomor-
137, 25 – 38. phologie, Supplementband 37, 120 – 147.
Jackli, H., 1957. Gegenwarts Geologie des Bundnerischen Rhein- Ryder, J.M., Thompson, B., 1986. Neoglaciation of the southern
gebietes. Beitrage zur Geologie der Schweiz, Geotech. Series, Coast Mountains of British Columbia: chronology prior to the
vol. 36. Kummerli and Frey, Bern, 135 pp. late neoglacial maximum. Canadian Journal of Earth Sciences
Joint Institute for the Study of Oceans and Atmosphere (JISOA), 24, 1294 – 1301.
2001. University of Washington. http://tao.atmos.washington. Schlesinger, W.H., Bruijnzeel, L.A., Bush, M.B., Klein, E.M.,
edu. Mace, K.A., Raikes, J.A., Whittaker, R.J., 1998. The biogeo-
Mann, M.E., Bradley, R.S., Hughes, M.K., 1999. Northern hemi- chemistry of phosphorus after the first century of soil develop-
sphere temperatures during the past millennium: inferences, un- ment on Rakata Island, Krakatau, Indonesia. Biogeochemistry
certainties and limitations. Geophysical Research Letters 26, 40, 37 – 55.
759 – 762. Slaymaker, O., 2000. Research developments in the hydrological
Menounos, B., Slaymaker, O., 2000. Beyond extrapolation: estimat- sciences in Canada (1995 – 1998): surface water quantity, quality
ing flood frequencies and using varved sediments (Paper H51A- and ecology. Hydrological Processes 14, 1539 – 1550.
02). Transactions of the American Geophysical Union, EOS Souch, C., 1994. Downvalley lake sediments as records of neo-
(Abstracts from Fall Meeting, 2000). glacial activity, Kwoiek Creek watershed, Coast Mountains,
Menounos, B., Slaymaker, O., Mazzucchi, D., 2000. Holocene cli- British Columbia. Geografiska Annaler 76A, 169 – 186.
mate change and its influence on sediment production. 8th In- Swanson, F.J., Janda, R.J., Dunne, T., Swanston, D.N., 1982. Sedi-
ternational Paleolimnology Symposium, August 20 – 24, 2000, ment budgets and routing in forested drainage basins. U.S. De-
Queens University, Kingston. Abstract Volume. partment of Agriculture, Forest Service, PNW Forest and Range
Oldfield, F., 1977. Lakes and their drainage basins as units of sedi- Experiment Station. General Technical Report, PNW-141, Port-
ment based ecological study. Progress in Physical Geography 1, land, OR, 165 pp.
460 – 504. Tiessen, H., Moir, J.O., 1993. Characterization of available phos-
Pellatt, M.G., Mathewes, R.W., 1997. Holocene tree line and cli- phorus by sequential extraction. In: Carter, M. (Ed.), Soil
mate change on the Queen Charlotte Islands, Canada. Quater- Sampling and Methods of Analysis. Lewis, Boca Raton, FL,
nary Research 48, 88 – 99. pp. 75 – 86.
Rapp, A., 1960. Recent development of mountain slopes in Karke- Vitousek, P.M., Chadwicj, O.A., Crews, T.E., Fownes, J.H., Hen-
vagge and surroundings. Geografiska Annaler 42A, 65 – 200. dricks, D.M., Herbert, D., 1997. Soil and ecosystem develop-
Reid, L.M., Dunne, T., 1996. Rapid Evaluation of Sediment Budg- ment across the Hawaiian Islands. GSA Today 7, 1 – 8.
ets. Catena Verlag, Reiskirchen, Germany, 164 pp. Walker, T.W., Syers, J.K., 1976. The fate of phosphorus during
Ryder, J.M., 1981. Geomorphology of the southern part of the pedogenesis. Geoderma 15, 1 – 19.
 Geomorphology 55 (2003) 317 – 328
www.elsevier.com/locate/geomorph

Debris flows in Glacier National Park, Montana: geomorphology

and hazards
Forrest D. Wilkerson *, Ginger L. Schmid
Department of Geography, Southwest Texas State University, 601 University Drive, San Marcos, TX 78666-4616, USA
Received 1 February 2002; received in revised form 8 October 2002; accepted 10 March 2003

Abstract

Debris flows are one of the many active slope-forming processes within Glacier National Park, Montana. Most debris flow
landforms exhibit classic morphology with a distinct failure scarp, incised channel, channel levees, and toe deposits that often
develop a lobate form. The Precambrian metasediments that dominate Glacier National Park’s geology weather into angular
clasts that range in size from platy gravels to boulders. Classic debris flows occur in areas where the topographic expression
provides a debris source from cliff faces and an accumulation of regolith, often in the form of talus slopes. Many of these debris
flows have long runout zones and can travel many hundreds of meters. Often they cross hiking trails or roads, including the
main east – west highway, Going-to-the-Sun Road. Debris flows impacting the road have resulted in several near fatalities, and
hikers have been forced to cross active debris flows to reach safe ground. The magnitude of debris flows varies between high
magnitude channel incising events and low magnitude channel filling and/or reworking events. The frequency of debris flow
events is irregular and appears to be controlled by the hydrology of triggering storms and antecedent moisture conditions, not by
the debris supply. As a result, debris flow magnitude is not a function of frequency, but is more closely related to the
characteristics of antecedent conditions and individual storms.
D 2003 Elsevier Science B.V. All rights reserved.

Keywords: Debris flows; Hazard; Magnitude; Frequency

1. Introduction interaction between the geomorphic and human com-

ponents are inevitable and debris flow hazards are an
Glacier National Park, Montana straddles the Con- active part of this mountain environment. For exam-
tinental Divide of the northern Rockies and reaches a ple, debris flows closed Going-to-the-Sun Road in
maximum elevation of 3190 m (Fig. 1). The geology 1997 and 1998. The 1998 event trapped several
and glaciated topography that make this national park vehicles between multiple debris flows (DeChano
unique also create a geomorphic environment prone to and Butler, 2001).
debris flows. With two million visitors per year, This paper describes investigations of the geomor-
phic environment associated with debris flows in
Glacier National Park. Our objective was to define
* Corresponding author. the magnitude and frequency of debris flows to
E-mail address: [email protected] (F.D. Wilkerson). determine the hazard they pose. During a 7-year study,

0169-555X/03/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0169-555X(03)00147-8
 318 F.D. Wilkerson, G.L. Schmid / Geomorphology 55 (2003) 317–328

Fig. 1. Glacier National Park. Numbers represent general drainage basin locations of debris flow study sites. (1) Iceberg Lake; (2) Ptarmigan
Lake; (3) Mt. Henkel; (4) Altyn Peak; (5) Apikuni Mountain; (6) Poia Lake – Windy Creek; (7) Cracker Lake; (8) Otokomi Lake; and (9)
Appisotki Creek.

a variety of methods were employed to gain a better nels, levees, and toe deposits (Hungr et al., 2001)
understanding of the magnitude and frequency of were included (Fig. 2).
debris flows. The study was limited to the east side
of the Park to reduce climatic differences encountered
on the west side of the Continental Divide. Nine 2. Study area
drainage basins with a total of 41 individual debris
flows were observed and measured from 1994 to 2001 The 4,144 km2 of Glacier National Park are domi-
(Fig. 1). Only classic debris flow landforms with nated by 5500 m of nearshore sediments of the Belt
identifiable, well-defined starting zones, incised chan- Supergroup. The accumulation of these sediments
 F.D. Wilkerson, G.L. Schmid / Geomorphology 55 (2003) 317–328 319

Fig. 2. Steep slopes with talus accumulation typical of Glacier National Park debris flow topography. Photograph taken from near Iceberg
Lake.

represents over 500 million years of deposition before (Earhart et al., 1989; Harris et al., 1997). With the
1100 Ma (Table 1). Most of these sediments consist of present-day physical and chemical weathering envi-
argillites and limestones that were slightly metamor- ronment of the northern Rockies, these metasediments
phosed during the formation of the Rocky Mountains are readily weathered into angular clasts ranging in

Table 1
Geologic formations of the Belt Supergroupa
Formation (age) Thickness Composition Number of
flows containing
specific lithology
Snow slip c 460 m Primarily argillite and siltite beds. 10
(c 1100 Ma) (1509 ft) Dolomite is more abundant in the upper formation.
Helena c 750 m Argillaceous dolomite, limestone, calcarenite, and quartzite in the lower 12
(c 1200 Ma) (2461 ft) formation. Also contains a Late Proterozoic intrusive dioritic sill.
Empire 200 – 300 m Argillite and siltite with many dolomitic beds in the upper formation. 38
(c 1300 Ma) (656 – 984 ft)
Grinnell c 750 m Grades from argillaceous facies in the north to quartz rich facies in the 33
(c 1400 Ma) (2461 ft) south.
Apikuni (Appekunny) c 1165 m Interbedded argillite and siltite. Western portion has little quartzite but does 33
(c 1500 Ma) (3822 ft) contain stromatolitic carbonate beds.
Altyn (c 1600 Ma) 300 m (984 ft) Primarily dolomites and stromatolitic limestone. Some argillaceous 4
dolomite and a few thin beds of argillite in the east.
(c 72 – 56 Ma) Lewis overthrust
Cretaceous shales c 3000 – 4000 m Shales, mud, silt, and sandstones of Early to Late Cretaceous age. 4
(c 144 – 60 Ma) (9842 – 13123 ft) Occasionally, sandy pebble – cobble conglomerates are interbedded.
a
Characteristics of the lithologic facies that affect the investigated debris flows. The formations are listed in their natural stratigraphical
order. (Source: Earhart et al., 1989.)
 320 F.D. Wilkerson, G.L. Schmid / Geomorphology 55 (2003) 317–328

size from platy gravels to boulders. Accumulation of (Butler and Walsh, 1994). Field observations since
weathered clasts on talus slopes at the base of cliffs these two studies support the importance of snow in
are characteristic of topography throughout the eastern both winter and summer debris flow triggering events
portions of the park (Fig. 3). (Butler and Wilkerson, 2001).
These talus slopes are the most common environ-
ments for debris flow activity in the study area. The
presence of late-lying snow in the talus and upper 3. Methods
channel reaches often provides near-saturated antece-
dent conditions for the initiation of debris flow move- The geomorphology of slope processes on the east
ment (Butler and Walsh, 1994). Triggering of most side of Glacier National Park has been intensely
debris flows results from one of two main types of studied and documented in a number of field and
hydrologic events: regional-scale frontal systems computer modeling studies (e.g., Butler and Walsh,
(September through June) or summer local-scale con- 1994; Butler and Malanson, 1996; Walsh and Butler,
vective thunderstorms (July and August) (Finklin, 1997; Butler et al., 1998). These investigations con-
1986; Butler and Malanson, 1996). The winter sidered different aspects of topography, distribution,
cyclonic events (September through June) often pro- and triggering mechanisms of debris flows specifi-
duce heavy rainfall onto an existing snow pack. July cally in Glacier National Park. The original database
and August summer thunderstorms produce localized constructed for the Butler and Walsh (1994) inves-
rainfall, which is often heavy enough to trigger debris tigation was used to define accessible debris flows for
flows (Butler and Malanson, 1996). The distribution this study. To characterize the magnitude and fre-
of snow patches within the park and their potential quency of debris flows in Glacier National Park, 41
role in adding increased water supply through snow separate debris flows in nine drainage basins have
melt during thunderstorms have been investigated been monitored since 1994. Only landforms that were

Fig. 3. Stratigraphic evidence of previous debris flow events exposed by incision. Photograph taken in Otokomi-1 debris flow channel.
 F.D. Wilkerson, G.L. Schmid / Geomorphology 55 (2003) 317–328 321

accessible through hiking and scrambling and ex- to investigate the topographic variables responsible
hibited a classic morphology were chosen. The topo- for the spatial distribution of debris flows. A total of
graphic characteristics varied widely among the 41 41 landforms was investigated in the current study.
landforms. Runout lengths varied from 600 m to as Only 12 in the original Butler and Walsh (1994)
little as 150 m, slope angles varied from a high of 33j database were selected on the basis of accessibility,
to a low of 10j, elevations ranged from 2100 to 1550 identification of smaller debris flows not visible in the
m, and aspects varied widely. A variety of geomorphic original air photos, and time constraints. Each of the
methods, including repeat photography, dendrogeo- 41 landforms was photographed from multiple terres-
morphology, lichenometry, plant succession, and trial positions between 1994 and 1996. Photographs
stratigraphic analyses, have been employed (Table 2). were taken between late June and late August each
summer using Kodachrome 64, and ASA 200 film.
3.1. Repeat photography: aerial and terrestrial Multiple exposures were taken from each vantage
approaches using a 28- to 70-mm zoom lens. Some problems
were encountered when fresh debris flows removed or
Repeat photography has proven to be a reliable buried photography sites. Photographs from at least
geomorphic method (e.g., Graf, 1978; Lambert, 1984; one perspective of each landform were repeated in
Byers, 1987; Ives, 1987; Butler, 1994). Photography 1997, 1999, and 2000. The Otokomi and Iceberg
can disclose important variations that are not neces- landforms were also rephotographed in 2001.
sarily apparent to visual inspection. Repeat photog-
raphy is useful for documenting movement within 3.2. Dendrogeomorphology
debris flow landforms, changes in morphology over
time, and the number of mobilizing events. Repeat Dendrogeomorphology utilizes changes in the
photography corroborated by eyewitness accounts and appearance of the annual growth rings of damaged
weather data provided absolute ages for several debris trees to provide a chronology for past geomorphic
flow events, and on one occasion was accurate to the events (Shroder, 1975; Giardino et al., 1984; Butler et
specific hour of occurrence (Bigler, 1999, personal al., 1987; Strunk, 1997; Yoshida et al., 1997; Fantucci
communication). Repeat photography is also useful and Sorriso-Valvo, 1999). Coniferous trees respond in
for developing coarse estimates of excavated and a variety of ways to traumatic events that remove the
deposited volumes of individual debris flows. bark and damage the cambium (e.g., Shroder, 1978,
Panchromatic air photos from 1966 and color 1980; Butler, 1987). Suppressed growth rings or
infrared (CIR) photos from 1984 served to establish reaction wood is suggestive of a past traumatic event.
an initial database of 157 debris flow landforms Many of the 41 debris flows in this study have
(Butler and Walsh, 1994). These landforms were used channels or toe deposits that penetrate alpine tree line,

Table 2
Methods applied to individual debris flows, arranged by drainage basin
Drainage Debris flows Method (number of flows applied)
basin (n = 9) measured Repeat Dendrogeomorphology Lichenometry Plant Stratigraphy
(n = 41)
photography succession
Altyn 3 3 – 1 3 1
Apikuni 4 4 2 2 4 2
Appistoki 7 7 2 3 7 1
Cracker 9 9 – 2 9 1
Henkel 3 3 – 1 3 1
Iceberg 5 5 – – 5 –
Otokomi 2 2 2 – 2 2
Poia 6 6 5 5 6 –
Windy 2 2 1 2 2 1
 322 F.D. Wilkerson, G.L. Schmid / Geomorphology 55 (2003) 317–328

and for several flows core samples supplied precise Table 3

dates for debris flow impact. In the 12 sites containing Methods applied to individual debris flows, arranged by drainage
basin
coniferous trees with impact damage, a combination
Drainage basin Debris flows Event chronology Method of
of 40 cores and 13 cross sections was removed.
(flow IDsa), measured, and typeb,c dating
Healthy, impacted trees were subject to coring using n=9 n = 41 (flow ID)
a Suunto Increment Borer, and these cores were used
Altyn (1, 1A, 2) 3 1955b (1A) lichen
primarily to determine the minimum age of debris Apikuni 4 1931b (BR) lichen
flow deposits. To minimize damage to living trees (DG, DGS, 1955b (DG) lichen
within the park, only unhealthy trees with obvious BR, LR) 1995b (BR, LR) photo
debris flow injuries, such as imbedded stones, were 1996c (BR, LR) photo
1998c (BR) photo
chosen for cross-sectional sampling. Trees that might
1999c (BR) photo
show confounding information from other types of Appistoki 7 1948b (5) lichen
impacts, such as snow avalanches were not included. (1, 2, 3, 1952b (7) lichen
(Butler, 1987). For example, trees were avoided if 4, 5, 6, 7) 1954b (1) lichen
they had branches trimmed by avalanches on their 1995b,c (1) photo
Cracker 9 1938b (CK) lichen
upstream side, because this type of damage can
(63, 64, 64E, 1944b (66) lichen
suppress growth rings, which might be misinterpreted 65, 66, CK, 1996c (SC) photo
as the result of a debris flow. CM, CMW, SC) 1999b photo
Analyzing the cores was complicated by problems (64E, 65, 66,
such as tree rot, multiple scars, and the slow growth CM, CMW)
Henkel (1, 2, 49) 3 1946b (49) lichen
rate of highly stressed trees living in marginal envi-
Iceberg 5 1997c (5) photo
ronments. Standard analytical techniques were (1, 2, 3, 4, 5) 1998b (3, 4) photo
employed (Shroder, 1975; Butler, 1987). The cores 1999b photo
were able to provide minimum ages for the debris (1, 2, 3, 4, 5)
flow deposits on which the trees grew and cross Otokomi (1, 2) 2 1955c (2) dendro
1990 – 1992 (1) photo
sections were able to define absolute ages for seven
1995 (z1,y2) photo
separate debris flows (Table 3). 1996c (1) photo
1997c (1) photo
3.3. Lichenometry 1999b (1, 2) photo
Poia (11, 12, 14, 6 1857b (19) dendro
15, 19, SH) 1947b (19) dendro
Lichenometry is a method for determining the age
1950b (11) lichen
of a rock surface by comparing the size of in situ 1952b (15) lichen
lichens with the established growth rates for that 1955b (14) lichen
lichen species. Lichenometry is used in geomorphic 1956b (12) lichen
studies in many areas that do not have historical 1958b (19) lichen
1964b (19) dendro
records or other methods for determining landform
1968b (19) dendro
age. It has been used to determine the age of debris 1979b (19) dendro
flows in many mountainous areas including the Cairn- Windy (23A, 23B) 2 1944b (23B) lichen
gorms of Scotland (Innes, 1983; Luckman, 1992), the 1955b (23B) dendro
High Tatras of Poland (Jonasson et al., 1991), and the 1958b (23A) lichen
a
Colorado Rockies (Curry, 1966). Other mass-wasting Debris flow IDs numbered 11 and higher refer to flows
events dated by lichenometry include stürzstrom included in the original database identified by Butler and Walsh
(1994). All others are debris flows identified by the authors. The
deposits in Glacier National Park (Oelfke and Butler,
exact location of individual flows is not provided in this publication,
1985; Butler et al., 1998) and rockfall recurrence rates but IDs are provided to allow comparison of repeat debris flow
in the Canadian Rockies (Luckman and Fiske, 1995). events within the same landform.
b
The lichen Xanthoria elegans is common through- Denotes channel-incising event.
c
out Glacier National Park and is particularly wide- Denotes channel-filling event.
spread on calcareous material (Debolt and McCune,
 F.D. Wilkerson, G.L. Schmid / Geomorphology 55 (2003) 317–328 323

1993). At each of 12 control sites, the maximum species diversity and abundance, and the number of
diameter of the 100 largest lichens was measured, species present and the density of growth for each
and the mean maximum thallus diameter for each of species were determined.
the control sites was calculated. Vernier calipers were The plant succession method does not provide
used to measure each thallus to the nearest millimeter. absolute dates for debris flows, but the data did
Coalescing thalli were excluded. The 12 control sites reliably distinguish the relative ages of debris flows
consisted of 10 masonry walls of known age (Deidre emanating from a single source. Species diversity and
Shaw, Park Archives, personal communication, 1997) density increased with landform age and was useful in
and two stürzstrom deposits of known age (Butler et determining the dynamics of events for five separate
al., 1998). The average maximum lichen diameter at debris flow landforms. The data were useful for
each site was plotted against the known age of the analyzing individual landforms, but the ambiguities
structure or deposit to develop a size-to-age curve created by varying lithologies and microclimatic con-
indicative of the species growth rate in the environ- ditions precluded extending the results between land-
ment of Glacier National Park. forms. Therefore, plant succession results are con-
Subsequently, 30 of the largest lichens were meas- sidered less reliable than the other methodologies
ured on 16 debris flows that had established lichen described above, and only the data from the more
colonies. Comparing the mean maximum thallus size reliable methods are reported.
for each flow with the size-to-age curve allowed
minimum dates for debris deposition to be estimated. 3.5. Stratigraphy
Sources of error include the prolonged ecesis rate on
unstable debris flow deposits, reworking of debris by An additional component of the chronological
subsequent events, microclimatic differences between methodology is the investigation of event stratigraphy
sites, and the nonlinear rate of growth of the X. (Mathews et al., 1997; Hinchcliffe et al., 1998).
elegans species (Locke et al., 1979). The maximum Channels excavated by debris flows reveal slope
standard deviation in lichen measurements was 1.2 cm stratigraphy that provides clues to past events and
on the debris flows and 1.4 cm on the control sites, the energy level of those events (Lewkowicz and
with a standard error of F 7 years for the size-to-age Hartshorn, 1998; Gomez-Villar and Garcia-Ruiz,
curve. The uncertainty associated with the standard 2000). Debris flows had impacted six other debris
deviation in lichen size and the regression line error flow landforms and produced incisions in the channel
lie within the 30-year ecesis interval of lichen growth wall or hillslope. The incisions disclosed a strati-
in this environment (Oelfke and Butler, 1985; Wilker- graphic sequence of slope deposits that demonstrated
son et al., 2002). the dynamic nature of past debris flow events and
their role in slope formation (Fig. 3). Exposed inci-
3.4. Plant succession sions were photographed with a grid reference in eight
landforms. In three of those, incision revealed up to 2
The impact of destructive debris flows on plant life m of stratified deposits that were subjected to rigorous
and the subsequent re-establishment and succession of sampling using standard Soil Survey Staff techniques
plant species is well documented (Gecy and Wilson, (Birkeland, 1999).
1989). Plant succession can provide a useful tool for
determining the recurrence interval of small-scale
debris flows. Succession was used to corroborate 4. Results
other methods, such as lichenometry and dendrogeo-
morphology, for determining the relative age of Results from five drainage basins are discussed to
deposits emanating from a single source. At least illustrate the application of these various dating meth-
two separate 1-m plots were placed in a stratified ods, along with their limitations, for characterizing the
manner in the deposits of all 41 debris flows; a total of magnitude and frequency of debris flows in Glacier
96 separate plots was placed. The plots were placed National Park. Due to the variability of debris flow
over the most highly vegetated areas to maximize activity within and between each basin, the results for
 324 F.D. Wilkerson, G.L. Schmid / Geomorphology 55 (2003) 317–328

each basin are outlined separately. Table 2 summa- snow and saturated conditions near the surface. This
rizes the methods applied, and Table 3 summarizes the storm event was responsible for debris flows in the
available data on the magnitude and frequency of Otokomi, Ptarmigan, Cracker Lake, and Iceberg Lake
debris flow events in the nine drainage basins studied. drainage basins. Volumetric measurements of toe
deposits were made on a single debris flow in the
4.1. Otokomi Lake drainage Iceberg drainage following this 1999 basin-wide,
channel-scouring event. Within the Iceberg drainage,
On-site observation (Butler, 1994, Department of this single storm initiated 13 separate flows that
Geography, Southwest Texas State University, per- crossed a 3.5-km stretch of the Iceberg Lake trail.
sonal communication) and repeat photography docu- Notably, the toe-deposit volume measurements were
ment that the Otokomi-1 debris flow originated conducted on a large flow that did not impact the
between the summers of 1990 and 1992. The cross- hiking trail.
sectional volume of a 100-m segment of channel was Original estimates of debris volume were made
measured in 1994 and again after a second channel- using August 2000 terrestrial photography in combi-
scouring event in 1999. The original channel-forming nation with 1:24 000 USGS topographic maps. Depos-
event scoured out a minimum of 1360 m3, while the its were outlined on the maps to determine surface
1999 event removed an additional 1880 m3. These area and objects of known size in photographs were
volumes were obtained by measuring the channel used to interpret the depth of the deposits. During
bottom width, levee height, and channel width parallel follow-up field measurements completed in August
to the slope along a transect, to provide a three- 2001, a 5-m grid was used to map the toe deposits for
dimensional volume of the debris flow channel. validation of the photographic estimates. Calculations
Repeat photography of the same channel shows that from photographs—and maps estimated the toe
these two channel-scouring events were separated by deposits at 1900 m3, while the 5-m grid field measure-
channel-filling and debris-reworking events in 1995, ments give an estimate of 1730 m3. These debris-
1996, and 1997 (Table 3). deposit volumes indicate that the magnitude of this
Stratigraphy revealed in adjacent slopes by the 1999 Iceberg debris flow is similar in magnitude to
original incision (1990 – 1992) shows a pattern of the channel-scouring events measured in the Otokomi
multiple debris pulses with distinct periods of stability basin.
marked by organic-rich surface layers (Wilkerson et
al., 2000) (Fig. 3). Although distinct debris deposits 4.3. Ptarmigan Lake drainage
and stable surface layers can be seen in the strati-
graphic record, the present-day pattern of multiple- The Ptarmigan Lake drainage was not included in
pulse activity within a single event re-emphasizes the the 1996 – 1998 field data collection years. In the
care that must be taken while interpreting the magni- original 1996 reconnaissance, the debris flows in the
tude and frequency of past events. A single debris Ptarmigan drainage were thought to be too erratic in
flow event can consist of multiple depositional and morphology to include in the study. Only two recog-
erosional pulses, and larger events can erase evidence nizable flows existed in the basin. One had recogniz-
of previous events. These processes complicate the able toe deposits but no distinct channel and no
interpretation of the stratigraphic sequence and pre- identifiable starting zone. The second had an incised
clude the use of stratigraphy for absolute dating. channel but no identifiable starting zone and no
recognizable toe deposits.
4.2. Iceberg Lake drainage The August 6, 1999 storm that initiated basin-wide,
channel-scouring events in the Iceberg drainage also
An August 6, 1999, thunderstorm initiated numer- impacted the adjacent Ptarmigan drainage. A total of
ous channel-scouring flows throughout the east side of 27 debris flows crossed 4.3-km of the Ptarmigan
Glacier National Park. Although snowfall during the Tunnel trail between the trail junction and the Ptarmi-
previous winter was below average, cool spring and gan Tunnel. Many of these flows displayed the same
summer temperatures resulted in abundant late-lying morphology observed in the basin in 1996; but several
 F.D. Wilkerson, G.L. Schmid / Geomorphology 55 (2003) 317–328 325

Fig. 4. Trail from Ptarmigan Lake to Ptarmigan Tunnel. (A) Sign on the trail reads ‘‘PLEASE STAY ON TRAIL. CUTTING SWITCHBACKS
IS DESTRUCTIVE AND ILLEGAL’’. (B) Photograph showing the threshold channel scouring debris flows initiated by the August 6, 1999
storm event. Arrow points to hikers standing at sign shown in A.
 326 F.D. Wilkerson, G.L. Schmid / Geomorphology 55 (2003) 317–328

of the 1999 flows exhibited classic starting zones, these events. In most cases, the irregular activity of
incised channels, levees, and toe-deposits. A ca. 1914 debris flows leads to management recognition of the
photograph shows the headwall of the Ptarmigan more regular but often less severe geomorphic events
cirque free of debris flows in an area now containing such as rockfall and trail erosion. Identifying the
13 flows from the 1999 storm. Clearly extreme storm hazard is usually limited to posting signs rather than
events can initiate multiple threshold channel-scour- issuing any type of interactive warnings to park
ing debris flows. visitors.

