Nat Hazards (2012) 60:831–864

DOI 10.1007/s11069-011-9872-y

ORIGINAL PAPER

Evaluating roadside rockmasses for rockfall hazards

using LiDAR data: optimizing data collection
and processing protocols

Matthew J. Lato • Mark S. Diederichs • D. Jean Hutchinson •

Rob Harrap

Received: 29 August 2010 / Accepted: 2 June 2011 / Published online: 18 June 2011
Ó Springer Science+Business Media B.V. 2011

Abstract Highways and railroads situated within rugged terrain are often subjected to the
hazard of rockfalls. The task of assessing roadside rockmasses for potential hazards typ-
ically involves an on-site visual investigation of the rockmass by an engineer or geologist.
At that time, numerous parameters associated with discontinuity orientations and spacing,
block size (volume) and shape distributions, slope geometry, and ditch profile are either
measured or estimated. Measurements are typically tallied according to a formal hazard
rating system, and a hazard level is determined for the site. This methodology often
involves direct exposure of the evaluating engineer to the hazard and can also create a
potentially non-unique record of the assessed slope based on the skill, knowledge and
background of the evaluating engineer. Light Detection and Ranging (LiDAR)–based
technologies have the capability to produce spatially accurate, high-resolution digital
models of physical objects, known as point clouds. Mobile terrestrial LiDAR equipment
can collect, at traffic speed, roadside data along highways and rail lines, scanning continual
distances of hundreds of kilometres per day. Through the use of mobile terrestrial LiDAR,
in conjunction with airborne and static systems for problem areas, rockfall hazard analysis
workflows can be modified and optimized to produce minimally biased, repeatable results.
Traditional rockfall hazard analysis inputs include two distinct, but related sets of variables
related to geological or geometric control. Geologically controlled inputs to hazard rating
systems include kinematic stability (joint identification/orientation) and rock block shape
and size distributions. Geometrically controlled inputs include outcrop shape and size,
road, ditch and outcrop profile, road curvature and vehicle line of sight. Inputs from both
categories can be extracted or calculated from LiDAR data, although there are some
limitations and special sampling and processing considerations related to structural char-
acter of the rockmass, as detailed in this paper.

Keywords Lidar
Rockfall
Rockmass
Characterization
Risk
Hazard

M. J. Lato
Norwegian Geotechnical Institute, Oslo, Norway

M. S. Diederichs (&)
D. J. Hutchinson
R. Harrap

Geological Sciences and Geological Engineering, Queen’s University, Kingston, ON, Canada
e-mail: [email protected]

123
832 Nat Hazards (2012) 60:831–864

1 Introduction: rockfall hazards

Geological hazards related to ground movement that directly affect the safety of motorist
and highway infrastructure include, but are not limited to rockfalls, rockslides, debris flows
and landslides. This paper specifically deals with hazards associated with rockfalls from
cut slopes. Natural slopes will be addressed but are not the direct focus of this research. A
rockfall is an event in which a rockmass fails, and a rock block(s) is released and falls
under gravity. The rock block will free fall, bounce or roll, depending on the gradient of the
slope (Derron et al. 2005). A rockfall is distinct from a rock avalanche, which typically
involves [105 m3 of rock material moving with flow-like characteristics (Holm and Jakob
2009). A rockfall hazard is a rockfall that places persons or infrastructure in potential peril.
The paper will begin with an introduction to the concept of rockfall hazard management
systems; discuss state-of-practice engineering assessment, the influence of geological
terrain and address LiDAR as a technology to aid in geotechnical engineering evaluations.
This will be demonstrated through the use of LiDAR equipment and LiDAR data as a
rockfall hazard management tool in different geological settings. The focus is on details
pertaining to optimization of LiDAR scanning and processing procedures for engineering-
based applications.

1.1 Nature of the problem

Modern highways and roadways in countries across the world traverse mountainous terrain
and are subject to roadside rockfall hazards (Abbott et al. 1998). Figure 1 displays
examples of highways in Canada, Italy and Norway in which passing motorists are subject
to roadside rockfall hazards. Rockfall events have the potential to cause obstruction,
damage and closure to the highway, as well as personal injury or death to motorists (Moore
1986; Bunce et al. 1997). Due to modern safety regulations and expectations of motorists,
rockfall hazard identification and mitigation are integral to transportation corridor design
and maintenance.
Roadside rockfall hazards typically manifest as rockmass failure along existing dis-
continuities, rock blocks as defined by the intersection of multiple discontinuities or dif-
ferential weathering and ravelling (Maerz 2000; Vanderwater, et al. 2005). Geological
discontinuities or structures that are typically present in any given rockmass are in the form
of randomly distributed joints or joint set families, faults, bedding or cleavage. It is the

Fig. 1 High rockfall hazard transportation corridors: a a highway near Vancouver, Canada; b a highway
near Turin, Italy; c a railway near Oslo, Norway

123
Nat Hazards (2012) 60:831–864 833

relationship between geological structure, lithology (including rockmass properties) and

slope configuration that determines potential kinematic instability in a rockmass. In
addition to physical conditions and rockmass properties, the environment, climate, seis-
micity and presence of water greatly influence the stability of a rockmass.

1.2 Hazard ratings and evaluation

Roadside rockfall hazard detection relies on visual inspection and direct measurement of
rockmass properties and face geometry by the evaluating engineer. Field measurements
typically involve joint orientation, length, persistence, roughness, aperture, filling and
shear strength (Hungr et al. 2003). However, field measurements and qualitative assess-
ment of rockmass properties can be subject to personal interpretation and therefore can
lead to discrepancies when comparing rockmasses evaluated by different engineers.
To aid in consistency of recording and evaluation, there exist numerous actively
employed management systems such as those developed for the Oregon Department of
Transportation—Rockfall Hazard Rating System (RHRS) (Pierson 1991), the State of
Colorado (Stover 1992), Canadian National Railway—Rockfall Hazard and Risk Assess-
ment (RHRA) (Bruce Geotechnical Consultants 1997; Pritchard et al. 2005) and the
Ontario Ministry of Transportation—Rockfall Hazard Rating—Ontario (RHRON)
(Franklin and Senior 1997). Modern 3D imaging and visualization techniques such as
Lidar can provide significant improvements in efficiency, repeatability and recording with
respect to any of these systems. Table 1 outlines the necessary measurements required to
perform a detailed rockfall analysis for a given rockmass based on the above rating
systems.
Rockfall hazard identification and management systems are fundamentally focused on
providing the field engineer with a methodology to discriminate zones of high hazard from
those of minimal hazard. This is done through the analysis of four main elements: situation,
source zone and accumulation, initiation, and runout. By establishing evaluation tech-
niques to quantify each element, rockfall evaluation methodologies can be rigorous and
comprehensive.
The spatial interaction between a potentially hazardous rockmass and a transportation
corridor is determined by assessing the slope, ditch and road geometries.
Rockfall source zone identification is an estimated parameter (used in the RHRS sys-
tem) based on observational evidence that may not always be available or immediately
visible. Periodically, physical evidence is visible in the form of unweathered rock, exposed
by recent rockfall activity, to aid in the identification of source zones. Similarly, accu-
mulation must be recent and visible to be observed by an evaluating field engineer. Where
recent activity has not occurred, source and accumulation zones may be observed, but the
frequency and intensity of the events will not be clear.
Potential for rockfall initiation is analysed through the use of stereonets and rudimentary
kinematic stability analyses. Either conducted in the field or on a computer, discontinuity
orientation measurements, in combination with an estimate of the rockmass strength,
indicate the location of potential kinematic instabilities.
Rockfall runout distance is a parameter that is typically estimated in the field through
observation of accumulated fallen rock fragments and further estimated through numerical
simulation. The accuracy of runout models is highly dependent on the quality and spatial
accuracy of the situational geometric data and developing an understanding of the dynamic
behaviour of the falling rocks.

