6 Landslide hazards
David Petley
6.1 Introduction
Landslides are naturally occurring phenomena in every
environment on Earth, including the tropics, the temperate
regions and the high latitudes, and in the oceans.
Unfortunately, this ubiquitous natural process represents a
substantial hazard to humans because people and structures
have a surprisingly low capacity to withstand the forces
generated by mobile soil and/or rock. In consequence, there
is a long recorded history of landslide disasters – for example,
Nihon Shoki (the ancient chronicle of Japan), which
was completed in the year AD 720, describes numerous
landslides and failures associated with the Hakuho earthquake
on 29 November AD 684, whilst the city of Helike in
Greece is believed to have been submerged and destroyed
as a result of a submarine landslide in 373 BC. Today,
landslides continue to inflict a substantial economic and
social toll, especially in mountainous, less developed countries,
and there is a widely held but admittedly poorly
quantified expert perception that the impacts associated
with mass movements are increasing rapidly with time.
The term landslide is unfortunately something of a misnomer
as many landslides do not in reality involve sliding.
The word landslide is used to describe a range of processes
that result in downward and outward movement of slopeforming
material composed of rock, soil and artificial materials.
In this context the term ‘mass movement’ might be
preferable, but here the term landslide will be retained as it
is in such common use in this context. In this chapter mass
movements that are mostly formed from snow or ice are
specifically excluded – these are discussed in Chapter 5.
Damaging landslides occur through a surprisingly wide
range of magnitudes and invoke a large number of mechanisms.For
example,thefallofa singlepiece of rockthe sizeof
a computer mouse can be enough to kill a person if it strikes
them on the head at terminal velocity. On the other hand, the
Seymareh landslide in the Zagros Mountains of Iran has a
deposit with a volume of about 20 km3
, whilst some submarine
landslides are now known to have a volume that may be as
much as two orders of magnitude as large again. For this
reason, landslide volumes are usually considered on a logarithmic
scale. The wide range of landslide types is usually
considered by classifying them according to the predominant
material involved and the movement type (Table 6.1). A
further refinement, which is important in the context of hazard
causation, is to classify the landslides by movement rate as
well. In general, rapid movement types are more likely to
cause loss of life than are slow movements, whilst even slow
rates of displacement (and low levels of total movement) can
cause extensive damage to buildings and infrastructure.
Figure 6.1 presents the distribution of fatal landslides
between January 2006 and December 2007 inclusive,
based upon the Durham fatal landslide database (see
Petley et al., 2005a). Care is needed in the interpretation
of such a map as the inclusion specifically of fatal landslides
(rather than all landslides or all landslides that impact
upon people) biases the data in particular ways. Most
importantly, the data are skewed towards less developed
countries where the level of mitigation against landslides
might be lower and where the density of vulnerable economic
assets might also be small. Nonetheless, it is clear
that the recorded landslides form very distinct clusters,
most notably in the following locations:
1. Along the southern edge of the Himalayan mountain
chain;
2. In Central China;
3. In SW India;
4. Along the western boundary of the Philippine Sea plate
through Japan, Taiwan and the Philippines;
5. In central Indonesia, particularly on the island of Java;
6. In the Caribbean and Central Mexico;
Geomorphological Hazards and Disaster Prevention, eds. Irasema Alcántara-Ayala and Andrew S. Goudie. Published by
Cambridge University Press. © Cambridge University Press 2010. 63
Downloaded from Cambridge Books Online by IP 147.251.65.55 on Thu Mar 17 10:25:08 GMT 2016.
http://dx.doi.org/10.1017/CBO9780511807527.006
Cambridge Books Online © Cambridge University Press, 2016
7. On the western edge of the northern part of South
America, especially in Colombia.
There is also a scattering of fatal landslides elsewhere,
for example through Europe, the tropical parts of Africa and
in North America. This distribution of intense fatal landslide
activity reflects the juxtaposition of three key factors:
1. The occurrence of tectonic processes that in particular
drive high rates of uplift and occasional seismic events;
2. The occurrence of high levels of precipitation, usually
including both high annual precipitation totals and high
short-term intensities;
3. The presence of a reasonably high population density.
Where one of these three factors is missing the occurrence
of fatal landslides reduces markedly – thus for example Iran
has a lower than expected landslide occurrence because of
the absence of precipitation; Alaska has a lower than
expected occurrence because of the absence of people; and
Central Africa has a lower than expected occurrence because
of the low levels of tectonic activity. A longer time-frame
would change this picture slightly as landsliding in arid,
tectonically active areas such as Iran is probably driven
primarily by seismic activity with an additional input from
low-frequency–high-magnitude rainfall events. Thus, if the
data were collected over a sufficiently long period to capture
a number of large seismic events in Iran then the maps would
take on a slightly different appearance.