4.4. Poia Lake and Windy Creek drainages 5.1. Hiking trails

Additional estimates of the frequency of debris Park management policies begin to address the
flow events have been possible with dendrochronol- lower levels of geomorphic hazard, such as off-trail
ogy on six debris flows in the Poia Lake and Windy erosion, with trail signs that warn hikers not to cut
Creek drainage basins. Several subalpine firs (Abies trail switchbacks to shorten their hiking time. These
lasiocarpa) in the runout zone of the Poia 19 debris signs are posted in an attempt to reduce erosion from
flow are protected from avalanche impact. Tree cores runoff that tends to follow off-trail hiking paths. One
provide a clear record of debris impact in 1947, 1964, such sign, posted at the base of a large switchback on
1968, and 1979 (Table 3). Additionally, a minimum the Ptarmigan Tunnel trail, now marks a 1- to 1.5-m-
age date for the Poia 19 debris flow of 1857 suggests deep channel (Fig. 4A,B). The channel, which was
that the originating event occurred at the completion incised over 90 m of elevation (200-m channel
of the Little Ice Age. Two debris flows located length), was formed during the August 1999 debris
directly across the Poia Lake drainage divide in the flow events.
Windy Creek drainage also had trees in the debris-toe The Iceberg– Ptarmigan trail complex serves an
deposits. Three tree cores and one cross section from average of 117 hikers per day from June through
the Windy Creek flows have impacts dating to 1955, September (Debbie Hervol, Park Resources Manage-
which correlates well with several of the Poia Lake ment, personal communication, 2002). The Iceberg
debris flows. branch typically hosts over 200 hikers per day during
Several debris flows in the Poia Lake and Windy the peak summer months of July and August. Table 4
Creek drainages also had measurable lichen growth. summarizes the impact of the August 6, 1999 storm
Debris flow dates reconstructed from the lichen data event on these two trails.
are often within 2– 4 years of dates established with
the dendrogeomorphology (Table 3). These two meth- 5.2. Going-to-the-Sun Road
ods could be identifying separate events, or the ecesis
interval for lichen growth may reduce the accuracy of With park visitation averaging two million people
that technique so that the two methods are actually per year, the main east – west access road hosts over
identifying the same event. Although there is some 3400 cars per day during an average week in early
uncertainty with combined lichenometric and dendro- August (Debbie Hervol, Park Resources Management,
geomorphic results, the two techniques in most cases
strengthen the overall results of debris flow dating by
Table 4
reducing error. Debris flow impact from a single storm event, August 6, 1999
Iceberg Ptarmigan
Lake trail Tunnel trail
5. Hazard
Trail length 3.5 km 4.3 km
Number of debris flows 9 27
Current management practices that might mitigate impacting trail
the hazard of debris flows to park visitors and Total linear impact 68.6 m 401.5 m
infrastructure are difficult and costly to implement Total percentage of 2% 9%
due to the irregularity and varying magnitude of trail impacted
 F.D. Wilkerson, G.L. Schmid / Geomorphology 55 (2003) 317–328 327

personal communication, 2002). The number of cars References

driving Going-to-the-Sun Road peaks on the week-
ends. For example, in August 2001, the average Birkeland, P.W., 1999. Soils and Geomorphology. Oxford Univ.
number of vehicles crossing Logan Pass was 395 Press, New York. 430 pp.
Butler, D.R., 1987. Teaching general principles and applications of
per hour, with the weekday travel averaging 374
dendrogeomorphology. Journal of Geological Education 35,
vehicles per hour. 64 – 70.
Most summer seasons have at least one temporary Butler, D.R., 1994. Repeat photography as a tool for emphasizing
road closure because of debris flows. The most sig- movement in physical geography. Journal of Geography 11,
nificant debris flow closure on Going-to-the-Sun 141 – 151.
Butler, D.R., Malanson, G.P., 1996. A major sediment pulse in a
Road in recent years occurred in 1998 when the road
subalpine river caused by debris flows in Montana, USA. Zeits-
was closed for nearly 24 h. Heavy thunderstorms on chrift für Geomorphologie 40, 525 – 535.
July 28, 1998 triggered three separate debris flows Butler, D.R., Walsh, S.J., 1994. Site characteristics of debris flows
that impacted the road. One debris flow deposited an and their relationship to alpine treeline. Physical Geography 15,
estimated 40 tons of rock onto the road, a second 181 – 199.
Butler, D.R., Wilkerson, F.D., 2001. Hazardous ice jams in north-
deposited an estimated 20 tons, and a third flow
western Montana: their prediction and effects on the landscape.
contributed only smaller rock and silt deposits. Park In: Tobin, G.A., Montz, B.E., Schoolmaster, F.A. (Eds.), Papers
road crews required more than an hour to clear an and Proceedings of the Applied Geography Conference, Fort
escape path for stranded vehicles, and the road was Worth, TX. Applied Geography Conferences Inc., Binghamton,
not re-opened until the afternoon of July 29 (Hungry NY, vol. 24, pp. 57 – 64.
Butler, D.R., Malanson, G.P., Oelfke, J.G., 1987. Tree-ring analysis
Horse News, 1998).
and natural hazard chronologies: minimum sample sizes and
Although the only recent fatality from summer index values. Professional Geographer 39, 41 – 47.
season geomorphic hazards on Going-to-the-Sun Butler, D.R., Malanson, G.P., Wilkerson, F.D., Schmid, G.L., 1998.
Road resulted from rockfall, the debris flows repeat- Late Holocene stürzstroms in Glacier National Park, Montana,
edly impact the road. It is only a matter of time until a USA. In: Kalvoda, J., Rosenfeld, C.L. (Eds.), Geomorphologi-
cal Hazards in High Mountain Areas. Kluwer Academic Pub-
debris flow will result in a fatality. A cartographic
lishing, Dordrecht, The Netherlands, pp. 149 – 166.
model is currently being developed to help delineate Byers, A., 1987. An assessment of landscape change in the Khum-
the spatial aspect of this hazard along Going-to-the- bu region of Nepal using repeat photography. Mountain Re-
Sun Road. search and Development 7, 77 – 81.
Curry, R.R., 1966. Observation of alpine mudflows in the Ten Mile
Range, central Colorado. Geological Society of America Bulle-
tin 77, 771 – 776.
6. Conclusions Debolt, A., McCune, B., 1993. Lichens of Glacier National Park,
Montana. Bryologist 96, 192 – 204.
A variety of geomorphic methods, including repeat DeChano, L.M., Butler, D.R., 2001. Analysis of public perception
photography, dendrogeomorphology, and lichenome- of debris flow hazard. Disaster Prevention and Management 10,
261 – 269.
try, have proved useful in monitoring debris flow
Earhart, R.L., Raup, O.B., Whipple, J.W., Isom, A.L., Davis,
activity and quantifying the magnitude and frequency G.A., 1989. Geologic maps, cross section, and photographs
of debris flows in Glacier National Park. The occur- of the central part of Glacier National Park, Montana. US
rence of debris flows in Glacier National Park is Geol. Surv. Misc. Investigation Series, MAP I-1508-B, Den-
irregular and does not appear to be controlled by ver, CO.
Fantucci, R., Sorriso-Valvo, M., 1999. Dendrogeomorphological
debris supply. Debris flow triggers are primarily
analysis of a slope near Lago, Clabria (Italy). Geomorphology
hydrologic events augmented by near-saturated ante- 30, 165 – 174.
cedent conditions (e.g., Zimmermann et al., 1997). Finklin, A.I., 1986. A climatic handbook for Glacier National Park
Therefore, the magnitude of debris flows is not a with data for Waterton Lakes National Park. US Department of
function of frequency but is more closely related to the Interior, US Forest Service General Technical Report INT-
204, Ogden, UT. 124 pp.
storm events. Monitoring of meteorologic conditions
Gecy, J.L., Wilson, M.V., 1989. Initial establishment of riparian
in areas prone to debris flow activity could help vegetation after disturbance by debris flows in Oregon. Ameri-
mitigate debris flow hazards within Glacier National can Midland Naturalist 123, 282 – 291.
Park. Giardino, J.R., Shroder, J.F., Lawson, M.P., 1984. Tree-ring analy-
 328 F.D. Wilkerson, G.L. Schmid / Geomorphology 55 (2003) 317–328

sis of movement of a rock-glacier complex on Mount Mestas, accretion rates by lichenometry. In: Slaymaker, O. (Ed.), Steep-
Colorado, USA. Arctic and Alpine Research 16, 299 – 309. land Geomorphology. Wiley, New York, pp. 233 – 255.
Gomez-Villar, A., Garcia-Ruiz, J.M., 2000. Surface sediment char- Mathews, J.A., Dahl, S.O., Berrisford, M.S., Nesje, A., Dresser,
acteristics and present dynamics in alluvial fans of the central P.Q., Dumayne-Peaty, L., 1997. A preliminary history of Holo-
Spanish Pyrenees. Geomorphology 34, 127 – 144. cene colluvial (debris-flow) activity, Leirdalen, Jotunheimen,
Graf, W.L., 1978. Fluvial adjustments to the spread of tamarisk in Norway. Journal of Quaternary Science 12, 117 – 129.
the Colorado Plateau region. Geological Society of America Oelfke, J.G., Butler, D.R., 1985. Lichenometric dating of calcare-
Bulletin 89, 1491 – 1501. ous landslide deposits, Glacier National Park, Montana. North-
Harris, A.G., Tuttle, E., Tuttle, S.D., 1997. Geology of National west Geology 14, 7 – 10.
Parks, 5th ed. Kendall/Hunt, Dubuque, IA, pp. 300 – 313. Shroder Jr., J.F., 1975. Dendrogeomorphologic analysis of mass
Hinchcliffe, S., Ballantyne, C.K., Walden, J., 1998. The structure movement. Proceedings of the Association of American Geog-
and sedimentology of relict talus, Trotternish, northern Skye, raphers 7, 222 – 226.
Scotland. Earth Surface Processes and Landforms 23, 545 – 560. Shroder Jr., J.F., 1978. Dendrogeomorphological analysis of mass
Hungr, O., Evans, S.G., Bovis, M.J., Hutchinson, J.N., 2001. A movement on Table Cliffs Plateau, Utah. Quaternary Research
review of the classification of landslides of the flow type. En- 9, 168 – 185.
vironmental and Engineering Geoscience 7, 221 – 238. Shroder Jr., J.F., 1980. Dendrogeomorphology: review and new
Hungry Horse News, 1998. Sun Road reopens after slides cleared. techniques of tree-ring dating. Progress in Physical Geography
July 30, 1998, No. 52, p. 5. 4, 161 – 188.
Innes, J.L., 1983. Lichenometric dating of debris-flow deposits in Strunk, H., 1997. Dating of geomorphological processes using den-
the Scottish Highlands. Earth Surface Processes and Landforms drogeomorphological methods. Catena 31, 137 – 151.
8, 579 – 588. Walsh, S.J., Butler, D.R., 1997. Morphometric and multispectral
Ives, J.D., 1987. Repeat photography of debris flows and agricul- image analysis of debris flows for natural hazard assessment.
tural terraces in the Middle Mountains, Nepal. Mountain Re- Geocarto International 12, 59 – 70.
search and Development 7, 82 – 86. Wilkerson, F.D., Schmid, G.L., Butler, D.R., 2000. Slope stratig-
Jonasson, C., Kot, M., Kotarba, J., 1991. Lichenometric studies and raphy revealed by debris flows incision Glacier National Park,
dating of debris flow deposits in the High Tatra Mountains, Montana. Assoc. Am. Geogr. Abstr., 96th Annual Meeting,
Poland. Geografiska Annaler 73, 141 – 146. Pittsburgh, PA. Association of American Geographers, Wash-
Lambert, W., 1984. Repeat photography and the repeat photography ington DC, pp. 773 – 774.
newsletter. Repeat Photography Newsletter 1, 1 – 3. Wilkerson, F.D., Schmid, G.L., Butler, D.R., 2002. Lichenometric
Lewkowicz, A.G., Hartshorn, J., 1998. Terrestrial record of rapid dating of debris flows, Glacier National Park, Montana. In:
mass movements in the Sawtooth Range, Ellesmere Island, Tobin, G.A., Montz, B.E., Schoolmaster, F.A. (Eds.), Papers
Northwest Territories, Canada. Canadian Journal of Earth Sci- and Proceedings of the Applied Geography Conference, Bing-
ences 35, 55 – 64. hamton, NY. Applied Geography Conference Inc., Binghamton,
Locke, W.W., Andrews, J.T., Webber, P.J., 1979. A manual for NY.
lichenometry. British Geomorphology Research Group Techni- Yoshida, K., Kikuchi, S., Nakamura, F., Noda, M., 1997. Dendro-
cal Bulletin 26 (47 pp.). chronological analysis of debris flow disturbance on Rishiri
Luckman, B.H., 1992. Debris flows and snow avalanche landforms Island. Geomorphology 20, 135 – 145.
in the Lairig Ghru, Cairngorm Mountains, Scotland. Geografis- Zimmermann, M., Mani, P., Romang, H., 1997. Magnitude – fre-
ka Annaler 74, 109 – 121. quency aspects of alpine debris flows. Ecologae Geologicae Hel-
Luckman, B.H., Fiske, C.J., 1995. Estimating long-term rockfall vetiae 90, 415 – 420.
 Geomorphology 55 (2003) 329 – 344
www.elsevier.com/locate/geomorph

Integrating soils and geomorphology in mountains—an

example from the Front Range of Colorado
P.W. Birkeland a,*, R.R. Shroba b, S.F. Burns c, A.B. Price d, P.J. Tonkin e
a
Department of Geological Sciences, University of Colorado, Boulder, CO 80309-0399, USA
b
US Geological Survey, MS 980, Box 25046, Denver Federal Center, Denver, CO 80225, USA
c
Department of Geology, Portland State University, Portland, OR, 97207, USA
d
US Department of Agriculture, Natural Resources Conservation Service, 655 Parfet Street, Lakewood, CO 80215, USA
e
Soil, Plant, and Ecological Science Division, Lincoln University, Canterbury, New Zealand
Received 25 April 2002; received in revised form 24 June 2002; accepted 10 March 2003

Abstract

Soil distribution in high mountains reflects the impact of several soil-forming factors. Soil geomorphologists use key
pedological properties to estimate ages of Quaternary deposits of various depositional environments, estimate long-term
stability and instability of landscapes, and make inferences on past climatic change. Once the influence of the soil-forming
factors is known, soils can be used to help interpret some aspects of landscape evolution that otherwise might go
undetected.
The Front Range of Colorado rises from the plains of the Colorado Piedmont at about 1700 m past a widespread, dissected
Tertiary erosion surface between 2300 and 2800 m up to an alpine Continental Divide at 3600 to over 4000 m. Pleistocene
valley glaciers reached the western edge of the erosion surface. Parent rocks are broadly uniform (granitic and gneissic).
Climate varies from 46 cm mean annual precipitation (MAP) and 11 jC mean annual temperature (MAT) in the plains to 102
cm and
4 jC, respectively, near the range crest. Vegetation follows climate with grassland in the plains, forest in the
mountains, and tundra above 3450 m. Soils reflect the bioclimatic transect from plains to divide: A/Bw or Bt/Bk or K
(grassland) to A/E/Bw or Bt/C (forest) to A/Bw/C (tundra). Corresponding soil pH values decrease from 8 to less than 5 with
increasing elevation. The pedogenic clay minerals dominant in each major vegetation zone are: smectite (grassland),
vermiculite (forest), and 1.0 – 1.8 nm mixed-layer clays (tundra). Within the lower forested zone, the topographic factor
(aspect) results in more leached, colder soils, with relatively thin O horizons, well-expressed E horizons and Bt horizons
(Alfisols) on N-facing slopes, whereas soils with thicker A horizons, less developed or no E horizons, and Bw or Bt horizons
(Mollisols) are more common on S-facing slopes. The topographic factor in the tundra results in soil patterns as a consequence
of wind-redistributed snow and the amount of time it lingers on the landscape. An important parent material factor is airborne
dust, which results in fine-grained surface horizons and, if infiltrated, contributes to clay accumulation in some Bt horizons.
The time factor is evaluated by soil chronosequence studies of Quaternary deposits in tundra, upper forest, and plains
grassland. Few soils in the study area are >10,000 years old in the tundra, >100,000 years old in the forest, and >2 million
years old in the grassland. Stages of granite weathering vary with distance from the Continental Divide and the best developed
is grus near the sedimentary/granitic rock contact just west of the mountain front. Grus takes a minimum of 100,000 years to
form.

* Corresponding author.

0169-555X/03/$ - see front matter D 2003 Elsevier B.V. All rights reserved.
doi:10.1016/S0169-555X(03)00148-X
330 P.W. Birkeland et al. / Geomorphology 55 (2003) 329–344

Some of the relations indicated by the soil map patterns are: (1) parts of the erosion surface have been stable for 100,000
years or more; (2) development of grus near the mountain front could be due in part to pre-Pennsylvanian weathering; (3) a
few soil properties reflect Quaternary paleoclimate; and (4) a correlation between soil development in the canyons and stream
incision rates.
D 2003 Elsevier B.V. All rights reserved.

Keywords: Geomorphology; Soils; Mountains; Front Range of Colorado

1. Introduction increase from east to west, and relate well to the soil
patterns. In the eastern part of the range, mean annual
Soils can provide information useful to the geo- precipitation (MAP) is relatively low ( < 100 cm),
morphological interpretation of mountainous terrain. uplift and erosion slight, and A/Bw/C soil profiles
Their distribution, laterally and vertically, results from on transport-limited slopes denote geomorphic stabil-
the impact of the soil-forming factors—climate, veg- ity during most of the Holocene. In areas with MAP
etation, topographic setting, parent material, and time between 140 and 200 cm, the soil pattern is complex
(Jenny, l980; Birkeland, 1999). The altitudinal distri- and consists of both eroded soils in the upper parts of
bution of soils shows the impact of the climate and the basins and one or more buried soils in the lower
vegetation factors. One of the more important roles of parts. In contrast, in the western part of the range, a
soils is in helping put surficial deposits of diverse high uplift rate combined with >1000 cm MAP results
origins (glacial, fluvial, colluvial, periglacial, etc.— in high erosion rates, A/R soil profiles on steep
the parent material factor) into age groups (time weathering-limited slopes, and soil residence times
factor). Richmond (1962) did this in his classic work of 100– 200 years.
in the La Sal Mountains of Utah. The topographic A few of the lessons learned in Southern Alps
factor needs more study, because if there are marked apply to the Front Range of Colorado. One is that the
differences in soils with topographic setting, one has precipitation in the latter area is similar to that of the
to determine if it is due to the time factor or to the drier part of the Southern Alps study area, and the
rates of pedological processes. landscape and soils in the Front Range of Colorado
This paper analyses the relation between soils and seem similarly relatively stable. A second is that
geomorphology in the Front Range of Colorado and basin-wide patterns of soil development seem to
adjacent plains. Birkeland studied the area in a recon- correspond with stream incision rates.
naissance fashion while taking students on field trips
for the past 35 years. Former students and colleagues
have contributed by their detailed published and 2. Regional setting
unpublished work on parts of the range and plains.
However, this study—a first approximation of a The Front Range of Colorado rises abruptly above
coherent soil-geomorphic history of the range—was the Colorado Piedmont section of the Great Plains
facilitated by soil maps and profile descriptions for the (Fig. 1). This paper focuses on a Colorado Piedmont-
Front Range west of Boulder by the U.S. Department to-Front Range of Colorado transect near the town of
Agriculture, U.S. Forest Service (2001, and written Boulder. The base of the mountain front is between
communications, 1999, respectively). 1700 and 1830 m, and the area west of the front rises
There are not many examples of integrating soils to about 2500 m over a distance of about 6 km.
and geomorphology at the scale of a mountain range. Farther west are remnants of a dissected Tertiary
One exception is the work of Tonkin and Basher erosion surface (Scott and Taylor, 1986; Bradley,
(1990) in the Southern Alps of New Zealand. Soil 1987). This surface is expressed by accordant ridges
distribution was well known from extensive mapping, and extensive areas of low-relief terrain that slope
and they did detailed studies in key basins across the gently upward to the west to about 2750 m over
range. Uplift, precipitation, and erosion rates all distances of about 10 – 20 km. West of the erosion
 P.W. Birkeland et al. / Geomorphology 55 (2003) 329–344 331

Fig. 1. Location map showing the Front Range of Colorado, mountain front, and Colorado Piedmont near Boulder, Colorado. Heavy solid lines
near Ward and Nederland indicate the approximate lower limit of Pleistocene glaciation; dotted pattern indicates till and related glacial deposits
(modified from Madole et al., 1998). Gray pattern shows the approximate extent of a Tertiary erosion surface (modified from Scott and Taylor,
1986).

surface, the mountains rise to the Continental Divide to its present elevation, or did it form at a lower
(3600 –4100 m) over distances of about 10 –15 km. elevation and subsequently been uplifted (Steven et
Several rivers drain the area. To the south is Coal al., 1997 and references cited therein). Whatever the
Creek, which heads about 28-km east of the Con- case, canyon cutting commenced about 5 my (million
tinental Divide; its headwaters were never glaciated. years) ago (Trimble, 1980; Steven et al., 1997).
To the north, both Boulder and Left Hand Creeks head
just east of the Continental Divide and their upper
reaches were glaciated during the Pleistocene. 3. Soil profiles
Boulder Creek was more extensively glaciated. The
glaciers in both of these drainages terminated at about Soil profiles are composed of various horizons, and
2500 – 2700 m, close to the western edge of the the kind of soil profile varies with the soil-forming
erosion surface. All of the above creeks flow in factors (Birkeland, 1999). The surface O and A
steep-sided canyons; canyon relief is high in the horizons are commonly high in organic matter and
glaciated mountains, minor near the western edge of have characteristically dark colors. Beneath the O and
the erosion surface, and increases to a maximum just (or) A, mostly in acidic forested conditions, is a light
west of the mountain front. Maximum relief in the colored E horizon. The color of the latter horizon
canyons is Coal Creek, 660 m; Boulder Creek, 840 m; indicates that much of the iron released in it has been
and Left Hand Creek, 350 m. translocated to the underlying B horizon. Beneath the
There is disagreement regarding the elevation at A and (or) E horizons in most environments, given
which the erosion surface formed—did it form close sufficient time for them to develop, is a B horizon.
332 P.W. Birkeland et al. / Geomorphology 55 (2003) 329–344

There are several types of B horizons. A Bw horizon weakly developed that it barely meets the criteria for
in this area is generally oxidized. An example would its recognition, it is followed by the letter j (e.g., Ej or
be a parent material of 2.5Y hue alters to a Bw with a Btj), a practice we have adopted.
10YR hue. The alteration is due to Fe release during
weathering. Given sufficient time, clay accumulation
is recognizable, and the horizon is designated Bt. The 4. Factors of soil formation, Front Range of
clay has either infiltrated from overlying horizon(s) or Colorado
formed in place. In drier moisture regimes (grasslands
in the plains of this study), soil moisture is insufficient Jenny (1980) formulated the general soil-forming
to leach carbonate from the soil, and it accumulates to equation
form Bk or K horizons; these horizons form beneath
the Bw or Bt, or can occur as an overprint on the Bt Soil ¼ f ðcl; o; r; p; t; . . . ; Þ
horizon (e.g., Btk). Six morphological stages in the
development of carbonate horizons are recognized in where cl is climate, o organisms, primarily vegetation,
the western USA, and the time to attain each stage r slope or topographic setting, p parent material, t time
varies with climate and the rate of influx of Ca- over which the soil has formed, and . . . unknown
enriched dust (Machette, 1985). The first two stages factors, a common one important to this study is
(I and II) are Bk horizons in which the color of the airborne dust. Each factor has a strong influence on
non-carbonate material is clearly visible. If carbonate the resulting soil, and knowledge of these factors can
accumulation is extensive and the carbonate and its be used to predict the soil properties at a particular site.
white color dominate the horizon, the horizon is Climate and vegetation vary with altitude in the
designated K (stages III through VI). At depth is the study area transect (Fig. 2). The Colorado Piedmont is
little altered surficial parent material referred to as the warm and dry with grassland vegetation. Westward
C horizon (Cox if slightly altered and Cu if unaltered). into the mountains, the mean annual temperature
Rock parent material is designated R, Cr if the rock is progressively decreases, MAP increases, and the veg-
weathered, and Crt if clay has accumulated. Horizons etation varies accordingly. Most of the range is in
with evidence of gleying are designated by the letter g forest, except for tundra above about 3450 m.
(e.g., Bg). In Canada (Soil Classification Working Bedrock parent material in much of the mountains
Group, 1998), if a particular type of horizon is so consists of a Precambrian core of igneous and high-

Fig. 2. Vegetation zones and climatic data for a transect from the Continental Divide to the Colorado Piedmont near Boulder (modified from
Veblen and Lorenz, 1991, Fig. 2, and Barry, 1973, Table 4 and p. 92). Elevation ranges are typical for the vegetation zones. Gaps between
adjacent elevation ranges represent ecotones between the zones.
 P.W. Birkeland et al. / Geomorphology 55 (2003) 329–344 333

grade metamorphic rocks (Lovering and Goddard, There is less agreement on the climatic control and the
1950; Tweto, 1979; Braddock and Cole, 1990). The timing of the older surficial deposits. The oldest
oldest rocks (>1.7 by [billion years] old) are biotite widespread deposit, the Rocky Flats Alluvium, is
and hornblende gneiss and schist; these were intruded about 2 Ma (million years old) (Birkeland et al.,
by granodiorite (1.7 by) and granite (1.4 by). Small 1996). The older mountain chronosequence is formed
intrusive bodies of Tertiary age (mainly monzonite in two widespread deposits—till of the Pinedale gla-
and quartz monzonite) comprise a minor amount of ciation (younger, and associated with marine isotope
the igneous rocks. The Colorado Mineral Belt of stages 2 and 4) and till of the Bull Lake glaciation
Tertiary mineralization trends NE –SW through the (mainly marine isotope stage 6). They are dated and
area and intersects the range front north of Boulder; correlated by the radiocarbon method, cosmogenic
hydrothermal alteration of the bedrock within this belt isotope exposure age estimates, soils, and rock-weath-
is common. A narrow outcrop of E-dipping sedimen- ering features (Madole,1986; Madole and Shroba,
tary rocks veneers the front of the range, and shale 1979; Madole et al., 1998; Nelson and Shroba, 1998;
commonly underlies the Quaternary fluvial and other Schildgen, 2000). Pertinent local ages are a radio-
surficial deposits in the plains. carbon age of 23.5 ky (thousand years old) for the
The parent material for most soils discussed here maximum extent of Pinedale ice, and a minimum
formed from rocks and surficial deposits high in cosmogenic isotope exposure age of 122 ky for the
feldspar, quartz, biotite, and hornblende. Most of the maximum extent of Bull Lake ice. A younger moun-
deposits are gravelly with a sandy matrix, except for tain chronosequence is formed in rock-glacier deposits
loess caps that are high in silt. and in tills deposited by glaciers that re-occupied
The ages of the soils throughout the study area can cirques after the disappearance of Pinedale ice. The
be estimated from stratigraphic and geochronologic oldest post-Pinedale till is about 12 ky (possible
studies in both the plains and mountains. A well- Younger Dryas equivalent), and there are at least three
expressed suite of river terrace deposits in the plains younger Holocene advances. Dating and correlating of
(Fig. 3) is dated and correlated by the radiocarbon and these deposits are by the radiocarbon method, lichen-
uranium-series methods, association with volcanic ash ometry, soils, and rock-weathering features (Benedict,
of known age, soils, and relative height above streams. 1985; Birkeland et al., 1987; Davis et al., 1992).
The lower terrace deposits are generally accepted as There are few data on the influence of topographic
having been deposited during the Holocene and the setting on soil development in the study area, as all the
Pinedale and Bull Lake glaciations (Madole, 1991). soil chronosequence study sites were on relatively flat

Fig. 3. Diagrammatic cross-section of alluvial units in the Colorado Piedmont near Denver (from Madole, 1991, Fig. 14). Approximate ages are:
post-Piney Creek, Piney Creek, and pre-Piney Creek alluvial units, Holocene; Broadway Alluvium, Late Pleistocene (Pinedale glaciation);
Louviers Alluvium, Late Pleistocene, or late Middle Pleistocene (Bull Lake glaciation); Slocum Alluvium, 240 ky; Verdos Alluvium, similar to
age of intercalated Lava Creek B ash, 640 ky; Rocky Flats Alluvium, about 2 Ma.
334 P.W. Birkeland et al. / Geomorphology 55 (2003) 329–344

surfaces. However, catena studies on Holocene and 6.1. Grassland soil chronosequence
Pleistocene moraines in other parts of the Rocky
Mountains serve as a model of soil development on Birkeland et al. (1996) provide the latest review of
slopes with time (Birkeland, 1999, p. 247). Aspect soils in the plains, much of which is based on the
influences soil development in the canyons because quantitative studies by Machette (1975, 1977, 1985)
they trend E – W, and the bioclimatic zones in the and Machette et al. (1976).
canyons are strongly influenced by aspect. At any A characteristic soil sequence is present through-
given elevation, the S-facing slopes are relatively out the plains of the study area (Fig. 4); carbonate
warm, and the N-facing cool, and these differences has accumulated in many of the soils and the
are reflected in the vegetation (Veblen and Lorenz, morphological stages of development of these soils
1991). are from Machette (1985, Table 2). Soils formed in
Holocene deposits (Piney Creek Alluvium and
younger) are A/Cu profiles in the youngest deposit,
5. General soil distribution, Front Range of and A/Bw/C with or without a Bk in the older
Colorado deposits. Overthickened (cumulic) A horizons, due
to pedogenesis keeping pace with floodplain sedi-
The general soil pattern of the study area reflects mentation, are common. Some of these latter A
the impact of the above soil-forming factors (U.S. horizons are several times thicker than the typical
Department Agriculture, Soil Conservation Service, (noncumulic) A horizons. Soils in Broadway Allu-
1975, with changes that correspond with the most vium (Pinedale age) have A/Bw/C profiles on coarse-
recent Soil Taxonomy, Soil Survey Staff, 1999). grained alluvium, but A/Bt/Bk profiles on fine-
Argiustolls and Paleustolls (A/Bw or Bt/Bk or K/C grained (silty) alluvium. Carbonate morphology is
profiles) dominate in the plains. The forested moun- stage I in post-Broadway and younger soils. Mor-
tains are mainly Cryalfs (cold soils with A/E/Bt/C phological differences with grain size are a good
profiles) at higher elevations and Ustalfs at lower example of the effect of the parent material factor, as
elevations. Cryepts (cold soils with A/Bw/C profiles) soils formed from the finer-grained (silty) materials
are above tree line. Soil pH decreases with elevation have sufficient primary clay for soil water to redis-
from about 8 in the grasslands to 5 and less in the tribute it downward to form a Bt horizon, and the
tundra (Netoff, 1977, written communication 1970; slight decrease in permeability is sufficient to cause
Birkeland et al., 1987). There are a wide variety of carbonate to precipitate rather than be carried in
clay minerals in the soils of the study area. Clays that solution through the soil. Soils formed in loess
are best associated with the overall bioclimate and considered to be the same age as the Broadway
geochemical conditions are smectite in the plains, Alluvium also have an A/Bt/Bk profile (Reheis,
vermiculite in the forest, and mixed layer 1.0 – 1.8- 1980). Soils formed in Louviers Alluvium (Bull Lake
nm clays above tree line (Netoff, 1977; Shroba and age) have an A/Bt/Bk profile with stage II – III
Birkeland, 1983; Birkeland et al., 1987). carbonate morphology. Soils in older alluvial depos-
its are progressively thicker, redder, more clay rich
(Fig. 5), and display higher stages of carbonate
6. Soil chronosequences morphology. The soil in the Rocky Flats Alluvium
(about 2 Ma) is about as clay rich, red (10R in the Bt
We have studied soil chronosequences in three horizon), and carbonate rich in the K horizon (94%)
major vegetation zones—the grasslands using mainly as any soil on Earth (Birkeland, 1999). Machette
fluvial deposits about 2 Ma and younger, the upper (1985) ranks the plains near Boulder as one of
montane using tills of the Pinedale and Bull Lake relatively low carbonate influx when compared to
glaciations (about 14 –47 and 120 – 160 ky, respec- other areas in the western USA. Pre-Broadway
tively; Nelson and Shroba, 1998), and in the alpine loesses are not present in the study area either
tundra using tills and rock-glacier deposits 12 ky and because they were deposited only in areas farther
younger. east (Muhs et al., 1999) or, if deposited, they were
 P.W. Birkeland et al. / Geomorphology 55 (2003) 329–344 335

Fig. 4. Diagram showing maximum development of soils formed in alluvial units and loess (from Madole, 1991, Fig. 15). Although developed
for the Lafayette area, Colorado (Machette, 1977), these soils are typical of those of the Colorado Piedmont in the Denver – Boulder area. Letters
denote soil horizons.