123
 834

123
Table 1 Rockfall hazard rating input parameters
Input variable Description/use RHRS RHRA RHRON Evaluation method

Field engineer Extracted

from LiDAR

Discontinuity structures Dip/dip direction measurements for kinematic evaluation X X X Measured Measured
Discontinuity spacing Distribution of interval between discontinuities X X X Measured/estimated Measured
Discontinuity persistence Persistence and interaction between discontinuity sets X X X Measured/estimated Measured
Block shape distribution Distribution of potential failed rock block shapes X Estimated Modelled
Block volume distribution Distribution of potential failed rock block volume X X Estimated Modelled
Curvature Corridor curvature and line of sight at any given point X X X Estimated Measured
Holistic sense of hazard potential Intuition or ‘‘feel’’ for the hazard X X X Visualized N/A
Rockfall runout distance Distance failed rock blocks will travel X X X Estimated Modelled
Slope angle and ditch configuration Slope geometry and corridor interaction X X X Basic profile Detailed profile
Zones of accumulation Zones where failed rock blocks accumulate or travel through X X X N/A Measured
Zones of rock block release Zones where failed blocks have been released from X N/A Measured
Nat Hazards (2012) 60:831–864
Nat Hazards (2012) 60:831–864 835

1.3 LiDAR technologies

LiDAR is an accurate (Mechelke et al. 2007) rapid distance measurement technique that
can be used to constrain 3D models of physical objects in their natural environment
(Buckley et al. 2008; Amann et al. 2001). LiDAR data are collected in the form of point
clouds where each point is referenced in an XYZ space. The 3D space can be in a locally
defined, site-specific coordinate system or a global coordinate system such as Universal
Transverse Mercator (UTM). LiDAR data for structural geological evaluation can be
collected from static terrestrial sensors mounted on tripods (Sturzenegger and Stead
2009a), low- and high-altitude aircraft (Janeras et al. 2004), mobile terrestrial vehicles
(Lato et al. 2009a) and mobile aquatic vehicles. If LiDAR data from different systems are
collected in common geographical coordinate systems, data can be fused and viewed (Lato
et al. 2009a; Iavarone and Vagners 2003).
In recent years, research has led to significant advancement for the employment of
LiDAR for geological applications (Pringle et al. 2004; McCaffrey et al. 2005; Bonnaffe
et al. 2007). The research of interest pertains to structural geological feature extraction
(Vosselman et al. 2004; Trinks et al. 2005; Feng and Röshoff 2004; Roncella and Forlani
2005; Sturzenegger and Stead 2009b), automated feature extraction (Lato et al. 2008),
rockmass deformation (Abellán et al. 2006, 2009; Rosser et al. 2005; Lemy et al. 2006;
Donovan and Raza 2008), stratigraphic modelling (Bellian et al. 2005), face mapping in
underground mines (Lemy and Hadjigeorgiou 2004) and rockfall evaluation and man-
agement (Turner et al. 2006; Janeras et al. 2004; Trevisani et al. 2009).
The potential uses of the LiDAR data in geotechnical hazard evaluation include geo-
mechanical analysis and differential monitoring, to support or augment traditional road
geometry surveys. The advantages of implementing LiDAR data for rockfall hazard
analysis are related to two features intrinsic to the system: full spatial coverage during data
collection creating a ‘permanent’ record of the rockmass at the time of the survey and
analytical flexibility during data processing. Furthermore, LiDAR data collection, whether
it is mobile or static, airborne or terrestrial, does not require the engineer to be directly
exposed to the hazard in challenging environments.
The research conducted to achieve the results presented in this paper used LiDAR data
collected by a static Leica HDS6000 (Leica 2007), a static Optech IlRIS 36D (Optech
2008), the mobile terrestrial unit Tactical Infrastructure and Terrain Acquisition Network
(TITAN) (Ambercore 2009) and a low-altitude helicopter airborne system (Ambercore
2007).

1.3.1 Static data

The static systems used in this project were specifically selected for high-resolution geo-
technical imaging capability. Both the Leica and Optech IlRIS 36D systems are capable of
generating high-resolution 3D scenes, which have photorealistic appearance from the
direction of the scan orientation. The data generated from these systems were used for
detailed structural analysis and differential monitoring.

1.3.2 Mobile terrestrial data

A mobile terrestrial system, TITAN, was used in this project to collect entire sections of
transportation corridors without disruption to regular track schedules or highway traffic.
The TITAN data were used for the identification of hazardous sections along the corridors

123
836 Nat Hazards (2012) 60:831–864

Fig. 2 a LiDAR data collection platforms: mobile terrestrial, static terrestrial and airborne with data c
generated from different data collection platforms; b oblique view of the data from a helicopter scanner;
c mobile terrestrial scanner, TITAN; d profile view from data collected from a mobile terrestrial data fused
with static terrestrial data, TITAN (surrounds) and HDS6000 (detailed path); e high-resolution static
terrestrial data, HDS6000

collected; preliminary geomechanical evaluations and rockfall simulations were conducted

using these data as geometric control. TITAN data were collected from a mobile platform
with survey grade Global Positioning System (GPS) for positioning (El-Rabbany 2006) and
Inertial Navigation System (INS) (Yang and Farrell 2003) for control when there was
insufficient view of the sky or available GPS satellites to constrain the positioning network
(Alshawa et al. 2007; Barber et al. 2008; Glennie 2007a, b).

1.3.3 Low-altitude helicopter data

The airborne system, ALMIS, was employed at low altitudes to generate data with max-
imum density on the ground surface, approximately 10 pts/m2. These data are used for
developing general overviews of the subject areas as well as identification of scarps,
vertical cliffs and possibly active rockfall zones.

1.4 Data fusion

Static terrestrial data can be collected with or without the aid of GPS. If the data are
collected with GPS, integration with other georeferenced sources should be fairly
straightforward. Static terrestrial data collected without GPS can be integrated with mobile
terrestrial or airborne methods, as demonstrated by Lato et al. (2009a) and illustrated in
Fig. 2 through the identification and matching of common features visible in both data sets.
Figure 2 illustrates an east-looking, oblique aerial view of a site under investigation.
This site was collected with TITAN and the Leica high-resolution static data. All data are
registered in a common UTM coordinate space. The data sets presented have been pro-
cessed using both Cyclone (Leica 2008) and PolyWorks (InnovMetric 2008) software.
The flow chart as shown in Fig. 3 demonstrates a typical rockfall risk management
procedure that uses inputs derived primarily from LiDAR data. The left side of the chart
shows input parameters essential to calculating probability of hazard, while the right side
illustrates the cumulative hazard probabilities that are necessary calculations for deter-
mining risk. The philosophy of this system is that through the calculation of risk values,
independent sites can be compared with one another and ranked in terms of severity of
rockfall potential so that effective mitigation programmes can be implemented. The system
presented in Fig. 3 is only a framework; there are no associated weighting factors.
Figure 3 identifies a set of rockfall hazard rating input parameters that can be measured,
calculated or modelled directly from sufficiently dense LiDAR data. Conducting digital
measurements on LiDAR data reduces the possibility of error in measurement technique,
recording and data transfer. Additionally, every measurement made from the LiDAR data
can be inspected or refined at any time in the future.
The input parameters defined as calculated in the legend of Fig. 3 are directly deter-
mined from the LiDAR data using algorithms commonly available in spatial analysis and
GIS tools such as ArcGIS (ESRI 2008). In the absence of LiDAR data, these would be
determined using commonly available, small-scale, 1:20,000 digital elevation models or
would be estimated in the field. Using spatial analysis on dense data ensures that the
calculated values are not subject to parameters or situations that could possibly introduce

123
Nat Hazards (2012) 60:831–864 837

123
838 Nat Hazards (2012) 60:831–864

Fig. 3 A typical rockfall hazard management workflow for railways. Inputs are shaded according to how
the information can be derived from LiDAR data

sensitivity or variable results, such as improper geometries as would be introduced when

using low-resolution 1:20,000 data.
The input parameters shown as modelled in Fig. 3 are determined based on the inte-
gration of the measured values into numerical models. These traditionally modelled values
are qualitatively estimated in the field through visual inspection of the rockmass. The use
of numerical models eliminates observer and access bias associated with field estimation.
As demonstrated in Table 1, there are benefits and limitations of each evaluation
method, including traditional field and LiDAR data analyses. This paper explains LiDAR
as a technology for geotechnical analysis and demonstrate, in fundamentally distinct
rockmasses, the ability to use LiDAR for hazard evaluation.
The main drawback of LiDAR data is its inability to provide an experienced evaluating
engineer with a ‘‘feel’’ of the hazard level. However, with advancements in photorealistic
LiDAR technologies, wherein still or digital imagery is automatically overlaid on the point
cloud, this will perhaps change.

2 Field sites

Rockfall hazard management systems must be able to identify hazards in different geo-
logical settings. In natural systems, lithology, exposure and type of rock cut all contribute

123
Nat Hazards (2012) 60:831–864 839

to the probability of a rock block fall. Rockmass lithology and setting determine fracture
style and in turn possible and actual failure mechanisms, whether toppling, sliding, wedge
or ravelling.
LiDAR sensors typically view an exposed rockmass from a location on or along the
highway or railway corridor. This can result in geomechanical features that inhibit the
ability of LiDAR to image the rockmass without occlusion.

2.1 Rockfall evaluation in geological domains

Traditional rockfall hazard management systems, through the use of field-based rockfall
evaluation protocols, are designed in a ‘one-size fits-all’ manner. This essentially means
that a ravelling limestone will be evaluated using the same techniques as massively jointed
granite or steeply dipping shale. This can pose difficulties if LiDAR is employed without
considering basic structural character.
Rockmasses can be rudimentarily classified into three distinct groups based on physical
conditions: layered sedimentary-, low anisotropy- and high horizontally layered sedi-
mentary-, low structural anisotropy-, and high structural anisotropy rockmasses. Although
each group may contain numerous subgroups, the rockmass behaviour and failure mode of
each general category share commonalities and can be evaluated using similar techniques.
When using digital methods, such as LiDAR, rockmasses can be classified in terms of
failure mechanics and evaluated according to optimize processing efficiency.

