See discussions, stats, and author profiles for this publication at: [Link]

net/publication/210252651

Land management impacts on runoff sources in small Amazon watersheds

Article in Hydrological Processes · June 2008

DOI: 10.1002/hyp.6803

CITATIONS READS

133 316

6 authors, including:

Joaquín Ernesto Chaves Christopher Neill

NASA Woodwell Climate Research Center
30 PUBLICATIONS 1,036 CITATIONS 185 PUBLICATIONS 9,658 CITATIONS

SEE PROFILE SEE PROFILE

Sonja Germer Alex V. Krusche

Ministry of Science Research and Culture Brandenburg Universidade Federal de São Paulo
54 PUBLICATIONS 1,876 CITATIONS 144 PUBLICATIONS 7,499 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by H. Elsenbeer on 25 October 2017.

The user has requested enhancement of the downloaded file.

HYDROLOGICAL PROCESSES
Hydrol. Process. 22, 1766– 1775 (2008)
Published online 24 September 2007 in Wiley InterScience
([Link]) DOI: 10.1002/hyp.6803

Land management impacts on runoff sources in small

Amazon watersheds
Joaquı́n Chaves,1 * Christopher Neill,1 Sonja Germer,2 Sérgio Gouveia Neto,3 Alex Krusche3
and Helmut Elsenbeer2
1 The Ecosystems Center, Marine Biological Laboratory, Woods Hole, MA 02543, USA
2 Universität Potsdam, Institut für Geoökologie, Potsdam, Germany
3 Laboratório de Ecologia Isotópica, CENA—USP, Piracicaba, SP, Brazil

Abstract:
Forest clearing and conversion to cattle pasture in the lowland Amazon region has been linked to soil compaction and increased
soil water storage, which combine to diminish soil infiltration, enhance quick lateral flows and increase the stream flow response
to precipitation. Quantifying the importance of quick surficial flow in response to this land use change requires identification
of water sources within catchments that contribute to stream flow. Using an end member mixing analysis approach, potential
contributing sources of stream flow were evaluated during an entire rainy season in a forest and a pasture watershed drained
by ephemeral-to-intermittent streams in the south-western Amazon. Water yield was 17% of precipitation in the pasture and
0Ð8% of precipitation in the forest. During the early rainy season, throughfall, groundwater, and soil water contributed 79%,
18%, and 3%, respectively, to total forest stream flow. Over the entire rainy season, throughfall, groundwater, and shallow soil
water provided 57%, 24%, and 19%, respectively, of stream flow. In the pasture watershed, overland flow dominated stream
flow both in the early (67%) and late (57%) rainy season, with a mean contribution of 60% overland flow, 35% groundwater,
and 5% soil water. The uncertainty associated with those estimates was studied using a Monte Carlo approach. In addition
to large changes in total surface flow, marked differences were found in the proportions of total stream flow in the second
half of the rainy season between the forest and pasture watershed. These results suggest that (1) there is great potential for
alteration of the hydrological budgets of larger watersheds as the proportion of deforested land in the Amazon increases, and
(2) as more rainfall is diverted into fast flowpaths to streams in established pastures, the potential to deliver water with higher
solute concentrations generated by erosion or by bypassing sites of solute removal increases. Copyright  2007 John Wiley
& Sons, Ltd.

KEY WORDS Amazon; hydrology; streams; deforestation; rain forest; pasture; EMMA

Received 14 November 2006; Accepted 1 May 2007

INTRODUCTION (Elsenbeer, 2001; Bonell, 2005). Recent work shows

that in many locations that support moist, lowland trop-
Precipitation falling in watersheds may take a num-
ber of pathways to streams. Soil physical character- ical forest, rainfall intensities may exceed permeability
istics and topography interact with rainfall intensity at shallow depths, which leads to shallow lateral flows
and amount to determine what flowpaths exist, what capable of triggering widespread overland and return
specific flowpaths are activated and what sources of flows (Elsenbeer and Lack, 1996; Schellekens et al.,
water contribute to stream flow (Dunne, 1978). Under- 2004; Godsey et al., 2004; Johnson et al., 2006). While
standing the origin of stream water and the paths that increasingly recognized as potentially important, a quan-
precipitation follows en route to streams is important titative understanding of the importance of these shal-
because the rates of water movement and the condi- low lateral flow contributions to stream flow across the
tions encountered by water along flowpaths control bio- lowland tropical landscapes is still lacking. Because the
geochemical transformations that ultimately determine properties of vegetation and soils play such a large role
stream water chemistry (Hill et al., 2000; McClain et al., in structuring the origin and flowpaths of water within
2003). catchments, changes to land use that influence soils and
In humid tropical forests, early perceptions of runoff vegetation have the potential to cause large changes to
generation were influenced by the notion that high infil- both the amount of water reaching streams and the flow-
trability and permeability in undisturbed soils led to paths of water in watersheds. This issue is particularly
predominantly vertical flowpaths, precluding fast lateral important in humid tropical regions. In these places,
flows such as overland flow except on steep hillslopes high rainfall intensities combine with the widespread
occurrence of soils with strong near-surface permeability
contrasts, creating the potential for widespread genera-
* Correspondence to: Joaquı́n Chaves, The Ecosystems Center, Marine
Biological Laboratory, Woods Hole, MA 02543, USA. tion of saturation overland flow, a potential that is apt
E-mail: jchaves@[Link] to increase upon disturbance (Bonell, 2005). Decreased

Copyright  2007 John Wiley & Sons, Ltd.