6.2 Landslide causes and triggers
Based upon the geomorphic distribution, it is unsurprising
that the vast majority of landslides are triggered by one or
more of three key factors: precipitation, seismicity or the
action of humans. However, for these processes to be able
to induce the landslide there must first have been a series of
other processes that have acted to prepare the slope for
failure. These processes, which are usually termed ‘causes’
as opposed to triggers, include the following.
6.2.1 Geological causes
Geological causes are factors that make the materials that
form a slope susceptible to failure. Key causes include, for
example, materials that are weak or that are weathered;
materials with strong joint sets, especially where they are
orientated in such a way that they allow sliding to occur;
and material combinations that cause water to be retained.
6.2.2 Morphological causes
The most obvious morphological cause is the slope angle.
The key parameter is the angle of the slope in comparison
with the strength of the material. Thus, it is not a straightforward
relationship in which steeper slopes are less stable –
in Norway, for example, slopes formed from unweathered
gneiss are able to form cliffs that can stand vertically to
elevations of many hundreds of metres along the margins of
fjords. On the other hand, quick clays close to the water’s
edge can fail at slope angles as low as 10° when disturbed.
A further key morphological factor can be the concavity or
convexity of the slope, which can serve to concentrate
water in key locations. Finally, in many high mountain
areas the loss of glacial ice leaves slopes unsupported and
thus prone to failure, whilst in coastal environments the
under-cutting of cliffs can lead to reductions in stability.
6.2.3 Physical causes
A third group of causes are related to physical processes. For
example, a slope might be more likely to fail if the groundwater
level has been elevated by previous prolonged rainfall
or by snowmelt. Similarly, the loss of tree cover may make a
slope far more susceptible to shallow landslides. In
California, for example, shallow landslides that sometimes
transition into destructive debris flows are a particular
problem in the wet season following large forest fires.
TABLE 6.1. A simplifed classification scheme for the main types of
landslide movements
Type of movement
Type of material
Rock Engineering soils
Coarse
grained
Fine
grained
Falls Rock fall Debris fall Earth fall
Topples Rock
topple
Debris
topple
Earth
topple
Slides Rotational Rock
slump
Debris
slump
Earth
slump
Translational Rockslide Debris
slide
Earth
slide
Lateral spreads Rock
spread
Debris
spread
Earth
spread
Flows Rock
flow
Debris
flow
Earth
flow
Complex slope movements (i.e. combinations of two or
more types)
After Varnes (1978)
David Petley64
Downloaded from Cambridge Books Online by IP 147.251.65.55 on Thu Mar 17 10:25:08 GMT 2016.
http://dx.doi.org/10.1017/CBO9780511807527.006
Cambridge Books Online © Cambridge University Press, 2016
6.2.4 Human causes
The final group of factors is centred around human activities
that can destabilise a slope. Examples include the
removal of forestry through logging or firewood collection;
over-steepening of slopes through cutting for road construction
or through quarrying; and the leakage of pipes
or swimming pools. For example, in Nepal the occurrence
of fatal landslides has increased markedly in recent years as
a result of the construction of rural roads with inadequate
levels of slope stabilisation and water management. Slopes
that have been destabilised by the road building then fail
during intense precipitation events associated with the summer
monsoon, blocking the road and causing substantial
and increasing levels of loss of life (Petley et al., 2007).
6.2.5 Causation vs. triggering
In most cases the final failure of a slope occurs as a result of
a clear trigger. On a day-to-day basis the most common
trigger is precipitation, sometimes supplemented by the
effects of snowmelt. Precipitation serves to increase the
pore pressures within a slope, reducing the resistance to
movement. In temperate and cold environments this occurs
primarily through increases in groundwater level. Thus,
unless there has been a marked change in a causal factor
as outlined above, landslides usually occur during precipitation
that is low frequency–high magnitude, in terms of
intensity and/or duration. However, if there has been a
major change to the causal factors – for example, if there
has been extensive recent deforestation – then extensive
landsliding can occur in non-exceptional rainfall events.
Thus, from a human perspective it is clear that landslide
hazard is associated with a complex range of interrelated
factors, which can be conceptualised as a ‘chain of events’
(Figure 6.2). For the landslide disaster to occur there must
be the juxtaposition of a number of causes plus the final
trigger and the element at risk. From a hazard management
perspective this can be helpful as the occurrence of the
damaging landslide can be prevented through any one of
a number of interventions, each of which addresses either a
key cause, the trigger or the elements at risk (Figure 6.2). Of
course this is a somewhat simplistic way to view landslide
hazard reduction, but in the real world the level of hazard
can often be reduced through a number of approaches.