2.1.1 Horizontally layered sediments

Horizontally layered sediments, such as undisturbed limestone, commonly exhibit exten-

sive and persistent horizontal jointing crosscut by two subvertical discontinuity sets. The
subvertical discontinuity sets commonly terminate at the interface with horizontal dis-
continuities (Priest 1993). These rockmasses, if undisturbed and where the primary dis-
continuity remains horizontal, tend to fail through a ravelling process. It is most important
to identify incompetent layers, source zones, overhangs and runout paths for these
rockmasses.

2.1.2 Low structural anisotropy

Low structural anisotropy rockmasses can be sedimentary, igneous or metamorphic in

origin. The most discriminating classification feature of such rocks is three discontinuity
sets with mutually orthogonal orientations. These orthogonal discontinuity sets tend to
produce blocky rockmasses. These rockmasses when observed on surface, such as, along
transportation corridors, typically fail on existing structure in a kinematically predictable
fashion. Thus, it is critically important for the evaluating engineer to measure discontinuity
orientations, spacing and persistence to assess the potential hazard.

2.1.3 High structural anisotropy

High structural anisotropy rockmasses are typically metamorphosed and can have a sed-
imentary or igneous protolith. These rockmasses typically display multiple discontinuity
sets at no particular orientation with respect to each other. It is also common for these
rockmasses to be dominated by one discontinuity set that is coplanar to a persistent

123
840 Nat Hazards (2012) 60:831–864

metamorphic fabric. These rockmasses must be critically evaluated for kinematic insta-
bilities as well as ravelling failure. High anisotropic rockmasses generally pose the greatest
challenge to the evaluating engineer due to the variable nature of the discontinuities and
rockmass conditions.

2.2 Case study locations

In order to determine optimal scanning practices and methodologies for different rock-
masses within a consistent management framework, five sites in Ontario, Canada were
selected for geotechnical investigation. Each of the geologically unique sites will be used
to demonstrate a component of the rockfall hazard management workflow illustrated in
Fig. 3. Sites 1–4 are located in the Frontenac Arch region. Precambrian basement rocks,
Paleozoic sedimentary units and Cenozoic sediments characterize the region. Site 5 is
located within the Superior Province of the Canadian Shield. All sites have undergone
previous glaciations and exhibit evidence of weathering.

2.2.1 Site 1: Orthogneiss: Sunbury, Ontario

The first unit under investigation is an 1,140–1,180 Ma upper-amphibolite-grade orthog-

neiss (Carr et al. 2000). It is exposed within the Frontenac Arch, roughly 30 km north of
Kingston, Ontario, and is part of the Frontenac Terrain of the Paleoproterozoic Grenville
Province. The protolith was andesitic to basaltic in composition. The rock displays no
visible evidence of folding although it has a very strong planar fabric (straight gneiss), and
folds may have been transposed. The fracturing within the outcrop is well defined
throughout by three main joint sets, two of which are near vertical and one horizontal, as
visible in Fig. 4. This represents a rockmass with low level of structural anisotropy.

2.2.2 Site 2: Mixed ortho- and paragneiss: Morton, Ontario

The second rock cut under investigation consists of a mixed ortho quartzofeldspathic
paragneiss, also located within the Frontenac Arch, roughly 60 km north of Kingston,

Fig. 4 Photograph of an orthogneiss outcrop located within the Frontenac Arch, 30 km north of Kingston,
Ontario along Provincial Highway 15

123
Nat Hazards (2012) 60:831–864 841

Fig. 5 Photograph of a quartzofeldspathic gneiss located within the Frontenac Arch, roughly 60 km north
of Kingston, Ontario along Provincial Highway 15

Ontario. The unit contains remnant pegmatite lenses and what appear to be quartzite layers.
It is highly sheared with no direct visual evidence of folding. Transverse dykes have been
transposed into the straight gneiss fabric. The 1,180–1,140 Ma unit is metamorphosed to
amphibolite facies (Carr et al. 2000). The rockmass is fractured by four main joint sets and
numerous random fractures, as visible in Fig. 5. This represents a rockmass with a high
level of structural anisotropy.

2.2.3 Site 3: Limestone on calcsilicate gneiss: Inverary, Ontario

The third rock cut under investigation is subdivided into two geologically distinct sections.
The outcrop itself is located roughly 30 km north of Kingston, Ontario near Inverary. The
upper unit is a Middle Ordovician limestone, a member of the Black River Group (McFall
1993). The lower unit is a calcsilicate paragneiss comprised of metamorphosed Grenville
sediments. The limestone unit lies, unconformably, on top of the Grenville Precambrian
basement rocks, representing a hiatus of *500 Ma. The lower gneiss layer is metamor-
phosed to the upper amphibolite facies, has constant planar fabrics and has no direct

123
842 Nat Hazards (2012) 60:831–864

Fig. 6 Grenville-aged calcsilicate gneiss overlain by Middle Ordovician–aged limestone, roughly 30 km

north of Kingston, Ontario

evidence of folding as displayed in Fig. 6. This represents a rockmass with a low level of
structural anisotropy.

2.2.4 Site 4: Limestone: Kingston, Ontario

The fourth rock cut under investigation is a south dipping, undeformed Middle Ordovician
limestone located in Kingston, Ontario; the outcrop is commonly known as the ‘‘Barrie-
field’’ outcrop. The section is a member of the Black River Group, a 73-m-thick unit that
can be subdivided into three formations: the 47-m-thick Pamelia, 18-m Lowville and the
8-m Chaumont Formation (McFarlane 1992). Both Sites 3 and 4 are members of the
Pamelia Formation, which is best described as pale grey dolostone and medium to thick
beds of carbonate mudstone and wackestone (Cushing 1908) as displayed in Fig. 7. This
represents a horizontally layered sedimentary rockmass.

2.2.5 Site 5: Felsic metavolcanic: Hearst, Ontario

The fifth rock cut under investigation is a greenschist to amphibolite-grade felsic meta-
volcanic located in the Superior Province 100 km north of Sault Ste. Marie, Ontario. The
region is dominated by northwest-southeast trending structure with parallel fault systems
that are crosscut by a minor fault system oriented in the orthogonal direction. The Agawa
Canyon hosts the Canadian National (CN) Algoma Central Railway (ACR) and as such has

123
Nat Hazards (2012) 60:831–864 843

Fig. 7 Middle Ordovician Black River Group Limestone,, located in Kingston, Ontario

different visibility constraints and logistical challenges. This represents a rockmass with a
high level of structural anisotropy (Fig. 8).

2.3 Geological implications for LiDAR scanning

While it is normally the goal of rockmass assessment systems to be general in nature, the
ability to analyse and extract valuable information from LiDAR data, with accuracy and
efficiency, relies on LiDAR data that is collected in a manner that is optimized to a certain
degree for a given situation and rockmass type. For simplicity, the three broad geological
classes, low structural anisotropy, high structural anisotropy and horizontally layered
sediments, are discussed here as they typically experience unique forms of instability.
Therefore, analysis of LiDAR data for these rockmasses requires specific data collection
and evaluation techniques. Employing scanning practices that will result in optimized data
for the foreseeable rockfall evaluation analyses will improve hazard management and risk
mitigation systems.

123
844 Nat Hazards (2012) 60:831–864

Fig. 8 a Typical terrain within the Superior Province, roughly 100 km north of Sault Ste. Marie, Ontario,
on the Algoma Central Railway; b accumulation of rockfall debris

Roadside rockmasses that exhibit a low degree of structural anisotropy typically fail
along multiple existing discontinuity features of varying orientation. The LiDAR data
should optimally be collected from multiple stations horizontally separated along the
length of the outcrop. This will reduce occlusion in the point cloud and therefore reduce
bias in the evaluation of the discontinuities and resultant kinematic analyses. Where
possible, scans from vertically separated vantage points are advisable as well.
Rockmasses that exhibit a high degree of structural anisotropy, similarly to low
anisotropic rockmasses, typically fail on existing discontinuities. However, these rock-
masses also behave in complex manners due to their anisotropic nature. These rockmasses
must be scanned from multiple spaced locations to reduce bias and occlusions although
these locations should be selected to optimize visibility of the specific fabric. In contrast to
low anisotropic rockmasses, the density of the point cloud must be greater to enable
visualization of small-scale features such as foliation or cleavage planes.
Horizontally layered sediments, such as a minimally disturbed limestone, exhibit
extensive horizontal bedding planes and vertical fractures that terminate on the horizontal
beds. These rockmasses typically fail by ravelling, which naturally leads to undercutting.
To effectively evaluate horizontally layered sedimentary outcrops with LiDAR data, it is
essential to produce minimally occluded point clouds. This can be achieved through
scanning of the outcrop from multiple horizontally and vertically spaced locations.