 RUNOFF SOURCES IN SMALL AMAZON WATERSHEDS 1767

interception and evapotranspiration (Wright et al., 1996), STUDY SITE

increased soil moisture (Hodnett et al., 1995; Wright
et al., 1996) and higher total runoff and groundwater Field measurements were carried out at Rancho Grande
recharge (Jipp et al., 1998; Bruijnzeel, 2004) associ- (10° 180 S, 62° 520 W, 143 m a.s.l.) in the Brazilian state
ated with tropical forest clearing have the potential to of Rondônia, in the south-western Brazilian Amazon
amplify the generation of surficial flows, a phenomenon basin (Figure 1). Mean annual precipitation, based
that may become widespread because the clearing of on daily rain totals collected since 1984 by the
humid tropical forest is now occurring rapidly over very landowners, is 2330 mm with a pronounced dry season
wide regions (Achard et al., 2002), such as the Amazon of <100 mm month
1 from June through August (H.
basin. Schmitz, unpublished data; Zimmermann et al., 2006).
The Amazon basin contains the earth’s largest area Adjacent forest (1Ð37 ha) and pasture (0Ð73 ha) research
of remaining tropical forest. This region is currently watersheds were selected, approximately 400 m apart
experiencing the highest absolute deforestation rates in (Figure 1). The forest watershed contains open moist
the world. The rates of forest clearing in the Brazil- tropical forest with a high density of palms typical
ian Amazon alone have ranged between 13 000 and of the region (Pires and Prance, 1986; Germer et al.,
2 9000 km2 yr
1 since 1988 (INPE, 2006). Cattle ranch- 2006). The pasture was cleared in 1985, planted to
ing has been the primary agricultural activity on defor- pasture in 1986 and has had continuous cover of African
ested Amazon lands and planted pastures cover about forage grass Brachiaria humidicola since that time. It
75% of the area converted from forest (Faminow, 1998; has been actively grazed at about 1 animal ha
1 . Both
Margulis, 2003). The Amazon cattle herd has doubled watersheds are drained by channels that convey water
from 15 to 30 million head since 1996 (Simon and in response to some storm events regardless of season,
Garagorry, 2005; Nepstad et al., 2006) and cattle ranch- and intermittently towards the end of the rainy season
ing will almost certainly be a major Amazon agricultural in some years. The area belongs to a morphostructural
sector for years to come. In several locations in the unit known as “Southern Amazon Dissected Highlands”
Amazon, forest clearing and conversion to cattle pasture (Planalto Dissecado Sul da Amazônia, Peixoto de Melo
has been shown to cause soil compaction that dimin- et al., 1978), which is characterized by a very pronounced
ishes infiltrability and enhances near-surface lateral flows topography with an altitudinal differential of up to
(Biggs et al., 2006; Moraes et al., 2006; Zimmermann
150 m; remnant ridges of Precambrian basement rock,
et al., 2006).
made up of granites and gneisses of the Complexo
Quantifying the importance of surficial flow in res-
Xingu (Leal et al., 1978), are separated by flat valley
ponse to land use change requires identification of water
floors of varying width. Soil orders associated with
sources within catchments that contribute to stream
this morphostructural unit are Ultisols, Oxisols, and
flow. End-member mixing analysis (EMMA) is a poten-
Inceptisols and Entisols (Soil Survey Staff, 1999) on
tially useful approach for inferring and quantifying the
steep slopes and along streams, respectively. Soils
relative contributions of different sources to stream
flow (Christophersen et al., 1990; Christophersen and underlying the watersheds are Kandiudults (Sobieraj
Hooper, 1992). This approach assumes that the chem- et al., 2002; Zimmermann et al., 2006), based on clay
istry of stream water is the product of a mixture of dis- content changes with depth and cation exchange capacity
crete “source” solutions within catchments whose solutes in the subsoil, which exceed 2 m in depth. Clay contents
behave conservatively as they travel to the stream and in both watersheds ranged from 20% in the A horizon
have extreme concentrations relative to stream flow. The to around 40% in the B horizon. Soil pH in water
application of EMMA has been used at temperate sites decreased from 5Ð8 to 5Ð3 in the forest and from 5Ð5
such as the Panola Mountain watershed (Burns et al., to 5Ð0 in the pasture between these two horizons. The
2001; Hooper, 2001) and the Walker Branch watershed corresponding cation exchange capacities on a clay basis
(Genereux et al., 1993; Mulholland, 1993) in the south- are 26 and 7 cmolc kg
1 in the forest, and 8 and
eastern USA. There are fewer examples of its appli- 9 cmolc kg
1 in the pasture (K. Adam, unpublished MSc
cation in tropical settings. In north-eastern Australia, thesis).
Elsenbeer et al. (1995) distinguished the contributions These watersheds were chosen after an extensive sur-
of groundwater, soil water and overland flow to storm vey of many potential sites. They are closely matched in
flow, ascribing a prominent contribution to overland geomorphology, soil characteristics and elevation and are
flow. representative of the soils and vegetation of Rondônia.
The objective of this study was to use EMMA to eval- Earlier surveys of soil hydraulic properties indicated:
uate the relative contribution of potential sources of water (1) that infiltration rates were significantly lower in pas-
to stream flow in adjacent catchments of open, moist ture, where average rates were 13 times lower than those
tropical forest and established cattle pasture. The results in forest; and (2) that infiltration rates decrease consid-
are used to infer how deforestation for cattle pasture erably at soil depths of 12Ð5–20 cm where an impeding
has altered both the magnitude and the relative impor- layer occurs, regardless of land cover; permeability at a
tance of different catchment sources to surface channel depth of 20 cm is up to two orders of magnitude lower
flows. than infiltrability (Zimmermann et al., 2006).

Copyright  2007 John Wiley & Sons, Ltd. Hydrol. Process. 22, 1766– 1775 (2008)
DOI: 10.1002/hyp
1768 J CHAVES ET AL.

Rondônia
(b) 1 Forest
2
3 0°
4

5 10° S
Brazil
6

x (a)
x
x 7
0 50 100 m
Pasture
8
4
1
x
x x
9 2
3
10

(c)

Throughfall collector x Overland flow collector H flume

Lysimeter nest Groundwater well Precipitation collector

Figure 1. Location of the study site: (a) marker shows the location of Rancho Grande cattle ranch in central Rondônia; (b) forest watershed; (c) pasture
watershed. Contour intervals are every 0Ð5 m; elevations are shown relative to the watershed outlets

MATERIAL AND METHODS The lysimeters were installed above and below the
texture break at the top of the B horizon (20, 100 cm).
Field measurements
Groundwater was sampled weekly from three wells at
Both watersheds were instrumented for hydrological each watershed. Wells were 5 m apart and 10 m from
and hydrochemical measurements beginning in August the outlet of each watershed and ranged 4–6 m in depth.
2004. A tipping bucket rain gauge (Hydrological Services Stream flow was collected from a 30Ð5 cm H-flume
P/L, Liverpool Australia) and a Campbell logger installed installed at the outlet of each watershed using ISCO
in the pasture recorded rainfall intensity. Incident rainfall samplers. Samples were collected every 5 min during the
volumes and rainfall samples for chemical analyses were first 80 min of flow and every 20 min for the following
collected after each precipitation event into a 20 L 160 min. Stage height was recorded every 5 min with
acid washed polyethylene container using a trough-type TruTrack loggers installed on each flume.
collector installed 1 m above ground (Germer et al., All field measurements and sampling were conducted
2006). Canopy throughfall was collected after each event between late August 2004 and late April 2005, excluding
at 20 similar trough collectors distributed throughout December 2004 due to personnel and logistical limita-
the catchment. Overland flow samplers (modified after tions, and encompassed an entire rainy season. The log-
Kirkby et al., 1976) were placed at one location in each ging equipment recorded precipitation intensity and stage
watershed during the first four months of sampling and heights continually for the duration of the study.
at three additional locations in the subsequent months
of sampling. Samples of soil solution from the vadose Laboratory analyses
zone were taken every ten days from tension lysimeters After collection, all samples were transported to a field
installed at two depths at ten locations in each watershed. laboratory located 500 m from the research watersheds