Thus, in Nepal, for example, in areas in which good quality
engineering input is utilised the slope hazard can be
reduced by using one or a combination of the following
approaches:
1. Selecting an alignment for the road that avoids areas of
known existing instability, or areas in which materials
are known to be problematic (black schist often causes
problems in humid tropical and subtropical environments
for example), or areas with a slope angle greater
than a pre-determined value;
FIGURE 6.1. Map showing the locations of fatal landslides in 2006 and 2007, as recorded in the Durham University Landslide Database.
Each black dot represents a single fatal landslide.
65Landslide hazards
Downloaded from Cambridge Books Online by IP 147.251.65.55 on Thu Mar 17 10:25:08 GMT 2016.
http://dx.doi.org/10.1017/CBO9780511807527.006
Cambridge Books Online © Cambridge University Press, 2016
2. Managing the water on the slope through the construction
of effective drainage;
3. The construction of walls to support slopes that have
been cut;
4. The planting of local vegetation species that can help to
increase the strength of the soil or to draw down the
water table. Vetiver grass (Chrysopogon zizanioides) is
widely used in this capacity for example as it rapidly
grows roots that extend to a depth of up to four metres.
Thus, key aspects of landslide hazard reduction are to
identify and understand the causes; to identify and understand
the triggering processes and the key thresholds at
which they occur; and to try to ensure that human elements
are not put at risk.
6.3 The role of geomorphology in
landslide hazard management
Landslide management is a well-developed science, and
where sufficient resources are available most small- to
medium-sized landslides can be managed or mitigated if
they have been identified and characterised properly and if
sufficient resources are available. The management of
landslides is a multi-disciplinary task, with key inputs
from geotechnical engineers, engineering geologists, biologists,
meteorologists, planners and others in addition to
geomorphologists. Indeed, landslide management is rarely
trusted to geomorphologists alone, instead the key role is to
act as part of a multi-disciplinary team. The nature of the
management of a landslide hazard depends to a large degree
on the nature and timing of the movement, and whether the
failure in question is a first time event or a reactivation of an
existing landslide. However, geomorphology is a key
aspect of all stages of the landslide management process.
6.4 Terrain mapping
A key aspect of landslide hazard management from a geomorphic
perspective is that of terrain mapping. Terrain mapping,
which is increasingly a key part of the early stages of
infrastructure projects, especially in mountainous terrain and
in areas affected by neotectonic processes, is most commonly
used to identify areas of existing or past slope instability. In
many cases, this provides a key input into route selection for
roads, railway lines and in particular for pipelines. For example,
route selection for the Dharan–Dhankuta highway in
Nepal was undertaken primarily on the basis of geomorphological
mapping (Brunsden et al., 1975). The value of this
approach has been demonstrated by the remarkable stability
of this road in comparison with similar projects for which this
approach was not adopted (Hearn, 2002). Sometimes it is
impossible to route the corridor away from areas with identified
instability, in which case the terrain mapping is used as
an input into more detailed geotechnical investigation.
Terrain mapping is also often used as a first order hazard
assessment technique, allowing the location of existing failures
that might threaten an asset. For example, this approach
was used to assess the likelihood of landslides along the
FIGURE 6.3. A terrain map used to identify the locations of existing
landslides along the Arniko Highway in Nepal. The grid squares are
1 km.
FIGURE 6.2. Conceptual diagram showing how the causes, a
trigger and the existence of a vulnerable element conspire to
create a slope accident. On the right, possible mitigating
approaches are shown. Any one of these, applied properly, can
prevent the accident from occurring.
David Petley66
Downloaded from Cambridge Books Online by IP 147.251.65.55 on Thu Mar 17 10:25:08 GMT 2016.
http://dx.doi.org/10.1017/CBO9780511807527.006
Cambridge Books Online © Cambridge University Press, 2016
access roads to Kathmandu in Nepal in the event of a large
earthquake (Figure 6.3).
In recent years, the availability of high resolution digital
aerial photography, satellite imagery, LiDAR data, and
digital terrain models has allowed increasing levels of
sophistication in terrain mapping. The development of
increasingly sophisticated automated feature extraction
will continue to enhance these capabilities. Despite this,
terrain mapping is an under-used tool in infrastructure
projects. Expensive failures, such as the reactivation of an
ancient landslide that was clearly visible on aerial photographs
at Ok Tedi in Papua New Guinea in 1984 (Griffiths
et al., 2004), show that there are substantial advantages in
the undertaking of good terrain mapping. The resultant
legal case, which reached the Supreme Court of Papua
New Guinea in 1989, was for approximately £575 million
(Griffiths et al., 2004). The case was settled out of court.