3 Geotechnical evaluation with LiDAR for rockfall investigation

LiDAR is a tool that will not replace the expertise of a geotechnical engineer. However,
LiDAR data can be used to conduct preliminary analyses and temporal monitoring when
determining the potential hazard due to rockfalls along transportation corridors while
allowing the field engineer to maintain a safe distance. LiDAR data can be used to aid in

123
Nat Hazards (2012) 60:831–864 845

the preliminary assessment and decision process leading to further investigation. In

addition, advanced geotechnical and geospatial modelling can be undertaken using inputs
derived or directly extracted from LiDAR data.

3.1 Geometric profile

Figure 9 illustrates a fully marked up cross-section derived from LiDAR data of the road,
shoulder, ditch and rockmass from Location 2. The cross-section is cut on a virtual plane
perpendicular to the direction of the road. Information derived from cross-sections such as
this can be directly input into a rockfall hazard management system, thus aiding in the
classification process. In the case of mobile scanning, this process can be automatically
executed every ‘X’ number of metres along the corridor. This simple method allows semi-
automatic characterization of rockmass geometry for slopes located close to the road that
require subsequent investigation and analysis.

Fig. 9 a A virtual cross-section (Site 2) cut through the raw point clout data and marked up according to
required rockfall hazard management measurements. b Cross-section through slopes at Site 4: Locations and
dates of data collection events in relation to the slope under investigation for ravelling failure

123
846 Nat Hazards (2012) 60:831–864

3.2 Source and accumulation

Monitoring active slopes for rockfall events is typically conducted through the use of
active technologies. These consist of continual video recording of the rock face (McHugh
2004), seismic monitoring (Zimmer et al. 2008) or trip wires to detect falling blocks
(Arattano and Marchi 2008). The drawback to all of these systems is that they are only
capable of detecting the failure; they are not capable of calculating the size or origin of the
failure. (Video recording methods can detect the origin of the failure, but often the reso-
lution is too low to qualitatively and quantitatively describe the failure.) Recent research
has proven the effective use of LiDAR data in monitoring landslides (McKean and Roering
2004) and potentially active rockfall slopes (Bauer et al. 2005; Lato et al. 2009a; Abellán
et al. 2009). The high-resolution data that LiDAR systems are capable of producing allow
temporal analysis of the data to detect zones that have changed or deformed when com-
paring two or more data sets.
The comparison of independent LiDAR data sets is contingent on two properties: the
data must sufficiently overlap, and there must be zones within the data sets that have not
displaced. The non-displaced zones are critical to facilitate the alignment of non-geore-
ferenced scans. Once these two qualifications have been accounted for and the data sets are
aligned, the differential analysis can be conducted. The ability of the user to discern actual
versus erroneous surface displacement or change is controlled by the accuracy of the
aligned data sources. This can be calibrated by isolating sections of the aligned data sets
that are assumed to be fixed in position.
The ability to detect zones of rock block failure is extremely important when ravelling is
the dominant failure mode within a rockmass. Site 4, described above, is an extensive
outcrop spanning over 250 m in length and 15 m in height. The horizontally layered
sediments typically fail along the bedding plane in a ravelling fashion. Site 4 was scanned
numerous times over a 3-year period to establish a baseline and assess the rockmass for
zones of failure and trends of failure along specify bedding planes. The rockmass at
Barriefield was scanned as illustrated in Fig. 9b.

Table 2 Error analysis of fused LiDAR data sets determined using three scanning sources (HDS6000,
IlRIS 36D and TITAN) over a period of 4 years
Comparison Reference Surface Number RMS error Max error SD
of pts (cm) (cm)
Year Scanner Year Scanner

2009 HDS6000 2006 IlRIS Rockmass 2,585 1.16 4.00 0.89

2009 HDS6000 2007 TITAN Rockmass 340 1.37 4.98 1.22
2009 HDS6000 2008 HDS6000 Rockmass 4,934 0.88 3.39 0.88
2008 HDS6000 2006 IlRIS Rockmass 2,657 1.04 5.70 0.89
2008 HDS6000 2007 TITAN Rockmass 338 1.48 5.00 1.33
2006 IlRIS 2007 TITAN Rockmass 340 1.51 7.73 1.52
2009 HDS6000 2006 IlRIS Sidewalk 3,715 1.15 3.09 0.31
2009 HDS6000 2007 TITAN Sidewalk 2,067 1.79 5.04 1.40
2009 HDS6000 2008 HDS6000 Sidewalk 4,702 0.70 2.17 0.23
2008 HDS6000 2006 IlRIS Sidewalk 3,749 0.56 2.30 0.29
2008 HDS6000 2007 TITAN Sidewalk 2,043 1.45 3.04 1.40
2006 IlRIS 2007 TITAN Sidewalk 2,055 1.43 4.00 1.43

123
Nat Hazards (2012) 60:831–864 847

To establish equipment sensitivity and accuracy, throughout the monitoring period, a

section of sidewalk and curb, and a section of rockmass, both, of which display visually
negligible movement were assessed for movement. The results from these tests are pre-
sented in Table 2. The most accurately matched data sets are those in which scans were
performed 1 year apart using the same system: the Leica HDS6000. The least accurately
matched data sets are those in which data from TITAN and either the HDS6000 or IlRIS
scanner are fused, in other words where there were fundamental differences in scale. As
displayed in the table, the error values reported for the isolated section of the rockmass are
greater than those reported for the sidewalk. This is due to the highly irregular shape of the
rockmass, and the greater likelihood that small rock blocks were released that created noise
when comparing the data sets.
The maximum error between the static scanners for the isolated section of sidewalk is
3.1 cm. This value can thus be used to determine the minimum detectable difference of
approximately 6.2 cm for rock blocks that have displaced or fallen, two times the maxi-
mum error (because errors occur in both the positive and negative directions). The mini-
mum detectability threshold used in the subsequent analysis was thus set at 8 cm, above the
possible ambiguity between noise and actual displacement.
Figure 10 illustrates three distinct rock blocks located on the southern side of the
Barriefield cliff (Site 4) that were released from the slope between 2006 and 2009. The
cliffs on both the north and south side of the road visually exhibit continual ravelling of
rock blocks between the size of 100 and 100,000 cm3. The cliffs were scanned in 2006
with an IlRIS 36D, in 2007 by TITAN, and in 2008 and 2009 with a HDS6000. The
locations of the scanners are illustrated in Fig. 9b, and the cross-section of the slope is
generated from mobile terrestrial (TITAN) data. The data from the different scanners are
aligned and compared using Polyworks V11.0.9 (InnovMetric 2008). All three static scans
were fused with the georeferenced TITAN data to assure spatially accurate and consistent
models.
The volume and mass of the released blocks were determined using two independent
methods. The first method simply measured the dimensions of the rock block and the
subsequent cavity to determine average dimensions; the results are displayed in Table 3.
The second method assumes the rock blocks to be a prolate spheroid, due to the abundance
of rounded to blocky shaped failed rock blocks commonly seen on-site. The cross-sectional
length and width of the spheroid are measured along the plane that would represent the
approximate middle of the failed rock block. The height (or in this case, depth) of the rock
block is determined by calculating the average distance from each of the points that
represent the scarp (in the LiDAR data) to their nearest neighbour in the LiDAR data that
depicts the rock block before failure. The length distance calculated from the LiDAR data
is given a range of ±10% to account for variability in the rockmass and incomplete LiDAR
data. The results of these calculations are displayed in Table 4. The results reported in
Tables 3 and 4 are comparable. Although these methods do not determine the exact vol-
ume of the detached rock block, their accuracy is acceptable for a rockfall hazard evalu-
ation. Typical volume estimations made by a field engineer using a hazard rating system
are divided into bins such as: \1 m3, [1 and \3 m3, [3 m3 (Bruce Geotechnical Con-
sultants 1997).

3.3 Structural kinematics

LiDAR data can be used to digitally measure geological structure. This is most commonly
completed through the analysis of static terrestrial data and mobile data to a lesser extent.

123
848 Nat Hazards (2012) 60:831–864

Fig. 10 Three independent rock block failures visually identified through analysis of LiDAR data collected
at different times (note the low noise visible in the comparative data), black zones represent occluded
sections within the LiDAR data

Geological structural feature identification and extraction from LiDAR is the most com-
monly researched and presented use of LiDAR data for geotechnical applications. Dis-
cussions concerning the application of geological feature extraction have been recently
presented by Pedrazzini and Jaboyedoff (2008), Kemeny and Turner (2008), Fekete et al.
(2009a, b) and Sturzenegger and Stead (2009b).