Copyright  2007 John Wiley & Sons, Ltd. Hydrol. Process. 22, 1766– 1775 (2008)
DOI: 10.1002/hyp
 RUNOFF SOURCES IN SMALL AMAZON WATERSHEDS 1769

and filtered through Whatman GF/F filters (0Ð7 µm pore resulting U space. Bivariate solute plots were also used
size). Samples were preserved with thymol and stored in evaluating tracers for use in the mixing analysis. In
frozen in 60 mL acid-washed polyethylene bottles. Con- the forest, the combination of KC , MgC2 , CaC2 , and NaC
centrations of cations and anions (NaC , KC , Ca2C , Mg2C , provided the most suitable mixing space. In the pasture,
Cl
, and SO4 2
) were determined by ion chromatogra- the combination of KC , MgC2 , and CaC2 was used. The
phy (Dionex DX-500) at the Centro de Energia Nuclear proportion of each end member in each stream water
na Agricultura, at the University of São Paulo, in Piraci- observation was obtained by solving the following system
caba, SP. of linear equations:

Data analysis 1DxCyCz
2

A multivariate PCA-based EMMA procedure (Christo- SWU1 D x EM1 U1 C y EM2 U1 C z EM3 U1
3
phersen and Hooper, 1992; Hooper, 2003) was used to
identify the sources that contributed to stream runoff dur- SWU2 D x EM1 U2 C y EM2 U2 C z EM3 U2
4
ing storm events, and calculate their relative contribution
where x, y and z are the unknown proportions of
to stream flow. The analysis was applied independently
each end member; SWU1 and SWU2 are the coefficients
for both forest and pasture. The purpose of the PCA is
in U space coordinates, U1 and U2, for a stream
to find a “lower-dimensional” space, U, where most of
water observation. Likewise, EMn U1 and EMn U2 are
the stream water observations are assumed to lie within
the coefficients in U space for the nth end member.
a certain accuracy. This approach allows the use of an
Equations (2)–(4) depict the case for a three end member
over-determined set of equations in which more tracers
mixing scenario. Because of various sources of error,
than necessary are used to solve for the end-member
such as non-conservative solute behavior, time-dependent
proportions, while incorporating most of the variance
end member variability, and/or analytical uncertainty,
associated with the tracers (Christophersen and Hooper,
some stream observations lie outside the mixing domain
1992; Hooper, 2003). The dimensionality of U space
defined by the end members chosen as sources of
is determined by the number of vectors (m) retained
stream flow. The solutions to the above equations in
from the PCA. The vectors with corresponding eigen-
those cases result in end member fractions for which
values greater than or equal to 1 (Reyment and Joreskog,
negative values are found. To circumvent that problem,
1993) were retained, which resulted in two-dimensional
the outlier observations were perpendicularly projected
mixing spaces for both watersheds. The vectors retained
explained 83% and 87% of the variance in forest and to the line joining the two non-zero end members and
pasture, respectively. solved geometrically in U space as binary mixtures of
The stream concentration data (n) for the solutes (p) these two end members (Liu et al., 2004).
were standardized by centering them about their means A Monte Carlo approach was used to compute con-
and dividing by their respective standard deviations. This fidence intervals for EMMA estimates following a pro-
ensured that all solutes had the same weight in the cedure similar to that used by Joerin et al. (2002). The
PCA. The median concentrations of each of the mea- uncertainty due to analytical error in measuring the con-
sured potential end-members were also standardized by centrations of solutes as well as that from the observed
subtracting the means and dividing by the standard devia- variability of the chosen end members was taken into
tions of the stream observations. The standardized stream account. The end members varied in U space according
data and potential end-member medians were projected to their observed probability density function, which was
onto the m-dimensional U space by the orthogonal pro- determined by maximum likelihood estimation using the
jection given by language R version 2Ð3Ð1 (Venables and Ripley, 2002).
U = XVT
1 The Monte Carlo procedure was performed 104 times for
each watershed.
where U is the n ð m projected data matrix, X is the
n ð p standardized data matrix, and V is the is the m ð p
matrix of the retained eigenvectors. The projected sources RESULTS
that best bounded the stream data in U space were chosen
as end members for the mixing models in each watershed. Rainfall, runoff
End members were calculated for both halves of the Total precipitation from August 2004 to the end of
study period, designated as “early” (August to December) April 2005 was 2184 mm (Table I). During the dry
and “late” (January to April) rainy season to examine period preceding the study (May–July 2004) precipita-
seasonal changes in both the variability of the end tion was 85 mm (H. Schmitz, unpublished data), which,
members and the contributions of water sources to when combined, placed the 2269 mm total for the year
stream flow. The procedures above were performed with close to the long-term average (2330 mm). In total, 163
all six solutes (Ca2C , Mg2C , KC , NaC , Cl
, SO4 2
) rain events were recorded. Of these, 44 events gener-
and all combinations of five, four and three solutes. ated flow in the forest watershed: 15 in the early rainy
This exercise was aimed at finding a combination of season and 29 in the late rainy season. In the pas-
solutes that bounded the stream flow observations in the ture, 55 events generated flow, 20 in the early rainy

Copyright  2007 John Wiley & Sons, Ltd. Hydrol. Process. 22, 1766– 1775 (2008)
DOI: 10.1002/hyp
1770 J CHAVES ET AL.