locally incorporated into soils on pre-Broadway 6.2. Forest soil chronosequence

deposits (Shroba and Carrara, 1996), or were sub-
sequently eroded. The forest chronosequence consists of soils formed
The Rocky Flats Alluvium has an estimated age of in tills of the Pinedale and Bull Lake glaciations in the
about 2 Ma, based in large part on landscape position, subalpine and upper montane forests, and therefore
minimum-limiting age of younger deposits, and spans the last 120 –160 ky (thousand years). Rich-
buried soils (Machette et al., 1976; Birkeland et al., mond (1960) did the early work on soils. In this study,
1996). The nearby Verdos Alluvium is dated by its we draw on the work by Madole (1969), Netoff
association with the Lava Creek B Ash (0.64 Ma; (1977), Madole and Shroba (1979), Shroba and Birke-
redated by Lanphere et al., 2002), and has an A/5YR land (1983), and Madole et al. (1998).
Bt/stage III – IV K profile. The period of time between The characteristic soil on till of the Pinedale glaci-
the deposition of the Verdos Alluvium and the Rocky ation has an O and (or) A/Ej or E/Bw or Btj/Cox
Flats Alluvium is locally represented by at least seven profile. Hue of the B horizon is usually 10YR, and
buried soils. Using the morphology of the latter, their most of the clasts within the soil are unweathered. Clay
clay and carbonate contents, and the fact that both content has a maximum content near 10% ( < 2 mm-
normal and reversed magnetic components are present size faction) and commonly decreases progressively
in these soils, the age of the Rocky Flats is likely to be with increasing depth. This distribution is thought to
about 2 Ma. reflect the downward infiltration of airborne dust.
336 P.W. Birkeland et al. / Geomorphology 55 (2003) 329–344

thin A/Cu profile. Pedogenesis in increasingly older

deposits results in an A/Cox/Cu profile in 2 ky, and an
A/Bw/Cox profile in 3.5 ky with a maximum hue of
10YR. Soil color (chroma and value) is used to
differentiate the Cox horizon from the Bw. The 12-
ky soil has an A/Bw or Bt/Cox profile. The Bt
horizons have a maximum clay content of 14%,
double that of the underlying horizon, and most have
a 7.5YR hue (versus 10YR for the Bw horizons).
Fine-grained material (fine sand and finer) is
prominent in the A horizons of soils 2 ky and older.
It is considered to be an eolian material mixed with
the uppermost part of the underlying till or rock-
glacier deposit. Geochemical data indicate that part
of the fine-grained material is derived from semiarid
basins upwind of the alpine soils (Muhs et al., 1992).
Fig. 5. Plot of clay content ( < 2 mm-size fraction) with depth in Locally, it has been translocated and produces silt
soils of the Colorado Piedmont in the Denver – Boulder area (from caps on clasts at depth (Burns, 1980) and contributes
Machette, 1975, and Machette et al., 1976). to the clay content in the Bt horizon.
Comparing the alpine and adjacent forest chrono-
sequences, one sees that some 12-ky alpine tundra
The characteristic soil on till of the Bull Lake soils can be as well, or better, developed than 20-ky
glaciation is an O and (or) A/E/Bt/Cox profile. The soils at lower elevations in the forest. Several factors
Bt commonly has a 7.5YR hue, clay content reaches a that might help explain this are (1) greater eolian
maximum of 16%, and about 20% of biotite-rich influx in the alpine, promoting greater availability of
granitic and gneissic clasts within the soil is weath- fines, (2) greater soil moisture in parts of the alpine
ered to grus. At one locality near Nederland, where due to wind redistribution of snow and the potential
there appears to be superimposed tills of the Pinedale for melt-season flushing by large amounts of water
and Bull Lake glaciations, the buried post-Bull Lake through the soil and possibly (3) greater local dis-
soil has a morphology similar to the above soil. We turbance of soils by tree throw in forested areas.
interpret this to mean that the major morphological
features of the post-Bull Lake soil formed prior to the 6.4. Soil catena chronosequences
time when the Pinedale ice was close to its maximum
extent. It is possible that much of this soil develop- Soil catena (toposequence) studies have not been
ment occurred during the last interglacial period when undertaken on moraines in the Rocky Mountains of
climate over much of North America was warmer than Colorado, but some have been done in other parts of
the present (Muhs et al., 2002). the Rocky Mountains and western USA (reviewed by
When one compares the soils younger than about Birkeland, 1999, chapter 9). A soil catena consists of a
160 ky in the mountains with those of similar age in number of soils aligned downslope on one landform
the plains, one sees many similarities in the relative to depict soil morphologies and properties at key
development of B-horizon properties (e.g., clay con- slope positions. If these soils are on moraines of
tent and color). different ages, they constitute a soil catena chronose-
quence. These studies can provide age estimates for
6.3. Alpine tundra soil chronosequence slopes and slope deposits throughout the Front Range
of Colorado.
The alpine chronosequence spans the time between Although many of our catena studies were under-
about 0.3 and 12 ky (Mahaney, 1974; Benedict, 1985; taken in sagebrush terrain in the western USA, Berry
Birkeland et al., 1987). Deposits that are 0.3 ky have a (1987) undertook one in the forested terrain in Idaho.
 P.W. Birkeland et al. / Geomorphology 55 (2003) 329–344 337

In general, the downslope soils have the same overall 7. Soil distribution in the Front Range of Colorado
profile morphology as those upslope on the nearly flat
moraine crest. In other words, the post-Pinedale soils The U.S. Department Agriculture, U.S. Forest
have mainly Bw horizons at all slope positions, and Service (2001) has prepared soil maps of the study
most of the post-Bull Lake soils have Bt horizons at area at a scale of 1:24,000, following a 3rd-order soil
all slope positions. Downslope changes are least in the survey (Soil Survey Division Staff, 1993). These
post-Pinedale soils, but in the post-Bull Lake soils, maps show a general relation between the major soils
both the Bt clay content and clay maximum are more and various elements of the landscape (Fig. 6).
pronounced downslope and pedogenic Fe content Most of the mountainous terrain west of Nederland
parallels the clay content with depth. and Ward has soils with A/Bw/C and (or) R profiles in
In some catena studies in the western USA, the the tundra and O and (or) A/E/Bw/C and (or) R
summit soil is the least developed of that catena. This profiles in the forest. Such profiles are due either to
is expected as soil moisture is least in the summit young deposits (Pinedale age and Holocene tills and
position. The relative development of some of the periglacial deposits) or to erosion under the present or
summit soils is so weak that we wondered if their glacial-age climate and vegetation. Rolling bedrock
weak development could be due in part to erosion. topography characterizes the unglaciated windswept
One possible answer came from a post-fire study of a tundra between glaciated valleys. Burns and Tonkin
forested catena in Wild Basin (McMillan, 1990), (1982) noted that soils on this undulating topography
about 35 km northwest of Boulder. The summit soil form a catena. The key to understanding these soils is
in burned areas has an A/Cox/Cu profile, whereas the wind redistribution of snow and dust (referred to as
downslope soils have A/E/Bw/Cox profiles. Erosion, mixed loess when it is incorporated into the under-
even on nearly flat surfaces, was widespread in lying soil). Both of them are minimal on the topo-
burned areas and over time could help explain these graphic highs and increase in thickness in the lee of
contrasting soils with catena positions. the highs. Hence, soils on the highs are mainly thin
and either lack or have a surface layer of mixed loess.
6.5. Application of chronosequence and catena Greater soil moisture downslope in the lee positions
studies to soils of the Front Range of Colorado results in progressively better developed and thicker
soils, as well as a mixed loess surface layer. At the
The above soil development relations can be used base of the slope, the soils show signs of poor drain-
to provide a time framework for the development of age (O/Bg profiles) and the thickest mixed loess,
soils in the Front Range of Colorado. For example, a transported from upslope chiefly during snowmelt.
soil with a mainly cumulic A horizon denotes for- Fir-spruce tree islands are a common feature in the
mation during the Holocene. Soils with a well-devel- tundra, just above tree line. They migrate downwind
oped Bw horizon or a weak Bt horizon suggest at a rate about 1 – 2 m/0.1 ky (Benedict, 1984).
formation over a period of between 10 ky and greater Holtmeier and Broll (1992) noted that the islands trap
than 20 ky. Soils with A/brown and sandy loam Bt/ both windblown snow and fine-grained eolian mate-
Cox profiles suggest soil development of at least 50 rial, resulting in underlying soils that are relatively
ky and perhaps as long as 150 ky. Finally, soils with leached A/Bw/C profiles and have a fine-grained
Bt horizons that are red (2.5YR) and have sandy clay surface layer. Burns has noted that the B-horizon
loam texture, or a texture class with even more clay, hue under these tree islands is 7.5YR versus 10YR
denote formation over several 100 ky. Interestingly, in the surrounding tundra. Only where the tree islands
these relations are generally true throughout the move slow enough is there sufficient time for E
range regardless of elevation. It could be that other horizons to form (Benedict, 2000).
factors locally influence soil development, such as Many of the soils on the erosion surface are O and
aspect, but consistently greater degree of soil devel- (or) A/Bt/C and (or) R profiles. These soils are
opment with stratigraphically older deposits under- extensive on wide tracts of the erosion surface. East-
scores the importance of time in pedogenesis in this ward, the erosion surface consists of accordant ridges,
area. and where the ridges are narrow the soils are A/Bw/C
338 P.W. Birkeland et al. / Geomorphology 55 (2003) 329–344

Fig. 6. Schematic perspective diagram showing relations between soils and landscape in the Front Range of Colorado and adjacent Colorado
Piedmont between Boulder and the Continental Divide. Cryolls and Ustolls shown are on S-facing slopes, but Cryalfs and Ustalfs on adjacent
N-facing slopes cannot be shown because of the orientation of the diagram. Cryepts commonly form in the steep terrain west of the erosion
surface; they have A/Bw/C and (or) R profiles (tundra) as well as O and (or) A/E/Bw/C and (or) R (forest) profiles. Cryalfs and Ustalfs
commonly form on the erosion surface east of Ward and Nederland. They have O and (or) A/E/Bt/C and (or) R profiles. Dotted patterns in two
small areas near Ward and Boulder are Cryalfs that have clayey, red (2.5YR hue) Bt horizons. Cryolls and Ustolls commonly form on S-facing
slopes in the canyons incised into the erosion surface; they have A/Ej/Bw/C and (or) R as well as A/Bw/C and (or) R profiles. Ustolls commonly
form in the plains; they have A/Bw/Bk and A/Bt/K profiles. Gray pattern along the mountain front indicates a large area of bedrock that is
locally mantled by surficial deposits and their associated soils. Soil distribution, classification, and horizon designations are modified from maps
(1:24,000 scale) and profile descriptions by the U.S. Department Agriculture, U.S. Forest Service (2001, and written communication, 1999,
respectively).

and (or) R profiles. Interestingly, these well-developed low temperature negates the effect of aspect. From the
soils with Bt horizons remain well expressed adjacent western edge of the erosion surface to the mountain
to areas formerly covered by Pinedale ice. We inter- front, however, N-facing soils are mainly O and (or)
pret this to mean that periglacial processes were not of A/E/Bt/Bw/C and (or) R profiles. In contrast, S-facing
sufficient intensity or duration to erode these soils. If soils have A/Ej/Bw/C and (or) R profiles and the A
this is correct, then these parts of the erosion surface horizons of these soils are thicker than the O and (or)
have been stable for at least 100 ky. On the erosion A horizons of the N-facing soils. These profile con-
surface near Ward (Fig. 6), there is a small area of soil trasts result in Cryalfs on the N-facing slopes and
with a morphology similar to the above, except that Cryolls on the S-facing slopes; if temperatures are
the Bt has a 2.5YR hue and a sandy clay texture. frigid at lower elevations, however, the soils are
Comparing this soil with those in the plains suggests classified as Ustalfs and Ustolls, respectively. This
several 100 ky of landscape stability to form such a contrast in soil-profile development could either be a
soil. function of age (N-facing slope deposits are more
Soils on the N-facing versus S-facing valley sides stable and older) or of process (Bt horizons form
(U.S. Department Agriculture, U.S. Forest Service, faster on N-facing slopes due chiefly to greater
2001) provide an opportunity to compare the influ- effective soil moisture).
ence of aspect on soil development (Fig. 7). In the Reconnaissance work by bicycle in many road cuts
tundra, as well as in the subalpine forest, soil maps along main and mining roads in the mountains indi-
show little difference in soil morphology between N- cates that grus development shows a strong relation
and S-facing slopes. Most of the above soils are with topographic setting and geological history. Clay-
Cryepts with a similar horizon sequence. Apparently, ton et al. (1979) have developed a seven-class classi-
the combination of relatively high precipitation and fication of granite weathering. Class 1 is unweathered
 P.W. Birkeland et al. / Geomorphology 55 (2003) 329–344 339

out the rest of the range, suggesting either moderate

erosion or a slow rate of weathering. In contrast, just
west of the contact between the sedimentary rocks and
the igneous– metamorphic complex is a narrow linear
band of rocks with class 4– 6 weathering. This could
be due to more intense weathering at the eastern
margin of the range, or weathering association with
the paleosol formed in granodiorite beneath the low-
ermost sedimentary rocks (Fountain Formation of
Permian and Pennsylvanian age), thoroughly studied
by Wahlstrom (1948). Our interpretation of the pale-
osol is that it is mainly a thick section of grus (class 6
maximum) (Birkeland et al., 1996). We suggest that
the linear band of class 4– 6 grus is either a lateral
expression of Wahlstrom’s paleosol (grus), or it is
some combination of pre-Fountain weathering of the
granodiorite enhanced by weathering during the Late
Cenozoic. Other areas of thick class 4 – 6 grus have
escaped erosion over the long-term for a variety of
reasons, one being the low rate of local steam incision.
The higher classes of granite weathering indicate
landscapes that were not rapidly eroded over long
periods of time. We do not know the rate of grus
formation, but we can make an estimate by consider-
ing the condition of boulders in tills. Those in tills of
Fig. 7. Generalized diagram showing soil profiles on N-facing Pinedale age are generally unweathered, but some in
versus S-facing slopes in the canyons near Boulder. Each profile is tills of Bull Lake age are sufficiently weathered to be
about 1 m thick. We estimate that S-facing slopes have the cryic – class 4– 6. Assuming this, grus formation can also
frigid soil – temperature-regime boundary within the upper montane take place in soil formed from rock and that most soils
vegetation zone. Some N-facing slopes can have a cryic soil –
are a minimum of 1 m thick, a 1-m thick layer of grus
temperature regime to the mountain front; if warmer at lower
elevations, however, the soil – temperature regime is frigid and the formed from granitic or gneissic bedrock probably
soils are Ustalfs. Modified from maps and profile descriptions by takes a minimum of about 100 ky to form.
the U.S. Department Agriculture, U.S. Forest Service (2001, and
written communication, 1999).
8. Some applications of soils of the Front Range of
rock; class 2 is oxidized rock with angular joint – plane Colorado
junctions. Classes 3– 6 are subdivided on the distinct-
ness and rounding of the joint plane junctions (gone Soils can be used to estimate incision rates of the
by class 5, and this is approximately where grus canyons that characterize the Front Range of Colo-
appears in the weathering sequence) and the ease with rado. One can use the ages of the fluvial terrace
which rock fragments disaggregate (easily disaggre- deposits at the mouths of the canyons—many esti-
gates in class 6). Class 7 exhibits rock fabric, but mated by their associated soils—to estimate incision
chemical weathering is advanced enough to form clay. rates, and then compare these soils with those on
The first task is to make sure that hydrothermal canyon side slopes. Using the former, Dethier et al.
alteration is not confused with surface weathering. (2000) calculated an incision rate of 0.04 m/ky in the
The landscape west of the erosion surface is generally plains at the mouth of Boulder Canyon. Schildgen
class 1, especially the part that was covered by Pine- (2000) and Schildgen and Dethier (2000) used soils to
dale glaciers. Class 2 weathering is common through- help guide a radiocarbon and cosmogenic isotope
340 P.W. Birkeland et al. / Geomorphology 55 (2003) 329–344

dating study within Boulder Canyon; they identified of the heavily mined sites did not respond drastically
deposits of Holocene, Pinedale, Bull Lake, and pre- to perturbations caused by human activities.
Bull Lake ages. At a major knickpoint within the Quaternary climate change can be deciphered in
canyon, they calculate an incision rate of 0.15 m/ky. some soils. Shroba and Birkeland (1983) note the
The contrast between this higher rate and the much presence of vermiculite and a high degree of illite
lower rate determined in the plains might be due in (mica) weathering in soils now 100 m above tree line.
part to the upstream migration of the knickpoint or to Because vermiculite is a forest indicator (Netoff,
the durability of the bedrock at or near the knickpoint. 1977; Shroba and Birkeland, 1983), these data suggest
Most soils on side slopes in Boulder Canyon seem to a tree line higher than present sometime during the
be fairly young, less than 20 ky. This seems fitting as Holocene. Organo-cutans are dark brown organic
the side slopes are steep, cliffs are common, and of the stains that form on the undersides of stones in the
three streams here compared, Boulder Creek has the lower part of the B horizon and in the C horizon of
highest discharge and was the most extensively gla- alpine soils that lack excessive snow cover (Burns,
ciated in its headwaters. The next major canyon to the 1980; Benedict, 2000, Appendix B). Burns and
north, Left Hand, has an incision rate in the plains Davenport (1988) observe these stains in forested
near the mountain front of 0.05 – 0.1 m/ky (range in soils at elevations as low as 2735 m, suggesting that
values is due to which fluvial deposit is used as a they formed in alpine soils during glacial-age low-
datum). In Left Hand Canyon, the headwaters were ering of upper tree line of greater than 700 m. This
not extensively glaciated, the canyon side slopes are interpretation is consistent with pollen data that sug-
more gentle than those in Boulder Canyon, and soils gest 500 – 800 m of tree line lowering during the
with Bt horizons on side slopes are more common. A Pinedale glaciation (Legg and Baker, 1980; Madole,
major canyon south of Boulder Canyon is the ungla- 1986). Strongly developed E horizons that persist to
ciated Coal Creek Canyon, where incision in the the mountain front might partly be a legacy of glacial
plains at the mountain front is 0.006 m/ky. Soils on climates, and the intact profiles suggest little tree
the valley sides near the canyon mouth commonly throw. Finally, Btk horizons are common in the plains;
have red, clay-rich Bt horizons that suggest they this mix of carbonate depletion and accumulation in
formed over several 100 ky (Birkeland et al., 1996). the same horizon might reflect leaching during glacial
Thus, soils on side slopes of the canyons correspond climate and accumulation during interglacial climate,
with incision rates. Coal Creek has the lowest incision respectively.
rate and the oldest soils on side slopes. Boulder Creek The rate of carbon accumulation and its storage in
has a high incision rate and the youngest soils on side soils are important issues in current biogeochemical
slopes. The differences in the development of soils on cycling studies. These trends are best studied by soil
side slopes in Boulder Canyon and Left Hand Canyon chronosequence studies. Bockheim et al. (1998) have
could be related to the higher discharge, more exten- taken data from our studies in the Front Range of
sive glacier activity, and knickpoint migration in the Colorado, the Wind River Range of Wyoming, and
former. the Southern Alps of New Zealand to make compa-
Veblen and Lorenz (1991) have documented the rative accumulation curves for alpine environments.
great change in vegetation in the Front Range of The New Zealand soils accumulate carbon faster and
Colorado during the last century. Many of the changes reach higher levels of carbon content, and those of the
are associated with mining, and the logging and Front Range of Colorado rank higher on both
burning that accompanied the mining. Many slopes accounts than those of the Wind River Range. We
were extensively devegetated. Reconnaissance study need more of these kinds of data for many other
of road cuts along many mining roads in the moun- environments, such as those for the forests of the
tains via bicycle reveals neither widespread extensive Front Range of Colorado.
erosion (enough to remove the steady-state A horizons Soil-profile development can be used to estimate
that take several thousand years to form) nor deposi- long-term (20 – 100 ky) rates of lowering of the
tion (buried soils) in these disturbed areas. These landscape (denudation). A/Bt/C profiles, such as
observations suggest that the mountain slopes outside those associated with the erosion surface, are best
 P.W. Birkeland et al. / Geomorphology 55 (2003) 329–344 341

for placing broad limits on rates of denudation. These accumulation of wind-blown fines, which produce a
soils take about 100 ky to form and typically the base result opposite to that of denudation. In any case, over
of the Bt horizon is close to 1-m depth. Therefore, to long periods of time, the lower of the two rates above
preserve such a profile, denudation had to be much would be compatible with the A/Bw/C soil profiles on
less than 1 m/100 ky (0.01 m/ky), and perhaps no Niwot Ridge, but the higher rate could not have been
more than 0.1 m/100 ky (0.001 m/ky). Those small sustained for even 10 ky and allow preservation of the
areas of soils with red, clay-rich Bt horizons would soils.
have an even lower rate of denudation. Comparing The above rates of denudation for the mountains
the latter denudation rate for the erosion surface west of the erosion surface (0.005 – 0.01 m/ky) and for
(0.001 m/ky) with the incision rate of Boulder Niwot Ridge (0.01 m/ky) seem reasonable when
Canyon mentioned earlier (0.04 m/ky) suggests that compared with other cosmogenic radionuclide erosion
relief in Boulder Canyon is increasing. The moun- rates in the nearby Front Range of Colorado. Small et
tains west of the erosion surface, outside of the areas al. (1997) calculate rates for tors and a large boulder
covered by Pinedale glaciers, are characterized by O of 0.007 –0.008 m/ky. These dated features are on
and (or) A/Bw/C and (or) R soil profiles. One summit surfaces above treeline, at 3,575 –3,734 m.
possible denudation scenario is that these soils The erosion rates for tors should be considered
formed in the last 10 –20 ky following sufficiently minimal because they are an low, isolated features
rapid denudation during the Pinedale glaciation to on low-relief terrain. The above rates are lower that
have removed the pre-existing soils. Assuming 1-m- those of Dethier et al. (2002) who determined 0.018 –
deep profiles, their preservation requires a denudation 0.030 m/ky for sediment removed from small (<50
rate much less than 0.05 –0.1 m/ky, perhaps as low as km2), non-glaciated catchment basins on granitic and
0.005 – 0.01 m/ky. This latter denudation rate is gneissic bedrock in the northern Front Range of
greater that that estimated for areas characterized by Colorado and the Laramie and Medicine Bow Ranges
soils with A/Bt/C profiles because Bw horizons form of southern Wyoming.
more rapidly than Bt horizons. The latter range
agrees with that calculated by Caine (1984; personal
communication in Thorne and Loewenherz, 1987) for 9. Summary
the tundra-covered Green Lakes Valley, just south of
Niwot Ridge. Soils reflect the geomorphic setting of the Front
Niwot Ridge is a broad, gently rolling ridge above Range of Colorado. The Front Range is an old range,
tree line, with a maximum elevation of about 3700 m. with canyons cut during the last 5 my. Glaciation
It is characterized by A/Bw/C soil profiles (Burns and affected the western quarter or third of the range east
Tonkin, 1982). Solifluction and frost creep have been of the Continental Divide. MAP doubles from the
active in moving surficial materials downslope during plains to the divide and MAT decreases 14 jC over
the Holocene (Benedict,1970). Bovis and Thorn the same distance; these environmental differences
(1981) measured contemporary erosion on Niwot result in contrasts in soil profiles. Soil chronosequence
Ridge and calculated denudation rates. If we extrap- studies help identify surfaces and deposits that have
olate their rates over a longer time frame, the rate for been stable during the Holocene, for about 20, 100,
the tundra meadow would be 0.01 m/ky (1 m/100 ky), and several 100 ky in the mountains, and for as long
and that for the dry meadow would be 0.1 m/ky (10 as about 2 my in the plains. Compared to a tectoni-
m/100 ky). These high values make us suspect that cally active range, such as the Southern Alps of New
these measured rates should not be extrapolated so far Zealand, soils in the Front Range of Colorado on
back in time. The previous authors as well as Caine predominately transport-limited slopes indicate that
(1974) used Benedict’s (1970) surface-material trans- the latter is geomorphically quite stable, and most
portation rates, most taken at sites that were quite similar to the drier parts of the Southern Alps. Future
active, to calculate a denudation rate of 0.01m/ky (1 work could concentrate on how soil-development
m/100 ky). All of these rates have problems associ- patterns relate to geomorphic activity with elevation
ated with them, and one important thing is the in the Front Range of Colorado (Caine, 1984). Appli-
342 P.W. Birkeland et al. / Geomorphology 55 (2003) 329–344

cations of soils include (1) helping to estimate stream soils in gneissic deposits, Colorado Front Range. U.S. Geol.
incision rates in the canyons; (2) linking the latter to Surv. Bull. 1590. 21 pp.
Birkeland, P.W., Miller, D.C., Patterson, P.E., Price, A.B., Shroba,
soil patterns on canyon side slopes following the R.R., 1996. Soil-geomorphic relationships near Rocky Flats,
model that Tonkin and Basher (1990) developed in Boulder and Golden, with a stop at the pre-Fountain Formation
the Southern Alps of New Zealand; (3) assessing paleosol of Wahlstrom (1948). Colo. Geol. Surv., Spec. Publ.
landscape stability following disruption accompany- (CD-ROM) 44. 13 pp.
Bockheim, J.G., Birkeland, P.W., Bland, W.L., 1998. Carbon stor-
ing mining; (4) assessing the effects of Quaternary
age and accumulation rates in alpine soils: evidence from Hol-
paleoclimate on the landscape as well as on soil ocene chronosequences. In: Lal, R., Kimble, J.M., Stewart, B.A.
development; and (5) estimating long-term denuda- (Eds.), Global Climate Change and Cold Regions Ecosystems.
tion rates. Lewis Publishers, Boca Raton, pp. 185 – 196.
Bovis, M.J., Thorn, C.E., 1981. Soil loss variation within a Colo-
rado alpine area. Earth Surf. Processes Landf. 6, 151 – 163.
Braddock, W.A., Cole, J.C., 1990. Geologic map of Rocky Moun-
Acknowledgements tain National Park and vicinity, Colorado. U.S. Geol. Surv.
Misc. Investigations Series Map I—1973.
Our understanding of various aspects of the soils Bradley, W.C., 1987. Erosion surfaces of the Colorado Front
and geomorphology of the Front Range and Colorado Range: a review. In: Graf, W.L. (Ed.), Geomorphic Systems
Piedmont was enhanced by conversations with many of North America. Geol. Soc. Amer., Centennial Spec., vol. 2,
pp. 215 – 220.
individuals during the past decades. They include Burns, S.F. 1980. Alpine soil distribution and development, Colo-
W.W. Atkinson, J.B. Benedict, W.C. Bradley, Nel rado Front Range. PhD Dissertation, Univ. Colorado, Boulder.
Caine, D.P. Dethier, M.N. Machette, R.F. Madole, 360 pp.
Gergely Markos, D.I. Netoff, P.E. Patterson, and John Burns, S.F., Davenport, R.E., 1988. Using organo-cutans in subal-
pine soils as indicators of a past lower treeline. Amer. Quaternary
Pitlick. D.R. Muhs, A.R. Nelson, and John Pitlick
Assoc., 10th Biennial Meeting, Program and Abstracts, p. 57.
provided helpful reviews of previous versions of this Burns, S.F., Tonkin, P.J., 1982. Soil geomorphic models and spatial
manuscript. M.A. Berger of the US Geological Survey distribution and development of alpine soils. In: Thorn, C.E.
prepared digital copies of the figures. P.W. Birkeland’s (Ed.), Space and Time in Geomorphology. Allen and Unwin,
attendance at the 2001 Binghamton Symposium was London, pp. 25 – 43.
funded by a Bill Hiss Creativity Award granted by the Caine, N., 1974. The geomorphic processes of the alpine environ-
ment. In: Ives, J.D., Barry, R.G. (Eds.), Arctic and Alpine En-
University of Colorado. vironments. Methuen, London, pp. 721 – 748.
Caine, N., 1984. Elevational contrasts in contemporary geomorphic
activity in the Colorado Front Range. Stud. Geomorphol. Car-
References patho-Balc. 18, 5 – 31.
Clayton, J.L., Megahan, W.F., Hampton, D., 1979. Soil and bedrock
Barry, R.G., 1973. A climatological transect on the east slope of the properties: weathering and alteration products and processes in
Front Range, Colorado. Arct. Alp. Res. 5, 89 – 110. the Idaho batholith. USDA For. Serv. Res. Pap. INT-237. 35 pp.
Benedict, J.B., 1970. Downslope soil movement in a Colorado Davis, P.T., Birkeland, P.W., Caine, N., Rodbell, D.T., 1992. New
alpine region: rates, processes, and climatic significance. Arct. radiocarbon ages from cirques in Colorado Front Range. Geol.
Alp. Res. 2, 165 – 226. Soc. Amer., Abstr. Programs 24 (7), A347.
Benedict, J.B., 1984. Rates of tree-island migration, Colorado Dethier, D.P., Schildgen, T.F., Bierman, P., Caffee, M., 2000. The
Rocky Mountains, USA. Ecology 27, 820 – 823. cosmogenic isotope record of late Pleistocene incision, Boulder
Benedict, J.B., 1985. Arapaho Pass. Glacial geology and archeol- Canyon, Colorado. Geol. Soc. Amer., Abstr. Programs 32 (7),
ogy at the crest of the Colorado Front Range. Center for Moun- A473.
tain Archeology, Research Report, 3. 197 pp. Dethier, D.P., Ouimet, W., Bierman, P., Finkel, R.C., 2002. Long-
Benedict, J.B., 2000. Game drives of the Devil’s Thumb Pass Area. term erosion rates derived from 10Be in sediment from small
In: Cassells, E.S. (Ed.), This Land of Shining Mountains. Center catchments, Northern Front Range and Southern Wyoming.
for Mountain Archeology, Research Report, 8, 18 – 94. Geol. Soc. Amer., Abstr. Programs 34 (6).
Berry, M.B., 1987. Morphological and chemical characteristics of Holtmeier, F., Broll, G., 1992. The influence of tree islands and
soil catenas on Pinedale and Bull Lake moraine slopes in the microtopography on pedoecological conditions in the forest –
Salmon River Mountains, Idaho. Quat. Res. 28, 210 – 225. alpine tundra ecotone on Niwot Ridge, Colorado Front Range,
Birkeland, P.W., 1999. Soils and Geomorphology. Oxford Univ. USA. Arct. Alp. Res. 24, 216 – 228.
Press, New York. 430 pp. Jenny, H., 1980. The Soil Resource—Origin and Behavior. Springer,
Birkeland, P.W., Burke, R.M., Shroba, R.R., 1987. Holocene alpine New York. 377 pp.
 P.W. Birkeland et al. / Geomorphology 55 (2003) 329–344 343