The greatest weakness of terrain analysis continues to be
the subjective nature of the process. Fookes and Dale (1992)
examined six independent interpretations of the pre-failure
condition of the Ok Tedi site described above, showing that
even with highly skilled geomorphologists substantially different
interpretations resulted. Similarly, when different analysts
looked at identical aerial images of the site, surprisingly
different interpretations resulted (Fookes et al., 1991).
Nonetheless, terrain mapping remains a core tool in the
development of infrastructure projects.
Brunsden (2002) advocated that terrain mapping should
be the basis of a much more integrated geomorphological
analysis of potential instability. He encouraged a move
from an essentially two-dimensional analysis of landslides –
i.e. the production of landslide maps – into a fourdimensional
analysis that used geomorphological tools to
understand the three-dimensional structure of an unstable
slope, plus its development through time. Thus, for example,
modern dating techniques can be used to understand
the evolution of the slope, which can then be benchmarked
against the increasingly high quality climatic and in some
cases seismic histories that are now available. In consequence
the relationship between movement and causal/
triggering factors can be elucidated, which gives an
improved ability to understand future behaviour. This is
becoming increasingly important in the context of climate
change, which means that existing magnitude/frequency
catalogues for triggering events may not be relevant.
6.5 Susceptibility analysis
The inherent subjectivity of terrain mapping has led to a
multitude of attempts to develop more reliable,
quantitative/semi-quantitative landslide hazard mapping
techniques. The development of GIS has undoubtedly
facilitated this approach as it hypothetically at least allows
a consistent, high resolution approach to be applied to large
areas. In all cases a simple algorithm is used to derive a
score that indicates the susceptibility of a slope to failure.
The simplest algorithms are based upon a small number of
parameters (for example slope angle and material), and the
parameters themselves are arranged in classes (e.g. slope
angles of 0° to 5°, 5° to 15°, etc.). Summation is commonly
used to derive an overall score that indicates the susceptibility
to failure. In the most basic applications the score is
determined on a slope by slope basis using a proforma that
is completed by an operator. However, GIS-based
approaches utilise algorithms that combine data from thematic
layers. More complex systems utilise multiple input
parameters and algorithms that are, for example, based
upon empirical models of slope behaviour, such as the
infinite slope equation and soil slope hydrology equations.
In recent years, as processing power has increased, the use
of probabilistic approaches, often based upon Monte Carlo
simulations, has become popular.
Geomorphology plays a key role in all aspects of the
development of susceptibility analyses. For example, geomorphologists
have been at the cutting edge of the development
of both the classification-based approach and the
more quantitative analyses, and in the interpretation of
their outcomes. Globally, the most commonly used
approach is the CHASM model developed by Malcolm
Anderson and colleagues at Bristol (e.g. Brooks et al.,
2004) and subsequently applied in many different parts of
the world. Furthermore, geomorphologists often implement
the model runs and analyse the outcomes. However, serious
FIGURE 6.4. Landslide frequency (here represented by a frequency
density function) plotted against magnitude (here represented by
number of fatalities per event) for Nepal. (After Petley et al. 2007.)
67Landslide hazards
Downloaded from Cambridge Books Online by IP 147.251.65.55 on Thu Mar 17 10:25:08 GMT 2016.
http://dx.doi.org/10.1017/CBO9780511807527.006
Cambridge Books Online © Cambridge University Press, 2016
questions about the reliability of this approach given the
simplifications that are inherent in the construction of any
algorithm, and in the input data, remain. The greatest concern
with such analyses lies in the ability to verify or refute
the outcomes. Thus, there is still much to do to improve the
approach.
6.6 Hazard and stability analyses
In some cases, susceptibility models are developed to full
hazard and even risk models. A hazard model requires that
the frequency (i.e. probability) of occurrence is quantified
and that the magnitude is also determined. In the case of
landslides, magnitude generally refers both to the size of the
landslide (i.e. the volume and surface area) and the rate of
movement, as both are critical to the actual impact of the
event. The development of magnitude–frequency relationships
for landslides remains a fascinating area of research,
not least because of the surprising similarities in these
relationships between different areas and as a result of
different triggers (see Malamud et al., 2004, for example).
The cause of this consistent pattern of self-organisation,
which is applicable to landslide losses as well as the landslides
themselves (Figure 6.4) remains unclear, but represents
an interesting potential input into hazard evaluation.