season and 35 in the late season (Table I). Of those that from the forest (Table I). Water yield was 17% of
events where flow developed, 22 events were sampled precipitation in the pasture and 0Ð8% of precipitation in
in the forest and 27 in the pasture (Table I). Precipita- the forest (Table I). Stream flow was not recorded in
tion during those sampled events accounted for approx- the forest or pasture streams until late September, after
imately 30% of the total precipitation recorded during 210 mm of rainfall in 12 events had been recorded since
the study. Stream flow in sampled events accounted 22 August. At both sites, the bulk of runoff occurred
for 16% of the total stream flow recorded in the for- during the second half of the rainy season (87% of the
est and 35% of the stream flow recorded in the pasture flow in the forest occurred between January and April,
(Table I). compared with 73% in the pasture). The proportion of
Over the entire study period, the total stream flow total precipitation during that same period was 53%,
from the pasture was more than 20-fold greater than similar to the proportion of rain events (55%). Baseflow,
considered here as flow rates <0Ð01 L s
1 , occurred in
Table I. Precipitation and runoff at the research watersheds the pasture watershed from early January onwards and
during the period of study comprised 1Ð3% of the total runoff. Baseflow did not
occur in the forest watershed.
Earlya Lateb Total

Precipitation events, 73 90 163

count Hydrochemistry
Precipitation, mm 1020 1164 2184
Runoff, mm (# Forest 2Ð2 (15) 14Ð5 (29) 16Ð7 (44) With the exception of NaC in groundwater, precipi-
events) Pasture 101 (20) 277 (35) 378 (55) tation and groundwater were generally depleted in all
Sampled events, count Forest 10 12 22 solutes relative to other water sources in both pasture
Pasture 11 16 27 and forest (Figure 2). Soil water, followed by through-
Precipitation during Forest 332 352 684 fall, had the widest range of concentrations spanning two
sampled events, mm Pasture 313 311 624 or three orders of magnitude for most solutes. Stream
Runoff during sampled Forest 0Ð6 2Ð1 2Ð6 flow and overland flow had a consistent overlap in solute
events, mm Pasture 29Ð1 104Ð4 133Ð5
concentrations, as suggested by the very close agreement
a Early rainy season, from 22 August to 31 December 2004. of their respective interquartile ranges for all solutes on
b Late rainy season, from 1 January 2005 to 25 April 2005. both forest and pasture (Figure 2).

Forest Pasture Forest Pasture

SF
TF
GW

SW

OF

SF
GW

SW

OF

SF
TF
GW

SW

OF

SF
GW

SW

OF
Prec.

Prec.

20 100 20 100 20 100 20 100

cm cm cm cm
1000
Cl- (a) K+ (d)
100

10

1
Concentration, µmol L-1

1000
SO4-2 (b) Mg+2 (e)
100

10

1000
Na+ (c) Ca+2 (f)
100

10

0.1
37 759 177 31 101 126 37 759 177 31 101 126
n=
273 103 97 513 139 45 273 103 97 513 139 45

Figure 2. Log scale box plots (minimum, 25%, median, 75%, and maximum) of concentration (µmol L
1 ) for (a) Cl
, (b) SO4
2 , (c) NaC , (d) KC ,
(e) MgC2 , and (f) CaC2 . Direct precipitation, Prec.; streamflow, SF; throughfall, TF; groundwater, GW; soil water, SW; and overland flow, OF

Copyright  2007 John Wiley & Sons, Ltd. Hydrol. Process. 22, 1766– 1775 (2008)
DOI: 10.1002/hyp
 RUNOFF SOURCES IN SMALL AMAZON WATERSHEDS 1771

Forest Pasture
(a) (c)
8 SW 100 cm 4
SW 20 cm
OF
2
4

Early
SW 20 cm
U2
GW

OF 0 SW 100 cm
0
TF GW
-2 Precip.
Precip.
-4
0 10 20 30 40 -4 0 4 8 12

(b) (d)
4 2

GW
OF OF
SW 20 cm SW 20 cm

Late
0
U2

0
SW 100 cm SW 100 cm
TF GW
-2
-4 Precip.
Precip.
-10 -5 0 5 10 15 -4 0 4 8
U1 U1
Figure 3. U space mixing diagrams for ‘early’ and ‘late’ rainy season for the forest and pasture watersheds. Stream flow observations are the gray
circles, pasture baseflow dark triangles, and end members medians dark squares. Error bars depict the interquartile range for each end member.
Abbreviations are as in Figure 2

End member mixing analysis Entire rainy season

20 (a) Forest (b)
The stream flow observations projected onto U space,
together with the potential water sources, provided evi- 19%
3%
dence of the hydrological functioning of forest and pas- 24%
18%
ture watersheds. Forest stream flow observations were 10
25%
79%
best bounded by groundwater, throughfall and shallow
(20 cm) soil water (Figure 3a, b). In the early rainy sea- 51% 24%
Runoff, mm

57%
son, most forest stream water observations were closer
0
to the throughfall end member, suggesting that through-
fall dominated stream flow at this time (Figure 3a). Also, 300 (c) Pasture (d)
in this initial period of the rainy season, overland flow
67% 57%
appeared as a simple mixture of throughfall and shallow 200 35%
soil water, since its location in U space placed it along 3%
the line joining these two sources (Figure 3a). 30% 6%
100
In the pasture during the early part of the rainy season, 60%
37%
stream flow was a mixture predominantly of overland 5%
flow and groundwater, with a small contribution from soil 0
Early Late
water (Figure 3c). There was considerable scatter around
overland flow, with many stream flow observations
falling outside the mixing domain determined by the
three end members. As the rainy season progressed, Throughfall Groundwater Soil water Overland flow
pasture stream flow remained a mixture of groundwater Figure 4. Early and late rainy season stream channel flow by source
and overland flow, although some of the variability according to EMMA for (a) forest and (c) pasture. Relative proportion of
sources contributing to channel flow for entire rainy season for (b) forest
shifted towards soil water, which suggests an increased and (d) pasture
contribution from soil water later in the rainy season
(Figure 3d). Soil water in pasture from depths 20 and
100 cm was chemically indistinguishable, so soil water (Figure 4a). As moisture increased towards the late rainy
contributions to stream flow could not be ascribed to a season and the forest catchment became more responsive
particular soil layer. to precipitation (Table I), throughfall remained the single
Based on the output from the solution of the EMMA most important source of stream flow (51%) but shallow
equations for the early rainy season, throughfall con- soil water became a more important source (24%), with
tributed 79% to total forest stream flow, groundwater groundwater providing the remainder (25%) (Figure 4a).
contributed 18% and shallow soil water contributed 3% Over the entire rainy season, throughfall provided 57%

Copyright  2007 John Wiley & Sons, Ltd. Hydrol. Process. 22, 1766– 1775 (2008)
DOI: 10.1002/hyp
1772 J CHAVES ET AL.