Lanphere, M.A., Champion, D.E., Christiansen, R.L., Izett, G.A., northeastern Colorado: part I—age and paleoclimatic signifi-
Obradovich, J.D., 2002. Revised ages for tuffs of the Yellow- cance. Geol. Soc. Amer. Bull. 111, 1861 – 1875.
stone Plateau volcanic field: assignment of the Huckleberry Muhs, D.R., Simmons, K.R., Steinke, B., 2002. Timing and warmth
Ridge Tuff to a new geomagnetic polarity event. Geol. Soc. of the last interglacial period: new U-series evidence from Ha-
Amer. Bull. 114, 559 – 568. waii and Bermuda and a new fossil compilation for North Amer-
Legg, T.E., Baker, R.G., 1980. Palynology of Pinedale sediments, ica. Quat. Sci. Rev. 21, 1355 – 1383.
Devlins Park, Boulder County, Colorado. Arct. Alp. Res. 12, Nelson, A.R., Shroba, R.R., 1998. Soil relative dating of moraine
319 – 333. and outwash – terrace sequences in the northern part of the upper
Lovering, T.S., Goddard, E.N., 1950. Geology and ore deposits of Arkansas Valley, central Colorado, U.S.A. Arct. Alp. Res. 30,
the Front Range, Colorado. U.S. Geol. Surv. Prof. Pap. 223. 349 – 361.
319 pp. Netoff, D.I., 1997. Soil clay mineralogy of Quaternary deposits in
Machette, M.N., 1975. The Quaternary geology of the Lafayette two Front Range – Piedmont transects, Colorado. PhD disserta-
Quadrangle, Colorado. Master’s thesis, Univ. Colorado, Boulder. tion, Univ. Colorado, Boulder. 169 pp.
105 pp. Reheis, M.C., 1980. Loess sources and loessial soil changes in a
Machette, M.N., 1977. Geologic map of the Lafayette Quadrangle, downwind transect, Boulder – Lafayette, Colorado. Mt. Geol.
Adams, Boulder, and Jefferson Counties, Colorado. U.S. Geol. 17, 7 – 12.
Surv. Geologic Quadrangle Map GQ-1392. Richmond, G.M., 1960. Glaciation of the east slope of Rocky
Machette, M.N., 1985. Calcic soils of the southwestern United Mountain National Park, Colorado. Geol. Soc. Amer. Bull. 71,
States. Geol. Soc. Amer. Spec. Pap. 203, 1 – 21. 1371 – 1382.
Machette, M.N., Birkeland, P.W., Markos, G., Guccione, M.J., 1976. Richmond, G.M., 1962. Quaternary Stratigraphy of the La Sal
Soil development in Quaternary deposits in the Golden-Boulder Mountains, Utah. U.S. Geol. Surv. Prof. Pap. 324. 135 pp.
portion of the Colorado Piedmont. Prof. Contrib. Colo. Sch. Schildgen, T.F., 2000. Fire and Ice: The Geomorphic History of
Mines (8), 217 – 259. Middle Boulder Creek as Determined by Isotopic Dating Tech-
Madole, R.F., 1969. Pinedale and Bull Lake glaciations in upper St. niques, Colorado Front Range. Honors thesis, Williams College,
Vrain drainage basin, Boulder County, Colorado. Arct. Alp. Res. Williamstown, MA. 103 pp.
1, 279 – 287. Schildgen, T.F., Dethier, D.P., 2000. Fire and ice: using isotopic
Madole, R.F., 1986. Lake Devlin and Pinedale glacial history, Front dating techniques to interpret the geomorphic history of Middle
Range, Colorado. Quat. Res. 25, 43 – 54. Boulder Creek, Colorado. Geol. Soc. Amer., Abstr. Programs 32
Madole, R.F., 1991. Colorado Piedmont section. In: Morrison, R.B. (7), A-18.
(Ed.), Quaternary Nonglacial Geology: Conterminous United Scott, G.R., Taylor, R.B., 1986. Map showing late Eocene erosion
States. Geol. Soc. Amer., The Geology of North America, surface, Oligocene – Miocene paleovalleys, and Tertiary depos-
vol. K-2, pp. 456 – 462. its in the Pueblo, Denver, and Greeley 1  2 quadrangles,
Madole, R.F., Shroba, R.R., 1979. Till sequence and soil develop- Colorado. U.S. Geol. Surv. Misc. Investigation Series Map
ment in the North St. Vrain drainage basin, east slope, Front I—1626.
Range, Colorado. In: Ethridge, F.G. (Ed.), Field Guide, North- Shroba, R.R., Birkeland, P.W., 1983. Trends in late-Quaternary soil
ern Front Range and Northwestern Denver Basin, Colorado. development in the Rocky Mountains and Sierra Nevada of the
Dept. Earth Resources, Colorado State University, Ft. Collins, western United States. In: Porter, S.C. (Ed.), Late Quaternary
pp. 123 – 178. Environments of the United States. The Late Pleistocene, vol. 1.
Madole, R.F., VanSistine, D.P., Michael, J.A., 1998. Pleistocene Univ. Minnesota Press, Minneapolis, pp. 145 – 156.
glaciation in the upper Platte River drainage basin, Colorado. Shroba, R.R., Carrara, P.E., 1996. Surficial geologic map of the
U.S. Geol. Surv. Geol. Investigations Series Map I—2644. Rocky Flats Environmental Technology Site and vicinity, Jef-
Mahaney, W.C., 1974. Soil stratigraphy and genesis of neoglacial ferson and Boulder Counties, Colorado. U.S. Geol. Surv. Misc.
deposits in the Arapaho and Henderson cirques, central Colo- Investigations Series Map I—2526.
rado Front Range. In: Mahaney, W.C. (Ed.), Quaternary Envi- Small, E.E., Anderson, R.S., Repka, J.L., Finkel, R., 1997. Erosion
ronments. Atkinson College/York University Monograph No. 5, rates of alpine bedrock summit surfaces deduced from in situ
Toronto, Canada, 197 – 240. 10Be and 26Al. Earth and Planet. Sci. Lett. 150, 413 – 425.
McMillan, M.E., 1990. Soil development on a subalpine Pinedale Soil Classification Working Group, 1998. The Canadian System of
moraine and erosion ten years after a burn, Rocky Mountain Soil Classification, 3rd ed. National Research Council Research
National Park, Colorado. Master’s thesis, Univ. of Colorado, Press, Ottawa Publ. 1646. 187 pp.
Boulder. 91 pp. Soil Survey Division Staff, 1993. Soil Survey Manual. U.S. Dept.
Muhs, D.R., Benedict, J.B., Evans, J., 1992. Sources of probable Agriculture Handbook, no. 18. 437 pp.
eolian sediments on late Quaternary alpine moraines, Colorado Soil Survey Staff, 1999. Soil Taxonomy, 2nd ed. U.S. Dept. Agri-
Front Range—evidence from trace element geochemistry. Amer. culture Handbook, no. 436. 869 pp.
Quaternary Assoc., 12th Biennial Meeting, Program and Ab- Steven, T.A., Evanoff, E., Yuhas, R.H., 1997. Middle and late
stracts, p. 73. Cenozoic tectonic and geomorphic development of the Front
Muhs, D.R., Aleinikoff, J.N., Stafford Jr., T.W., Kihl, R., Been, J., Range of Colorado. In: Bolyard, D.W., Sonnenberg, S.A.
Mahan, S.A., Cowherd, S., 1999. Late Quaternary loess in (Eds.), Geologic History of the Colorado Front Range. Rocky
344 P.W. Birkeland et al. / Geomorphology 55 (2003) 329–344

Mountain Assoc. Geologists Guidebook, Field Trip no. 7, U.S. Department Agriculture, Soil Conservation Service, 1975.
115 – 124. General Soil Map—Colorado, Portland, Oregon.
Thorne, C.E., Loewenherz, D.S., 1987. Alpine mass wasting in the U.S. Department Agriculture, U.S. Forest Service, 2001. Soil Sur-
Indian Peaks area, Front Range, Colorado. In: Graf, W.L. (Ed.), vey of Roosevelt – Arapahoe National Forest, Colorado, Interim
Geomorphic Systems of North America. Geol. Soc. Amer., Cen- Version [unpublished]. Forest Supervisor’s Office, U.S. Forest
tennial Spec., vol. 2, pp. 238 – 247. Service, Ft. Collins, CO.
Tonkin, P.J., Basher, L.R., 1990. Soil-stratigraphic techniques in the Veblen, T.T., Lorenz, D.C., 1991. The Colorado Front Range, A
study of soil and landform evolution across the Southern Alps, Century of Ecological Change. Univ. Utah Press, Salt Lake City.
New Zealand. Geomorphology 3, 547 – 575. 186 pp.
Trimble, D.E., 1980. Cenozoic tectonic history of the Great Plains Wahlstrom, E.E., 1948. Pre-Fountain and recent weathering on
contrasted with that of the Southern Rocky Mountains: a syn- Flagstaff Mountain near Boulder, Colorado. Geol. Soc. Amer.
thesis. Mt. Geol. 17, 59 – 69. Bull. 59, 1173 – 1189.
Tweto, O., 1979. Geologic map of Colorado. U.S. Geol. Surv.
 Geomorphology 55 (2003) 345 – 361
www.elsevier.com/locate/geomorph

Remote sensing and geomorphometry for studying relief

production in high mountains
Michael P. Bishop a,*, John F. Shroder Jr. a, Jeffrey D. Colby b
a
Department of Geography and Geology, University of Nebraska at Omaha, Omaha, NE 68182, USA
b
Department of Geography, East Carolina University, Greenville, NC 27858, USA
Received 20 October 2001; received in revised form 28 May 2002; accepted 10 March 2003

Abstract

Mountain topography is the result of highly scale-dependent interactions involving climatic, tectonic, and surface processes.
No complete understanding of the geodynamics of mountain building and topographic evolution yet exists, although numerous
conceptual and physical models indicate that surficial erosion plays a significant role. Mapping and assessing landforms and
erosion in mountain environments is essential in order to understand landscape denudation and complex feedback mechanisms.
This requires the development and evaluation of new approaches in remote sensing and geomorphometry. The research herein
evaluates the problem of topographic normalization of satellite imagery and demonstrates the use of terrain analysis using a
digital elevation model (DEM) to evaluate the relief structure of the landscape in the western Himalaya. We specifically
evaluated the Cosine-correction and Minnaert-correction methods to reduce spectral variation in imagery caused by the
topography. Semivariogram analyses of the topography were used to examine the relationships between relief and surface
processes. Remote-sensing results indicate that the Minnaert-correction method can be used to reduce the ‘‘topographic effect’’
in satellite imagery for mapping, although extreme radiance values are the result of not accounting for the diffuse-skylight and
adjacent-terrain irradiance. Geomorphometry results indicate that river incision and glaciation can generate extreme relief,
although the greatest mesoscale relief is produced by glaciation at high altitudes. At intermediate altitudes, warm-based
glaciation was found to decrease relief. Our results indicate that glaciation can have a differential influence on the relief
structure of the landscape. Collectively, our results indicate that scale-dependent analysis of the topography is required to
address radiation transfer issues and the polygenetic nature of landscape denudation and relief production.
D 2003 Elsevier B.V. All rights reserved.

Keywords: Remote sensing; Geomorphometry; Relief production

1. Introduction man, 1992; Avouac and Burov, 1996). This hypothesis

suggests that enhanced glaciation should result in
Globally significant interactions between climate, increased erosion, uplift and relief production. Numer-
surface processes, and tectonics have recently been ous studies have examined mountain topography to
proposed to explain climate change and mountain buil- understand the influence of surface processes on relief
ding (Molnar and England, 1990; Raymo and Ruddi- production (e.g., Burbank et al., 1996; Brozovik et al.,
1997; Bishop and Shroder, 2000; Bishop et al., 2002;
* Corresponding author. Brocklehurst, 2002; Finlayson et al., 2002), although
E-mail address: [email protected] (M.P. Bishop). the influence of glaciation on the relief structure of the

0169-555X/03/$ - see front matter D 2003 Elsevier B.V. All rights reserved.
doi:10.1016/S0169-555X(03)00149-1
346 M.P. Bishop et al. / Geomorphology 55 (2003) 345–361

landscape is complex and not always accounted for in metric calibration procedures and analysis of the top-
conceptual models and numerical simulations. ography at Nanga Parbat, in northern Pakistan, to
Understanding the differential influence of glacia- better understand radiation – transfer processes and
tion on relief production will require research to relief production. Specifically, our objectives were to
address the issues of polygenetic topographic evolu- (i) compare the Cosine- and Minnaert-correction
tion, identifying and mapping the operational scale of models for radiometric calibration of satellite imagery;
surface processes, identifying and mapping the opera- and (ii) examine the mesocale relief structure of the
tional scale at which landscape denudation affects the landscape on the north side of Nanga Parbat.
macroscale tectonic influx of mass (Bishop and
Shroder, 2000; Bishop et al., 2002), and developing
and evaluating new parameterizations for erosion 2. Study area
modeling. Some of these issues can be investigated
using satellite imagery and digital elevation models The Nanga Parbat summit (8125 m) is part of a
(DEMs), as valuable spatial and temporal information knife-edged ridge that extends in a NE – SW direction
can be extracted. Producing reliable geomorphologi- (Fig. 1). Structurally, the Nanga Parbat massif occurs
cal information from satellite data, however, is diffi- within transpressive shear zones that are spatially and
cult because numerous environmental factors such as temporally linked with youthful granite plutonism.
the atmosphere, topography, and land cover control
the irradiant and radiant flux. Other factors such as
solar and sensor geometry must also be taken into
consideration, such that the magnitude of the surface-
radiant flux varies in all directions (anisotropic reflec-
tance). Consequently, satellite imagery must be radio-
metrically calibrated to account for these factors so
that variations in image radiance are representative of
biophysical variations of the landscape. An opera-
tional solution to this problem has yet to materialize
(Bishop and Colby, 2002).
Analysis of the topography, however, is increas-
ingly being used to study surface processes and
erosion due to the availability of DEMs from remote
sensing. Geoscientists routinely examine the first-
order statistical nature of the topography to study relief
production. The nature of the problem, however,
requires that ‘‘issues of scale’’ be accounted for when
characterizing geomorphometric properties and con-
ducting spatial analysis (Shroder and Bishop, 1998;
Bishop and Shroder, 2000). Furthermore, an important
goal is to accurately characterize the organization of
the topography in relation to erosion and internal
forcings so that statistical metrics/indices or model
parameters can be calibrated with measured erosion
rates (Howard, 1996). Consequently, new spatial anal-
ysis procedures and models need to be developed and
evaluated (Small, 1999; Bishop and Shroder, 2000). Fig. 1. SPOT 3 ortho-rectified NIR image of Nanga Parbat, in
Northern Pakistan. The Indus and Astor Rivers are located in the
The purpose of this research was to examine NW and NE portions of the image respectively. The knife-edged
various issues associated with information extraction Nanga Parbat ridge is oriented in a NE – SW direction and is visible
from satellite imagery and DEMs. We evaluate radio- in the lower part of the image.
 M.P. Bishop et al. / Geomorphology 55 (2003) 345–361 347

The NW Raikot and SE Stak fault shear zones define a River seems to have initiated uplift of the massif. This
crustal-scale, antiformal pop-up structure that resem- newly recognized combination of fluvial superposition
bles an asymmetrically upward moving, cork-like plug followed by focused tectonism leads to a new under-
that began f 12– 10 Ma (Schneider et al., 1999). standing of supercedance in which capture of a super-
Shroder and Bishop (2000) noted that the initial posed stream caused antecedant uplift across the
denudation cascade at Nanga Parbat probably began course of the master Indus and Astor rivers. The rising
f 11 Ma when the Indus River was captured from its supercedant massif then caused consequent radial
west-flowing position on Kohistan rocks overlying drainages from its flanks, one of which seems to have
Nanga Parbat gneisses and diverted abruptly south eventually captured the upper Astor – Rupal and
across what is now the north and west sides of the resulted in annular encirclement of half the range.
massif. This time is also approximately coincident with Eventually, the mountain rose high enough and the
possible structural disturbance of the drainage and the climate grew cold enough so that glaciers developed
onset of the first monsoonal precipitation in the region, and glacial – drainage diversions would have been
which suggest further causation for the capture. Sub- common. It is unlikely, however, that records exist of
sequently from the late Miocene to early –middle Pleis- all these geomorphic events since the late Miocene that
tocene, the accelerated erosion of the Nanga Parbat would enable detailed reconstructions of topographic
region has left little record of the processes involved. evolution of the massif. Instead, interpretation of uplift
The development of glacial and fluvial drainages progress and landform development must be based
that dominate the Nanga Parbat massif is the result of upon examination of the existing landscape and geo-
complex geologic evolution reflecting a sequence of morphometric properties that reflect the polygenetic
temporally linked drainage types (Table 1). Beginning nature of topographic evolution.
with superposition from the Kohistan cover sequence, Bishop et al. (1998b) and Bishop and Shroder
the newly captured and rapidly downcutting Indus (2000) used satellite imagery to examine erosion and
deposition features at Nanga Parbat to identify geo-
morphic events, and to better understand the role of
Table 1 surface processes in the denudation cascade. They
Sequence of geomorphic events over the past 12 Ma leading to the
existing drainage network on Nanga Parbat
initially found that reliable information produced from
satellite imagery is dependent upon reducing spectral
1.. Paleo-Indus River flowing WNW on cover volcanic rocks of
Kohistan sequence that overlie gneisses of Indian Plate rocks.
variation caused by the topography and land cover.
. NW-flowing, paleo-Astor River is tributary to the Indus well Furthermore, they indicated that more advanced forms
NW of its present junction and also flowing on cover rocks. of analysis and modeling are required to produce useful
. Ancestral SW-flowing Nagri River eroding headward into information.
Kohistan cover rocks to eventually capture upper Indus River. In general, investigators have attempted to correct
2.. Capture of Indus River and superposition of paleo-Indus and
paleo-Astor down to expose Indian plate rocks.
for the influence of topography on upward radiance by
. Incipient uplift zone established between Nanga Parbat and accounting for the nature of surface reflectance (Lam-
Haramosh because of erosional unloading. bertian or non-Lambertian) and the local topographic
3.. Superposition + antecedance = supercedant drainage established. conditions (Colby, 1991; Ekstrand, 1996; Colby and
. Accelerating uplift of Nanga Parbat massif.
Keating, 1998). Semi-empirical approaches include
. Development of consequent radial drainage on massif.
4.. Subsequent, radial drainage development. Cosine-correction (Smith et al., 1980), Minnaert-cor-
. The Rupal – upper Astor River system erodes further as a rection (Colby, 1991), the c-correction model (Teillet
subsequent annular stream with apparent conformity to et al., 1982), and other empirical corrections that make
metamorphic isograds, perhaps in response to updoming. use of a DEM to account for the local illumination
. Some brittle fault displacement of Indus River along Raikot
conditions for each pixel. These models have been
Fault.
. Waxing and waning Pleistocene glaciations and possible
widely applied given their relative simplicity and ease
diversions cause further minor development of drainages on of implementation. They have not been appropriately
Nanga Parbat. evaluated in rugged terrain such as the Himalayas.
5.. Holocene development of the modern drainage net. Research findings indicate that these approaches
. Continued Raikot Fault displacement of Indus River.
may work only for a given range of topographic
348 M.P. Bishop et al. / Geomorphology 55 (2003) 345–361

3. Methodology

3.1. Digital elevation model

A SPOT 3 panchromatic stereo-pair was acquired

on 27 October 1996 and 28 October 1996. A DEM was
generated by SPOT Data using the imagery and the

Fig. 2. SPOT 3 ortho-rectified NIR image depicting sample locations

for scatterplots and two-dimensional semivariogram analysis. Each
sample location represents a 100
100 pixel area of relatively
homogeneous land cover.

conditions (Smith et al., 1980; Richter, 1997). The

Cosine- and Minnaert-correction procedures are fre-
quently used and Bishop et al. (1998a) found that the
Minneart-correction procedure has the potential to be
used to study the problem, provided that issues asso-
ciated with computation and topography are addressed.
Topographic analysis of the Nanga Parbat massif
has been conducted to characterize the hypsometry,
slope distributions, and first-order relief. Burbank et al.
(1996) concluded that landsliding was primarily re-
sponsible for the relief production in the western
Himalaya, while Brozovik et al. (1997) concluded that
glaciation limits the production of relief there. Subse-
quent analyses by Bishop and Shroder (2000) and
Bishop et al. (2002) indicated that glaciation has had
a significant influence on the landscape and that Fig. 3. SPOT 3 ortho-rectified panchromatic image of the digital
glaciation plays a significant role in relief production elevation model subarea that was used for analysis of the
in the Nanga Parbat Himalaya. topography.
 M.P. Bishop et al. / Geomorphology 55 (2003) 345–361 349

stereo-autocorrelation method. Ground control points

were obtained from a 1:50,000 topographic map. This
approach requires the accurate identification of a pixel
in each stereo-pair image that represents the same
ground location. The stereo-autocorrelation method
requires adequate spectral variation to enable correla-
tion and registration of pixels. Sensor radiometric
resolution and highly reflective features such as snow
and ice can cause spectral saturation in images, and
cast shadows can also cause a significant reduction in
spectral variation thereby causing correlation failure.
Topographic map coverage was used to construct the
topography at the summit where spectral saturation did
occur. This area was extremely small and was not
included in our topographic analysis for assessing
relief. The ground control points were used to ensure
quality control. The resulting DEM exhibited a meas-
urement resolution of 20 m with a vertical accuracy of
F 8– 12 m.

3.2. Image preprocessing

Satellite imagery in mountainous terrain must be

geometrically and radiometrically calibrated to account
for geometric error, sensor system effects, and relief
displacement (Chavez, 1996; Bishop et al., 1998a). Fig. 4. Normalized SPOT 3 NIR image using the Cosine-correction
SPOT 3 multispectral data were ortho-rectified using method. Extreme radiance values are found throughout the image.
the DEM. SPOT Data produced an ortho-rectified
panchromatic image for the Nanga Parbat area using
metric calibration was achieved using the following
map control [1:50,000 German topographic map by
equation:
Finsterwalder, 1936]. The multispectral imagery were
rectified using control points that could easily be iden- DNE
tified in the multispectral imagery and ortho-rectified LE ¼ ; ð1Þ
aE
panchromatic imagery. The total root mean squared
error was 5.0 m, based on the use of 20 control points. where LE is the radiance in W m 2 sr 1 Am 1, DN
The satellite imagery were radiometrically cali- represents the digital number (0– 255), and aE repre-
brated using SPOT’s calibration coefficients. Radio- sents SPOT’s absolute calibration coefficient. Atmos-

Table 2
Summary statistics for SPOT 3 NIR image and normalized NIR imagery
Image Min Max Median Mean S.D. Range
SPOT 3 NIR 17.73 300.24 85393 120.95 62.51 282.51
Cosine correction 19.70 2.48
107 118.82 401.18 55 958.03 2.48
107
Minnaert correctiona 18.69 50 879 90.42 128.64 144.66 50 860.31
Minnaert correctionb 17.95 26 399 78.22 123.19 86.62 26 381.05
a
Globally computed k.
b
Land-cover computed ks.
350 M.P. Bishop et al. / Geomorphology 55 (2003) 345–361

pheric correction procedures were not applied to slope aspect angle of the terrain. The Cosine-correction
correct for the additive path-radiance component, method is computed as
given the high altitude of the study area.
LE
LnE ¼ ð3Þ
3.3. Anisotropic-reflectance correction cosi

where LEn represents normalized radiance resulting

The extreme relief and complex topography at
from the correction procedure, and LE represents the
Nanga Parbat produces high and low, spatial – fre-
radiance values from Eq. (1). This method has been
quency, spectral variation in satellite images. We eval-
found to be useful, given high solar elevation angles and
uated the Cosine-correction method, which assumes
low slope angles (Smith et al., 1980; Ekstrand, 1996).
that surface reflectance is isotropic (reflectance is equal
We also evaluated the Minnaert-correction proce-
in all directions). Research indicates that this assump-
dure using two different implementations. This pro-
tion is not valid in mountain terrain as the topography
cedure was implemented using the general backwards
and land cover cause anisotropic reflectance (magni-
radiance-correction procedure described by Colby
tude of radiance varies with direction) (Smith et al.,
(1991):
1980; Hugli and Frei, 1983). This procedure accounts
for the local topographic conditions that influence the
LnE ¼ ðLE coseÞ=ðcosk i cosk eÞ ð4Þ
direct solar irradiance. It is based upon the use of the
cosine of the incidence angle (i), which is proportional where e is the excitant angle (e = bt) and k is the
to the direct solar irradiance such that Minnaert constant that ranges from 0.0 – 1.0. The
coefficient is not a true constant, but attempts to
cosi ¼ coshs cosbt þ sinhs sinbt cosð/t  /s Þ ð2Þ characterize the departure of reflectance from that of
a Lambertian surface (i.e., isotropic reflectance). A k
where hs is the solar zenith angle, bt is the slope angle value of 1 indicates isotropic reflectance. A lower
of the terrain, /s is the solar azimuth angle, and /t is the coefficient represents departure from isotropic reflec-

Fig. 5. Scatterplot of normalized (Cosine-correction method) NIR radiance and cos i for sample location 2 in Fig. 2. The method ‘‘overcorrects’’
and produces extreme radiance values where steep slopes angles reduce the direct solar irradiance.
 M.P. Bishop et al. / Geomorphology 55 (2003) 345–361 351

tance and represents the degree of anistrophy. It is also gravitative energy gradients that are responsible for
important to note that k is wavelength-dependent. The much geomorphic work by the different processes. We
Minnaert constant was calculated using least-squares differentiate five geomorphic-process zones on the
regression on the variables x and y, where x = log(cos i northern slopes of Nanga Parbat that reflect temporal
cos e) and y = log(LEcos e). The slope of the regression and spatial differences in lithology, structure, and
equation represents k. process: (i) a lowermost zone of fault scarps and river
One way to evaluate the Minnaert-correction pro- incision; (ii) a zone of past glacial erosion and deposi-
cedure is to compute a global k value for an image. tion; (iii) a lower zone of superimposed past and
This approach assumes that the anisotropic nature of modern-day warm-based ice; (iv) an upper zone of
reflectance is homogeneous over the study area. This superimposed past and modern-day, warm-based ice;
assumption is known to be invalid as topographic and and (v) a high-altitude zone of glacier erosion (equili-
land-cover variation will cause k to vary spatially, such brium line altitude (ELA)/paleoELA-based), cold-
that a global value may not produce spatially consis- based ice.
tent normalization results. Numerous investigators, We systematically sampled these zones on the north
however, have indicated that this approach can reduce side of Nanga Parbat (Fig. 3). We specifically limited
the ‘‘topographic effect’’ in satellite imagery. the extent of the DEM subarea in the east – west
It is reasonable to assume that land cover exhibits a direction so that statistical characterization along a
strong influence on anisotropic reflectance such that k north –south profile might show variations related to
should be computed for land-cover classes. To test this,
an image segmentation approach was investigated.
The fundamental land-cover structure within the scene
was produced using NIR image thresholding and the
ISODATA clustering algorithm on a NIR/RED ratio
image. This resulted in a land-cover map (three clas-
ses) for Nanga Parbat.
Snow, vegetation, and nonvegetation features were
easily mapped as these land-cover classes are known to
exhibit different anisotropic-reflectance characteristics
(Hugli and Frei, 1983). The purpose of the classifica-
tion is not to accurately map these features, but to
segment the image into regions so that three k values
can be computed that characterize the influence of land
cover on anisotropic reflectance.
We tested these procedures on the near-infrared
image because the influence of topography on the
direct irradiance is evident (Fig. 1). We also performed
two-dimensional semivariogram analysis on several
100
100 pixel subsections to compare the spectral
variance of original and normalized imagery (Fig. 2).
The locations for testing were selected on the basis of
homogeneous land cover. Effective topographic nor-
malization should result in decreased spectral variance
(lower semivariance values).

3.4. Topographic analysis

Fig. 6. Normalized SPOT 3 NIR image using the Minnaert-
correction method (globally computed k). Spectral variance visually
The 7 km of vertical relief of the Nanga Parbat appears to be reduced, however, extreme radiance values are found
massif produces exceptionally strong orographic and at the base of steep slopes.
352 M.P. Bishop et al. / Geomorphology 55 (2003) 345–361

surface processes and erosion with altitude. In order to We recognize that this approach characterizes the
conduct one- and two-dimensional analyses of the relief in an east – west direction, and that two-dimen-
relief along this profile, each zone was of equal size sional analysis is required to accurately characterize
(f 10
4 km). the topography. Results from summary statistics
Relief statistics were calculated for east – west clearly depicted erosion– process zones. Consequent-
swaths across the DEM subarea over a 10-km distance. ly, we conducted two-dimensional semivariogram
This equates to calculating statistics for altitude for analysis for each zone, where every grid cell was
every row of our DEM subarea, using every column of compared to every other grid cell within each zone.
that row. This analysis enables the production of an Within an nx
ny grid there will be N point pairs,
aggregate north –south profile where we examine how where N =n x2 (n y2  1)/2. The variance component
the relief varies as a function of surface processes. is Dz=(z1  z2)2,and the horizontal distance is Dx=
Similarly, we examined the scale-dependent char- [(x1  x2)2+( y1  y2)2]0.5. The variance was summed
acteristics of the topography in relation to surface over a binned horizontal distance interval (20 m) and
processes. We conducted one-dimensional semivario- divided by the number of observations within each bin
gram analysis of each row, where the average semi- to calculate the average semivariance (S̄2).
variance (S̄2) is computed as follows: Our spatial analysis on the north side of Nanga
Parbat is not spatially constrained to the Raikot basin.
1 X m
S̄ 2 ðhÞ ¼ m ðzi  ziþh Þ2 ð5Þ Basin analysis is a very popular approach that can be
2 i¼1
used to assess relief characteristics. From a geophys-
where z is the altitude from a single grid cell, h ical perspective, however, we know that the influence
represents the sampling lag distance, and m is the of denudational unloading on tectonic uplift at Nanga
number of sampled pairs per lag. This analysis resulted Parbat is not spatially constrained at the basin scale
in the production of a semivariogram for each row of and that fluctuations in the influx of mass are depend-
the DEM subarea. By selecting a particular lag dis- ent upon denudational unloading at the regional
tance, a north – south profile of the semivariance can be scale, which may be spatially constrained by the
examined. Raikot and Stak faults (Zeitler et al., 2001a,b). There-

Fig. 7. Scatterplot of normalized NIR radiance (Minnaert-correction method and globally computed k) and cos i for sample location 3 in Fig. 2.
The method ‘‘overcorrects’’ and produces relatively high radiance values.
 M.P. Bishop et al. / Geomorphology 55 (2003) 345–361 353

fore, scale-dependent relief characteristics may be

used to assess surface processes and provide a frame-
work for estimating erosion rates (Bishop and Shroder,
2000).