A key aspect of the evaluation of the frequency of landslides
is the reconstruction of landslide chronologies using
archival data. The principle is that our conventional scientific
records are too short to provide a proper representation
of the occurrence of landslides, and that the occurrence of
landslides under environments that are different to the
present can only be properly evaluated using long-term
records. A range of approaches have been developed by
the geomorphology community to allow this, and these
continue to improve with time as, in particular, dating
methods become better constrained. The use of long-term
archive records of movement extracted from non-scientific
reports is a key approach – for example, the relationship
between movement and rainfall patterns for the Ventnor
landslide on the Isle of Wight has been constrained using
such a technique (Ibsen and Brunsden, 1996). Such datasets
are limited by human records, which in a geomorphological
sense are short. Extension of these records has been
achieved using various dating techniques. For example,
Borgatti and Soldati (2002) used carbon dating of wood
entrained in landslide debris to establish the temporal
occurrence of landslides in the Alps that could then be
related to palaeoclimate reconstructions for the Holocene,
establishing the link between climate and landslides. Whilst
being very powerful, such techniques are expensive and
time-consuming as many dates are required. Furthermore,
uncertainties remain as landslides tend to be triggered by
weather events whereas palaeoclimate data provide an
indication of climate. Similar studies using other dating
FIGURE 6.5. The town of Taihape in North Island, New Zealand, which is built upon a slow moving earthflow. Monitoring is used to
ensure that the town remains secure.
David Petley68
Downloaded from Cambridge Books Online by IP 147.251.65.55 on Thu Mar 17 10:25:08 GMT 2016.
http://dx.doi.org/10.1017/CBO9780511807527.006
Cambridge Books Online © Cambridge University Press, 2016
techniques have been used to establish landslide chronologies,
including lichenometry and dendrochronology (see
Lang, 1999, for a detailed review). In the last few years
the availability of cosmogenic isotope dating, which allows
the time of unroofing of rock surfaces to be established, has
allowed the dating of large rock avalanches and rockfalls
(Mitchell et al., 2007). There is increasing interest in the use
of this technique for the determination of the chronology of
large earthquakes using rock avalanches as the indicator of
the seismic event (Korup et al., 2004).
Estimating the likely volume or surface area of a potential
failure remains very troublesome, especially if runout
distance is to be included. Rates of movement are generally
analysed using numerical models. This is easier for a single
landslide in a site that is well characterised, but is very
difficult where hazard is being assessed for a larger area.
Even in the case of a well-constrained specific site, reliable
modelling of known failures still requires that the model is
tuned with specific parameters that are not physically representative.
Thus, predictive modelling has a poor level of
reliability.
6.7 Monitoring, behaviour prediction and
warning systems
In many parts of the world, people live on or are affected by
landslides that cannot be easily mitigated. Examples include
settlements that are located on slow moving or inactive landslides
(the towns of Ventnor in the UK and Taihape in New
Zealand (Figure 6.5) are examples where this is the case);
transportation routes that cross landslides (for example the
Ashcroft landslides in Canada), and landslides that threaten
other economic assets, such as reservoir bank failures that
imperil dams through displacement waves. Frequently the
landslides are either too large or too numerous to be effectively
mitigated and the assets cannot be relocated. In this
case an increasingly common approach to the management of
the slope hazards is the use of monitoring systems, some of
which are used to provide warnings.
Warning systems over large areas tend to be based upon
measurement and analysis of the thresholds at which potential
triggers start to induce landslide movement. Considerable
work has been undertaken in particular on the development
of thresholds of rainfall for landslide activation. Generally the
approach used is to examine rainfall events that are known to
have caused landslides and to compare them with events in
which no landslides have occurred. Usually, the best relationships
are found by looking at a combination of medium-term
precipitation (perhaps rainfall total over the previous month)
and short-term rainfall (over the last few hours). Thus, the
threshold for movement in terms of short-term rainfall is
usually lower if the previous few days have been wet than if
they have been dry. Such approaches are the basis of warning
systems in a number of places – for example Japan and Hong
Kong both operate large-scale systems for landslide warning.