Forest Pasture
(a) 100% Early (c) Early

80%

60%
Proportion of stream runoff

40%

20%

0%

(b) 100% Late (d) Late

80%

60%

40%

20%

0%
Throughfall Groundwater Soil Water Groundwater Soil Water Overland flow

Figure 5. Boxplots (5%, 25%, median, 75%, and 95%) of 104

Monte Carlo EMMA computations of the proportion of total stream runoff attributed
to each source during the early and late rainy season in forest and pasture. Gray bars depict the estimated proportions derived from the EMMA
shown on Figure 4a, c

of stream flow, groundwater provided 24% and shallow Grande (17% versus 0Ð8% of rainfall) was consistent with
soil water provided 19% (Figure 4b). reports of increased stream flow from pasture relative to
In the pasture, the relative proportions of sources forest basins of similar scale at other locations in the
contributing to stream flow varied less between early and lowland Amazon. In approximately 0Ð5 ha watersheds
late rainy season (Figure 4c). Overland flow dominated on plinthic Oxisols in the eastern Amazon state of Pará,
stream flow in the pasture watershed both in the early Moraes et al. (2006) found that stream flows accounted
(67%) and late (57%) rainy seasons, while groundwater for 17% of rainfall in pasture but only 2Ð7% of rainfall in
contributed 30% in the early rainy season and 37% forest. The occurrence of a strong reduction of hydraulic
in the last rainy season (Figure 4c). Over the entire conductivity at the plinthite layer at that site contributed
rainy season, overland flow supplied 60% of pasture to maintaining the soil close to saturation during most
stream flow, groundwater supplied 35% and soil water of the rainy season and led to some saturation overland
contributed the remaining 5% (Figure 4d). flow under forest but much more frequent saturation
overland flow in pasture (Moraes et al., 2006). During
Uncertainty analysis sampling of 10 rainstorms in a 3Ð9-ha watershed draining
Results from the Monte Carlo simulations showed cattle pasture on Ultisols in central Rondônia, Biggs
the same trend in stream flow contribution estimates as et al. (2006) found that stream flow was 16% of rainfall,
the EMMA estimates (Figure 5). High solute variability with infiltration excess overland flow accounting for half
of the throughfall end member (Figure 2) resulted in a of the flow. Although results from these three sites,
wide range of estimates for this source as contributor along with the results at Rancho Grande, capture only
to stream flow in the forest watershed (Figure 5a, b). a small range of the soil and rainfall conditions that
Much more tightly constrained were the estimates of soil occur over the larger Amazon region where clearing for
water contribution to flow at both sites. For the forest site, pasture is widespread, they suggest there is potential for
the simulation replicated the considerable increase in soil a large amount of rainfall in pastures to be diverted
water contribution to flow in the late rainy season relative to stream flow via overland flow on hillslopes and
to the early period (Figure 5b). In the pasture, overland small zero-order watersheds compared with the original
flow still appeared as the main contributor in the early forest.
part of the season, although its relative role in the late In addition to large changes in total stream flow,
season appeared statistically indistinguishable from that marked differences were found in the proportions of flow
of groundwater (Figure 5c, d). in the second half of the rainy season between the for-
est and pasture watershed. While the amount of stream
flow as a fraction of precipitation changed between peri-
DISCUSSION
ods in the pasture by a factor of 2Ð5, forest stream flow
Effect of forest conversion to pasture increased by a factor of 6 during the second half of
The greatly increased amount of water exiting the the rainy season (Table I). This pattern suggested that
pasture watershed as surface stream flow at Rancho antecedent soil moisture played a relatively small role

Copyright  2007 John Wiley & Sons, Ltd. Hydrol. Process. 22, 1766– 1775 (2008)
DOI: 10.1002/hyp
 RUNOFF SOURCES IN SMALL AMAZON WATERSHEDS 1773