4. Results

4.1. Anisotropic-reflectance correction

Statistical analysis of the original and normalized

imagery indicated that the correction methods did not
produce acceptable results for all pixels (Table 2).
Radiance values were frequently high and not repre-
sentative of the land-cover classes. This ‘‘overcorrec-
tion’’ (extreme radiance values) resulted in an increase
in spectral variance in the normalized imagery.
The Cosine-correction method did not effectively
normalize the SPOT NIR image (Fig. 4). Extreme
radiance values exist where steep slopes reduce the
magnitude of the direct irradiance, such that the
magnitude of cos i values do not accurately estimate
the magnitude of the irradiance at the surface. This is
demonstrated in Fig. 5, where we sampled a homoge-
neous land-cover class (snow) and plotted the relation-
ship between normalized radiance and cos i. The
nonlinear relationship is the result of steep slopes that
receive additional irradiance from the diffuse-skylight
and adjacent-terrain irradiance components. Fig. 8. Semivariograms from sample location 1 in Fig. 2 depicting
the spectral variance in the original NIR image (NIR), Minnaert-
Results from the Minnaert-correction method (glo- corrected NIR image using a globally computed k (MC1k), and
bally computed k) yielded a Minnaert constant and r2 Minnaert-corrected NIR image using land cover-computed k’s
value of 0.5234 and 0.123, respectively. Statistical (MC3k). A greater reduction in the ‘‘topographic effect’’ results
analysis and visual interpretation indicated that the from using multiple k values.
influence of topography was reduced, although high
radiance values were found at the base of steep slopes
(Fig. 6). Similar to Cosine-correction results, we found ple Minnaert constants can be computed with high r2
a nonlinear relationship between normalized radiance values.
values and cos i (Fig. 7). Semivariogram analysis Results from the Minnaert-correction method
revealed that the local spectral variance was dramati- (land-cover computed k’s) indicated that it is possible
cally reduced over snow-covered portions of the land- to produce multiple k values that reasonably character-
scape (Fig. 8), even though the k value does not ize the influence of land cover and topography on
uniquely characterize the influence of snow/firn and reflectance (Table 3). The normalized image is visu-
topography on surface radiance (low r2). These results ally similar to Fig. 6, however, this implementation
indicate that a globally derived Minnaert constant does reduced spectral variance better than the other meth-
not characterize the relationship between topography, ods over the entire image (Table 2) and for homoge-
land cover and surface radiance in complex scenes neous land-cover classes (Fig. 8). The decrease in
exhibiting multiple land-cover classes. It does, how- semivariance represents a decrease in spectral varia-
ever, demonstrate the potential of the method if multi- tion caused by the topography.
354 M.P. Bishop et al. / Geomorphology 55 (2003) 345–361

Table 3 incision. The relief in this area systematically increases

Results of land-cover stratified regression analysis to a maximum of f 2600 m at the southernmost
Class k r2 n boundary of the fault zone. The maximum relief then
Nonvegetation 0.4052 0.4081 1 266 086 decreases, as the influence of past glaciation alters the
Vegetation 0.4437 0.5182 418 362 topography.
Snow/firn 0.4840 0.5757 958 465
This rapid decrease in relief is associated with a
monsoon-driven glacial event that occurred approxi-
4.2. Topographic relief mately 65 K year 1 (Phillips et al., 2000). At this time,
glaciation covered most of the landscape. Farther along
Swath profile analysis revealed the extreme relief the profile at intermediate altitudes, the collective in-
on the northern side of Nanga Parbat (Fig. 9). In fluence of glaciation accounts for the relatively low
general, the altitude systematically increases towards maximum relief conditions (f 1800 m). Farther south
the south, although three distinct altitude trends can be at higher altitudes, the maximum relief systematically
identified based upon the average rate of change in increases, indicating the erosive nature of temperate al-
altitude. These trends represent (i) the lower altitude pine glaciers. At a profile distance of f 17 and 24 km,
Raikot fault zone, which exhibits steep slopes and V- deep glacier valley erosion is associated with f 3.1
shaped valleys transverse to the fault; (ii) intermediate and 3.4 km of relief, respectively. The decrease in relief
altitudes where past and modern, warm-based glacier portrayed in Fig. 10 at a profile distance of f 20– 23
erosion and deposition has significantly altered the km is the result of the orientation of transects with
landscape; and (iii) an upper region where cold-based respect to the orientation of the knife-edge ridge (i.e.,
and protective glaciation dominates. orientations of the transects and the ridge are nearly
The relief profile provides additional insight into equivalent, so that the relief decreases).
the nature of surface processes that erode the land- Examination of the variation in altitude along the
scape (Fig. 10). In the Raikot Fault zone (0 –5 km profile also characterizes the role of past and modern
profile range), rapid river incision produced deep glaciation on the landscape hypsometry (Fig. 11). An
gorges that exhibit strath terraces caused by rapid river increase in variability in altitude would be expected

Fig. 9. Swath profile of the altitude on the northern side of Nanga Parbat.
 M.P. Bishop et al. / Geomorphology 55 (2003) 345–361 355

Fig. 10. Swath profile of the relief on the northern side of Nanga Parbat. The decrease in relief at f 20 – 23 km is the result of the sampling
design, such that the metric does not effectively characterize relief for comparison purposes.

where active river incision and recent uplift produce profile, and the landscape begins to record the influence
steep slopes and extreme relief. As the current influ- of past glaciations, we see a sharp decrease in varia-
ence of these processes decrease to the south along the bility. The start of this decline coincides exactly where

Fig. 11. Swath profile of the variability of altitude on the northern side of Nanga Parbat. The decrease in the variability at f 20 – 23 km is the
result of the sampling design, such that the metric does not effectively characterize variability for comparison purposes.
356 M.P. Bishop et al. / Geomorphology 55 (2003) 345–361

Fig. 12. Swath profile of the semivariance at two lag distances on the northern side of Nanga Parbat. The decrease in semivariance at f 23 km
is the result of the sampling design, such that the metric does not effectively characterize relief for comparison purposes.

Fig. 13. Semivariograms for different surface process zones on the northern side of Nanga Parbat. Each semivariogram was computed by
comparing the altitude of every pixel in a zone. Zone 1 is a lowermost zone of fault scarps and river incision. Zone 2 represents a zone of past
glacial erosion and deposition. Zone 3 is the lowermost zone of superimposed past and modern-day warm-based ice. Zone 4 is the uppermost
zone of superimposed past and modern-day warm-based ice. Zone 5 represents the high altitude zone of glacier erosion (ELA/paleoELA-based),
cold-based ice.
 M.P. Bishop et al. / Geomorphology 55 (2003) 345–361 357

the Bezar Gali glacial erosion surface occurs. Further- V 1 km, the maximum relief on the landscape is
more, as modern glacial processes leave an imprint on associated with active river incision in the Raikot
the topography, the variability along the profile fault zone and at the highest altitudes where tectonic
increases, such that a local variability maximum is uplift and multiple glaciations are responsible for
associated with the location of modern-day alpine ffiffiffiffiffi intermediate altitudes, a
extreme relief (Fig. 12).pAt
2
glaciers. These results initially suggest that glaciation systematic decrease in S values (relief) is caused
plays an important role in mesocale relief production. by glacier erosion and redistribution of sediment. This
It is problematic to make conclusions based upon result suggests that as modern glaciation dominates
the use of first-order statistics that do not account for the landscape at intermediate altitudes, glacial pro-
the scale-dependent nature of relief. One-dimensional cesses operating at this scale do not produce extreme
semivariogram analysis revealed that at lag distances relief. We might expect this result as river incision

Fig. 14. SPOT 3 false-color composite image draped over the digital elevation model. The 3-D perspective view is from the north looking south
at Nanga Parbat. The Raikot basin is in the center of the scene. Vertical exaggeration is 2
to visually enhance mesoscale relief and display the
leveling of the topography by glaciation at intermediate altitudes.
358 M.P. Bishop et al. / Geomorphology 55 (2003) 345–361

operates at a very different scale compared to glaci- trine deposits, catastrophic floods) and the topography.
ation on the north side of Nanga Parbat. Warm-based Quantitative information about surficial materials and
glaciers are known to be an effective erosion agent; landforms is difficult to obtain due to complex radia-
therefore, if they are capable of producing extreme tion transfer processes involving the atmosphere and
relief, we must examine altitude differences at greater topography.
distances that represent the operational scale of glacier Our radiometric calibration results indicate that it is
erosion. possible to significantly reduce the spectral variation in
Results of semivariogram analysis for lag distances imagery using the Minnaert-correction method, al-
z 2.5 km indicated that glaciation can produce com- though numerous issues remain. Our simplistic proce-
parable, if not greater mesoscale relief than processes dure of image-segmentation to generate accurate k
of river incision and slope-failure (Fig. 12; 3.6-km lag values is problematic because it initially requires
pffiffiffiffiffi
distance is most representative of extreme relief). land-cover information. Colby (1991) indicated that
2
Periodic fluctuations in S values at 3.6 km lag locally computed k’s may have value in reducing the
distance reflect the overprinting of glacial events along ‘‘topographic effect’’ by accounting for local land-
the profile caused by glaciation at intermediatepand ffiffiffiffiffi cover conditions.
2
high altitudes. The localized maximum peaks of S Our results also indicate that the calibration prob-
values at the profile distances of f 14 and 17 km (Fig. lem is not limited to the computation of accurate k
12) are approximately associated with the ELA zones values. Extreme radiance values are the result of
of the Buldar and Raikot glaciers, respectively. The inadequate modeling of the irradiance as cos i does
maximum relief, at a profile distance of 24 km, is the not account for cast shadows and the diffuse-skylight
result of glacier erosion at lower altitudes closer to the and adjacent-terrain irradiance. These radiation trans-
ELA zones on both sides of the main Nanga Parbat fer processes are highly scale-dependent and hemi-
ridge. If we take into consideration the depth of glacier spherical topographic analysis is required to account
ice in the ELA zones, the relief is greater than what our for their varying influence over the landscape. Con-
analysis of topography presents. sequently, this semi-empirical radiation transfer model
We compare these results to two-dimensional anal- requires additional topographic parameters to account
ysis before making conclusions. In the tectonically for complex atmosphere –topography, radiation inter-
active Raikot Fault zone, surface processes are actively actions. The development of improved radiation –
adjusting to tectonic forcing (Bishop et al., 2002), transfer models will be valuable in remote sensing
producing a considerable amount of relief in zone pffiffiffiffiffi
1 and geomorphology studies, and will enable the
2
(Fig. 13). At intermediate altitudes (zones 2 –4), S investigation of the spatial distribution of climate
pffiffiffiffiffi at all lag distances. Zone 5
values generally decrease forcing on surface processes (e.g., snowmelt, runoff,
2
exhibits the greatest S values at all lag distances. ablation) and erosion.
These results suggest that the product of glaciation at Nanga Parbat has undergone extreme environmen-
intermediate altitudes is one of decreasing relief and an tal change over a very short period of geological time
overall leveling of the landscape. This can be clearly (Shroder and Bishop, 2000; Bishop et al., 2002).
seen in the middle of Fig. 14. Bishop et al. (2002) have shown that glacial, fluvial
and slope processes all play important roles at different
times, and that they can do so sequentially on the same
5. Discussion portion of the landscape. Investigators have attempted
to characterize the individual roles of river incision and
Geoscientists have predominantly used qualitative glaciation in relief production (Burbank et al., 1996;
methods to extract geomorphological information Brozovik et al., 1997; Whipple and Tucker, 1999;
from satellite imagery. Scientific visualization techni- Brocklehurst, 2002; Finlayson et al., 2002). These
ques can be used to examine two- and three-dimen- studies and others have provided us with a better
sional perspectives, thus providing important spatial understanding of river incision, glaciation, and modern
information about geomorphic features and events process rates. The influence of glaciation on the un-
(e.g., high-altitude erosion surfaces, glaciers, lacus- loading history of the massif, however, is still rela-
 M.P. Bishop et al. / Geomorphology 55 (2003) 345–361 359

tively unknown. We know that the major valleys, morphometry research that addresses the modeling and
including the Indus, have undergone extensive and analysis of the hierarchical organization of topography
periodic glaciation (Shroder et al., 1989). Similarly, may provide new insights.
systematic and catastrophic geomorphic events are To many, it may not appear that radiation transfer
also responsible for the denudational unloading. and geomorphometry are related research topics for
Therefore, modern-day river or glacier incision rates understanding relief production. Both, however, are
alone cannot be applied over millions of years to important for studying surface processes and erosion.
characterize denudational unloading at Nanga Parbat. For example, complex radiation transfer processes that
The influence of glacial erosion on the relief struc- include atmosphere – topography interactions govern
ture of the landscape is an important question. Our the radiation budget. Radiative forcing directly and
results for intermediate altitudes concur with Brozovik indirectly affects glacial ablation, meltwater produc-
et al. (1997), indicating that glaciation can limit relief tion, the generation of fluvial and glacial landforms,
up to some altitude near the equilibrium line altitude and ultimately denudation and relief production. If
(ELA). Our results also indicate that glaciation can erosion models are to accurately simulate topographic
produce greater mesoscale relief than river incision, as evolution, glacier mass balance and glacial erosion
glaciation in concert with uplift is primarily respon- must be taken into consideration. Currently, the mul-
sible for relief production at high altitudes. Given the ti-scale topographic effects on glacier mass balance
temporal and spatial fluctuations of surface processes are not always considered. Topographic, solar – radia-
in topographic evolution, it is difficult to precisely tion, transfer models can be used to study radiative
determine the degree to which river incision and alpine forcing on surface processes. Similarly, solving the
glaciation erode the landscape and produce relief. anisotropic-reflectance correction problem will enable
At intermediate to higher altitude, deep valley gla- object-oriented landform analysis, such that topo-
cier erosion and headwall erosion can effectively graphic information from DEMs can be integrated
remove rock and sediment, while protective cold-based with image information to examine the spatial and
ice on the highest peaks permits ridge and peak contextual relationships between landforms. Such
formation. Consequently, in high mountain environ- multidisciplinary research involving remote sensing
ments, glaciation directly and indirectly alters the relief science and geomorphology should provide new infor-
structure of the landscape differently with altitude. mation regarding surface processes, erosion, and relief
These differences can be attributed to the degree of production.
temporal overprinting and spatial overlap of glacial
events, such that at intermediate altitudes, the degree of
spatial overlap is less than at high altitudes. More 6. Conclusions
research into providing a more detailed explanation
of this differential influence of glaciation on topogra- Remote sensing and geomorphometry provide new
phy is required. opportunities to study mountain environments and
A challenge will be to develop better approaches relief production. Satellite imagery must be radiomet-
that can be used to relate the properties of topography rically calibrated to account for atmospheric and topo-
to the magnitude of erosion (Howard, 1996; Small, graphic effects. Our results indicate that the ‘‘topo-
1999). Currently, there is no widely acceptable method graphic effect’’ in satellite imagery can be reduced
of such estimation, and scientists would like to use using the Minnaert-correction method, although ex-
DEMs to estimate erosion rates provided that geo- treme radiance values are the result of the model’s
morphometric properties can be calibrated with meas- inability to account for the diffuse-skylight and adja-
ured rates. Our results indicate that the topography cent-terrain irradiance. Additional topographic analy-
does contain information/patterns that reflect changes sis and parameterizations that characterize complex
in erosion processes with altitude. We speculate that atmosphere – topography radiation interactions are
calibration will be difficult to achieve without charac- required. Such new parameterizations are needed to
terizing variations in the topography that represent the improve information extraction from satellite imagery,
spatial constraints for surface processes. More geo- and in geomorphological studies to investigate micro-
360 M.P. Bishop et al. / Geomorphology 55 (2003) 345–361

climate, surface runoff, and glacial processes which Brocklehurst, S.H., 2002. Glacial erosion and relief production in
influence erosion and relief production. the Eastern Sierra Nevada, California. Geomorphology 42,
1 – 24.
Our topographic analysis revealed that the relief Brozovik, N., Burbank, D.W., Meigs, A.J., 1997. Climatic limits on
structure at Nanga Parbat is spatially variable and landscape development in the northwestern Himalaya. Science
correlated to geomorphic events and dominant surface 276, 571 – 574.
processes. River incision and glaciation are both re- Burbank, D., Leland, J., Fielding, E., Anderson, R.S., Brozovik, N.,
Reid, M.R., Duncan, C., 1996. Bedrock incision, rock uplift and
sponsible for relief production at Nanga Parbat, al-
threshold hillslopes in the northwestern Himalaya. Nature 379,
though our results show that glaciation generates the 505 – 510.
greatest mesoscale relief at high altitudes. At inter- Chavez Jr., P.S., 1996. Image-based atmospheric corrections—re-
mediate altitudes, warm-based glaciation was found to visited and improved. Photogrammetric Engineering and Re-
decrease relief. These results indicate that glaciation mote Sensing 62 (9), 1025 – 1036.
can have a differential influence on the relief structure Colby, J.D., 1991. Topographic normalization in rugged terrain.
Photogrammetric Engineering and Remote Sensing 57 (5),
of the landscape. More research is required to develop 531 – 537.
and formalize a spatio-temporal theory on mountain Colby, J.D., Keating, P.L., 1998. Land cover classification using
topographic organization so that geomorphometric Landsat TM imagery in the tropical highlands: the influence of
parameters can be used to estimate erosion rates. This anisotropic reflectance. International Journal of Remote Sensing
will require spatial and temporal information acquired 19 (8), 1459 – 1500.
Ekstrand, S., 1996. Landsat TM-based forest damage assessment:
from satellite imagery and DEMs. correction for topographic effects. Photogrammetric Engineer-
ing and Remote Sensing 62 (2), 151 – 161.
Finlayson, D.R., Montgomery, D.R., Hallet, B., 2002. Spatial co-
Acknowledgements incidence of rapid inferred erosion with young metamorphic
massifs in the Himalayas. Geology, 219 – 222.
This work was funded by the University Commit- Finsterwalder, R., 1936. Carte der Nanga Parbat gruppe (1:50,000).
tee on Research at the University of Nebraska at Deutsche Himalaya Expedition 1934.
Howard, A., 1996. The ephemeral mountains. Science 379,
Omaha and the National Science Foundation (grant
488 – 489.
no. EAR 9418839 and EPS-9720643). Hugli, H., Frei, W., 1983. Understanding anistropic reflectance in
mountainous terrain. Photogrammetric Engineering and Remote
Sensing 49 (5), 671 – 683.
References Molnar, P., England, P., 1990. Late Cenozoic uplift of mountain
ranges and global climate change: chicken or egg? Nature 346,
Avouac, J.P., Burov, E.B., 1996. Erosion as a driving mechanism of 29 – 34.
intracontinental mountain growth. Journal of Geophysical Re- Phillips, W.M., Sloan, V.F., Shroder Jr., J.F., Sharma, P., Clarke,
search 101, 747 – 769. M.L., Rendell, H.M., 2000. Asynchronous glaciation at Nanga
Bishop, M.P., Colby, J.D., 2002. Anisotropic reflectance correction Parbat, northwestern Himalaya Mountains, Pakistan. Geology
of SPOT-3 HRV imagery. International Journal of Remote Sens- 28, 431 – 434.
ing 23 (10), 219 – 222. Raymo, M.E., Ruddiman, W.F., 1992. Tectonic forcing of the late
Bishop, M.P., Shroder Jr., J.F., 2000. Remote sensing and geomor- Cenozoic climate. Nature 359, 117 – 122.
phometric assessment of topographic complexity and erosion Richter, R., 1997. Correction of atmosphere and topographic effects
dynamics in the Nanga Parbat massif. In: Khan, M.A., Treloar, for high spatial resolution satellite imagery. International Journal
P.J., Searle, M.P., Jan, M.Q. (Eds.), Tectonics of the Nanga of Remote Sensing 18 (5), 1099 – 1111.
Parbat Syntaxis and the Western Himalaya, pp. 181 – 200. Schneider, D.A., Edwards, M.A., Kidd, W.S.F., Khan, M.A., Seeber,
Bishop, M.P., Shroder Jr., J.F., Hickman, B.L., Copland, L., 1998a. L., Zeitler, P.K., 1999. Tectonics of Nanga Parbat, western
Scale dependent analysis of satellite imagery for characteriza- Himalaya: synkinematic plutonism within the doubly vergent
tion of glacier surfaces in the Karakoram Himalaya. Geomor- shear zones of a crustal-scale pop-up structure. Geology 27,
phology 21, 217 – 232. 999 – 1002.
Bishop, M.P., Shroder Jr., J.F., Sloan, V.F., Copland, L., Colby, J.D., Shroder Jr., J.F., Bishop, M.P., 1998. Mass movement in the Hima-
1998b. Remote sensing and GIS technology for studying litho- laya: new insights and research directions. Geomorphology 26,
spheric processes in a mountain environment. Geocarto Interna- 13 – 35.
tional 13 (4), 75 – 87. Shroder Jr., J.F., Bishop, M.P., 2000. Unroofing of the Nanga Parbat
Bishop, M.P., Shroder Jr., J.F., Bonk, R., Olsenholler, J., 2002. Himalaya. In: Khan, M.A., Treloar, P.J., Searle, M.P., Jan, M.Q.
Geomorphic change in high mountains: a western Himalayan (Eds.), Tectonics of the Nanga Parbat Syntaxis and the Western
perspective. Global and Planetary Change 32, 311 – 329. Himalaya, pp. 163 – 179.
 M.P. Bishop et al. / Geomorphology 55 (2003) 345–361 361

Shroder Jr., J.F., Khan, M.A., Lawrence, R.D., Madin, I.P., Hig- river incision model: implications for height limits of mountain
gins, S.E., 1989. Quaternary glacier chronology and neotec- ranges, landscape response timescales, and research needs. Jour-
tonics in the Himalaya of northern Pakistan. In: Malinconico, nal of Geophysical Research 104, 17661 – 17674.
L.L., Lillie, R.J. (Eds.), Tectonics of the Western Himalayas. Zeitler, P.K., Koons, P.O., Bishop, M.P., Chamberlain, C.P., Craw,
Special Paper - Geological Society of America, vol. 232, D., Edwards, M.A., Hamidullah, S., Jan, M.Q., Khan, M.A.,
pp. 275 – 294. Khattak, M.U.K., Kidd, W.S.F., Mackie, R.L., Meltzer, A.S.,
Small, E., 1999. Does global cooling reduce relief? Nature 401, Park, S.K., Pecher, A., Poage, M.A., Sarker, G., Schneider,
31 – 33. D.A., Seeber, L., Shroder Jr., J.F., 2001a. Crustal reworking at
Smith, J.A., Lin, T.L., Ranson, K.J., 1980. The Lambertian assump- Nanga Parbat, Pakistan: metamorphic consequences of ther-
tion and Landsat data. Photogrammetric Engineering and Re- mal – mechanical coupling facilitated by erosion. Tectonics 20
mote Sensing 46 (9), 1183 – 1189. (5), 712 – 728.
Teillet, P.M., Guindon, B., Goodenough, D.G., 1982. On the slope- Zeitler, P.K., Meltzer, A.S., Koons, P.O., Craw, D., Hallet, B., Cham-
aspect correction of multi-spectral scanner data. Canadian Jour- berlain, C.P., Kidd, W.S., Park, S.K., Seeber, L., Bishop, M.P.,
nal of Remote Sensing 8 (2), 733 – 741. Shroder Jr., J.F., 2001b. Erosion, Himalayan geodynamics, and
Whipple, K.X., Tucker, G.E., 1999. Dynamics of the stream-power the geomorphology of metamorphism. GSA Today 11, 4 – 8.
 Geomorphology 55 (2003) 363 – 380
www.elsevier.com/locate/geomorph

High spatial resolution hyperspectral mapping of in-stream

habitats, depths, and woody debris in mountain streams
W. Andrew Marcus a,*, Carl J. Legleiter b, Richard J. Aspinall b,
Joseph W. Boardman c, Robert L. Crabtree d
a
Department of Geography, University of Oregon, Eugene, OR 97403-1251, USA
b
Department of Earth Sciences, Montana State University, Bozeman, MT 59717, USA
c
Analytical Imaging and Geophysics, 4450 Arapahoe Ave, Suite 100, Boulder, CO 80303, USA
d
Yellowstone Ecological Research Center, 7500 Jarmen Circle, Suite #2, Bozeman, MT 59715, USA
Received 29 January 2002; received in revised form 10 May 2002; accepted 10 March 2003

Abstract

This article evaluates the potential of 1-m resolution, 128-band hyperspectral imagery for mapping in-stream habitats,
depths, and woody debris in third- to fifth-order streams in the northern Yellowstone region. Maximum likelihood supervised
classification using principal component images provided overall classification accuracies for in-stream habitats (glides, riffles,
pools, and eddy drop zones) ranging from 69% for third-order streams to 86% for fifth-order streams. This scale dependency of
classification accuracy was probably driven by the greater proportion of transitional boundary areas in the smaller streams.
Multiple regressions of measured depths ( y) versus principal component scores (x1, x2,. . ., xn) generated R2 values ranging from
67% for high-gradient riffles to 99% for glides in a fifth-order reach. R2 values were lower in third-order reaches, ranging from
28% for runs and glides to 94% for pools. The less accurate depth estimates obtained for smaller streams probably resulted from
the relative increase in the number of mixed pixels, where a wide range of depths and surface turbulence occurred within a
single pixel. Matched filter (MF) mapping of woody debris generated overall accuracies of 83% in the fifth-order Lamar River.
Accuracy figures for the in-stream habitat and wood mapping may have been misleadingly low because the fine-resolution
imagery captured fine-scale variations not mapped by field teams, which in turn generated false ‘‘misclassifications’’ when the
image and field maps were compared.
The use of high spatial resolution hyperspectral (HSRH) imagery for stream mapping is limited by the need for clear water to
measure depth, by any tree cover obscuring the stream, and by the limited availability of airborne hyperspectral sensors.
Nonetheless, the high accuracies achieved in northern Yellowstone streams indicate that HSRH imagery can be a powerful tool
for watershed-wide mapping, monitoring, and modeling of streams.
D 2003 Elsevier Science B.V. All rights reserved.

Keywords: Remote sensing; In-stream habitat; Morphology; Woody debris; Depth; Mountain stream

1. Introduction

* Corresponding author. Fax: +1-541-346-2067. Mapping variations in stream habitat, water depth,
E-mail address: [email protected] (W.A. Marcus). and woody debris throughout entire watersheds can be

0169-555X/03/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0169-555X(03)00150-8
364 W.A. Marcus et al. / Geomorphology 55 (2003) 363–380

notoriously difficult in mountain environments. The glides, and riffles—critical habitats for benthic fauna
paucity or absence of fine spatial resolution, basin- (Orth and Maughan, 1983). These morphologic units
wide maps of stream characteristics poses a major also provide a template for predicting where contam-
obstacle to understanding how mountain stream pro- inants might accumulate in stream sediments (Ladd et
cesses and responses are integrated into broader, al., 1998). Hardy et al. (1994) found ‘‘close’’ agree-
watershed-wide geomorphic systems and are modified ment between ground- and image-based maps of
by human impacts. hydraulic features such as pools, eddies, and runs on
High spatial resolution hyperspectral (HSRH) the Green River, UT, using multispectral video
imagery is a tool potentially capable of acquiring imagery at spatial resolutions of 0.25 –3 m. In con-
detailed maps over a stream’s entire length. High trast, in the much smaller third- and fourth-order Soda
spatial resolution ( V 5 m) provides the detail neces- Butte and Cache Creeks of Montana and Wyoming,
sary to capture key features in the narrow headwater Wright et al. (2000) obtained poor results using 4-
streams typical of mountain environments. The spec- band, 1-m imagery to map stream habitats, with
tral coverage and narrow bandwidths associated with standard supervised classification procedures produc-
hyperspectral imagery (generally z 64 bands) allow ing overall accuracies ranging from 10% to 53%. The
the user to differentiate features of interest that could inadequacy of their results was largely attributed to
not be distinguished otherwise. This article evaluates coregistration problems, although the need for greater
the ability of 1-m, 128-band imagery to map in-stream spectral resolution was also noted. To avoid these
habitats, stream depths, and woody debris in third- to coregistration issues, Marcus (2002) mapped stream
fifth-order streams of the northern Yellowstone microhabitats in the fifth-order Lamar River of
region. If a single remote sensing instrument could Wyoming directly to hardcopy printouts of 1-m,
be used to map this suite of features at watershed 128-band hyperspectral imagery to ensure a perfect
scales, it would simplify data collection, improve match between the classified imagery and field maps.
access to remote areas, reduce costs, and enable better Marcus used a maximum likelihood classification
monitoring and modeling of river habitat in mountain approach to achieve an overall classification accuracy
watersheds. of 85%. Using the same imagery and the same field
data, Goovaerts (2002) was able to improve the in-
stream habitat classification accuracy to 99% using a
2. Previous research co-kriging classification scheme.
Both the improved spectral and enhanced spatial
The early applications of digital remote sensing resolution provided by HSRH imagery appear neces-
technology in the context of fluvial systems focused sary to accurately map in-stream habitats. Marcus
on large rivers or reservoirs, where the water could be (2002) found that using 1 m rather than 5-m pixel
clearly delineated using the coarse 10 – 80-m pixel resolution increased overall accuracies by 19.4% and
resolution available from satellites. Since approxi- that using 128 band rather than 4-band, 1-m imagery
mately 1990, however, the development and accessi- improved accuracies by 17.6%. Similarly, Legleiter et
bility of airborne sensors (e.g., AVIRIS); improved al. (2002) noted that overall accuracies increased by
technology for converting film-based images to digital 7.2% when using 128-band HSRH imagery rather
form; and high spatial resolution satellite-based sen- than 8-band imagery and by 4.7% when using 1 m
sors (e.g., IKONOS) have created new opportunities rather than 2.5-m imagery.
for applying digital remote sensing to small streams Research on remote measurement of stream depth
typical of mountain environments. Fourteen sensors has also received increasing attention from the remote
with spatial resolutions of V 5 m are scheduled for sensing community in recent years. Lyon et al. (1992)
launch by the year 2005 and imagery from nine of classified five depth categories with 95% accuracy
these instruments will be commercially available using a radiative transfer model with four visible light
(Ustin and Costick, 2000). bands in the relatively large (compared to our study
Research on remote sensing of in-stream features sites) Saint Mary’s River of Michigan. The radiative
has largely focused on microhabitats, such as pools, transfer model and resultant depth estimates, however,
 W.A. Marcus et al. / Geomorphology 55 (2003) 363–380 365

were dependent on Secchi Disk measurements of the research has focused on using imagery to map one
extinction coefficients, an approach that would be variable, usually within a reach of limited size. Eval-
difficult or impossible to apply in the comparatively uating whether a single instrument can map a suite of
shallow and often turbulent conditions found in critical habitat characteristics indicates how effec-
mountain rivers. Gilvear et al. (1995) used scanned tively these approaches can simplify and improve
panchromatic photos with f 1-m resolution to differ- watershed-wide mapping of streams. Quantifying var-
entiate deep versus shallow water, but did not attempt iations in mapping accuracy is also required to under-
precise depth estimates. A quantitative study by stand the performance of these techniques at varying
Winterbottom and Gilvear (1997) succeeded in esti- stream scales. This article evaluates the potential for
mating depths with 2-m, 9-band simulated Daedalus using high spatial resolution hyperspectral imagery
AADS imagery, generating R2 values of 67% for (HSRH imagery) to map stream microhabitats and
water depths up to 1.2 m. Their cross-section analysis depths in third-, fourth-, and fifth-order channels and
indicated that the 67% figure probably underrepre- wood within and on the bars of a fifth-order channel.
sented the quality of the depth estimates, which
produced cross-sections very similar to those sur-
veyed at the same sites. 3. Field area
Active sensors, including radar, have shown greater
potential for accurate depth measurements than have We collected hyperspectral imagery and mapped
passive, reflected signals. In a variety of stream con- in-stream habitats and depths in August 1999 along
ditions with depths up to 5 m, ground-penetrating radar two reaches in Soda Butte Creek and one reach in the
produced an overall R2 value of 97% for measured Lamar River (Fig. 1). Because of time constraints,
versus estimated depths (Spicer et al., 1997). Ground- woody debris was only surveyed in the Lamar Reach.
penetrating radar also shows great promise as a techni- Approximate reach lengths were 2 km for the Foot-
que for estimating stream velocities (Spicer et al., 1997; bridge Reach in Soda Butte Creek and the Lamar
Costa et al., 2000). River and f 5 km for the Cooke City Reach. These
In comparison to work on microhabitats and stream stream reaches were selected because they are part of
depth, remote sensing of wood has received relatively long-term studies on stream disturbance and change
little research attention. Marcus et al. (2002) had due to mining (Ladd et al., 1998; Marcus et al., 2001),
relatively poor success using 1-m, 4-band imagery fire and the associated introduction of woody debris
to map wood in Cache Creek in northern Yellowstone (Minshall et al., 1998; Marcus et al., 2002), and
Park, largely because of spectral confusion between flooding (Meyer, 2001). In addition, portions of these
gravel and wood in mixed pixels and coregistration reaches and nearby areas have been used to test the
problems similar to those encountered by Wright et al. ability of 4-band, 1-m resolution imagery to map in-
(2000). Although readily apparent to the naked eye, stream habitats (Wright et al., 2000) and woody debris
the configuration of woody debris (especially when (Marcus et al., 2002). The following description of
present as isolated logs or small sets of logs) made up these reaches focuses on key factors controlling spec-
such small portions of individual pixels that it often tral response and classification accuracy.
could not be distinguished from the surrounding All the reaches displayed pool/riffle morphology
background, even with 1-m imagery. Marcus et al. (Montgomery and Buffington, 1997) with single- to
(2002) recommended using either finer resolution multichannel configurations and well-developed
imagery, which is difficult to acquire, or hyperspectral gravel bars. Variations in stream hydraulics were also
imagery, which enables objects with a clear spectral similar across reaches, with surface turbulence rang-
signal (like wood) to be distinguished even when they ing from areas of mixed white water and surface
make up only a fraction of a pixel (Boardman, 1989, waves in riffles to mirror-like surfaces in glides.
1993). Stream depths in all reaches typically varied
Previous research has demonstrated the potential of between 0 and 0.6 m, with measured maximum
digital remote sensing as a tool for mapping key stream depths up to 1.6 m in the Lamar River and
characteristics of small streams. To date, however, 1.2 and 1.3 m in the Footbridge and Cooke City
366 W.A. Marcus et al. / Geomorphology 55 (2003) 363–380

Fig. 1. Location of the three study reaches. Figure modified from Marcus et al. (2001).