However, implementation of such systems is rather complex
as rainfall is very spatially variable. In Hong Kong, which has
a surface area of just 1,092 km2
, the warning system is based
upon 110 rain gauges plus the use of a Doppler rain radar
system. Even then, the system is operated with caveats, in
particular that unexpected intense rainfall, perhaps associated
with a rapidly developing convective system, can induce
landslides before a warning can be issued. Nonetheless, the
development of appropriate rainfall thresholds has been and
continues to be a key task for geomorphologists. In Malaysia
for example, which has a surface area of about 330,000 km2
,
thereare plans to ultimatelycreateanationwideslopewarning
system based upon rainfall thresholds, although this will take
years to implement and will be expensive to maintain. Similar
warning systems for earthquake-induced slope hazards are in
their infancy, but the new Taiwan High Speed Rail Line has
earthquake acceleration sensors mounted along the length of
the track that automatically stop the trains if ground accelerations
exceed 40 gals, partly to ensure that trains do not hit
failures induced in the earthworks alongside the track. Given
the lowshakingthreshold,itcan behoped that in theevent of a
large earthquake the sensors would initiate the stopping
sequence before the main earthquake waves arrive at the
track, and thus before any embankment failures, occurred.
However, such systems have rarely been tested by large earthquakes,
so we wait to see their effectiveness.
Warning systems on individual slopes generally take a
quite different approach, in this case being based upon the
detection of landslide movement. Three approaches are
generally adopted:
1. Sensors, often based upon vibration or using echo sounders
to detect changes in sediment volume, are placed in
the path along which a landslide is expected to move,
allowing a warning signal to be issued. Such warning
systems generally provide a warning over only a very
short period (seconds to minutes), allowing emergency
evacuations along pre-determined routes or the closure
of key transportation routes such as roads.
Geomorphology plays a key role in all aspects of the
development of such systems, including the identification
of hazardous slopes that require warning systems;
the determination of the optimum monitoring locations
and sensor type; the selection of appropriate thresholds;
and the determination of safe escape routes and zones.
Such warning systems have proven to be successful in
69Landslide hazards
Downloaded from Cambridge Books Online by IP 147.251.65.55 on Thu Mar 17 10:25:08 GMT 2016.
http://dx.doi.org/10.1017/CBO9780511807527.006
Cambridge Books Online © Cambridge University Press, 2016
many locations – for example, the town of Funes in the
Dolomite mountains of northern Italy has been protected
for over a decade against catastrophic debris flows using
such a system (Petley et al., 2005b).
2. An alternative approach is to measure the conditions in
the slope that are considered likely to trigger failure. In
most cases this is based upon a calculation of the stability
of the slope using a static equilibrium (i.e. factor of
safety) calculation. This allows the critical groundwater
level (or pressure condition) at which failure is considered
likely to be determined. This is often backed up
with records of movement, analysed in conjunction with
measurements or models of the groundwater conditions
at the time. Monitoring of the groundwater level using
piezometers then allows a warning to be made when the
groundwater approaches this critical depth.
3. The final, increasingly popular, approach is to detect the
early (precursory) stages of landslide movement, generally
using movement sensors. The rationale is that catastrophic
failure is usually presaged by a period of
accelerating movements of the slope. Thus, once a predetermined
movement rate is reached, or when a specific
pattern of acceleration is observed, a warning is sounded.
Early systems focussed upon the use of inclinometers
located in boreholes, which measure the movement
between the base of the landslide and the underlying
bed, or extensometers that measure movement across the
back scar at the landslide head. Increasingly, however,
technologically based approaches are being adopted, typically
using robotised theodolites in conjunction with
reflective targets located on the landslide body (Petley
et al., 2005b) or high resolution differential GPS receivers
located on the landslide mass (Tagliavini et al., 2007).
These approaches are proving capable of providing warnings,
and also have the advantage of generating detailed
datasets on the movement patterns of the landslide that
allows a better understanding of its dynamics. A series of
new experimental approaches are also being developed,
most notably slope radar, which is increasingly being used
to provide warnings on large open-cast pit slopes, and
satellite-based radar interferometry (InSAR) (Catani et al.,
2005). The latter shows some potential but remains problematic,
not least because displacement–time plots generally
have an unrealistic linear trend, which suggests that
crucial elements of movement are not being resolved.
A key challenge for the geomorphological community is
the development of enhanced understanding of warning
thresholds and the development of new techniques for the
analysis of movement patterns. In particular, we still understand
precursory movement patterns very poorly. An even
greater challenge is that of behaviour prediction – i.e. can
we determine the likely future movement of a landslide?
Generally such problems are associated with large mass,
slow moving movements that seem to have the potential to
move rapidly. The slow movement implies a factor of safety
very close to unity, so increased movement often seems
very possible. An example is the Ventnor landslide on the
Isle of Wight in southern England. This landslide has built
upon it a town with a permanent population of 7,000 people
and economic assets with a replacement cost of over £150
million. The landslide moves continually at a rate of a few
millimetres to centimetres per annum, and the geomorphology
evolves as a result. Opinion remains divided as to
whether the slope is likely to evolve into a large-scale,
rapid failure and if so, under what circumstances.