in controlling the generation of surface flows in pas- 2001). Detection of shallow surface flows in the forest
ture, presumably because saturation excess overland flow watershed at Rancho Grande was consistent with previ-
caused by reduced near-surface permeability (Zimmer- ous measurements of saturated soil hydraulic conductivity
mann et al., 2006) dominates generation of stream flow below 20 cm in Ultisols under forest at Rancho Grande
in pasture under most soil moisture conditions. The inten- that indicated the potential for generation of saturation
sification of forest stream flow only after soils became overland flowpaths during hourly rainfall intensities that
thoroughly wet in the later portion of the rainy season have recurrences of 10 and 30 times per year (Zim-
arises from reduced permeability that occurs deeper in mermann et al., 2006). The dramatic increase in stream
the forest soil than in the pasture. flow that was observed in pasture relative to forest was
Antecedent moisture conditions are important deter- also consistent with the greatly reduced hydraulic con-
minants of catchment runoff response to precipitation ductivity down from the soil surface observed in pas-
(Mulholland et al., 1990). Pasture and forest soils in tures on Ultisols at Rancho Grande, which should result
the Amazon region have been documented to exhibit in saturation-excess overland flow under a much wider
major differences in soil moisture and water storage over range of commonly-experienced rainfall intensities (Zim-
an annual cycle, especially in locations with a marked mermann et al., 2006).
dry season (Hodnett et al., 1995; Moraes et al., 2006).
Decreases in evaporation, interception and deep root EMMA-derived source contributions to stream flows
water uptake caused by forest removal result in increased
storage and earlier soil saturation and rise of the water Throughfall clearly dominated forest stream flow dur-
table after the start of the rainy season under pasture soils ing the early part of the rainy season (Figure 4a). The
(Peck and Williamson, 1987; Hodnett et al., 1995). The rapid detection of water matching the chemical signature
observed onset of baseflow exclusively in the Rancho of throughfall in stream flow early in the rainy season
Grande pasture watershed was consistent with increased may imply that shallow fast flow pathways, presumably
water storage under pasture and a link between increased overland flow, deliver this water to the stream chan-
soil moisture and a heightened runoff response in the nel. Although, some throughfall must fall directly on the
pasture watershed. stream channel, this was not considered the dominant
The increase in surface runoff and the differences pathway for this source in reaching the stream. Based
in stream flow responses between the early and late purely on the EMMA, one cannot make inferences about
rainy season observed at Rancho Grande are interpreted flowpaths in the absence of ancillary hydrometric infor-
to be effects of the land cover conversion from forest mation. Overland flow as an important contributor to
to pasture. The watersheds have nearly identical soil stream flow has been documented in tropical forest sites
type, topography, underlying geology, slope and other in the Amazon basin and elsewhere (Bonell and Gilmour,
geomorphologic features. Because of their proximity, 1978; Elsenbeer and Vertessy, 2000; Godsey et al., 2004;
they receive nearly identical rainfall amounts in each Johnson et al., 2006, Moraes et al., 2006), and has been
storm. The chronology and history of land use conversion indicated for the Rancho Grande forest based on the verti-
and further agricultural uses are well documented at this cal profile of hydraulic conductivity (Zimmermann et al.,
location and are similar to those across a wide area of 2006).
the cattle ranching region of the western Amazon (Neill Overland flow was not used as an explicit end member
et al., 1997). in the mixing analysis for the forest catchment because
the mixing diagrams suggested that this flowpath could
Role of soil hydraulic properties be explained as a mixture of other water sources. In the
Soil hydraulic properties are particularly sensitive to early rainy season, overland flow appeared as a mixture
land cover modification, and soil infiltrability is typically of predominantly throughfall with some contribution of
reduced after forest clearing (Bonell, 2005). It is well shallow soil water (Figure 3a), thus the assertion that
established that soil hydraulic properties are a major fac- the path to the stream for throughfall is overland flow.
tor in controlling storm flow pathways (Dunne, 1978). Later in the season overland flow adopted a composition
Dominant soil-landscape units exert an important con- essentially equal to that of stream flow, with greater
trol on the proportion of storm flow that is diverted into contributions of soil and groundwater relative to the early
horizontal versus vertical flowpaths because soil charac- wet period (Figure 3b).
teristics, such as horizons of reduced hydraulic conductiv- The mixture of water sources that contributed to stream
ity, control the generation of surficial “quick” flowpaths flow in the pasture was simpler than that in the forest.
versus deep “slow” flowpaths (Elsenbeer, 2001). Ulti- Over the entire rainy season, more precipitation was
sols, many of which exhibit sharp downward increases in directed through fast pathways to the stream channel,
clay content and decreases in macroporosity and therefore while the relative proportions of all the sources remained
low permeabilities, have been associated with predomi- almost unchanged (Figure 4c). This appeared to follow
nantly lateral, near surface flow pathways, while Oxisols, directly from greatly reduced permeabilities of pasture
many of which have high permeabilities throughout the soils (Zimmermann et al., 2006).
soil profile despite high clay contents, have been asso- Overland flow was used as an end member contributing
ciated with predominantly vertical flowpaths (Elsenbeer, to pasture stream flow because this flowpath could not

Copyright  2007 John Wiley & Sons, Ltd. Hydrol. Process. 22, 1766– 1775 (2008)
DOI: 10.1002/hyp
1774 J CHAVES ET AL.

be described, according to the EMMA, as a mixture of sources and flowpath structure have an important influ-
two or more discrete sources such as precipitation and ence on biogeochemical transformations and watershed