Reaches, respectively, of Soda Butte Creek. Water The similarities among stream reaches suggest that
conditions were clear when the spectral imagery was major differences in classification accuracies between
collected, and field crews could easily see the stream stream reaches noted in this article most likely result
bottom at all sites, except in white water riffles. from scale differences and not from gross differences
The substrate in all reaches varied from sand size in morphology, surface turbulence, substrate, or water
in eddy drop zones to cobbles in scour pools or high- column characteristics. The Lamar River reach is
gradient riffles. Substrate composition in all reaches fifth-order as determined from USGS 1:24,000-scale
primarily consisted of sediments from the Eocene topographic quadrangles. Soda Butte Creek is third-
volcanics that dominate the northern Yellowstone order at the Cooke City site and fourth-order at the
landscape. Substrate materials from granitic gneisses Footbridge Reach (Fig. 1). Variations in the scale of
and sedimentary rocks above Cooke City (Metesh et in-stream features (e.g., the size of riffles or gravel
al., 1999) were also present in Soda Butte Creek, bars) paralleled the variations in stream order, with
particularly in the Cooke City Reach. The substrate larger features in higher-order channels. Based on
was, for the most part, free of algae when imagery was USGS gage records farther downstream, discharge
acquired. in the Lamar River was f 7.1 m3/s (250 cfs) on 3
 W.A. Marcus et al. / Geomorphology 55 (2003) 363–380 367

August 1999 when imagery was collected. Discharge trum (0.438 – 2.507 Am) with a spectral bandwidth of
at a USGS gage on the Footbridge Reach of Soda 0.016 – 0.020 Am.
Butte Creek was 3.9 m3/s (139 cfs) on 2 August 1999 The digital data were downloaded the day of the
when imagery was collected. Based on a ratio of flight and printed out as true color composites. Field
discharge to basin area, the discharge at the head of teams mapped depths and in-stream habitats directly
the Cooke City Reach would have been 20% of that at to the imagery within 10 days of the overflight. The
the Footbridge Reach, or f 0.79 m3/s (28 cfs). field teams mapped seven types of habitat units based
on the Bisson et al. (1982) scheme that is widely used
by fisheries biologists and management agencies. In
4. Methods addition, in the Footbridge and Cooke City Reaches,
we mapped a category we called standing water,
4.1. Data collection which referred to shallow pools cut off from the main
channel. We later merged the high- and low-gradient
Hyperspectral data were collected on 3 August riffles into one category called riffles, merged glides,
1999 for the Lamar River and the Cooke City Reach runs, and rough-water runs into one unit called glides,
of Soda Butte Creek, and on 2 August 1999 for the and merged standing water and eddy drop zones into
Footbridge Reach of Soda Butte Creek. Imagery was one unit called eddy drop zones, leaving us with four
collected using a Probe-1 sensor mounted on an A- in-stream habitats: pools, glides, riffles, and eddy drop
Star Aerospatiale helicopter flying 600 m above the zones (Fig. 2). The use of this simplified four-unit
ground surface. The Probe-1 measured reflected classification system significantly reduced field map-
energy across 128 contiguous bands covering the ping errors that resulted when different surveyors
visible to shortwave –infrared portions of the spec- mapped the same type of unit (e.g., a run) in different

Fig. 2. The four types of in-stream habitats mapped in Soda Butte Creek and the Lamar River. The boundaries between units are often
transitional and usually require subjective mapping decisions by the field teams. Photo from Soda Butte Creek immediately downstream of
Pebble Creek, November 1999.
368 W.A. Marcus et al. / Geomorphology 55 (2003) 363–380

ways (e.g., as a run or as a glide or as a rough-water remote sensing software package. Field maps were
run) (Marcus, 2002). Removing these disagreements digitized on-screen from the hard copies used in the
in field mapping was critical because such inconsis- field. Because imagery for the various reaches was
tency would have confused subsequent accuracy collected at different times of day with different
assessments of image-based maps and made it diffi- lighting conditions, we based analyses for each reach
cult to determine whether classification errors indi- on training data from the same reach. Using training
cated limitations of the imagery or subjectivity in the sites from Cooke City, for example, to generate
ground ‘‘truth’’ maps produced by field teams. More classifications in the Lamar River was not appro-
detailed descriptions of these unit types are provided priate given the variations in sun angle between
in Ladd et al. (1998), Wright et al. (2000), Marcus et images. Classifications were entirely spectrally dri-
al. (2002), and Legleiter et al. (2002). ven; other types of data (e.g., texture measurements)
Depths were mapped at points that could be clearly were not used. Analyses specific to in-stream hab-
located on both the image printouts and in the field. itats, depths, and wood are presented in the follow-
We attempted to measure at least one depth in each ing sections.
individual habitat unit (e.g., one point in every pool in
a given reach). Depths were measured using a ski pole 4.2.1. Image preprocessing
calibrated in 5-cm increments. Because depths and in-stream habitats were map-
Woody debris in the Lamar River was mapped on ped directly to printouts of the raw imagery within
10 – 12 November 1999. Time and logistical con- several weeks of the flight, we did not attempt to
straints prevented field mapping of wood in other georeference the imagery to correct for image distor-
reaches. No flows that were capable of moving the tion. Subsequent coregistration of field data and
wood occurred between the time of image acquisition imagery used the pixel x, y coordinates of the imagery.
and field mapping. Wood was mapped directly to true The absence of geographic coordinates did not affect
color printouts of the images to ensure precise cor- the field mapping or classification techniques we
egistration of imagery and maps. We only mapped used, but would be an obstacle to integrating the
those pieces of wood longer than 2 m, >10 cm in classification maps into GIS analyses with other data
diameter, and clearly visible on the imagery. layers. Following straight flight lines paralleling the
stream’s course and continuously running the Probe-1
4.2. Data analysis sensor’s cross track scanner provided images 2 – 6 km
in length. Each study reach, therefore, fell entirely
All remote sensing analysis was conducted in within a single image and, unlike applications using
Environment for Visualizing Images (ENVI), a along track scanners, we did not need to mosaic a

Fig. 3. Eigenvalue plots of the eigenvalues for the 128 principal components for the Lamar River (dashed line) and the Cooke City Reach (solid
line with squares) and Footbridge Reach (solid line) of Soda Butte Creek.
 W.A. Marcus et al. / Geomorphology 55 (2003) 363–380 369

required for a maximum likelihood supervised classi-

fication was 129 with 128 spectral bands as input
(Richards, 1994).
Eigenvalue plots (Fig. 3) indicated that over 100 of
the 128 principal components contributed significant
information (i.e., eigenvalue >1.0), but the highest
overall classification accuracies and kappa coeffi-
cients were generated using only the first 25 PC bands
in the Lamar River, the first 20 in the Cooke City
Reach, and the first 15 in the Footbridge Reach.
Classification accuracies gradually decreased as addi-
Fig. 4. A schematic diagram of the general relation of morphologic tional PC bands beyond this optimal number were
units to substrate size and surface turbulence of the water. used for the classification (Marcus, 2002), presumably
because these additional bands included increasing
number of separate scenes. The data were not atmos- amounts of residual noise from the original images.
pherically corrected.
Table 1
4.2.2. In-stream habitats Error matrix for maximum likelihood supervised classifications of
Field crews mapped in-stream habitats as polygons in-stream habitats using principal component images derived from
that covered the entire stream. This article reports on 128-band hyperspectral imagerya
the results for classifying all the stream pixels that were Image Ground truth (number of pixels)
mapped as well as a spatially buffered version of the classification
Glides Riffles Pools Eddy drop Total
original field map polygons. Spatial buffers were zones
applied because most boundaries between units were Third-order, Cooke City Reach
fuzzy, with one unit (e.g., a riffle) gradually transition- Glides 2574 424 51 7 3056
ing into the next (e.g., a glide) over a distance of several Riffles 496 550 115 21 1182
meters. Removing a buffer zone where these transitions Pools 191 30 95 2 318
occurred better represented the reality of the stream Eddy drop zones 161 56 19 62 298
Total 3422 1060 280 92 4854
environment.
Generating image-based classifications required Fourth-order, Footbridge Reach
several steps. All nonstream areas were first masked Glides 6152 561 25 10 6748
out. We then transformed the image data using a Riffles 2020 1047 11 0 3078
nonstandardized, principal component algorithm to Pools 151 16 94 0 261
Eddy drop zones 7 0 0 22 29
minimize spectral noise and to reduce the number of Total 8330 1624 130 32 10,116
bands required for classification, a standard technique
in hyperspectral analysis. We used variance-based Fifth-order Lamar River Reach
PCA images rather than correlation-based PCA Glides 6083 389 12 77 6561
because the variance-based images generated higher Riffles 174 2653 6 23 2856
Pools 323 313 132 1 769
classification accuracies. Classifications for each Eddy drop zones 139 142 2 564 847
reach were conducted using a maximum likelihood Total 6719 3497 152 665 11,033
supervised classification based on training sites from a
A two-pixel (f2 m) spatial buffer has been removed from the
the same reach. This maximum likelihood principal perimeter of each unit prior to accuracy assessment. Columns show
component approach generated the best results among the number of pixels in a given class as mapped by the field team,
the numerous techniques we investigated and is, and the rows show the number of pixels in a given unit as classified
by the image. For example, 3422 total pixels were field mapped as
therefore, the focus of this article.
glides, of which 2572 were classified as glides on the image, 496 as
The classification was developed using 129 ran- riffles, 191 as pools, and 161 as eddy drop zones. Data for the
domly chosen training sites from each feature type in Lamar River from Marcus (2002). Bolded values indicate number of
each reach. The minimum number of training sites pixels correctly classified in each category.
370 W.A. Marcus et al. / Geomorphology 55 (2003) 363–380

Table 2
Summary of classification accuracies for Soda Butte Creek and the Lamar Rivera
Reach Buffer Number of Overall Kappa Producer’s and user’s accuracies (%) for:
size (m) PC bands accuracy coefficient Glides Riffles Eddy drop zones Pools
(%)
Producer User Producer User Producer User Producer User
Cooke City 2 20 67.6 0.35 75.2 84.2 51.9 46.5 33.9 29.9 67.4 20.8
Cooke City No buffer NA NA NA NA NA NA NA NA NA NA NA
Footbridge 2 15 72.3 0.31 73.8 91.2 64.5 34.0 68.8 75.9 72.3 36.0
Footbridge No buffer 15 62.0 0.29 67.1 67.1 49.7 49.7 77.9 15.8 33.6 30.2
Lamar 2 25 85.5 0.74 90.5 92.7 75.9 92.9 84.8 66.6 86.8 17.2
Lamar No buffer 25 65.6 0.40 67.6 90.1 59.5 67.2 84.7 7.8 68.9 10.2
a
Data for the Lamar River classification from Marcus (2002) using a 2-m buffer zone. Producer’s accuracy refers to the accuracy that the
producer of the original field maps can calculate by overlaying the pixels classified as a particular feature (e.g., glides) on the ground map of that
feature. The producer’s accuracy is the percent of the ground truth map that is correctly classified for all ground-mapped pixels of a given
feature. The user of the classified image does not have the original ground truth map and therefore must check the accuracy by going to all sites
in the field classified as a specific feature type (e.g., a glide). The user’s accuracy is the percent of the image map that is correctly classified for a
given feature. See Congalton and Green (1999) for a more detailed explanation.

Table 3
Step-wise multiple regression results for depth ( y) versus values of different principal component scores (x1, x2,. . .,xn) for each pixel where
depth was measureda
Location of depth Number Order of principal component bands Adjusted Predicted
measurements of sites included in regression R2 (%) R2 (%)
Lamar Reach, 5th order
All sites combined 175 7, 3, 22, 15, 25, 28, 11, 21, 23, 4, 26 44.0 38.3
Eddy drop zones 9 23, 29, 4 93.2 88.9
Pools 23 23, 225, 15, 28, 12, 10, 13 85.1 78.4
Glides 60 7, 19, 3, 11, 27, 25, 17, 21 41.6 28.7
Runs 33 20, 15, 14, 13, 26, 25, 224, 24, 22, 27, 28, 17 88.3 79.4
Low-gradient riffles 29 3, 22, 12, 4, 17, 20, 9, 1 79.8 66.3
High-gradient riffles 16 11 20.0 0.00

Footbridge Reach, 4th order

All sites combined 144 4, 2, 11, 5, 7, 6 27.9 23.9
Standing water 11 10, 20, 19, 22, 14, 15 99.6 99.1
Eddy drop zones 13 4, 11, 20, 17, 10, 9 91.0 79.2
Pools 7 21 45.3 21.9
Glides 22 5, 13 42.6 31.5
Runs 50 26, 30, 22, 20, 7, 6, 8, 0, 21 43.6 26.6
Low gradient riffles 21 7, 3, 24, 28, 11, 19, 2, 30 79.3 54.5
High gradient riffles 15 11, 16 72.2 66.0

Cooke City Reach, 3rd order

All sites combined 105 2, 4, 1, 10, 9, 3, 11, 8, 5 59.0 53.4
Eddy drop zones 17 2, 4, 9, 11, 14, 1, 13, 8, 10 93.8 80.9
Pools 5 1 72.3 0.0
Glides 17 1, 7, 3, 5, 13, 15, 6, 10, 2, 9, 8 98.6 90.0
Runs 9 2, 1, 5, 3 93.1 76.7
Rough water runs 10 3, 2, 9 82.5 55.1
Low gradient riffles 11 2, 8, 5, 1, 3 91.7 83.0
High gradient riffles 35 2, 1, 4, 17, 9, 18 67.0 55.4
a
Only PC scores significant at p V 0.10 are entered into the regression. High R2 values are suspect because of small sample sizes (i.e., small
number of sites), but results indicate that stratifying by morphologic units has potential as a means for improving depth predictions with HSRH
data.
 W.A. Marcus et al. / Geomorphology 55 (2003) 363–380 371

Classification accuracies reported in this article rep- mine the strength of the relationship between depth
resent the maximum accuracies achieved when the and spectral reflectance and to develop equations for
optimal number of PC bands were used. estimating depths throughout the stream. Because
Accuracy assessment followed standard confusion variations in spectral reflectance within a reach at a
matrix procedures, as outlined by Congalton and given time are a function of surface turbulence and
Green (1999). Pixels used in the training set were substrate composition as well as depth, we used the
not used for evaluating classification accuracy. spectral classification of in-stream habitats to stratify
the sample by morphologic unit. High-gradient riffles,
4.2.3. Depth analysis for example, tend to have similar surface turbulence
Depth analysis was conducted on the same princi- and substrate characteristics at all locations within a
pal component images used for the in-stream habitat reach, thus reducing the variation in those factors and
analysis. All depth sites within each reach were isolating depth as the primary driver of the observed
entered into a stepwise multiple regression to deter- spectral reflectance.

Fig. 5. Woody debris mapping in the upper 0.75 km of the Lamar River Reach. (A) Field teams mapped wood directly to a printout of the
original image. Training pixels (129) were randomly selected from the (B) field map and a matched filter process was applied to generate
threshold images. Shown in white are pixels that equal or exceed a threshold value of (C) 0.18, (D) 0.50, or (E) 1.00 on the matched filter image.
Choice of the appropriate threshold value is partly a subjective decision, as is discussed in the text. The solid white box in (A) corresponds to the
location of the image shown in Fig. 6; the dashed box shows the location of the depth map in Fig. 7.
372 W.A. Marcus et al. / Geomorphology 55 (2003) 363–380

For purposes of depth analysis, we used the orig- identify single feature types. The MF technique cre-
inal eight in-stream habitats (high- and low-gradient ated an image that assigned values to each pixel, with
riffles, eddy drop zones, standing water, rough-water high values indicating more wood-like features.
runs, runs, glides, and pool) rather than merging the Normally, one selects a pure pixel or set of pure
units into riffles, glides, pools, and eddy drop zones as pixels to train a matched filter image, but this gen-
was done for the mapping of in-stream units. Use of erated poor classification results for wood. We there-
the eight units better stratified the stream by surface fore selected 129 pixels that were equally distributed
turbulence and substrate size, although some overlap around the image in wood of all sizes, ranging from
between units was present (Fig. 4). Stratifying the individual logs to large debris jams.
sample using the image-based classification of in-
stream units rather than our field maps provided a
better indication of the potential of the technique to be 5. Results
applied in areas where field teams are not available to
generate detailed ground maps prior to using the Table 1 shows error matrices for in-stream habitat
imagery for depth analysis. mapping using a 2-m spatial buffer and principal
component images. Table 2 depicts the variations in
4.2.4. Wood analysis overall accuracies, kappa coefficient values, and pro-
Procedures for classifying wood initially followed ducer’s and user’s accuracies among the three reaches
the same techniques used for in-stream habitats. A and between buffered and unbuffered units. Overall
mask was used to remove features outside the exposed classification accuracies are 67.6% for the third-order
bars where wood was located. We then performed a reach, 72.3% for the fourth-order reach, and 85.5% for
principal component transformation on the remaining the fifth-order reach. Use of the spatial buffer to remove
portion of the image. transitional areas along unit boundaries increases accu-
We used the matched filter (MF) approach in ENVI racies by 10.3% in the fourth-order Footbridge Reach
to identify woody debris in the principal component and 19.9% in the fifth-order Lamar Reach.
image. The MF algorithm performs a partial unmixing The results of the stepwise multiple regression
of spectra to estimate the relative quantity of a given relating depths to non-normalized principal compo-
material in a pixel based on user-defined spectral end nent scores for the hyperspectral data are shown in
members (wood in this case) (Harsanyi and Chang, Table 3. R2 values when all the depth data are pooled
1994; Boardman et al., 1995). Matched filtering has together for each reach range from 27.9% in the
the advantage of not requiring knowledge of all end Footbridge Reach to 59.0% in the Lamar River. When
members within an image scene and so can be used to the depth data are stratified by in-stream habitat type,

Table 4
Changes in wood mapping accuracy in the Lamar River Reach as different matched filter minimum threshold values are applieda
Minimum Overall Accuracy measures for wood only
threshold accuracy Error of Error of Producer’s User’s # of pixels
value (%) commission omission accuracy accuracy classified
(%) (%) (%) (%) as wood
mf>0.18 93.4 39.7 15.3 84.7 68.1 12,638
mf>0.20 94.1 33.5 16.3 83.7 71.4 11,897
mf>0.23 94.7 26.5 18.2 81.8 75.5 11,002
mf>0.25 95.0 22.7 19.5 80.5 78.0 10,477
mf>0.35 95.4 12.0 26.9 73.1 85.9 8638
mf>0.50 94.9 3.9 38.7 61.3 94.0 6620
mf>0.75 93.0 0.6 57.8 42.2 98.5 4354
mf>1.00 91.2 0.2 73.4 26.6 99.3 2722
a
Only two classes were mapped for purposes of this analysis: wood and nonwood. The field team mapped 10,153 wood pixels on the
ground truth image (Fig. 4).
 W.A. Marcus et al. / Geomorphology 55 (2003) 363–380 373

adjusted R2 values vary widely, ranging from 20.0% One approach to resolving the discrepancies
for high-gradient riffles in the Cooke City Reach to between homogenous ground truth maps and hetero-
99.6% for standing water in the Footbridge Reach. geneous classifications would be to use post-classi-
Fig. 5 portrays a black-and-white composite image fication filters that ‘‘clump and sieve’’ the classified
of the area where wood was mapped, the field map of pixels to create more cohesive classification units.
wood, and wood classifications based on matched This post-classification fix would ignore, however,
filter analysis using different minimum thresholds. the possibility that the concentration of pixel-scale
Table 4 shows the changes in different accuracy heterogeneity at unit boundaries on the image maps is
measures that occur as different threshold values are real; i.e., that the imagery may provide more accurate
used. The implications of using different thresholds and precise maps than the ground teams (Marcus,
are discussed below. 2002; Legleiter et al., 2002). For example, close
examination of Fig. 6 shows a scattering of riffle
and glide pixels near unit boundaries, suggesting that
6. Discussion the imagery detected subtle variations where surface
roughness is locally riffle-like, but glide-like only a
The following sections on in-stream habitats, few centimeters away. Such variations are indeed
depth, and woody debris discuss issues specific to observed in the field, but ground teams could not
each parameter. We also describe factors affecting map at this scale because of time limitations and the
mapping accuracy, possibilities for improving classi- difficulty of precisely locating the pixel on the image.
fication results, extensions of HSRH mapping to The field team’s placement of unit boundaries through
watershed scales, and future research directions. The these transitional areas was also somewhat arbitrary.
section concludes with some general comments that The image classification also indicates shoreline
apply to HSRH mapping of these three important boundary effects that make sense from a hydraulic
stream parameters. perspective. Areas mapped as riffles by the field team
are shown by the imagery to have glide pixels near the
6.1. In-stream habitats shoreline; a feature that occurs as water shallows and
slows at stream edges (Fig. 6). Eddy drop zones are
The ability of HSRH imagery to classify in-stream shown immediately adjacent to the shore, where water
habitats varied with stream scale. Overall and produc- velocity drops to nearly zero and sands and silts are
er’s accuracies consistently improved as stream order deposited along the banks. These features, however,
increased (Table 2). The greater accuracy in the fifth- were too small to be mapped in the field.
order Lamar River probably occurred because larger Two primary sources of ground survey error there-
streams feature more sizable, more homogenous units fore probably led to discrepancies between field maps
than do smaller streams. This, in turn, significantly and image-based maps: (i) inconsistent identification
reduced the proportion of the stream occupied by of units by field crews; and (ii) lumping of small
transitional boundary areas between in-stream habitats regions of variability (e.g., a small riffle) into a larger
as well as the amount of internal variability displayed more homogenous region (e.g., a glide). Based on this
within a single unit (e.g., small riffle-like areas within rationale and on our experience with the field sites, we
a glide). believe that the HSRH maps were more consistent and
The application of the 2-m buffer improved clas- precise in their characterization of unit boundaries and
sification accuracies (Table 2) by removing a portion heterogeneity than the large, homogenous polygons
of the transitional boundary zone that is difficult to created by the field teams. If this is the case, then the
map in the field. Even with the 2-m buffer, however, poorer classification results in the Cooke City Reach
image-based classifications often displayed a mix of may not reflect a weakness of the classification capa-
unit types on the margins of individual units. These bilities of HSRH imagery in lower-order streams, but
regions generated apparent classification errors when a weakness of the techniques used to field map and
compared to the homogenous ground-truth polygons validate the image classifications in these tremendous-
mapped by field teams. ly variable environments.
374 W.A. Marcus et al. / Geomorphology 55 (2003) 363–380

Fig. 6. (A) The field map of in-stream habitats, (B) a photo looking downstream at the same location 3 months later when water levels had
dropped and the extent of the riffles had decreased, and (C) the HSRH classification of the same location. Eddy drop zones were not mapped by
the field team in this area because they were below the minimum mapping size, but the HSRH imagery designated shallow shoreline areas that
were often covered with silt and sand as eddy drop zones. The riffle on the right channel is portrayed as having a mix of riffle, glide, and deeper
water pool units, which is also a realistic portrayal of that particular location. The location of these images is shown by the solid white rectangle
in Fig. 5A.

Future work could evaluate the role of surveyor number of tarps needed to establish adequate ground
subjectivity by comparing maps generated by differ- control can be daunting, especially in a popular, yet
ent individuals at the same location, or by revisiting relatively pristine locale such as Yellowstone National
sites using a GPS to pinpoint locations, assuming that Park, where special care must be taken to avoid
precision geocoding can generate meter-scale accu- disrupting either the perceived or real natural environ-
racy in image coordinates. Evaluating the role of ment. In sites like Yellowstone, registration panels
spatial heterogeneity in driving classification error cannot be left in the field for long periods. There are
would be more difficult, requiring mapping of stream significant logistical costs, personnel requirements,
habitats at submeter precision and coregistration of and disruptions of team morale caused by emplacing
these ground maps to imagery with submeter preci- and removing registration panels on a constant basis
sion. This level of precision might be obtained using as the team awaits flights that are rescheduled or
registration panels laid out in the field prior to the canceled for various reasons (Aspinall et al., 2002).
image collection flight. We attempted to establish Although we believe that the HSRH classifications
ground control points of this type using 2  2 m blue provide relatively accurate maps of the in-stream
plastic tarps. Unfortunately, the portions of the images habitats at all stream scales, the validation results
showing the tarps were unusable because of shadow- (Tables 1 and 2) call into question whether this
ing, image swirl, and line skip. Furthermore, the technique is ready for watershed-wide applications,
 W.A. Marcus et al. / Geomorphology 55 (2003) 363–380 375

especially in smaller streams. In particular, although for mapping the transitional boundaries and character-
the producer’s accuracies for the 2-m buffered units istics of in-stream habitats. Finally, more sophisticated
typically approached or exceeded the 85% criteria classification algorithms that take advantage of the
often considered acceptable for remote sensing clas- tremendous spatial information available in HSRH
sifications, the user’s accuracies were often notably imagery will improve classification results for in-
lower, especially for the smaller units (pools and eddy stream units (Goovaerts, 2002; Maruca and Jacquez,
drop zones) that make up a small portion of the total 2002).
stream area (Table 2). Unfortunately, these small pool
and eddy drop zone units are often most important as 6.2. Stream depths
fish habitat and sites of contaminated sediment accu-
mulation (Ladd et al., 1998), making these the fea- The adjusted R2 values for depth predictions strati-
tures managers are most interested in mapping. fied by in-stream habitat (Table 3) are generally higher
A number of options are available for improving than those previously obtained with passive sensors,
the classification accuracies at all scales. Scaling the although lower than those achieved with radar (Spicer
size of the spatial buffer to the stream size, for et al., 1997; Costa et al., 2000). Except for the glides
example, could remove more of the confusion caused and high-gradient riffles in the Cooke City Reach, the
by transitional zones between units. For example, one R2 values in all three reaches were higher for individ-
might apply a 4-m buffer in the Lamar, a 3-m buffer in ual units than when all sites were combined, suggest-
the Footbridge Reach, and a 2-m buffer in the Cooke ing that stratification may be useful in other settings as
City Reach. This has the disadvantage, however, of well.
not classifying increasingly larger areas of the river as The high R2 values suggest that HSRH imagery
stream size grows. Developing clearer criteria for field has promise as a tool for estimating depths in clear
mapping boundaries of habitat units might also water streams with no overstory. But the small sample
improve accuracies by ensuring consistency amongst sizes (Table 3) call into question the validity of the
field maps and subsequent selection of training sites analysis from a statistical perspective. This concern
for image classification. Fuzzy classifications (e.g., can be partially addressed by examining a depth map
Wright et al., 2000) may be particularly appropriate generated from the regression equations (Fig. 7). This