However, Carey et al. (2007) used a combination of geomorphological
analysis, interpretation of movement and
piezometer data, and novel laboratory testing to examine
the mechanisms of movement and thus to forecast likely
behaviour, suggesting that a very rapid failure event is
unlikely. Such approaches will increasingly represent a
frontier of landslide research that builds upon the availability
of good-quality, real-time movement data and, increasingly,
models that reliably represent the full range of
processes that occur within a slope.
6.8 Secondary hazards and sediment
production
Landslides are frequently considered to be secondary hazards
associated with a primary event such as a typhoon or an
earthquake. However, it is increasingly clear that landslides
generate their own set of comparatively poorly understood
secondary hazards, most notably dam-break floods and tsunamis.
The former was strongly demonstrated by the 2008
Wenchuan earthquake in Sichuan province in China, which
induced a very large number of landslides that caused catastrophic
damage – for example, the town of Beichuan was
almost completely destroyed by two rock slope large failures.
In the aftermath of the earthquake, however, there was great
concern associated with the existence of 44 valley-blocking
landslides, each of which had the capacity to cause a substantial
and very destructive dam-break flood. The mitigation
of such sites requires a high level of input from geomorphologists,
most notably through:
1. The identification of sites in which there is high potential
for the occurrence of valley-blocking landslides,
most notably through terrain analysis (see above);
2. The analysis of the dynamics of a potential dam-break
flood. Two key approaches are used: first, using
David Petley70
Downloaded from Cambridge Books Online by IP 147.251.65.55 on Thu Mar 17 10:25:08 GMT 2016.
http://dx.doi.org/10.1017/CBO9780511807527.006
Cambridge Books Online © Cambridge University Press, 2016
statistical relationships derived from previous events;
and second, through flood modelling. In both cases
information on the dimensional and material properties
of the dam is critical;
3. The identification of safe locations for populations
downstream that need to be relocated;
4. The design of safe mitigation measures, most notably
the construction of a spillway to safely drain the lake.
In the case of Wenchuan, the authorities successfully
drained all of the dangerous landslide lakes without the
reported loss of a single life. However, the experience has
caused many other earthquake-prone countries to reflect
upon their own capabilities in this area. It is likely that
there will need to be considerable investment in this field
to allow a repeat of the Chinese achievements elsewhere.
The further substantial secondary hazard is that of the
generation of a tsunami from a terrestrial or a submarine
landslide. Here the threat is very real, as a number of welldocumented
cases demonstrate (Bardet et al., 2003).
Unfortunately, the discussion of these hazards is sometimes
sidetracked by over-blown descriptions of extreme landslide
events for which there is no real physical evidence, in
which single landslides are postulated as having the potential
to generate tsunamis that could devastate whole ocean
basins. Although such scenarios have little or no scientific
credibility, the potential for serious localised impacts of
large landslides into water bodies is well established. For
example, in 1997 a tsunami associated with a Mw = 7.0
earthquake in Papua New Guinea struck the Sissano
Lagoon, killing over 1,000 people. It is now clear that the
source of this tsunami was a submarine slump triggered by
the earthquake (Lynett et al., 2003). Currently, our understanding
of the likelihood of such tsunamis is comparatively
poor, not least because they are high magnitude but
low frequency events. Considerable work is now being
undertaken both to model the occurrence of such events
and to map and date deposits left by them.
The 1999 Chi-Chi earthquake in Taiwan probably represents
the most intensely studied landslide event on record
(e.g. Chang et al., 2007). One aspect of these landslides that
has been particularly interesting is the emphasis placed on
understanding the ways in which the mass failures contribute
to the erosional mass balance of tectonically active mountain
chains (Lin et al., 2006). There have been two key components
of this work. First, considerable effort has been
expended in trying to understand how patterns of landsliding
evolve in the aftermath of large earthquakes. In Taiwan it has
been clear in a number of catchments that whilst the number
of landslides associated with the initial earthquake is high, the
number increases dramatically in the first extreme rainfall
event after the temblor. In many catchments the number of
landslides more than doubled. This of course has profound
implications for hazard management in the aftermath of the
earthquake – in Taiwan the Central Cross Island Highway,
which is the main arterial route across the Central Mountain
chain, has been repeatedly destroyed by landslides in the
aftermath of the earthquake, at considerable cost. Nine years
after the earthquake the occurrence of landslides is still well
above the background level. Similar effects have been noted
elsewhere (Keefer, 1994), but considerable further work is
required to understand this process properly.