123
Nat Hazards (2012) 60:831–864 849

Table 3 Estimated block mass determined through measuring visible edges of the block in the LiDAR data
Block number Measured dimensions Volume (cm3) Mass (kg)

D-1 (cm) D-2 (cm) D-3 (cm)

1 16 31 18
18 30 21
21 26
Average 18 31 22 12,000 33
2 27 76 41
30 80 33
21 34
Average 26 78 36 73,000 196
3 80 34 29
82 36 27
25 18
Average 81 31 25 63,000 170

Table 4 Estimated block mass determined through measuring cross-sectional values and surface differ-
entials, assuming a prolate spheroid cavity shape
Block number Measured dimensions Depth differential Volume (cm3) Mass (kg)

D-1 (cm) D-2 (cm) Mean - 10% Mean ? 10% Low High Low High

1 52 24 18 22 11,000 14,000 32 39
2 115 50 19.8 24.2 60,000 73,000 161 197
3 106 46 21.7 25.3 55,000 64,000 150 174

3.3.1 Discontinuity spacing

Discontinuity spacing can be measured in the field, measured from images collected using
standard photography methods or stereo-photographic methods; from LiDAR data directed
or screen images of processed data (Fekete et al. 2009a). Standard photographic analysis
techniques are limited due to the lighting direction that is controlled by either shadows
from the sun or manual lighting angles. Conversely, minimally occluded LiDAR data when
meshed can be artificially lit from any user-defined direction (Figs. 11, 12).
The analysis of discontinuity spacing is extremely important for all potentially unstable
rockmasses. The ability to determine discontinuity spacing enables the engineer to assess
the likely size of potentially unstable blocks, assuming the failure is constrained to a single
block. In a rockmass such as the horizontally layered sediment at Site 4, knowledge of the
bedding thickness will directly enable an estimation of the size of a potentially unstable
block. This assumes the failure is limited to a single bed.
Figure 11 illustrates two LiDAR-based techniques that can be employed to measure
discontinuity spacing. The first method, illustrated on the left-hand side of the figure,
represents a 3D LiDAR model lit from above at a downward angle. The lighting was
positioned to accent the horizontal oriented discontinuity set. The 2D image was ‘sharp-
ened’ and then processed to automatically detect linear traces, in this case horizontal
discontinuities. The flattened image is taken in an orthographically projected space and

123
850 Nat Hazards (2012) 60:831–864

Fig. 11 Discontinuity spacing evaluation using 2D semi-automated trace detection and 3D LiDAR data
evaluation

contains no spatial distortion. As well, the image is taken at an angle that is in direct line of
sight of the discontinuity edge, therefore eliminating the requirement for an angular cor-
rection. The second evaluation method illustrated in the right-hand side of the figure
directly measures the orientation frequency and spacing in the LiDAR data. This process is
completed by digitizing visible discontinuity surfaces in the 3D data and directly mea-
suring the spacing. The results of both methods are presented in Table 5.
Both methods generate similar results; however, the image processing method presents
a powerful option to individually process one discontinuity set at a time with high accuracy
through semi-automated means. The LiDAR processing option provides an excellent visual
demonstration of discontinuity configuration.
A second non-horizontal discontinuity set was analysed using the image analysis
method to demonstrate the use of variable lighting directions. Figure 12 demonstrates two
rockmasses, Site 3 and Site 4, lit from user-defined angles to accentuate non-vertical
discontinuity sets. The left-hand side of the figure identifies vertical discontinuities in a

123
Nat Hazards (2012) 60:831–864 851

Fig. 12 Discontinuities identified and traced using automated detection algorithms, input image is
generated by lighting meshed LiDAR data at selected angles

Table 5 Analysis results of discontinuity detection along 1.75-m sample lines

Sample line Number of discontinuities Average spacing (cm)

2D image 3D LiDAR 2D image 3D LiDAR

analysis analysis analysis analysis

1 6 5 29 35
2 7 7 25 25
3 5 5 35 35

limestone. As expected, the data is comprised of numerous discontinuities terminating on

the primary horizontal discontinuity.
Meshed LiDAR data in combination with user-defined light sources can be used to
discriminate discontinuity sets (Pedrazzini and Jaboyedoff 2008). The upper and lower
models in Fig. 13 are each ‘‘lit’’ by three independently coloured light sources. Each
virtual light source is oriented parallel to the normal vector of each of the main discon-
tinuity sets. The upper image illustrates a section of Site 4. The red light is virtually
positioned below the outcrop pointing in the up direction. This light highlights the hori-
zontal discontinuity surface and accentuates all overhanging sections of the rockmass. The
other two light sources identify the two subvertically dipping minimally persistent

123
852 Nat Hazards (2012) 60:831–864

Fig. 13 Meshed high-resolution LiDAR data uniquely lit to identify discontinuity surfaces. The upper
image (Site 4) accents the overhang (coloured red), while the lower image (Site 3) accents the distinction
between the major discontinuity sets

discontinuity surfaces. This is a powerful tool to assess the instability and potential hazard
of a layered sedimentary outcrop.
The lower image in Fig. 13 illustrates a section of Site 3. The three light sources
uniquely colour the three main discontinuity sets. This image can be used to aid in the
identification of discontinuity sets, as well as measure spacing, and determine discontinuity
persistence through the direct visualization of cross-cutting relationships. Both identifi-
cation of sets and discontinuity persistence are critical to the development of numerical
models.

3.3.2 Structural evaluation

Structural geological features, such as discontinuities or fractures, when their orientations

are measured at the outcrop must be corrected for sampling bias due to the observed
versus true spacing of the discontinuity (Terzaghi 1965; Priest and Hudson 1976; Park and
West 2002). This correction must be made for data collected in the field or derived from
LiDAR data. Similarly, geotechnical information derived from LiDAR data must account
for biases inherent in the collection method employed. As discussed above, LiDAR data
are a line-of-sight technology that produces data with zones of occlusion (i.e. zones
shadowed from the look angle of the scanner) (Sturzenegger et al. 2007). Using mathe-
matically derived geometric correction factors, based on scanner-specific properties, the
bias associated with structural discontinuity measurements can be reduced (Lato et al.
2009b).

123
Nat Hazards (2012) 60:831–864 853

Fig. 14 Meshed LiDAR data with discontinuity surfaces identified within the calcsilicate gneiss unit,
Site #3

Structural information is readily extracted from raw LiDAR data as well as from
meshed data. The true orientation of discontinuity surfaces is extracted in the form of
normal vectors and subsequently converted to geological information such as strike and
dip. Figure 14 illustrates meshed LiDAR data of a competent limestone and calcsilicate
gneiss outcrop in Inverary (Site 3). The calcsilicate gneiss is a low anisotropic rock; thus, it
is prone to sliding and wedge type failure modes. This enables the engineer to directly
assess the rockmass for kinematic instabilities through the investigation of discontinuity
orientations and interactions.
Due to the unconformity in the outcrop at Site 3, the geomechanical analysis was
divided between the limestone and calcsilicate units. The boundary between these is
clearly visible in the LiDAR data. This division of the geomechanical analysis is necessary
because the geological structures are rock type dependant. However, in reality, these
systems are linked: if there is a failure in the lower calcsilicate unit, the upper limestone
will likely then become unstable.

123
854 Nat Hazards (2012) 60:831–864

Fig. 15 Stereonet analysis of the lower calcsilicate unit at Inverary, a original measurements;
b measurements corrected for line of sight and outcrop orientation (sampling) bias

The LiDAR data were collected from three locations along the face of the outcrop to
minimize horizontal occlusion; a Leica HDS6000 was used to collect the data. Over 120
structural discontinuities present in the calcsilicate unit have been digitized from the
LiDAR data and are plotted on a stereonet in Fig. 15 using Dips (Rocscience 2006). The
structural discontinuity surfaces are identified by selecting points assumed to lie on the
surface, and planes are subsequently fit to the points. The structural data has undergone a
detailed bias correction in accordance with correction factors proposed by Lato et al.
(2009a), related to line-of-sight bias, and Terzaghi (1965), related to outcrop exposure bias,
as presented in Fig. 15a (uncorrected) and b (corrected). Both of these correction factors
are calculated based on the dimensions, exposure and orientation of the outcrop under
investigation.
Upon completion of the structural discontinuity evaluation from the LiDAR data, the
information plotted on the stereonet can be evaluated for possible kinematic instabilities.
Figure 16a–c illustrates the potential for topple, planar and wedge failure, respectively.

Fig. 16 Stereographic kinematic failure mode analysis, a topple, b wedge and c planar sliding

123
Nat Hazards (2012) 60:831–864 855

The ability to measure discontinuity orientations in an unbiased manner provides a

robust methodology to enable the unbiased kinematic geotechnical evaluation of a rock-
mass using remotely sensed data.

3.3.3 Rockmass geometry and discrete block modelling

Rockfall hazard management systems such as the Canadian National Railway Rockfall
Hazard and Risk Assessment System (Bruce Geotechnical Consulting 1997) require the
assessment of both the volume of a potential failure and the shape of the rock blocks.
Volume of historical failure can be determined using differential modelling as outlined
previously pending sufficient coverage by high-resolution LiDAR data. To accurately
forecast rock block and shape failure, size requires advanced modelling.
The rockmass at Site 1 is a low anisotropic rockmass that exhibits large block failures.
To assess the rockmass in terms of how future failures may occur can be done through 3D
discrete modelling programs. The LiDAR data are used to constrain the topology of the
model, thus creating a spatially accurate numerical model. Using structural discontinuity
information calculated from the LiDAR data, accurately orientated and spaced fracture
networks can be implemented within 3DEC. The model monitors the rock blocks as they
disengage from the rockmass and accurately distinguishes block shape based on the
internal physical parameters of the rock block and measures the block volume (Kalenchuk
et al. 2006). Figure 17 illustrates a simplified workflow from LiDAR data, to 3DEC, to the
result containing information regarding expected block shape and size distributions (Lato
et al. 2007).