soil water (Figure 3c, d). The mixing diagram indicated export, the use of EMMA to differentiate water arriv-
that the solute content in precipitation was modified as ing at streams and how water sources are altered by
it reached the soil surface and that as overland flow deforestation has the potential to significantly increase
developed it acquired a chemical fingerprint different our understanding of the watershed-scale biogeochemi-
from that of any other watershed source. The mechanism cal consequences of this increasingly important land use
that controls the development of that distinct signature change.
is not known but may involve incorporation of solutes
from dry deposition (Germer et al., 2007) or evaporites
from the soil surface. A considerable fraction of pasture ACKNOWLEDGEMENTS
stream flow observations lay close to the overland flow We thank Alexandra A. Montebelo and Gustavo G. Baldi
end member on the mixing diagram but outside the at the Centro de Energia Nuclear na Agricultura in Piraci-
mixing domain established by the chosen end members. caba for conducting the chemical analyses. Shelby Hay-
In an application of EMMA in north-western Australia, hoe, Sonya Remington, Mathew Shamey, Tobias Vetter,
storm flow observations exhibited a similar pattern, with and Lisa Werther helped during the field work phase of
extreme solute content relative to median overland flow this study. Luis F. Chaves provided valuable guidance
composition, not explainable by a combination of other with Monte Carlo simulations and likelihood analyses.
water sources (Elsenbeer et al., 1995). This was attributed Three anonymous reviewers provided helpful criticism
to the “flushing effect”, which is analogous to the and insights that greatly improved the final product. Sup-
incorporation of solutes from the soil surface (Elsenbeer port for this study was provided by the US National Sci-
et al., 1995). It is possible that some variability in solute ence Foundation (DEB-0315656), the NASA LBA Pro-
content that overland flow displayed in the pasture caused gram (NCC5-690) and by grants from Brazilian agencies
by this or other mechanisms, was missed. FAPESP (# 03/13172-2) and CNPq (# 420199/2005-5).
The basic hydrological measurements indicated that Sonja Germer acknowledges travel support by the DAAD
sources contributing to stream flow changed considerably (German Academic Exchange Service). Special thanks to
following deforestation. Surface flow increased signif- the Schmitz family for allowing us to work on their land.
icantly after the establishment of pasture, a transition
that was accompanied by a shift towards predominantly
REFERENCES
overland flow as a source of streamflow in increasingly
responsive catchments. Given the current rates of forest Achard F, Eva HD, Stibig HJ, Mayaux P, Gallego J, Richards T,
Malingreau JP. 2002. Determination of deforestation rates of the
conversion to cattle pasture, and the persistence of this world’s humid tropical forests. Science 297: 999– 1002. DOI:
modified land cover in the Amazon region over decades, 10Ð1126/science.1070656.
this result suggests there is great potential for alteration of Adam K. 2001. Soil–landscape relationship in Rondônia, Brazil. MSc
thesis, University of Bern, Switzerland.
the hydrological budgets of larger watersheds as defor- Biggs TW, Dunne T, Muraoka T. 2006. Transport of water, solutes and
ested land occupies larger fractions of total watershed nutrients from a pasture hillslope, southwestern Brazilian Amazon.
area in an increasing number of locations (Costa et al., Hydrological Processes 20: 2527– 2547. DOI: 10Ð1002/hyp.6214.
Bonell M. 2005. Runoff generation in tropical forests. In Forests, Water
2003; Costa, 2005). and People in The Humid Tropics, Bonell M, Bruijnzeel LA (eds).
Deforestation’s effect of altering the sources of water Cambridge University Press: Cambridge; 314– 406.
Bonell M, Gilmour DA. 1978. The development of overland flow in a
to streams may have other important consequences. The tropical rainforest catchment. Journal of Hydrology 39: 365– 382.
routes that rainfall follows en route to streams con- Boyer EW, Hornberger G, Bencala E, McKnight DM. 1997. Response
trols the amount and form of solutes in stream water characteristics of DOC flushing in an alpine catchment. Hydrological
Processes 11: 1635– 1647.
because flowpaths differ in water volume, flow veloc- Bruijnzeel LA. 2004. Hydrological functions of tropical forests: not
ity and contact with reactive surfaces, redox condi- seeing the soil for the trees? Agriculture, Ecosystems & Environment
tions and chemical environments (Creed et al., 1996; 104: 185– 228. DOI: 10Ð1016/[Link].2004Ð01Ð015.
Burns DA, McDonnell JJ, Hooper RP, Peters NE, Freer JE, Kendall C,
Boyer et al., 1997; Findlay et al., 2001; McGlynn and Beven K. 2001. Quantifying contributions to storm runoff through
McDonnell, 2003). Overland flow has the potential to end-member mixing analysis and hydrologic measurements at the
deliver water with higher solute concentrations gener- Panola Mountain Research Watershed (Georgia, USA). Hydrological
Processes 15: 1903– 1924. DOI: 10Ð1002/hyp.246.
ated by erosion or by bypassing sites of solute retention Christophersen N, Hooper RP. 1992. Multivariate analysis of streamwa-
or removal in deep flowpaths. At the Nossa Senhora ter chemical data: The use of principal components analysis for the
end-member mixing problem. Water Resources Research 28: 99–107.
ranch, for example, also in central Rondônia, export Christophersen N, Neal C, Hooper RP, Vogt RD, Andersen S. 1990.
of water in shallow flowpaths in pasture led to water- Modelling streamwater chemistry as a mixture of soilwater end-
shed export of phosphorus associated with soil particu- members–a step towards second-generation acidification models.
Journal of Hydrology 116: 307–320.
lates (Biggs et al., 2006). In the eastern Amazon, greater Costa MH. 2005. Large-scale hydrological impacts of tropical
water delivery in shallow flowpaths has been hypothe- forest conversion. In Forests, Water and People in the Humid
sized to lead to greater delivery of cations from shal- Tropics, Bonell M, Bruijnzeel LA (eds). Cambridge University Press:
Cambridge; 590– 597.
low soil layers in watersheds that contain significant Costa MH, Botta A, Cardille JA. 2003. Effects of large-scale changes
deforested area (Markewitz et al., 2001). Because water in land cover on the discharge of the Tocantins River, Southeastern