Fig. 7. A depth map of the upper end of the Lamar River Reach, WY. Depths are derived from the stepwise multiple regression equations that
relate depth to principal component values for the hyperspectral imagery. Apparent gaps in the stream channel are areas where woody debris
overlies the channel and blocks the sensor view of the water. The location of this reach in the upper Lamar River is shown by the dashed white
line in Fig. 5A.
376 W.A. Marcus et al. / Geomorphology 55 (2003) 363–380

map is hydraulically reasonable, with depths that are could be accomplished by coregistering depth maps
consistently deepest in the thalweg, cut banks of with high-resolution DEM-based maps of bed slope
meander bends, or areas of flow convergence and or, alternatively, by measuring water surface slopes
channel scour. Likewise, the map shows shallow from radar or lidar imagery. Such maps developed at
depths along the inside of bends and in areas of flow watershed scales would be remarkably valuable tools
divergence. Depth ranges displayed by the map are in for understanding stream habitats and disturbance
close agreement with those that occurred in the impacts, predicting channel change, and providing
stream. the empirical data needed to create and evaluate
The accuracy of depth estimates did not vary in a thermodynamic theories about the distribution of
consistent manner between the stream reaches, energy expenditure in fluvial systems.
although the poorest R2 values all occurred in the
lower-order streams. The lower values in smaller 6.3. Woody debris
streams may have been driven by the relative increase
in the number of mixed pixels, where a wide range of The data in Fig. 5 and Table 4 indicate that wood
depths and surface turbulence occurred within a single can be readily identified on hyperspectral imagery.
pixel. Stream size will put a bounding limit on the When evaluating a two-class system (wood and
ability of this depth detection technique to work nonwood in this case), no single-matched filter
because it cannot be applied in streams that are small threshold value provided the ‘‘best’’ accuracy. Data
enough for banks or overstory to block the sensor’s from Table 4 suggest, for example, that 1.0 would be
view of the water. the best threshold value for minimizing errors of
The greatest influence on the accuracy of the commission, which occur when nonwood areas are
technique appears to be surface turbulence, with classified as wood. A value of 1.0, however, max-
greater surface turbulence interfering with the sen- imizes the errors of omission, which occur when
sor’s ability to penetrate the water and provide ac- wood pixels are classified as nonwood. Similar
curate depth estimates. For the most part, units with arguments can be made for producer’s and user’s
little surface turbulence (pools, eddy drop zones, and accuracy, where choosing a high threshold value of
standing water) displayed higher R2 values. R2 values 1.0 ensures that almost every image map pixel
for other units did not vary in a consistent fashion, visited by a user will be wood, but also misses
although high-gradient riffles (the most turbulent of many of the wood pixels mapped by field teams
the units) had the lowest R2 values in the Cooke City (i.e., the producers of the data). The choice of a
and Lamar Reaches. threshold will ultimately depend on the map maker’s
The principal component scores most strongly and map user’s willingness to accept certain types of
correlated with depth (i.e., the first scores entered into error.
the stepwise regression) did not follow the rank order We believe that wood, because of its strong spec-
of the principal components. Apparently, the signal tral signature, can be detected even when it covers
related to depth is captured within subtle variations in only a portion of a pixel. For example, a pixel with a
the spectral reflectance of the water. Combined with matched filter threshold value of 0.75 and above
the large number of principal component scores might represent a pixel covered entirely by wood,
required to achieve high R2 values, this emphasizes while a value >0.18 but < 0.25 might represent a pixel
the desirability of hyperspectral rather than multi- with 20% wood coverage. This raises the strong
spectral imagery. In addition, HSRH imagery is nec- possibility that many pixels identified as errors of
essary to accurately classify the in-stream habitats commission in the accuracy analysis (Table 4) were in
(Marcus, 2002; Legleiter et al., 2002), a key step in fact pixels that contained wood, but wood that was
stratifying the stream data in order to improve accu- smaller than the minimum mapping unit used by field
racy. teams (Fig. 8). As is the case with the in-stream
Finally, the creation of continuous depth maps like habitats, this suggests that the image map may be
Fig. 7 raises the exciting possibility of creating con- more accurate than the field map that was used to
tinuous maps for shear stress and stream power. This ground truth the imagery.
 W.A. Marcus et al. / Geomorphology 55 (2003) 363–380 377

the image (Aspinall et al., 2002). This concern was

partially bypassed by only mapping wood that could
be clearly identified on printouts of the imagery. This
approach, however, did not permit field mapping of
wood smaller than what could be seen on the imagery,
which in turn limited the effectiveness of accuracy
assessment techniques. In particular, we could not
match areas where subpixel wood coverage was
indicated by the matched filter image (e.g., Fig. 5C)
to corresponding ground sites (e.g., Fig. 8), because
the image and ground coordinates could not be
coregistered to pixel-scale resolution. Future wood
mapping should be done using high quality geo-
corrected imagery. Corrections at this precision can
be accomplished using ray tracing methods for post-
correction (Boardman, 1999).
The wood classification worked best when using
pixels scattered throughout the image as training sites,
rather than when using one or a small number of pure
pixels, which is the more standard approach with
matched filters. We speculate that the scattered pixel
approach worked better because (i) woody debris is
heterogeneous, with some that has bark, some that is
burned, some that has foliage, and so on; and (ii) the
1-m imagery was not corrected for bidirectional
reflectance, so that a series of points from across the
entire image captured the range of image spectra for
Fig. 8. Wood in the Lamar River Reach that was smaller than the
wood better than spectra from just one or two sites.
minimum size (2 m long and 10-cm diameter) mapped by field
teams. Even though wood of this size was not mapped by field This suggests that segregating the wood by attribute
teams, it was probably sensed by the hyperspectral imagery and (with or without bark, burned or not burned, with or
shows up on matched filter maps (Fig. 5), creating false errors of without foliage, etc.) should improve wood detection
commission in the accuracy analysis. The footprint of a hypothetical by providing purer spectral signatures on which to
1-m pixel is shown in white.
train the matched filter. In addition, atmospheric
correction and removal of bidirectional reflectance
The results depicted in Fig. 5 and Table 4 indicate both might improve the detectability of wood.
that hyperspectral imagery has great potential as a tool
for mapping woody debris, especially at watershed 6.4. General considerations
scales where the personnel, cost, and time required for
ground mapping can be prohibitive. The only signifi- Regardless of the parameter being mapped, we
cant logistical obstacles to image-based mapping of encountered a number of difficulties with image
wood occur when wood is obstructed by water, acquisition and processing that are likely to plague
vegetation, or shadows. any research team using HSRH imagery (Aspinall et
Several issues should be addressed, however, to al., 2002). Coordination of overflights with field
improve validation procedures and classification ac- teams was remarkably difficult given the dependence
curacies. The greatest difficulties we encountered re- on weather, stream conditions, availability of hyper-
sulted from line skip, line oversampling, and image spectral sensors, and the large number of people that
swirl created by platform motion, all of which com- must be mobilized to collect sufficient ground data in
bined to make it impossible to accurately georeference a timely fashion. Accurate, pixel-scale georeferencing
378 W.A. Marcus et al. / Geomorphology 55 (2003) 363–380

and coregistration of imagery and ground data were Finally, visual examination of the spectral curves
not possible with submeter-scale precision. Finally, and of unsupervised classifications for both Soda
bidirectional reflectance created particular problems Butte Creek and the Lamar River shows that the
with high spatial resolution imagery because the hyperspectral sensor is capturing far more information
narrow swath width of the cross track scanner about the stream environment than can be conveyed in
( f 500 m at 1-m pixel resolution with the PROBE- a simple four-unit classification or a wood/nonwood
1 sensor) made it difficult to keep the streams near the dichotomy. HSRH-driven unsupervised classifications
center of the flight line. of the stream environment coupled with subsequent
The difficulty in realistically evaluating the accu- ground investigations might well open our eyes to
racy of pixel-scale variations shown on the imagery critical components of the stream system that have
for in-stream habitats and for subpixel variations in previously been overlooked.
wood indicates that alternative means for accuracy
assessment are needed. Accuracy assessment for both
wood and in-stream habitat mapping could benefit 7. Summary and conclusions
from (i) pixel-scale ground mapping coupled with use
of techniques that provide pixel-scale accuracy in Overall classification accuracies for in-stream hab-
coregistering images and field maps (e.g., Clark et itats (glides, riffles, pools, and eddy drop zones)
al., 1998; Boardman, 1999); and (ii) the development ranged from 69% for third-order streams to 86%
of techniques that enable realistic comparisons for fifth-order streams (Table 2). R2 values for com-
between the pixel- and subpixel-scale variations parisons of measured and estimated depths ranged
shown on images and the field maps with their large, from 20% for high-gradient riffles in third-order
homogenous polygons or one class per pixel portrayal reaches to 99% for glides in the fifth-order reach
of the surface (Aspinall, 2002). Mapping of in-stream (Table 3). The accuracy of woody debris mapping
habitats in particular will benefit from spatial models covered a wide range, depending on the threshold
that aggregate pixel-scale variations into larger nearby value chosen for the matched filter (Table 4). Visual
units, thus generating an image classification that analysis of image-based maps of in-stream habitats
more closely matches the homogenous polygons (Fig. 6), depths (Fig. 7), and wood (Fig. 5) all
mapped by field teams (e.g., Goovaerts, 2002; Maruca suggests that HSRH imagery generated results that
and Jacquez, 2002). are more accurate and useful than the validation
The approach taken in this study was ‘‘top-down,’’ statistics alone seem to suggest.
meaning that spectra from the airborne image were Given clear water and an unobstructed view of the
used to drive the classification. These image spectra, stream, our results indicate that there is great potential
however, represent a mixture of all the factors affect- for HSRH mapping and monitoring of streams. In a
ing the electromagnetic radiation as it travels from the number of cases, accuracies approached or exceeded
target to the sensor. The reflectance spectra for in- the 85% value typically expected for remote sensing
stream habitats, for example, represent variations in mapping. The ability to remotely map and monitor
reflectance because of surface turbulence, substrate stream channels at spatial scales of 1 m raises exciting
size, substrate color and composition, periphyton, prospects for the advancement of stream studies. In
turbidity, and depth, not to mention atmospheric particular, HSRH has the potential to enable exami-
effects. The development of deterministic, ‘‘bottom- nation of the pattern, process, and scale so critical to
up’’ models [such as that devised by Lyon and understanding fluvial ecosystems (Walsh et al., 1998)
Hutchinson (1995)] could be valuable for developing at resolutions previously thought to be obtainable only
classification schemes that can be applied across with localized ground-based monitoring (Muller et al.,
multiple watersheds and multiple images without 1993).
requiring such extensive ground truth collection Methodological hurdles remain to be overcome,
efforts. These models would also provide an important however, before we can be confident in HSRH-based
explanatory tool for understanding variations in spec- stream maps. In particular, relative to our study,
tral signals in different settings. improved procedures should be used or developed
 W.A. Marcus et al. / Geomorphology 55 (2003) 363–380 379

for georectifying images, removing bidirectional Laboratory Airborne Geoscience Workshop, Jet Propulsion Lab-
reflectance, assessing accuracy (including the field oratory Publication 93-26. 14 pp. http://popo.jpl.nasa.gov/docs/
workshops/93_docs/4.pdf.
mapping component), and analyzing the tremendous Boardman, J.W., 1999. Precision geocoding of low altitude AVIRIS
spatial as well as spectral information in HSRH data: lessons learned in 1998. AVIRIS 1999 Proceedings, Jet
imagery. These methodological advances need to be Propulsion Laboratory, CA. 6 pp. http://makalu.jpl.nasa.gov/
coupled with a greater understanding of the spectral docs/workshops/99_docs/7.pdf.
Boardman, J.W., Kruse, F.A., Green, R.O., 1995. Mapping target
characteristics of streams in order to explain classi-
signatures via partial unmixing of AVIRIS data. Summaries, Fifth
fication results and their variation among sites. Only if JPL Airborne Earth Science Workshop, Jet Propulsion Labora-
these studies are undertaken and new working tools tory Publication 95-1, pp. 23 – 26. http://popo.jpl.nasa.gov/docs/
developed will the full potential of HSRH mapping of workshops/95_docs/7.pdf.
streams be realized. Clark, R.N., Livo, K.E., Kokaly, R.F., 1998. Geometric correction
of AVIRIS imagery using on-board navigation and engineering
data. AVIRIS 1998 Proceedings, Jet Propulsion Laboratory, CA.
9 pp. http://makalu.jpl.nasa.gov/docs/workshops/98_docs/9.pdf.
Acknowledgements Congalton, R.G., Green, K., 1999. Assessing the Accuracy of Re-
motely Sensed Data: Principles and Practices. Lewis Publishers,
This research was supported by the NASA EOCAP New York. 137 pp.
program, Stennis Space Flight Center, Mississippi. Costa, J.E., Spicer, K.R., Cheng, R.T., Haeni, F.P., Melcher, N.B.,
Thurman, M.E., Plant, W.J., Keller, W.C., 2000. Measuring
William Graham, Joseph Spruce, and Greg Terrie of stream discharge by non-contact methods: a proof of concept
Stennis Space Flight Center all provided important experiment. Geophysical Research Letters 27 (4), 553 – 556.
logistical support and technical advice. Robert Ahl, Gilvear, D.J., Waters, T.M., Milner, A.M., 1995. Image analysis of
Kerry Halligan, and Jim Rasmussen collected ground aerial photography to quantify changes in channel morphology
and instream habitat following placer mining in interior Alaska.
truth data. The senior author wishes to thank Samuel
Freshwater Biology 34, 389 – 398.
Goward for introducing him to the field of remote Goovaerts, P., 2002. Geostatistical incorporation of spatial coordi-
sensing and supporting his early research efforts to nates into supervised classification of hyperspectral data. Jour-
apply remote sensing technology to watershed nal of Geographical Systems 4 (1), 99 – 111.
measurement and modeling. Hardy, T.B., Anderson, P.C., Neale, M.U., Stevens, D.K., 1994.
Application of multispectral videography for the delineation of
riverine depths and mesoscale hydraulic features. In: Marston,
R., Hasfurther, V. (Eds.), Effects of Human-Induced Change on
References Hydrologic Systems. Proceedings Annual Water Resources As-
sociation, Jackson Hole, WY, pp. 445 – 454.
Aspinall, R.J., 2002. Use of logistic regression for validation of Harsanyi, J.C., Chang, C.I., 1994. Hyperspectral image classifica-
maps of the spatial distribution of vegetation derived from high tion and dimensionality reduction: an orthogonal subspace pro-
spatial resolution hyperspectral remotely sensed data. Ecological jection approach. IEEE Transactions on Geoscience and Remote
Modelling. Special Issue on Advances in Generalized Linear/ Sensing 32, 779 – 785.
Generalized Additive Modelling: from Species’ Distribution to Ladd, S., Marcus, W.A., Cherry, S., 1998. Trace metal segregation
Environmental Management 157 (2 – 3), 301 – 312. within morphologic units. Environmental Geology and Water
Aspinall, R., Marcus, W.A., Boardman, J.W., 2002. Considerations Sciences 36, 195 – 206.
in collecting, processing, and analysing high spatial resolution Legleiter, C.J., Marcus, W.A., Lawrence, R., 2002. Effects of sensor
hyperspectral data for environmental investigations. Journal of resolution on mapping in-stream habitats. Photogrammetric En-
Geographical Systems 4 (1), 15 – 29. gineering and Remote Sensing 68, 801 – 807.
Bisson, P.A., Nielson, J.L., Palmalson, R.A., Grove, L.E., 1982. Lyon, J.G., Hutchinson, W.S., 1995. Application of a radiometric
A system of naming habitat types in small streams, with ex- model for evaluation of water depths and verification of results
amples of habitat utilization by salmonids during low stream with airborne scanner data. Photogrammetric Engineering and
flow. In: Armantrout, N.B. (Ed.), Acquisition and Utilization Remote Sensing 61, 161 – 166.
of Aquatic Habitat Inventory Information. Proceedings Ame- Lyon, J.G., Lunetta, R.S., Williams, D.C., 1992. Airborne multi-
rican Fisheries Society, Portland, OR, pp. 62 – 73. spectral scanner data for evaluating bottom sediment types and
Boardman, J.W., 1989. Inversion of imaging spectrometry data water depths of the St. Marys River, Michigan. Photogrammet-
using singular value decomposition. Proceedings, IGARSS’89, ric Engineering and Remote Sensing 58, 951 – 956.
12th Canadian Symposium on Remote Sensing 4, 2069 – 2072. Marcus, W.A., 2002. Mapping of stream microhabitats with high
Boardman, J.W., 1993. Automated spectral unmixing of AVIRIS data spatial resolution hyperspectral imagery. Journal of Geographi-
using convex geometry concepts. Summaries, 4th Jet Propulsion cal Systems 4 (1), 113 – 126.
380 W.A. Marcus et al. / Geomorphology 55 (2003) 363–380

Marcus, W.A., Meyer, G.A., Nimmo, D.R., 2001. Geomorphic con- Orth, D.J., Maughan, O.E., 1983. Microhabitat preferences of
trol on long-term persistence of mining impacts, Soda Butte benthic fauna in a woodland stream. Hydrobiologica 106,
Creek, Yellowstone National Park. Geology 29, 355 – 358. 157 – 168.
Marcus, W.A., Marston, R.A., Colvard Jr., C.R., Gray, R.D., 2002. Richards, J.A., 1994. Remote Sensing Digital Image Analysis.
Mapping the spatial and temporal distributions of large woody Springer, Berlin. 340 pp.
debris in rivers of the Greater Yellowstone Ecosystem. U.S.A., Spicer, K.R., Costa, J.E., Placzek, G., 1997. Measuring flood dis-
Geomorphology 44 (3 – 4), 323 – 335. charge in unstable stream channels using ground penetrating
Maruca, S.L., Jacquez, G.M., 2002. Area-based tests for association radar. Geology 25 (5), 423 – 426.
between spatial patterns. Journal of Geographical Systems 4 (1), Ustin, S.L., Costick, L., 2000. Multispectral remote sensing over
69 – 83. semi-arid landscapes for resource management. In: Hill, M.J.,
Metesh, J., English, A., Lonn, J., Kendy, E., Parrett, C., 1999. Aspinall, R.J. (Eds.), Spatial Information for Land Use Manage-
Hydrogeology of the upper Soda Butte Creek basin, Montana. ment. Gordon and Breach, London, pp. 97 – 109. Chapter 8.
Montana Bureau of Mines and Geology Report of Investigation, Walsh, S.J., Butler, D., Malanson, G.P., 1998. An overview of scale,
vol. 7. 66 pp. pattern, process relationships in geomorphology: a remote sens-
Meyer, G.A., 2001. Recent large-magnitude floods and their impact ing and GIS perspective. Geomorphology 21, 183 – 205.
on valley-floor environments of northeastern Yellowstone. Geo- Winterbottom, S.J., Gilvear, D.J., 1997. Quantification of channel-
morphology 40 (3 – 4), 271 – 290. bed morphology in gravel-bed rivers using airborne multispec-
Minshall, G.W., Robinson, C.T., Robinson, T.V., Royer, T.V., 1998. tral imagery and aerial photography. Regulated Rivers: Research
Stream ecosystem responses to the 1988 wildfires. Yellowstone and Management 13, 489 – 499.
Science 6 (3), 15 – 22. Wright, A., Marcus, W.A., Aspinall, R.J., 2000. Applications and
Montgomery, D.R., Buffington, J.M., 1997. Channel-reach mor- limitations of using multispectral digital imagery to map geo-
phology in mountain drainage basins. Geological Society of morphic stream units in a lower order stream. Geomorphology
American Bulletin 109 (5), 596 – 611. 33 (1 – 2), 107 – 120.
Muller, E., Decamps, H., Dobson, M.K., 1993. Contribution of
space remote sensing to river studies. Freshwater Biology 29,
301 – 312.
 Geomorphology 55 (2003) 381 – 398
www.elsevier.com/locate/geomorph

Solifluction steps and risers, Lee Ridge, Glacier National Park,

Montana, USA: a scale and pattern analysis
Stephen J. Walsh a,*, Ling Bian b, Sean McKnight a,
Daniel G. Brown c, Evan S. Hammer a
a
Department of Geography, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599-3220, USA
b
Department of Geography, State University of New York at Buffalo, Amherst, NY 14261-0023, USA
c
School of Natural Resources and Environment, University of Michigan, Ann Arbor, MI 48109, USA

Received 28 January 2002; received in revised form 29 June 2002; accepted 10 March 2003

Abstract

Solifluction patterns on Lee Ridge, Glacier National Park, Montana, USA, were characterized through a combination of field
transects, enhanced, digital aircraft data, and geostatistics. The objectives were to (i) assess the role of high spatial resolution
digital imagery for mapping the areal extent and spatial pattern of solifluction steps and risers through digital enhancements
[i.e., principal components analysis (PCA) and wavelet transforms (WTFs)] of ADAR-5500 aircraft imagery; (ii) describe the
pattern of solifluction steps and risers through the placement of twelve 50-m field transects oriented normal to the steps and
risers and organized along an elevational gradient extending from highest to lowest positions along the ridge; (iii) define scale
dependent relationships of the solifluction steps and risers mapped along one-dimensional field transects and two-dimensional
remote sensing images using indicator semivariograms to define the structure of the spatial data across a range of sampled
scales; and (iv) relate the solifluction patterns observed through the transects and the remote sensing interpretations to
topographic trends associated with distance from the dominant peak, a proxy for elevation change along Lee Ridge. Findings
suggest that the semivariance pattern of the transects changes with elevation, that is the semivariograms of the lower transects
show smaller amplitudes and shorter wavelengths than those at higher elevations, and the pattern of steps and risers from the
enhanced ADAR-5500 images indicate less variation in the semivariance with elevation than represented in the field transects.
D 2003 Elsevier Science B.V. All rights reserved.

Keywords: Solifluction patterns; Remote sensing; Principal components analysis; Wavelet transform; Semivariograms; Scale dependence

1. Introduction steps and risers on Lee Ridge, Glacier National Park,

Montana, USA, through a research design that inte-
The basic intent of this research was to examine the grated field techniques and spatial digital technologies
spatial pattern and scale dependence of solifluction with an emphasis on global positioning systems,
remote sensing, and geostatistics. The design incorpo-
rated (i) field data at twelve 50-m transects collected as
* Corresponding author. Tel.: +1-919-962-3867; fax: +1-919-
one-dimensional control data for the solifluction steps
962-1537. and risers; (ii) ADAR-5500 digital aircraft data to map
E-mail address: [email protected] (S.J. Walsh). the two-dimensional patterns of the steps and risers for

0169-555X/03/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0169-555X(03)00151-X
382 S.J. Walsh et al. / Geomorphology 55 (2003) 381–398

the areas corresponding to the twelve 50-m field 2. Background

transects through a combination of digital enhance-
ment techniques; (iii) indicator semivariograms to Glacier National Park is a very appropriate labo-
assess the variance of the steps and risers for specified ratory to assess the application of remote sensing for
sampling ranges for the one- and two-dimensional characterizing the alpine environment. Prior research
analyses; and (iv) the pattern and scale dependent in the Park, conducted by the Mountain GeoDynamics
relationships found in the transect and ADAR data Research Group, has tended to focus on (i) resource
were interpreted relative to the distance from Gable assessments such as snow avalanche paths and debris
Mountain, a surrogate for elevation variation along flows (e.g., Butler and Walsh, 1990; Walsh et al.,
Lee Ridge, as a preliminary study of the morphometric 1994a; Walsh and Butler, 1997), assessment of lake
controls of solifluction patterns. The research design turbidity and basin morphometry (e.g., Brown and
was developed to assess the effectiveness of remote Walsh, 1991, 1992), identification of deltaic wetlands
sensing methodologies in characterizing solifluction (e.g., Butler et al., 1991a,b), and evaluation of forest
steps and risers over extensive and inaccessible areas fire potential (e.g., Butler et al., 1991a,b); (ii) exami-
where in situ data collection may be restricted. nation of the scale dependence of plant biomass and
The following describes the data collection proto- topography (e.g., Bian and Walsh, 1993; Walsh et al.,
cols implemented, analyses performed, and findings 1997); and (iii) characterization of the alpine tree line
and interpretations derived. Beginning with a descrip- ecotone (e.g., Allen and Walsh, 1993, 1996) and the
tion of Glacier National Park, Montana, USA, as the development of cartographic, empirical, and mecha-
regional context, Lee Ridge is described as the specific nistic models that assess the variability of the alpine
setting of the research. Literature is described that tree line ecotone over space and time scales (e.g.,
emphasizes some of the prior applications of remote Walsh et al., 1990a,b, in press; Brown et al., 1994).
sensing in Glacier National Park for mapping alpine Currently, the use of spatially explicit digital technol-
tree line patterns and their relationships to biophysical ogies are being examined to study the effects of
controls, as well as the use of geostatistical analyses geomorphic processes and patterns influencing the
for characterizing their scale dependence. The mission spatial organization and variability of the alpine tree
and system specifications of the ADAR-5500 digital line ecotone that extends from closed canopy forest to
aircraft data are described, as well as the pre- and post- open canopy forest, meadow-tree complexes, tundra,
processing that were performed to geometrically cor- and rocks, snow, and ice surfaces (e.g., Walsh et al.,
rect the data and transform the row/column coordi- 1994b). The central issues of the research involve the
nates of the 1  1-m pixels to Universal Transverse sensitivity of tundra to invasion by trees, i.e., the
Mercator (UTM) coordinates. The digital enhance- ecotone as an indicator of climate change (e.g.,
ment techniques applied to the ADAR data [i.e., Malanson et al., 2000). Prior work has illustrated
principal components analysis (PCA) and wavelet how the elevation and position of alpine tree line in
transform (WTF)] for mapping the solifluction steps the Park is constrained in many locations by lithologic
and risers are then considered. A description of the and structural controls, geomorphic processes—the
collection and analysis of field transect data follows. locations and nature of which are influenced by
Analyses are described that involve the assessment of lithology and structure, topographic site conditions
solifluction patterns from the field transects and the and associated soil and moisture conditions, and
linking of that data to the enhanced remote sensing patterns of snow distribution. Remote sensing, GIS,
images. The scale dependence of the transect data (i.e., spatial analysis and modeling, and data visualizations
one-dimensional analysis) and the remote sensing support the analyses. The work reported here is a
images (i.e., two-dimensional analysis) are then natural continuation of that research.
assessed through a discussion of indicator semivario-
grams and their use for defining the ranges of spatial 2.1. Study area
scales that are autocorrelated, periodic frequencies and
amplitudes, and the breaks in the slope of the semi- Glacier National Park, a UNESCO-designated Bio-
variogram in which independence is suggested. sphere Reserve, is located in NW Montana, USA,
 S.J. Walsh et al. / Geomorphology 55 (2003) 381–398 383

bordering Canada along the Park’s northern boundary. steps and risers that extend throughout the ridge. The
The Park is f 400,000 ha in size, has peaks with solifluction steps and risers are characterized by com-
elevations of nearly 3200 m, and sits astride the pacted terrace surfaces supporting large-diameter Xan-
Continental Divide, which follows the Livingston thoria elegans lichens, indicative of the relict nature of
Range in the north and the Lewis Range in the east the solifluction processes occurring on the ridge. The
(Fig. 1). The Lewis Overthrust is the dominant struc- vegetation of the ridge is dominated by a dense canopy
tural feature in the Park. It is responsible for the of lodgepole pine (Pinus contorta) at lower elevations,
emplacement of Precambrian Belt Series formations grading to an open canopy of lodgepole pine and
overtop of relatively incompetent Cretaceous shales localized patches and forested fingers of lodgepole
and mudstones. Areas above tree line in the Living- pine, subalpine fir (Abies lasiocarpa), Engelmann
ston Range tend to be steeper and geomorphically spruce (Picea engelmannii), and five-needle pines
more active, with little colluvial cover. (i.e., Pinus albicaulis and/or Pinus flexilis). On the
Lee Ridge, the study area for this research, is a lower portions of Lee Ridge, upright trees appear to
north-facing slope located in the NE portion of Glacier have invaded a former expanse of tundra, and similar
National Park, and hence on the east side of the patterns are observed at higher places on the ridge as
Continental Divide. The ridge is comprised of soli- well. Solifluction treads and risers appear to increase
flucted colluvium derived primarily from the Altyn in significance with elevation—their distinctiveness,
Formation that forms the cliffs of Gable Mountain, just while evident throughout the ridge, appears to become
above and to the south of Lee Ridge. Relict solifluc- more striking with increasing elevation in which the
tion lobes and terraces dominate the study area. Fig. 2 width and clarity of alternating sequences of risers
is a helicopter view of the ridge from the NE that (vegetation material, particularly Dryas octopetala) in-
indicates the widespread nature of the solifluction creases between f 1825 to 2150 m, is oriented from

Fig. 1. Regional study area location: Glacier National Park, Montana, USA.
384 S.J. Walsh et al. / Geomorphology 55 (2003) 381–398

Fig. 2. Specific study area location: Lee Ridge, Glacier National Park.

N –S with increasing elevation towards Gable Moun- active frost shattering above tree line (Butler and
tain, and extends f 5 km from the closed canopy Malanson, 1999), whereas the turf-banked terraces
forest south to Gable Mountain. (Hansen-Bristow and Price, 1985) were attributed to
Bamberg and Major (1968) provided a detailed varying severities of gelifluction frost creep during the
description of the periglacial landforms, soils, and postglacial (Benedict, 1970, 1976).
alpine vegetation in the Park, focussing on Siyeh Pass
and the vegetation and soils associated with calcare-
ous parent material. Choate and Habeck (1967) and 3. Data collection
McGregor (1998) provided a description of the veg-
etation of the tundra environment in Glacier National During the summer of 1999, ADAR-5500 digital
Park. Butler and Malanson (1989) described patterned aircraft data were acquired for a number of test sites
ground in Glacier National Park as a surrogate indi- distributed along the Continental Divide in Glacier
cator of Holocene climate change (Carrara, 1987; National Park, Montana, and its eastern slopes. The
Dutton and Marrett, 1997). They reported on three data collection mission was designed to support a
populations of patterned groups: (i) miniature-sorted broad set of research goals associated with the study
polygons; (ii) currently inactive alpine turf-banked of alpine tree line in the Park and its hypothesized
terraces, active during the most recent stage of the geomorphic, biogeographic, and lithologic controls.
Neoglacial (Gardner and Jones, 1985); and (iii) larger ADAR data were acquired for Lee Ridge, a tree line
turf-banked terraces located slightly below present- test site that was previously instrumented and assessed
day tree line and in well-sheltered Pleistocene cirques. by members of the Mountain GeoDynamics Research
The sorted polygons were attributed to currently Group, a team of research collaborators from US
 S.J. Walsh et al. / Geomorphology 55 (2003) 381–398 385

universities with links to the Biological Resources points were collected from USGS Digital Orthophoto
Division of the U.S. Geological Survey. Quarter Quadrangles (DOQQs). Features present on
During the summer of 2001, field data were the ADAR images and the DOQQs were identified
collected on Lee Ridge through a set of twelve 50- and used for geometric rectification. An existing 7.5-
m transects that were used to record the start and stop min, 30-m USGS digital elevation model (DEM) was
distances of solifluction steps and risers along the also used for graphical display of image data and as
longitudinally positioned transects, which were organ- the basis for illumination corrections to the ADAR-
ized into general elevation zones that ranged from the 5500 data.
highest to lowest regions of Lee Ridge. Below, we Specification of the aerial overflight included (i)
describe transects and specifics of the remote sensing stereo coverage through a nominal forward lap of 60%
data collection mission and the specifications of the and a nominal side lap of 30%, (ii) < 10% clouds and
ADAR-5500 sensor system. shadows at the time of image acquisition, (iii) a high
sun acquisition period of F 1.5 h of solar noon, (iv)
3.1. ADAR-5500 digital aircraft imagery mean flying height of 2225 m above the ground, (v)
exposure setting of ‘‘maximum contrast’’ using tundra
Solifluction steps and risers were mapped through as the focal target, and (vi) on-board sensor specifi-
an acquired set of geometrically and radiometrically cations set for a ‘‘camera model’’ correction that
rectified ADAR-5500 digital aircraft images. The incorporated sensor specifications and flight parame-
ADAR system operates in four spectral channels ters in image rectification.
extending from the visible into the near-infrared The ADAR-5500 system is a second generation,
wavelengths. A spatial resolution of 1  1-m was used charge-couple device frame camera system. It oper-
to characterize the study area through a series of ates in four channels extending from the visible to the
image frames measuring 1500  1000 pixels in extent. near-infrared spectral regions. Specifically, channel 1
A combination of the instantaneous-field-of-view extends from 460 to 550 nm, channel 2 from 520 to
(IFOV) and the flying height above the ground of a 610 nm, channel 3 from 610 to 700 nm, and channel 4
sensor determine the spatial resolution, which is from 780 to 920 nm. The across-track field of view
defined as the smallest distinguishable spatial unit was 39j, and the radiometric resolution was 8 bits.
recorded on an image or the ground resolution ele- Postflight processing performed (i) a vignette correc-
ment of the sensor (Walsh et al., 2003). tion for exposure variation imposed by the internal
The individual frames of ADAR were combined effects of the sensor and its optics, (ii) channel to
into a seamless image mosaic through the use of a channel co-registration of the four acquired multi-
geodetic control network acquired throughout the spectral images, and (iii) an ERDAS Imagine file
study area and integrated using ERDAS Imagine format was specified for direct importation of the
software. An on-board GPS combined with UTM acquired data into our image processing software.
coordinates were used to reference the imagery to a Following the application of standard geometric
geographic base for preprocessing and subsequent and radiometric corrections to the ADAR data to
analyses. UTM coordinates were secured from reduce the effects of geometric errors, terrain-imposed
ground-based rover units and collected at rock out- illumination variance, and atmospheric inequalities
crops and semipermanent snow patches, trail inter- (Walsh et al., 1998; Jensen, 2000), the ADAR
sections, survey markers, and constructed and imagery was processed to digitally enhance the soli-
distributed registration panels (i.e., 3  3-m white fluction patterns on Lee Ridge. Principal components
plastic panels placed on the ground in a ‘‘cross-hair’’ analysis (PCA) and wavelet transform (WLT) were
pattern). The UTM coordinates collected by the GPS applied as image enhancement approaches.
units were differentially corrected through the use of a
nearby base station that improved the spatial accuracy 3.2. Principal components analysis
of the data to F 5 m and corrected the spatial align-
ment of the ADAR imagery to f 0.5 pixel, or 0.5 m Principal components analysis (PCA) is a data
through the georeferencing procedure. Ground control compression and enhancement technique that gener-
386 S.J. Walsh et al. / Geomorphology 55 (2003) 381–398

ates independent, orthogonal axes using the ADAR three visible channels (blue – green – red, respectively)
spectral channels as input. The PCA was calculated and the near-infrared channels of the ADAR-5500
using a covariance matrix derived from the four sensor to the eigenvectors of the four principal
channels of ADAR data. The eigenvalue for PCA 1 components.
was 5041.77, which accounted for 94.30% of the
overall image variance; PCA 2 had an eigenvalue of 3.3. Wavelet transform
218.67 (4.09% of the overall variance); PCA 3 had an
eigenvalue of 80.38 (1.50% of the overall variance); Wavelet transform (WLT) was used to enhance
and PCA 4 had an eigenvalue of 5.47 (0.12% of the both ‘‘trend’’ and ‘‘detail’’ of image properties (e.g.,
overall variance). Although PCA 4 and 3 accounted shape, tone, texture, and pattern) expressed as numeric
for very small portions of the overall variance, they indices (e.g., mean, standard deviation, and number of
provided the best characterization of the solifluction edge pixels) that were applied as multilevel vertical
steps (i.e., rock) and risers (i.e., vegetation) occurring and horizontal transforms. Fig. 3 shows how the first
on Lee Ridge, respectively. Low variance accounted and second horizontal and vertical transforms are
for by the derived components does not suggest the generated from the original image using a simple Harr
absence of information. In this case, the solifluction wavelet transform for a 4  4-pixel image. Com-
steps and risers represented relatively little of the monly, the Harr and Daubechies transforms are the
overall image variance, controlled primarily by topo- two types of WLTs applied. With the simplest Harr
graphic variation and solar angle, and the lingering wavelet, the variability of the image is decomposed
presence of snow patches that occurred in gullies and into three directional subimages—vertical, horizontal,
other partially shadowed areas in 1999, an above- and diagonal. In Fig. 3 transform, two directional
average snow year in the Park. filters are applied to the image. First, the horizontal
While the eigenvalues indicate the amount of filter averages every nonoverlapping pair of adjacent
variance explained by each component, the eigenvec- pixels along the horizontal direction. The average
tors relate the input spectral channels to each of the values form the approximation subimage. The differ-
derived principal components, which are sign inde- ences between the original and the average values form
pendent (Walsh et al., 1990b). Eigenvectors for the a detail subimage. The vertical filter is then applied to
four principal components are presented in Table 1. both the approximation and the detail subimages to
By relating the ADAR channels to vegetation char- produce four subimages, denoted as the LL, HL, LH,
acteristics that are most responsible for influencing and HH subimages, respectively. Of the four subim-
spectral responses in particular wavelengths, defining ages, the LL subimage is the approximation of the
spectral – biophysical relationships suggested by the original image, while the remaining three subimages
eigenvectors is relatively simple. For instance, the are details in horizontal, vertical, and diagonal direc-
visible wavelengths are most associated with plant tions, respectively. This WLT is applied again to the
pigmentation, the near-infrared wavelengths to chlor- approximation subimage resulting in four additional
ophyll content, and the middle-infrared wavelengths quarter-size second-level subimages (Bian, 2003).
(not represented on the ADAR-5500 sensor system) Wavelet transforms were developed as a frame-
to moisture content of the leaf. Table 1 relates the first work for decomposing images into their component
parts at different resolutions. Mallat (1989a,b) used
Table 1 the WLT to reorganize image information into a set of
Eigenvectors and the associated ADAR channels for the derived detailed characteristics. Within the context of wavelet
principal components decomposition, Csillag and Kabos (1996) and Csillag
ADAR Principal components (1997) applied quadtree image procedures for map-
channels
1 2 3 4 ping a complex landscape; they indicated how the
1 0.563 0.110 0.344 0.742 relationship between mapping units and spatial vari-
2 0.540 0.006 0.526 0.655 ability could be assessed. Li and Shao (1994) used
3 0.623 0.055 0.772 0.106 wavelet to enhance the edges of image features and to
4 0.031 0.992 0.085 0.084
extract object characteristics from the derived image.
 S.J. Walsh et al. / Geomorphology 55 (2003) 381–398 387

Fig. 3. Wavelet transform: schematic of multilevel iteration of horizontal and vertical transforms.