3.4 Rockfall runout

Rockfall modelling for mountainous terrain such as the terrain that surrounds the Algoma
Central Railway (ACR), in Northern Ontario, is dependent on Digital Elevation Model
(DEM) quality, source zone identification, terrain classification, material behaviour and
properties. The Algoma Central Railway (Site 5) is a highly anisotropic rockmass with
complex jointing. The railway alignment is subject to hazards from both cut slopes and
natural slopes and requires extensive and detailed analyses. Primary analysis of cut slopes
for kinematic instabilities has been conducted from mobile terrestrial LiDAR data along
the ACR (Lato et al. 2009a). The probabilistic results of the kinematic analysis can be used
in conjunction with the rockfall modelling results to aid in the calculation of overall hazard
levels.
The capability of mobile terrestrial LiDAR, such as TITAN, is optimized for cut slopes
that are visible from the track. Large mountainous natural slopes are occluded in the
TITAN data set and must be scanned via airborne methods. Airborne methods, as discussed
and illustrated above, are not suitable for detailed geomechanical analysis.
Rockfall modelling is traditionally completed using 2-dimensional (2D) profile models;
new state-of-the-art models such as Rockfall Analyst (Lan et al. 2007, Agliardi and Crosta
2003) incorporate 3D DEM data. However, their use is not optimized for rugged forested
terrain. The difference between readily available 1:20,000 DEM information derived from
air photo interpretation and airborne LiDAR fused with TITAN data is illustrated in
Fig. 18. The 1:20,000 DEM is smooth and does not properly represent vertical cliffs,
horizontal sections such as the track and certainly not features such as ditches. The use of
small-scale DEM data will lead to misrepresentation of potential rockfall hazards.

123
856 Nat Hazards (2012) 60:831–864

Fig. 17 a Minimally occluded photo-overlaid LiDAR point cloud data; b & c 3DEC model with rockmass
face topology controlled by LiDAR data; d estimated block shape and block volume distributions

A rockfall model involves the release of virtual rock blocks of a known size (weight)
down a slope. The rocks are released from a specific location set by the user. The model is
constrained by the known/estimated properties of the slope, which are dependent on
material coefficients of restitution and friction (Perret et al. 2004; Dorren 2003; Bourrier
et al. 2009). Coefficients of restitution control the horizontal and vertical ‘bounciness’ as
the rock as it interacts with the terrain. Failed rock blocks will fall, bounce or roll down a
slope as a function of slope gradient (Ritchie 1963). Accurate modelling of rockfall
depends on sufficient surface model detail with respect to slope profile, as well as dynamic
modelling parameters (Chau et al. 2002).

123
Nat Hazards (2012) 60:831–864 857

Fig. 18 TINs based on two different DEM data sources: 1:20 000 and fused airborne and TITAN LiDAR

3.4.1 Profile rockfall modelling

Traditional 2D profile rockfall models were completed for a hazardous section of track
near Mile 94 on the ACR (Site 5). The 2D ‘‘fall-line’’ was extracted from a Triangulated
Irregular Network (TIN) developed from the LiDAR data using ArcGIS. The TIN used was
developed using two different data sets, as illustrated in Fig. 19. The first data set contains
only bare-earth airborne LiDAR data (Fig. 19b), and the second data set contains bare-
earth airborne data fused with bare-earth TITAN data (Fig. 19c). The observed topo-
graphical differences in the cross-sections are a result of the point clouds used, which are
affected by the data resolution and by the direction of the scan.
Due to the resolution of the airborne data, and the perspective from above, the ditch is
not accurately observed. Rockfall modelling for both data sets allows an evaluation of the
effect of the two data sets on the modelling results and of the effectiveness of the ditch. The
model that incorporates the TITAN data allows for a more detailed model within the
proximity of the track, as illustrated in Fig. 19b, c.
RocFall (Rocscience 2007) was used to complete the rockfall modelling in 2D. Fig-
ure 19d, e illustrates the track profile. The model simulated the release of 500 rock blocks
weighing 100 kg each. Default values for friction and restitution coefficients were used, as

123
858 Nat Hazards (2012) 60:831–864

Fig. 19 Profile rockfall models generated by cutting sections through 3-dimensional data sets. b Illustrates
data collected from a helicopter, while c illustrates the fusion of helicopter and TITAN data. d and
e Illustrate the effect of data resolution on rockfall modelling

123
Nat Hazards (2012) 60:831–864 859

determining site-specific values was not the focus of this study. The end point of each rock
block is determined and plotted against slope position and is defined as being located in the
ditch, on the track or past the track. Table 6 reports the determined end locations of all
blocks released during the two simulations. The first notable result of the comparison is the
percentage of blocks that come to rest in the ditch reported using only helicopter collected
LiDAR data (0%) versus the percentage of blocks that come to rest in the ditch when using
the fused data (23%). A second notable point is that the increase in the number of blocks
that stop in the ditch, this directly reduces the number of blocks that stop on the track. This
illustrates the importance of a well-engineered ditch. This model statistically and visually
demonstrates the benefit of having a second set of data taken from track level to give the
detail of the rockcut and ditch data required for geomechanical rockfall analysis: in this
case, high-resolution mobile terrestrial LiDAR.

3.5 Line-of-sight modelling

Hazard evaluation systems primarily rely on the physical site characteristics, observable
geological inputs, slope activity and rockfall path calculations. However, there is one
further input that is conventionally estimated and has a substantial influence on the hazard
level for a given site: line of sight. Where hazardous zones are obscured from sight, the
train operator does not have the opportunity to observe a rock on the track and therefore
cannot slow or stop the train. In reality, in these areas, the train operators are often obliged
to run more slowly than at normal track speed.
An additional benefit of 3D modelling of landscapes is that it is fairly simple to analyse
the visibility of portions of the landscape model from other points within the model. The
calculation of visibility takes into account all 3D obstructions present within the landscape
model.
The 3D Analyst extension to ArcGIS includes a viewshed function, which, for a chosen
point, indicates which parts of a 3D TIN are visible and which parts are not. The viewshed
function requires that the model data be represented as a raster; this was achieved by using
the same raster created from the fused TITAN and Airborne TIN model that was used for
the 2D rockfall model in Fig. 19.
Three points along the track were chosen, and a viewshed was calculated for each point.
The viewable distance along the track, looking approximately north, was measured for
each point based on the calculated viewshed. The results are shown in Fig. 20. The
viewable distance along the track is shown for each point; an arrow marks the position of
each point and the look direction of the observer. Also shown is the view of the TITAN

Table 6 Profile rockfall model-

End point LiDAR data collection method
ling end point location with
respect to a transportation Helicopter Helicopter ? TITAN
corridor
500 Rocks released in simulation

In ditch 0% 0 Rocks 23% 115 Rocks

On track 28% 140 Rocks 5% 25 Rocks
Past track 16% 80 Rocks 17% 85 Rocks
% In or through danger 44% 220 Rocks 22% 110 Rocks
zone

123
860 Nat Hazards (2012) 60:831–864

Fig. 20 Line-of-sight visibility models from a given point along the track corridor. As illustrated, visibility
decreases as the location approaches the bend in the track and then increases as the turn is exited

data from the same point. These views show clearly how the 3D rock face surfaces obscure
the ability to see the track in this area.
This analysis represents a static approach to investigate locations of low visibility as
determined by the user. Dynamic calculations to assess the line of sight of the train
conductor at any possible location along the track require customized software such as the
program that has been developed by Mekni et al. (2008) at Laval University, Canada in a
related research project.