Copyright  2007 John Wiley & Sons, Ltd. Hydrol. Process. 22, 1766– 1775 (2008)
DOI: 10.1002/hyp
 RUNOFF SOURCES IN SMALL AMAZON WATERSHEDS 1775

Amazonia. Journal of Hydrology 283: 206–217. DOI:10Ð1016/S0022- Leal JWL, Silva GH, dos Santos DB, Texeira W, de Lima MIC,
1694(03)00267-1. Fernandes CAC, do C. Pinto A. 1978. Geologia—Folha SC.20 Porto
Creed IF, Band LE, Foster NW, Morrison IK, Nicolson JA, Semkin RS, Velho. Projeto Radambrasil, 17: 219–355.
Jeffries DS. 1996. Regulation of nitrate-N release from temperate Liu F, Williams MW, Caine N. 2004. Source waters and flow paths
forests: A test of the N flushing hypothesis. Water Resources Research in an alpine catchment, Colorado Front Range, United States. Water
32: 3337– 3354. Resources Research 40: W09401. DOI: 10Ð1029/2004wr003076.
Dunne T. 1978. Field studies of hillslope flow processes. In Hillslope Margulis S. 2003. Causes of Deforestation of the Brazilian Amazon. The
Hydrology, Kirkby M (ed.). Wiley-Interscience: New York; 227– 293. World Bank: Washington, DC.
Elsenbeer H. 2001. Hydrologic flowpaths in tropical rainforest Markewitz D, Davidson EA, Figueiredo RdO, Victoria RL, Krusche AV.
soilscapes—a review. Hydrological Processes 15: 1751– 1759. DOI: 2001. Control of cation concentrations in stream waters by surface soil
10Ð1002/hyp.237. processes in an Amazonian watershed. Nature 410: 802– 805. DOI:
Elsenbeer H, Lack A. 1996. Hydrometric and hydrochemical evidence for 10Ð1038/35071052.
fast flowpaths at La Cuenca, Western Amazonia. Journal of Hydrology McClain ME, Boyer EW, Dent CL, Gergel SE, Grimm NB, Groff-
180: 237–250. man PM, et al. 2003. Biogeochemical hot spots and hot moments at the
Elsenbeer H, Lorieri D, Bonell M. 1995. Mixing model approaches to interface of terrestrial and aquatic ecosystems. Ecosystems 6: 301–312.
estimate storm flow sources in an overland flow-dominated tropical DOI: 10Ð1007/s10021-003-0161-9.
rain forest catchment. Water Resources Research 31: 2267– 2278. McGlynn B, McDonnell J. 2003. Role of discrete landscape units in
Elsenbeer H, Vertessy RA. 2000. Stormflow generation and flowpath controlling catchment dissolved organic carbon dynamics. Water
characteristics in an Amazonian rainforest catchment. Hydrological Resources Research 39: 1090. DOI:10Ð1029/2002WR001525.
Processes 14: 2367– 2381. DOI: 10Ð1002/1099-1085(20001015) Moraes JM, Schuler AE, Dunne T, Figueiredo RO, Victoria R. 2006.
14 : 14<2367:AID-HYP107>3Ð[Link];2-H. Water storage and runoff processes in plinthic soils under forest and
Faminow MD. 1998. Cattle, Deforestation and Development in the pasture in Eastern Amazonia. Hydrological Processes 20: 2509– 2526.
Amazon—an Economic, Agronomic and Environmental Perspective. DOI: 10Ð1002/hyp.6213.
CAB International: New York. Mulholland PJ. 1993. Hydrometric and stream chemistry evidence of
Findlay S, Quinn JM, Hickey CW, Burrell G, Downes M. 2001. Effects three storm flowpaths in Walker Branch watershed. Journal of
of land use and riparian flowpath on delivery of dissolved organic Hydrology 151: 291– 316.
carbon to streams. Limnology and Oceanography 46: 345–355. Mulholland PJ, Wilson GV, Jardine PM. 1990. Hydrogeochemical
Genereux DP, Hemond HF, Mulholland PJ. 1993. Use of radon-222 and response of a forested watershed to storms: effects of preferential
calcium as tracers in a three-end-member mixing model for streamflow flow along shallow and deep pathways. Water Resources Research 26:
generation on the West Fork of Walker Branch Watershed. Journal of 3021– 3036.
Hydrology 142: 167– 211. Neill C, Melillo JM, Steudler PA, Cerri CC, De Moraes JM, Pic-
Germer S, Elsenbeer H, Moraes JM. 2006. Throughfall and temporal colo MC, Brito M. 1997. Soil carbon and nitrogen stocks following
trends of rainfall redistribution in an open tropical rainforest, south- forest clearing for pasture in the southwestern Brazilian Amazon. Eco-
western Amazonia (Rondônia, Brazil). Hydrology and Earth System logical Applications 7: 1216– 1225.
Sciences 10: 383–393. Nepstad D, Stickler CM, Almeida OT. 2006. Globalization of the Ama-
Germer S, Neill C, Krusche AV, Gouveia-Neto SC, Elsenbeer H. 2007. zon soy and beef industries: opportunities for conservation. Conserva-
Seasonal and within-event dymamics of rainfall and throughfall tion Biology 20: 1595– 1603. DOI:10.1111/j.1523-1739.2006.00510.x.
chemistry in an open tropical rainforest in Rondônia, Brazil. Peck AJ, Williamson DR. 1987. Effects of forest clearing on
Biogeochemistry DOI:10.1007/s10533-007-9152-9. groundwater. Journal of Hydrology 94: 47–65.
Godsey S, Elsenbeer H, Stallard R. 2004. Overland flow generation in Peixoto de Melo D, da Costa RC, Natali Filho T. 1978. Geomorfologia—
two lithologically distinct rainforest catchments. Journal of Hydrology Folha SC.20 Porto Velho. Projeto Radambrasil, 16: 187–250.
295: 276–290. DOI:10Ð1016/[Link].2004Ð03Ð014. Pires JM, Prance GT. 1986. The vegetation types of the Brazilian
Hill AR, Devito KJ, Campagnolo S, Sanmugadas K. 2000. Subsurface Amazon. In Key Environments: Amazônia, Prance GT, Lovejoy TM
denitrification in a forest riparian zone: Interactions between hydrology (eds). Pergamon: Oxford; 109–129.
and supplies of nitrate and organic carbon. Biogeochemistry 51: Reyment RA, Joreskog KG. 1993. Applied Factor Analysis in the Natural
193– 223. Sciences. Cambridge University Press; 384.
Hodnett MG, da Silva LP, da Rocha HR, Cruz Senna R. 1995. Seasonal Schellekens J, Scatena FN, Bruijnzeel LA, van Dijk AIJM, Groen MMA,
soil water storage changes beneath central Amazonian rainforest and van Hogezand RJP. 2004. Stormflow generation in a small rainforest
pasture. Journal of Hydrology 170: 233– 254. catchment in the Luquillo Experimental Forest, Puerto Rico. Hydro-
Hooper RP. 2001. Applying the scientific method to small catchment logical Processes 18: 505– 530. DOI: 10Ð1002/hyp.1335.
studies: a review of the Panola Mountain experience. Hydrological Simon MF, Garagorry FL. 2005. The expansion of agriculture in
Processes 15: 2039– 2050. DOI: 10Ð1002/hyp.255. the Brazilian Amazon. Environmental Conservation 32: 203–212.
Hooper RP. 2003. Diagnostic tools for mixing models of stream DOI:10Ð1017/S0376892905002201.
water chemistry. Water Resources Research 39: 1055. DOI:10Ð1029/ Sobieraj JA, Elsenbeer H, Coelho RM, Newton B. 2002. Spatial
2002WR001528. variability of soil hydraulic conductivity along a tropical rainforest
INPE. 2006. Monitoramento da floresta Amazônica Brasileira por satelite: catena. Geoderma 108: 79–90.
Projeto PRODES. Instituto Nacional de Pesquisas Espacias. Retrieved Soil Survey Staff. 1999. Soil Taxonomy—A Basic System of Soil
February 12, 2006, from [Link] Classification for Making and Interpreting Soil Surveys, 2nd edn. US
Jipp PH, Nepstad DC, Cassel DK, Reis de Carvalho C. 1998. Deep soil Government Printing Office: Washington, DC.
moisture storage and transpiration in forests and pastures of seasonally- Venables WN, Ripley BD. 2002. Modern Applied Statistics with S .
dry Amazonia. Climatic Change 39: 395– 412. Springer: New York.
Joerin C, Beven KJ, Iorgulescu I, Musy A. 2002. Uncertainty in Wright IR, Gash JHC, da Rocha HR, Roberts JM. 1996. Modelling
hydrograph separations based on geochemical mixing models. Journal surface conductance for Amazonian pasture and forest. In Amazonian
of Hydrology 255: 90–106. Deforestation and Climate, Gash JHC, Nobre CA, Roberts JM,
Johnson MS, Lehmann J, Couto EG, Filho JPN, Riha SJ. 2006. DOC Victoria R (eds). Wiley: Chichester; 437–458.
and DIC in flowpaths of Amazonian headwater catchments Zimmermann B, Elsenbeer H, De Moraes JM. 2006. The influence
with hydrologically contrasting soils. Biogeochemistry 81: 45. of land-use changes on soil hydraulic properties: Implications for
DOI:10Ð1007/s10533-006-9029-3. runoff generation. Forest Ecology and Management 222: 29– 38.
Kirkby M, Callan J, Weyman D, Wood J. 1976. Measurement and DOI:10Ð1016/[Link].2005Ð10Ð070.
modeling of dynamic contributing areas in very small catchments
Working Paper No. 167, University of Leeds, School of Geography,
Leeds, p. 39.

Copyright  2007 John Wiley & Sons, Ltd. Hydrol. Process. 22, 1766– 1775 (2008)
DOI: 10.1002/hyp

View publication stats