Fig. 4. Position of the twelve 50-m field transects on Lee Ridge.

388 S.J. Walsh et al. / Geomorphology 55 (2003) 381–398

Yocky (1996) applied wavelet decomposition to set of three, three, two, two, and two transects
merge satellite image data from sensors of differing positioned in a descending elevational pattern in each
spatial resolutions, and Mohanty (1997) applied WLTs of five elevation zones. In general, the upper slopes of
for general image enhancement. Here, we apply the Lee Ridge are the steepest, the intermediate slopes are
WLTs to the derived principal components found to be less steep, and the lowest slopes are least steep. The
most effective in the enhancement of the solifluction highest areas of Lee Ridge have the widest solifluc-
steps and risers (i.e., PCA 4 and 3) that occur on Lee tion extents, and so three transects were collected at
Ridge as an approach for further emphasizing step and the higher elevations and two transects at the lower
riser edges and overall feature trends. sections of Lee Ridge. At the ends of each of the 50-m
transects, a GPS unit was used to geographically
3.4. Field transects of solifluction patterns reference their position using UTM coordinates. The
coordinates were gathered so that the position of
During the summer of 2001, a field team posi- transects could be superimposed on the ADAR image
tioned and then measured the distance and sequential data as well as on other GIS coverages of Lee Ridge
pattern of alternating solifluction steps and risers for subsequent analyses. The run distance of the
along twelve 50-m transects using a line intercept alternating solifluction steps and risers was noted by
approach. The transects were longitudinally arrayed recording their start/stop distances on the measure-
parallel to the slope of Lee Ridge and organized as a ment tape. Fig. 4 shows the position of the 12 field

Fig. 5. The third and fourth principal components of the ADAR-5500 spectral channels.
 S.J. Walsh et al. / Geomorphology 55 (2003) 381–398 389

transects on a digital elevation model of Lee Ridge. wave pattern, which has been enhanced through
The transects are referenced as 1A – 1B – 1C for the interpretive tracings for the sake of presentation.
uppermost sites; followed by 2A – 2B –2C, 3A –3B, Fig. 7 shows transects 1A, 1B, and 1C overlaid on a
4A – 4B, and 5A – 5B in a descending elevational PCA composite of components 4, 2, and 1. The
sequence extending to the lowest elevations of the solifluction steps and risers are well represented at
ridge where solifluction patterns still persisted. the highest transect sample zone and less so at
successively lower sites because of the reduced sizes
of the steps and risers and the shift in the orientation
4. Data analysis of the solifluction pattern relative to the acquisition
angle of the ADAR-5500 imagery.
4.1. Digital enhancements of ADAR-5500 imagery
4.2. Field transects of solifluction patterns
Two digital enhancements of the ADAR-5500 air-
craft imagery were performed. Fig. 5 presents the Figs. 8 and 9 are box plots of the transects that
third and fourth principal components, the compo- indicate intraclass variability and interclass differen-
nents that best characterized the solifluction steps and ces for vegetation (the risers) and rocks (the steps),
risers. The wavelet transform of principal components respectively. The boxes represent the interquartile
3 and 4 were not sensitive to the step/riser edges at the distance that is the distance between the first and
1-m scale and therefore did not improve upon the third quartiles. The line in the middle of the box
derived PCA images as a stand-alone enhancement. represents the median of the data. The whiskers are
However, the WLTs did enhance a coarser wave 1.5 times the interquartile distance. Anything above or
pattern on which the finer-grained steps and risers below that value is considered an outlier and is
are composited. This coarser wave pattern was pre- indicated by an ‘‘x.’’ The vertical axis is the distance
viously unidentified. Fig. 6 shows this newly defined in meters of the extent of the solifluction steps and

Fig. 6. Wavelet transforms of the third and fourth principal components for the ADAR-5500 spectral channels; interpretive lines added for
presentation.
390 S.J. Walsh et al. / Geomorphology 55 (2003) 381–398

Fig. 7. Overlay of field transects 1A – 1B – 1C (highest sample zone on Lee Ridge) on a PCA 4-2-1 composite.

risers, while the horizontal axis represents the twelve image scale on landscape characterization. Collins and
50-m transects arrayed from highest to lowest eleva- Woodcock (1999) suggested using a variance estima-
tions (i.e., left to right on the page). Transects are tion to better retain resolution-dependent variance as
dominated by risers, where variability is also more coarser images are derived from finer images. Walsh et
pronounced than that associated with the steps. al. (1997) suggested that subimage statistics are useful
in discerning image characteristics at successively
4.3. Semivariograms of solifluction patterns from the degraded images—a series of images known as image
transects pyramid layers because they are built by degrading the
resolution at level n 1 to generate level n.
Previous research has reported on the importance of Lam and Quattrochi (1992), Bian and Walsh (1993),
spatial scale and the use of semivariance and spatial Xia and Clarke (1997), and Walsh et al. (1997) used
autocorrelation for the characterization of scale across a fractal analysis for assessing the effects of scale on
sample range. For example, Marceau et al. (1994), Bian landscape analyses involving the degrading of the
and Butler (1999), and Collins and Woodcock (1999) spatial resolution of remotely sensed data. Spatial
reported on the importance of spatial resolution and autocorrelation is an important factor in considering
 S.J. Walsh et al. / Geomorphology 55 (2003) 381–398 391

Fig. 8. Box plots of the solifluction risers for the field transects on Lee Ridge.

spatial resolution and image models. The semivario- the slope and the breaks in the slope of the plotted
gram is the most commonly used measure for assess- semivariance (Curran, 1988; Jupp et al., 1989a,b; Bian
ing the nature of spatial autocorrelation by considering and Walsh, 1993). Goovaerts (1997) and de Bruin

Fig. 9. Box plots of the solifluction steps for the field transects on Lee Ridge.
392 S.J. Walsh et al. / Geomorphology 55 (2003) 381–398

(2000) used categorical indicator semivariograms to nounced on the landscape with elevation and less well
characterize the spatial pattern of land cover. Here, we characterized by the enhanced remote sensing images
also use indicator semivariograms to describe the even at a 1-m spatial resolution. As a consequence of
frequency of transitions between steps and risers. degrading the spatial resolution from the 1-m base
Because the twelve 50-m transects were organized scale, the solifluction pattern becomes even less dis-
into five groups in which the field transects were tinct. As the width and distinctiveness of the steps and
positioned and systematically located on Lee Ridge risers degrade with elevation, the ability to discern the
from upper to lower slopes, similarities in the semi- pattern of steps and risers is also reduced.
variance within each group and dissimilarities between
groups were examined. The semivariance was com- 4.4. Semivariograms of ADAR imagery of solifluction
puted by plotting the lagged distance of steps and risers patterns
using a 0.1-m resolution to summarize the actual run-
lengths derived from each of the 50-m transects. Semivariograms were computed for the areally
Fig. 10 shows plots of the semivariance for transects bounded regions centered on the five groups of field
1A, 3A, and 5A, representing the highest, intermediate, transects. These areas were extracted and the semi-
and lowest elevations on Lee Ridge, respectively. In variograms computed to assess the variance in soli-
general, the semivariance decreases with elevation, as fluction patterns associated with the steps and risers
does the amplitude of the wave and its wavelength. The when characterized at a range of scales 1 –20 m using a
solifluction pattern of steps and risers is less pro- 1-m interval. Fig. 11 shows the plots of the semi-

Fig. 10. Semivariograms for transects 1A, 2A, 3A, 4A, and 5A (representing the highest, intermediate, and lowest sample locations) on Lee Ridge.
 S.J. Walsh et al. / Geomorphology 55 (2003) 381–398 393

Fig. 11. Semivariograms for ADAR-5500 images for elevation zones 1A – 1B – 1C, 2A – 2B – 2C, 3A – 3B, 4A – 4B, and 5A – 5B (representing
the highest, intermediate, and lowest sample areas) on Lee Ridge.

variance against the 1-m sampling interval, extending 2A – 2B – 2C) and for the two intermediate transect
to a 20-m pixel. Semivariograms are only presented for groups (3A – 3B and 4A – 4B); the lowest transect
the highest, intermediate, and lowest elevations cen- group (5A –5B) is most dissimilar. The plots suggest
tered on transects 1A – 1B –1C, 3A –3B, and 5A –5B, that (i) the mean distance of the risers is greater than the
respectively. The semivariance was computed for all mean distance of the steps for the two uppermost
PCA images, but only component 4 is presented—the transect groups, perhaps reflecting the steeper slopes
one that best enhanced the solifluction steps and risers. associated with these transects; the percent of the
In general, the plots appear similar in form and pattern transects for this group is dominated by risers, which
regardless of the elevation zone that they represent. also shows a greater variance in the distance of the
risers. (ii) The transects for the two intermediate groups
indicate that the pattern of the steps and risers vary
5. Discussion across Lee Ridge from W – E. The transects located
closer to the western escarpment that topographically
5.1. Semivariograms of transect solifluction patterns defines Lee Ridge have generally more step/riser seg-
ments over the 50-m length of the transect—the trans-
A review of the step and riser lengths (Figs. 8 and 9) ects are dominated by risers that have a larger mean
indicates a similarity in the step and riser patterns for distance and a greater variance; the transects located
the two uppermost transect groups (1A –1B – 1C and further away from the escarpment show a decrease in
394 S.J. Walsh et al. / Geomorphology 55 (2003) 381–398

the number of segments represented over the transect ess. The wavelength of the middle transects, which is
length and an increase in the mean distance and approximately equal to the resolution of the sensor
variance. (iii) The lowest transects (5A – 5B) have the and, therefore, the variograms calculated from the
fewest step/riser segments for the 50-m transect length, images for those slopes, do not exhibit the strong
have the largest mean distance and variance of the periodicity that the upper and lower slopes do. This
risers, and have the highest percentage of the transect result is consistent with the general theory of reso-
characterized by risers; distance from the escarpment lution in remote sensing, which states that objects
and the mean distance of the riser segments increases as must be at least twice the size of the image resolution
well as the percent of the transect dominated by risers. to be detectable on the image.
Because of the site and situation of Lee Ridge, the As part of the digital enhancement research, we
distance from the western escarpment of Lee Ridge applied the WLTs on the third and fourth principal
indicates a decrease in the distance to upright trees components, the components that best represented the
retained in forested patches and fingers. Also, related to solifluction steps and risers. We hypothesized that the
distance from the escarpment is a decrease in the E – W- transform would further accentuate the trend in the
trending elevational gradient, the most dominated solifluction pattern as well as the detail in the areal
gradient being S –N from Gable Mountain. extent of the steps and risers across Lee Ridge. What
The semivariograms for the five elevation zones we found was something else—a coarser grained trend,
and 12 transects (Figs. 10 and 11) all exhibit a actually a set of repetitive waves on which the finer-
periodic structure of alternately increasing and grained steps and risers were superimposed. The
decreasing semivariance with increasing lag distance. coarse-grained trends appeared to be about 100 m in
The periodic structure suggests the lack of a single width and arranged in a harmonic pattern that proceeds
characteristic scale of pattern of variation, but also downslope, oriented from S – N. The same pattern
that the scale of the patterns in the treads and risers appears on the adjacent basin that generally faces west.
can be characterized by the wavelength of the periodic These waves tend to bend down slope, with the crest of
structure. The differences among variograms suggest the wave arching slightly upslope (Fig. 8). Hypotheses
that the total variance in the variograms decreases for the observed pattern include the presence of stream
with decreasing elevation. On average, the wave- channels, a pulsing green wave associated with the
length, or the distance between similar components melting of late-lying snow, repetitive pattern of relict
(e.g., from tread to tread or from riser to riser) on the landslides, underlying pattern of cyclicity within the
uppermost slopes, is about 4 m, but on the lowermost bedrock, larger ancient solifluction activity, structural
slopes, the wavelength is about one-half that distance. controls with the intersection of the topographic slope
The middle slopes have a less well-developed periodic and dip (D.R. Butler, Southwest Texas State Univer-
structure, but at greater lag distances, the wavelength sity, and G.P. Malanson, University of Iowa, personal
appears to be about 1 m. The wavelength may well be communication, 2001), and artifacts of our processing.
related to the energy available for the solifluction While only speculative at this point, the pattern does
process, which is determined by the slope steepness. appear to be real. Whether viewed from the vertical,
horizontal, or a vertical – horizontal composite, the
5.2. Semivariograms of ADAR imagery solifluction pattern persists and is quite distinctive. Using a 30-m
patterns DEM of Lee Ridge and altering its representation to
view the ridge in a profile mode, the undulating and
The variograms calculated from the PCA 4 trans- scalloped pattern of the ridge is consistently apparent
form of the ADAR imagery (Fig. 11) indicate a strong across the ridge. This suggests the occurrence of a
degree of correspondence with those based on the number of intersecting and overlapping patterns that
field transects, especially for the upper- and lower- exist on Lee Ridge and vicinity. Additional analyses are
most transects. This result suggests that imagery of in progress that will further examine the semivariance
sufficient resolution, processed appropriately, can be structure observed in the wavelet-enhanced principal
used to successfully map the spatial structure of the component images. The pattern is reported here
treads and risers resulting from the solifluction proc- because of the demonstrated relevance of WLTs in
 S.J. Walsh et al. / Geomorphology 55 (2003) 381–398 395

Fig. 12. Mean variance of the field transects plotted against the mean transect distance from the peak, Gable Mountain.

defining scale dependent patterns associated with geo- remote sensing resolutions—spatial, spectral, tempo-
morphic activities, in the case of Lee Ridge, not ral, and radiometric. Here, we used a high spatial
previously reported. resolution sensor and acquisition systems, the ADAR-
5500 sensor, and subsequently degraded the image
5.3. Semivariance and terrain relationships through spatial resampling from its base spatial reso-
lution of 1– 20 m using a 1-m sampling interval. More
To examine preliminary relationships between the and more high spatial resolution data sets acquired
semivariance of the twelve 50-m transects and terrain from both aircraft and satellite systems are becoming
characteristics, the mean (and standard deviation) available to researchers. For example, Ikonos and
distance from the peak, Gable Mountain, was com- Indian Remote Sensing (IRS) satellite systems now
puted for each of the twelve transects. This variable is complement the ADAR-5500 system in their spatial
a proxy for elevation change—Lee Ridge is oriented and spectral resolutions, and Digital Orthophoto Quar-
from south to north from Gable Mountain in a pattern ter Quadrangle (DOQQ) data have become another
of decreasing elevation. Fig. 12 shows the mean important airborne source of information. The inter-
semivariance of transects plotted against the mean play between the size and complexity of the target
distance from Gable Mountain, the central peak at the feature, the type of the environment, kinds of infor-
upper elevations of Lee Ridge. The trend of the point mation to be extracted, and the image analysis tech-
distribution indicates a decrease in the mean semi- niques used all impact the selection of the appropriate
variance with an increasing distance from the peak, spatial resolution for landscape analyses.
thereby suggesting a finer grain pattern and structure On Lee Ridge, the solifluction steps and risers are
of the steps and risers at higher elevations. striking in their areal extent and persistent in their
pattern. They exhibit patterns formed under relict
conditions. As part of the program of research being
6. Conclusions conducted by the Mountain GeoDynamics Research
Group, we are attempting to discern the scale – pat-
One of the fundamental considerations in the use of tern –process relationships in Glacier National Park
remote sensing for landscape studies involves the four that are associated with the alpine tree line ecotone.
396 S.J. Walsh et al. / Geomorphology 55 (2003) 381–398

Traditionally, we have relied upon Landsat Thematic Acknowledgements

Mapper data (30-m pixels), and to a lesser extent,
SPOT Panchromatic (10-m pixel) and Multispectral Appreciation is expressed to the U.S. Geological
(20-m pixel) data, in addition to air photo interpreta- Survey, Biological Resources Division (Daniel Fa-
tion, for characterizing the alpine tree line. Higher gre), for the support of this research, and to the
spatial resolution data, such as ADAR-5500, offer research team of the Mountain GeoDynamics Re-
views of the landscape previously relegated to in situ search Group.
data collection, ground- and aircraft-based videogra-
phy, and occasionally from analog aerial photography.
Field work conducted in remote and in hospitable References
environments offers its own set of challenges and
limitations, and the use of previously acquired aerial Allen, T.R., Walsh, S.J., 1993. Characterizing multitemporal alpine
photography may be plagued by inappropriate mission snowmelt patterns for ecological inferences. Photogrammetric
specifications including the date and time of image Engineering and Remote Sensing 59 (10), 1521 – 1529.
Allen, T.R., Walsh, S.J., 1996. Spatial and compositional structure
acquisition, photo scale, and spectral sensitivities.
of the alpine treeline ecotone. Photogrammetric Engineering and
The ADAR-5500 data used in this research af- Remote Sensing 62 (11), 1261 – 1268.
forded us an opportunity to collected data to service a Bamberg, S.A., Major, J., 1968. Ecology of the vegetation and soils
diversity of studies, mostly clustered around biogeo- associated with calcareous parent materials in three alpine re-
graphic and geomorphic relationships centered on the gions of Montana. Ecological Monographs 38, 127 – 167.
alpine environment. The spatial and spectral resolu- Benedict, J.B., 1970. Downslope soil movement in a Colorado
alpine region: rates, processes, and climatic significance. Arctic
tions of ADAR give us the opportunity to consider the and Alpine Research 2, 165 – 226.
study area in substantial spatial detail and in spectral Benedict, J.B., 1976. Frost creep and gelifluction features: a review.
resolutions that have a proven history of performance. Quaternary Research 6, 55 – 76.
The application of remote sensing technology to the Bian, L., 2003. Retrieving urban objects using a wavelet trans-
study of pattern and process relationships have form approach. Photogrammetric Engineering and Remote Sens-
ing 69 (2), 133 – 142.
allowed us to map broad geographic areas and to Bian, L., Butler, R., 1999. Comparing effects of aggregation methods
apply image processing methods not customarily used on statistical and spatial properties of simulated spatial data. Pho-
in geomorphology. Newly emergent high spatial res- togrammetric Engineering and Remote Sensing 65 (1), 73 – 84.
olution, satellite systems such as Ikonos and QUICK- Bian, L., Walsh, S.J., 1993. Scale dependencies of vegetation and
topography in a mountainous environment of Montana. The
BIRD further offer geomorphologists the capability to
Professional Geographer 45 (1), 1 – 11.
extend field measurements into remote locations and Brown, D.G., Walsh, S.J., 1991. Compatibility of non-synchronous
to acquire alternate views of landscape processes and in-situ water quality data and remotely sensed spectral informa-
feature distributions by acquiring data from orbital tion for assessing lake turbidity levels in complex and inacces-
systems that offer variations in solar illumination and sible terrain. GeoCarto International 6 (2), 5 – 11.
landscape phenologies. Brown, D.G., Walsh, S.J., 1992. Relationship between the morph-
ometry of alpine and sub-alpine basins and remotely sensed
This paper was designed to explore the pattern estimates of lake turbidity, Glacier National Park, Montana,
analysis capabilities of remote sensing, accentuated USA. Physical Geography 13 (3), 250 – 272.
through image enhancement applications, and the use Brown, D.G., Cairns, D.M., Malanson, G.P., Walsh, S.J., Butler,
of semivariance for examining the structure of patterns D.R., 1994. Remote sensing and GIS techniques for spatial and
across spatial scales. Follow-on papers will emphasize biophysical analyses of alpine treeline through process and em-
pirical models. In: Michener, W.K., Brunt, J.W., Stafford, S.G.
the geomorphic significance and biophysical controls (Eds.), Environmental Information Management and Analysis.
on the mapped pattern of solifluction steps and risers. Taylor and Francis, London, pp. 453 – 481.
As used here, integrated science and integrated meth- Butler, D.R., Malanson, G.P., 1989. Periglacial patterned ground,
odologies offer new landscape views through scale and Waterton-Glacier International Peace Park Canada and USA.
pattern relationships. Linking processes to the new Zeitschrift fur Geomorphologie 33 (1), 43 – 57.
Butler, D.R., Malanson, G.P., 1999. Site locations and character-
image characterizations offer challenges as well as istics of miniature patterned ground, eastern Glacier National
opportunities that geomorphologists are examining in Park, Montana, USA. Landform Analysis 2 (1), 45 – 49.
mountain environments and elsewhere. Butler, D.R., Walsh, S.J., 1990. Lithologic, structural, and topo-
 S.J. Walsh et al. / Geomorphology 55 (2003) 381–398 397

graphic controls of snow-avalanche path location, eastern Gla- Lam, N.S.-N., Quattrochi, D.A., 1992. On the issues of scale, res-
cier National Park, Montana. Annals of the Association of olution, and fractal analysis in the mapping sciences. Professio-
American Geographers 80 (3), 362 – 378. nal Geographer 44 (1), 88 – 98.
Butler, D.R., Malanson, G.P., Walsh, S.J., 1991a. Identification of a Li, D., Shao, J., 1994. Wavelet theory and its application in image
deltaic environment in an alpine finger lake. Environmental edge detection. International Journal of Photogrammetry and
Professional 13, 352 – 362. Remote Sensing 49, 4 – 12.
Butler, D.R., Walsh, S.J., Malanson, G.P., 1991b. GIS applica- Malanson, G.P., Xiao, N., Alftine, K., Bekker, M., Butler, D.R.,
tions to the indirect effects of forest fires in mountainous Brown, D.G., Cairns, D.M., Fagre, D., Walsh, S.J., 2000.
terrain. In: Nodvin, S.C., Waldrop, T.A. (Eds.), Fire and the Abiotic and biotic controls of spatial pattern at alpine tree-
Environment: Ecological and Cultural Perspectives. Proceed- line. Proceedings GIS and Environmental Modeling, Banff,
ings of an International Symposium, USDA Forest Service, Canada. University of Colorado, CIRES, USA.
Southeastern Forest Experiment Station General Technical Re- Mallat, S., 1989a. A theory for multiresolution signal decomposi-
port SE-69. Knoxville, TN, pp. 202 – 211. tion: the wavelet representation. IEEE Transactions on Pattern
Carrara, P.E., 1987. Holocene and latest Pleistocene glacial chro- Analysis and Machine Intelligence 11 (7), 674 – 693.
nology, Glacier National Park, Montana. Canadian Journal of Mallat, S., 1989b. Multifrequency channel decompositions of im-
Earth Sciences 24, 387 – 395. ages and wavelet models. IEEE Transactions on Acoustics,
Choate, C.M., Habeck, J.R., 1967. Alpine plant communities at Speech, and Signal Processing 37 (12), 2091 – 2110.
Logan Pass, Glacier National Park. Proceedings of the Montana Marceau, D.J., Howarth, P.J., Gratton, D.J., 1994. Remote sensing
Academy of Sciences 27, 36 – 54. and the measurement of geographical entities in a forest envi-
Collins, J.B., Woodcock, C.E., 1999. Geostatistical estimation of ronment. 1: the scale and spatial aggregation problem. Remote
resolution-dependent variance in remotely sensed image. Photo- Sensing of Environment 49, 93 – 104.
grammetric Engineering and Remote Sensing 65 (1), 41 – 51. McGregor, S.J., 1998. An integrated geographic information system
Csillag, F., 1997. Quadtrees: hierarchical multiresolution data struc- approach for modeling the suitability of conifer habitat in an
tures for analysis of digital images. In: Quattrochi, D.A., Good- alpine environment. Geomorphology 21, 265 – 280.
child, M.F. (Eds.), Scale in Remote Sensing and GIS. Lewis Mohanty, K.K., 1997. The wavelet transform for local image en-
Publishers, Boca Raton, FL, pp. 247 – 272. hancement. International Journal of Remote Sensing 18 (1),
Csillag, F., Kabos, S., 1996. Hierarchical decomposition of variance 213 – 219.
with applications in environmental mapping based on satellite Walsh, S.J., Butler, D.R., 1997. Morphometric and spectral analyses
images. Mathematical Geology 28 (4), 385 – 405. of debris flows: components of a natural hazards methodology.
Curran, P.J., 1988. The semivariogram in remote sensing: an intro- GeoCarto International 12 (1), 59 – 70.
duction. Remote Sensing of Environment 24, 493 – 507. Walsh, S.J., Butler, D.R., Brown, D.G., Bian, L., 1990a. Carto-
de Bruin, S., 2000. Spatial uncertainty in estimates of the areal graphic modeling of snow avalanche path location in Glacier
extent of land cover types. Proceedings Fourth International National Park, Montana. Photogrammetric Engineering and Re-
Symposium on Spatial Accuracy Assessment in Natural Resour- mote Sensing 56 (5), 615 – 621.
ces and Environmental Sciences, Delft University Press, Am- Walsh, S.J., Cooper, J.W., Von Essen, I.E., Gallager, K.R.,
sterdam, The Netherlands, pp. 37 – 144. 1990b. Image enhancement of Landsat Thematic Mapper data
Dutton, B.L., Marrett, D.J., 1997. Soils of Glacier National Park and GIS integration for evaluation of resource characteristics.
East of the Continental Divide. Land and Water Consulting, Photogrammetric Engineering and Remote Sensing 54 (8),
Missoula, MT. 1135 – 1141.
Gardner, J.S., Jones, N.K., 1985. Evidence for a Neoglacial advance Walsh, S.J., Butler, D.R., Allen, T.R., Malanson, G.P., 1994a.
of the Boundary Glacier, Banff National Park, Alberta. Cana- Influence of snow patterns and snow avalanches on the
dian Journal of Earth Sciences 22, 1753 – 1755. alpine treeline ecotone. Journal of Vegetation Science 5,
Goovaerts, P., 1997. Geostatistics for Natural Resources Evalua- 657 – 672.
tions. Oxford, New York. Walsh, S.J., Butler, D.R., Brown, D.G., Bian, L., 1994b. Form and
Hansen-Bristow, K.J., Price, L.W., 1985. Turf-banked terraces in pattern in the alpine environment: an integrated approach to
the Olympic Mountains, Washington, USA. Arctic and Alpine spatial analysis and modelling in Glacier National Park, USA.
Research 17, 261 – 270. In: Heywood, D.I., Price, M.F. (Eds.), Mountain Environments
Jensen, J.R., 2000. Remote Sensing of the Environment: An Earth and GIS. Taylor and Francis, London, pp. 189 – 216.
Resource Perspective. Prentice Hall, Upper Saddle River, NJ. Walsh, S.J., Moody, A., Allen, T.R., Brown, D.G., 1997. Scale
Jupp, D.L.B., Strahler, A.H., Woodcock, C.E., 1989a. Autocorre- dependence of NDVI and its relationship to mountainous ter-
lation and regularization in digital images. I: basic theory. rain. In: Quattrochi, D.A., Goodchild, M.F. (Eds.), Scaling in
IEEE Transactions on Geoscience and Remote Sensing 26 (4), Remote Sensing and GIS. Lewis Publishers, Boca Raton, FL,
463 – 473. pp. 27 – 55.
Jupp, D.L.B., Strahler, A.H., Woodcock, C.E., 1989b. Autocorre- Walsh, S.J., Butler, D.R., Malanson, G.P., 1998. An overview of
lation and regularization in digital images. II: simple image scale, pattern, and process relationships in geomorphology: a
models. IEEE Transactions on Geoscience and Remote Sensing remote sensing and GIS perspective. Geomorphology 21 (3 – 4),
27 (3), 247 – 258. 183 – 205.
398 S.J. Walsh et al. / Geomorphology 55 (2003) 381–398

Walsh, S.J., Butler, D.R., Malanson, G.P., Crews-Meyer, K.A., Remote Sensing and GIS. Lewis Publisher, Boca Raton, FL,
Messina, J.P., Xiao, N., 2003. Mapping, modeling, and visual- pp. 309 – 360.
ization of the influences of geomorphic processes on the alpine Yocky, D.A., 1996. Multiresolution wavelet decomposition image
treeline ecotone, Glacier National Park, Montana, USA. Geo- merger of Landsat Thematic Mapper and SPOT Panchromatic
morphology, (in press). data. Photogrammetric Engineering and Remote Sensing 62 (9),
Xia, Z., Clarke, K.C., 1997. Approaches to scaling of geo-spatial 1067 – 1074.
data. In: Quattrochi, D.A., Goodchild, M.F. (Eds.), Scale in