4 Discussion

LiDAR technologies, as demonstrated in this paper and by papers referenced herein, is

rapidly developing and emerging as an essential geotechnical investigation tool. However,
there are numerous research topics in the fields of both LiDAR development and geo-
technical data processing that are essential before LiDAR can become a widely accepted
state-of-practice technology. Currently, the most significant limitation is the lack of tools
available to efficiently transform LiDAR data into geotechnical information. Vulcan
(Maptek 2008) is capable of numerous geotechnical calculations and measurements from
LiDAR data; however, it is limited to mining environments and incapable of many pro-
cessing options and modelling tools.
With respect to feature extraction, such as fracture discontinuities, information can be
readily extracted from LiDAR data as demonstrated here. However, guidelines and stan-
dards must be developed to enable users to properly delineate discontinuities and measure
the orientations. Currently, numerous methodologies exist to determine the strike and dip
of a plane (often within a single program), each giving slightly different results.
As well, the majority of rockmass evaluations completed using LiDAR data employ a
standard rockmass rating scheme such as Q (Barton et al. 1974) or RMR (Bieniawski
1989). These systems are optimized for field-based investigations. Extracting geotechnical
information from LiDAR data has different strengths and limitations in comparison with

123
Nat Hazards (2012) 60:831–864 861

field-based approaches; these must be accounted for in the rating system. A new LiDAR-
based rockmass rating system is essential to the development of LiDAR as a geotechnical
evaluation tool.
The key next step for advanced application is to develop highly accurate digital terrain
models. If rockfall models are going to properly exploit the data resolution and accuracy,
the models must account for rock block fracture, physical impediments such as trees, and
true 3D geometries. This will perhaps force a different modelling approach than is cur-
rently available on the industry standard 2.5D ArcGIS platform.
As LiDAR systems continue to develop, as data processing and feature extraction
become simplified and automated and as data quality increases, there will be a shift in how
geological field work will be completed. However, there continues to be a growing need
for calibration of such tools and confidence in their capability within the geotechnical
community. Through continued research, publications and presentations, future geotech-
nical engineering projects will greatly benefit from the spatial and temporal knowledge that
can be extracted from LiDAR data.
Finally, LiDAR data are capable of producing spatially accurate virtual 3D models of
rockmass and surface topology. However, the technology is limited to surface exposure. To
develop a true understanding of rockmass behaviour, LiDAR should be integrated with
other geophysical techniques that are capable of imaging the subsurface. The coupling of
such technologies will enable an unprecedented clarity in rockmass visualization capa-
bilities and investigation possibilities.

5 Conclusions

The geotechnical applications of LiDAR data have been developed and proven by
numerous engineers and geo-scientists. The focus of numerous research campaigns has
been to adapt scanning technologies and processing methodologies to augment previously
established investigatory or evaluation protocols. However, what these research campaigns
have not achieved, for the most part, is new workflows to establish evaluation protocols
that exploit the advantages of LiDAR scanning and processing technologies. The theories
and workflows presented in this paper attempt to develop new methodologies for rockfall
hazard management systems.
The processing workflows presented in this paper demonstrate the applicability of
LiDAR-based field investigations into geologically and geographically distinct domains
and environments. The approach of discriminating potentially unstable rockmasses in
terms of geological setting and failure modes, as described in this paper, present the
evaluating engineer with a vast library of analytical techniques to accurately and quanti-
tatively assess the hazard for a particular outcrop.
The aforementioned methodology pertains to all rockfall hazard evaluation management
programmes, whether the target area be individual outcrops or entire transportation sys-
tems. The development of complex and dynamic rating systems for unstable rockmasses
will evolve through the testing of LiDAR data in different environments and geological
settings.

Acknowledgments The authors would like to extend gratitude to Kathy Kalenchuk and Dave Ball from
Queen’s University. The authors would like to thank the support of Paul Mrstik, Craig Sheriff and Kresimir
Kusevic from Ambercore for mobile terrestrial and helicopter LiDAR data collection and knowledgeable
support. The authors would also like to thank Isabel Coderre for her countless hours of editing. As well, the

123
862 Nat Hazards (2012) 60:831–864

authors would like to thank InnovMetric for the consultations regarding processing geological LiDAR data
in PolyWorks. This research has been generously funded since 2006 by the NSERC, GEOIDE and PREA
agencies with tactical support from CN and CPRail.

References

Abbott B, Bruce I, Keegan T, Oboni F, Savigny W (1998) A methodology for the assessment of rockfall
hazard and risk along linear transportation corridors. In: 8th Congress, international association of
engineering geology, a global view from the Pacific rim, Vancouver, vol 2. A. A. Balkema, Vancouver,
pp 1195–1200
Abellán A, Jaboyedoff M, Oppikofer T, Vilaplana JM (2009) Detection of millimetric deformation using a
terrestrial laser scanner: experiment and application to a rockfall event. Nat Hazards Earth Syst Sci
9:365–372
Abellán A, Vilaplana JM, Martı́nez J (2006) Application of a long-range Terrestrial Laser Scanner to a
detailed rockfall study at Vall de Núria (Eastern Pyrenees, Spain). Eng Geol 88(3–4):136–148
Agliardi F, Crosta GB (2003) High resolution three-dimensional numerical modelling of rockfalls. Int J
Rock Mech Min Sci 40:455–471
Alshawa M, Smigiel E, Grussenmeyer P, Landes T (2007). Integration of a terrestrial LiDAR on a mobile
mapping platform: first experiences. In: 5th International symposium on mobile mapping technology,
Italy, p 6
Amann M-C, Bosch T, Myllyla R, Rioux M (2001) Laser ranging: a critical review of usual techniques for
distance measurement. Opt Eng 40(1):10–19
Ambercore (2007) ALMIS. Ottawa, Canada
Ambercore (2009) TITAN. Ottawa, Canada
Arattano M, Marchi L (2008) Systems and sensors for debris-flow monitoring and warning. Sensors
8:2436–2452
Barber D, Mills J, Smith-Voysey S (2008) Geometric validation of a ground-based mobile laser scanning
system. J Photogramm Remote Sens 63:128–141
Barton N, Lien R, Lunde J (1974) Engineering classification of rock masses for the design of rock support.
Rock Mech 6:189–236
Bauer A, Paar G, Kaltenböck A (2005) Mass movement monitoring using terrestrial laser scanner for rock
fall management. In: Geo-information for disaster management, vol 5. Springer, Berlin, pp 393–406
Bellian JA, Kerans C, Jennette DC (2005) Digital outcrop models: applications of terrestrial scanning
LiDAR technology in stratigraphic modelling. J Sediment Res 2(75):166–176
Bieniawski ZT (1989) Engineering rock mass classification. Wiley, New York
Bonnaffe F, Jennette D, Andrews J (2007) A method for acquiring and processing ground-based LiDAR data
in difficult-to-access outcrops for use in three-dimensional, virtual-reality models. Geosphere
3(6):501–510
Bourrier F, Dorren L, Nicot F, Berger F, Darve F (2009) Toward objective rockfall trajectory simulation
using a stochastic impact model. Geomorphology 110:68–79
Bruce Geotechnical Consultants Inc. in association with Oboni Associates Inc (1997) Canadian National
Railway Rockfall Hazard and Risk Assessment System. Field Implementation Manual
Buckley SJ, Howell JA, Enge HD, Kurz TH (2008) Terrestrial laser scanning in geology: data acquisition,
processing and accuracy considerations. J Geol Soc 165:625–638
Bunce CM, Cruden DM, Morgenstern NR (1997) Assessment of the hazard from rock fall on a highway.
Can Geotech J 34:344–356
Carr SD, Easton RM, Jamieson RA, Culshaw NG (2000) Geologic transect across the Grenville Orogen of
Ontario and New York. Can J Earth Sci 37:193–216
Chau KT, Wong RH, Wu JJ (2002) Coefficient of restitution and rotational motions of rockfall impacts. Int J
Rock Mech Min Sci 39:69–77
Cushing HP (1908) Lower portion of the Paleozoic section in northwestern New York. Geol Soc Am Bull
19:155–176
Derron MH, Jaboyedoff M, Blikra LH (2005) Preliminary assessment of rockslide and rockfall hazards
using a DEM (Oppstadhornet, Norway). Nat Hazards Earth Syst Sci 5:285–292
Donovan J, Raza AW (2008) A change detection method for slope monitoring and identification of potential
rockfall using three-dimensional imaging. American Rock Mechanics Association, San Francisco
El-Rabbany A (2006) Introduction to GPS: the global positioning system, 2nd edn. Artech House Publishers,
Norwood

123
Nat Hazards (2012) 60:831–864 863

ESRI (2008) ArcGIS V9.3 (Build 1770)

Fekete S, Diederichs MS, Lato M (2009) Geotechnical and operational applications for dimensional laser
scanning in drill and blast tunnels. Tunn Undergr Space Technol; submitted Oct 2009 (manuscript 18
pages, 28 figures)
Fekete S, Diederichs M, Lato M (2009b) Geotechnical applications of LiDAR scanning in tunnelling,
RockEng. Diederichs and Grasselli, Toronto, pp 1–12
Feng QH, Röshoff K (2004) In Situ mapping and documentation of rock faces using a full-coverage 3D laser
scanning technique. Int J Rock Mech Min Sci 41(3):1–6
Franklin JA, Senior SA (1997) Outline of RHRON, the Ontario rockfall hazard rating system. In: Pro-
ceedings of the international symposium on engineering geology and the environment, Athens,
pp 647–656
Glennie C (2007a) Reign of point clouds: a kinematic terrestrial LiDAR scanning system. In: InsideGNSS,
pp 21–31
Glennie C (2007b) Rigorous 3D error analysis of kinematic scanning LiDAR systems. J Appl Geodesy
1:147–157
Holm K, Jakob M (2009) Long rockfall runout, Pascua, Chile. Can Geotech J 46:225–230
Hungr O, Fletcher L, Jakob M, MacKay C, Evans SG (2003) A system of rock fall and rock slide hazard
rating for a railway. In: Geohazards, Edmonton, pp 277–283
Iavarone A, Vagners D (2003) Sensor fusion: generating 3D by combining airborne and tripod mounted
LiDAR data. In: International archives of photogrammetry, remote sensing and spatial information
sciences, vol XXXIV. ISPRS, Regensburg, pp 1–7
InnovMetric (2008) PolyWorks V11.0.1. Quebec City, Canada
Janeras M, Navarro M, Ruiz A, Kornus W, Talaya J, Barbera M et al. (2004) LiDAR applications to rock
fall hazard assessment in Vall de Nuria. In: 4th ICA Mountain Cartography Workshop, Spain, pp 1–14
Kalenchuk K, Diederichs MS, McKinnon S (2006) Characterizing block geometry in jointed rockmasses. Int
J Rock Mech Min Sci 43:1212–1225
Kemeny J, Turner K (2008) Ground-based LiDAR rock slope mapping and assessment. Federal Lands
Highway Technology Deployment Initiatives and Partnership Program. Federal Highway Adminis-
tration, Lakewood
Lan H, Martin CD, Lim CH (2007) Rockfall analyst: a GIS extension for three-dimensional and spatially
distributed rockfall hazard monitoring. Comput Geosci 33:262–279
Lato M, Hutchinson J, Diederichs M, Kalenchuk K (2007) Evaluating block shape and block volume
distributions of rock faces using LiDAR and 3DEC. Geophys Res Abstr (European Geosciences
Union) 9:1–2
Lato M, Hutchinson DJ, Diederichs MS, Harrap R (2008) Optimization of LiDAR scanning and processing
for automated structural evaluation of discontinuities in rockmasses. Int J Rock Mech Min Sci
46:194–199
Lato M, Diederichs MS, Hutchinson DJ (2009a) Bias correction for static LiDAR scanning of rock outcrops
for structural characterization. Rock Mech Rock Eng 23:5–7. doi:10.1007/s00603-010-0086-4
Lato M, Hutchinson DJ, Diederichs MS, Ball D, Harrap R (2009b) Engineering monitoring of rockfall
hazards along transportation corridors: using mobile terrestrial LiDAR. Nat Hazards Earth Syst Sci
9:935–946
Leica Geosystems (2007) HDS6000. Heerbrugg, Switzerland
Leica Geosystems HDS LLC (2008) Cyclone 6.0.2 Build 980. Heerbrugg, Switzerland
Lemy F, Hadjigeorgiou J (2004) A digital face mapping case study in an underground hard rock mine. Can
Geotech J 41:1011–1025
Lemy F, Yong S, Schultz T (2006) A case study of monitoring tunnel wall displacement using laser scanning
technology. In: IAEG. The Geological Society of London, Nottingham, pp 1–11
Maerz NH (2000). Highway rock cut stability assessment in rock massed not conducive to stability cal-
culations. In: Proceedings of the 51st annual highway geology symposium, Seattle, pp 249–259
Maptek (2007) I-Site 4400. Denver, Co, USA
McCaffrey KW, Jones R, Holdsworth R, Wilson R, Clegg P, Imber J (2005) Unlocking the spatial
dimension: digital technologies and the future of geoscience fieldwork. J Geol Soc Lond 162:927–938
McFall GH (1993) Structural Elements and Neotectonics of Prince Edward County, Southern Ontario.
Geographie Physique et Quaternaire 47(3):303–312
McFarlane RB (1992) Stratigraphy, paleoenvironmental interpretation, and sequences of the Middle
Ordovician Black River Group. Queen’s University at Kingston, Kingston (Publication No. AAT
MM80627)
McHugh EL (2004) Video motion detection for real-time hazard warnings in surface mines. In: SME annual
meeting. Society for Mining, Metallurgy, and Exploration, Denver, p 10

123
864 Nat Hazards (2012) 60:831–864

McKean J, Roering J (2004) Objective landslide detection and surface morphology mapping using high-
resolution airborne laser altimetry. Geomorphology 57:331–351
Mechelke K, Kersten TP, Lindstaedt M (2007) Comparative investigation into the accuracy behaviour of the
new generation of terrestrial laser scanning systems. In: Optical 3D measurement techniques, vol VIII.
Gruen and Kahmen, Zurich, pp 319–327
Mekni M, Sahli N, Moulin B (2008) A geosimulation approach involving spatially-aware agents: a case
study on the identification of risky areas for trains. SpringSim 1-56555-319-5, 37-45
Moore HL (1986) Wedge rockfalls along Tennessee highways in the Appalachian region: their occurrence
and correction. Bull As Eng Geol 23(4):441–460
Optech (2008) Lynx Mobile Mapper. Toronto
Park HJ, West TR (2002) Sampling bias of discontinuity orientation caused by linear sampling bias. Eng
Geol 66:99–110
Pedrazzini A, Jaboyedoff M (2008). Structures and failure mechanisms analysis of turtle mountain. In: Locat
DPJ (ed) Geohazards, Quebec City, pp 349–356
Perret S, Dolf F, Kienholz H (2004) Rockfalls into forests: analysis and simulation of rockfall trajectories—
considerations with respect to mountainous forests in Switzerland. Landslides 1:123–130
Pierson LA (1991) ‘‘Rockfall hazard rating system’’ final report. Federal Highway Administration, Oregon
State Highway Division, Salem
Priest SD (1993) Discontinuity analysis for rock engineering. Chapman and Hall, London
Priest SD, Hudson JA (1976) Discontinuity spacings in rock. Int J Rock Mech Min Sci 13:135–148
Pringle J, Gardiner A, Westerman R (2004) Virtual geological outcrops—fieldwork and analysis made less
exhaustive. Geol Today 20(2):67–72
Pritchard M, Porter M, Savigny KW, Bruce I, Oboni F, Keegan T (2005) CN rockfall hazard risk man-
agement system: Experience, enhancements, and future direction. The American Railway Engineering
and Maintenance-of-Way Association, Chicago
Ritchie AM (1963) Highway Research Record 17: evaluation of rockfall and its control. Highway Research
Board, National Research Council. Transportation Research Board, Washington
RocScience (2006) Dips 5.1. Toronto
RocScience (2007) RocFall 4.045. Toronto, Canada
Roncella R, Forlani G (2005) Extraction of planar patches from point clouds to retrieve dip and dip direction
of rock discontinuities. ISPRS WG III/3, III/4, V/3 Workshop ‘‘Laser scanning 2005’’ Enschede
Rosser N, Dunning SA, Lim M, Petley DN (2005) Terresterial laser scanning for quantitative rockfall hazard
assessment. In: Hungr, Fell, Couture, Eberhardt (eds) Landslide risk management, Paper 91. Balekema,
Rotterdam
Stover BK (1992) Highway rockfall research report. Geological Survey Special Publication, Denver
Sturzenegger M, Stead D (2009a) Quantifying discontinuity orientation and persistence on high mountain
rock slopes and large landslides using terrestrial remote sensing techniques. Nat Hazards Earth Syst Sci
9:267–287
Sturzenegger M, Stead D (2009b) Close-range terrestrial digital photogrammetry and terrestrial laser
scanning for discontinuity characterization on rock cuts. Eng Geol 106:163–182
Sturzenegger M, Yan M, Stead D, Elmo D (2007) Applications and limitations of ground-based laser
scanning in rock slope characterization. In: Eberhardt E, Stead D, Morrison T (eds) Proceedings of the
first Canadian US rock mechanics symposium, vol 1. Taylor and Francis, London, pp 29–36
Terzaghi RD (1965) Source of error in joint surveys. Geotechnique 15:287–304
Trevisani S, Cavalli M, Marchi L (2009) Variogram maps from LiDAR data as fingerprints of surface
morphology on scree slopes. Nat Hazards Earth Syst Sci 9:129–133
Trinks I, Clegg P, McCaffrey K, Jones R, Hobbs R, Holdsworth B (2005) Mapping and analysing virtual
outcrops. Vis Geosci 1–7
Turner AK, Kemeny J, Slob S, Hack R (2006) Evaluation, and management of unstable rock slopes by 3D
laser scanning. In; International association for engineering geology and the environment. The Geo-
logical Society of London, pp 1–11
Vanderwater CJ, Dunne WM, Mauldon M, Drumm EC, Bateman V (2005) Classifying and assessing the
geologic contribution to rockfall hazard. Environ Eng Geosci XI(2):141–154
Vosselman G, Gorte B, Sithole G, Rabbani T (2004) Recognising structure in laser scanner point clouds. In:
International archives of photogrammetry, remote sensing and spatial information sciences vol 46.
ISPRS, Freiberg, pp 33–38
Yang Y, Farrell JA (2003) Magnetometer and differential carrier phase GPS-aided INS for advanced vehicle
control. IEEE Trans Robot Autom 19(2):269–282
Zimmer VL, Stock GM, Sitar N (2008) Seismic monitoring of rock falls in Yosemite National Park.
American Geophysical Union, San Francisco, abstract #H51F-0897

123