A linked issue is that of sediment mobility. The landslides
triggered by large events can release huge quantities
of sediment into the fluvial system. At the same time however
large valley-blocking landslides can cause sediment to
be deposited and stored within the channel. Understanding
the interaction between these two processes, and their relationship
to periods that are not affected by recent large
events, is a key aspect of work in geomorphology. In the
case of Taiwan, sediment transport increased dramatically
in the aftermath of the earthquake, especially during large
flood events. The rivers responded by aggrading – in places
the river bed level has increased by as much as 30 m.
Dadson et al. (2003) demonstrated that across Taiwan as a
whole there is a correlation between seismic moment
(i.e. earthquake energy release) and sediment production,
presumably as a result of the occurrence of landslides.
These variations in sediment release and transportation
associated with earthquakes have important implications
for the understanding of the evolution of hazards in affected
areas – in the Tachia River Valley in Taiwan the landslides
associated with the Chi-Chi earthquake and the subsequent
sediment disasters are estimated to have cost a total of about
US$968 million (Table 6.2).
TABLE 6.2. Estimated costs of landslide and sediment induced damage
in the aftermath of the 1999 Chi-Chi earthquake
Item
Cost (US
dollars)
Repairs to hydroelectric power
infrastructure
$300 million
Additional power generation costs $320 million
Initial road repairs $53 million
Additional transportation costs for
agriculture
$95 million
Estimated costs to rebuild Central Cross
Island Highway
$200 million
Total $968 million
71Landslide hazards
Downloaded from Cambridge Books Online by IP 147.251.65.55 on Thu Mar 17 10:25:08 GMT 2016.
http://dx.doi.org/10.1017/CBO9780511807527.006
Cambridge Books Online © Cambridge University Press, 2016
6.9 Conclusions
Landsliding is a natural geomorphological process that acts
primarily to balance uplift. Human activities exacerbate this
situation considerably, increasing the spatial density and
temporal frequency of failures. As such, landslides represent
a hazard in all inhabited areas with slopes. Geomorphologists
play a key role in the management of these hazards. Indeed,
in recent years many UK-based engineering consultants have
formed geomorphology units specifically to make use of the
expertise that geomorphologists can bring to infrastructure
projects in potentially unstable areas. However, many challenges
remain, not least to gain a better understanding of the
mechanisms of landslides in mountainous, tropical environments
and to find effective ways to manage landslides in less
developed countries.
References
Bardet, J.-P., Synolakis, C. E., Davies, H. L., Imamura, F. and
Okal, E. A. (2003). Landslide tsunamis: recent findings and
research directions. Pure and Applied Geophysics, 160,
1793–1809.
Borgatti L. and Soldati M. (2002). The influence of Holocene
climatic changes on landslide occurrence in Europe. In
J. Rybar, J. Stemberk and P. Wagner (eds.), Landslides.
Rotterdam: Balkema, pp. 111–116.
Brooks, S. M., Crozier, M. J., Glade, T. W. and Anderson, M. G.
(2004). Towards establishing climatic thresholds for slope
instability: use of a physically-based combined soil
hydrology-slope stability model. Pure and Applied
Geophysics, 161, 881–905.
Brunsden, D. (2002). Geomorphological roulette for engineers
and planners: some insights into an old game. Quarterly
Journal of Engineering Geology and Hydrogeology, 35,
101–142.
Brunsden, D., Doornkamp, J. C., Fookes, P. G., Jones, D. K. C.
and Kelly, J. M. H. (1975). Large scale geomorphological
mapping and highway engineering design. Quarterly
Journal of Engineering Geology, 8, 227–225.
Carey, J. M., Moore, R., Petley, D. N. and Siddle, H. J. (2007).
Pre-failure behaviour of slope materials and their significance
in the progressive failure of landslides. In R. McInnes,
J. Jakeways, H. Fairbank and E. Mathie (eds.), Landslides
and Climate Change: Challenges and Solutions. London:
Taylor and Francis, pp. 207–215.
Catani, F., Farina. P., Moretti, S., Nico, G. and Strozzi, T. (2005).
On the application of SAR interferometry to geomorphological
studies: estimation of landform attributes and mass movements.
Geomorphology, 66, 119–131.
Chang, K. T., Chiang, S. H. and Hsu, M. L. (2007). Modeling
typhoon- and earthquake-induced landslides in a mountainous
watershed using logistic regression. Geomorphology, 89,
335–347.
Dadson, S., Hovius, N., Chen, H., et al. (2003). Links between
erosion, runoff variability and seismicity in the Taiwan orogen.
Nature, 426, 648–651.
Fookes, P. G. and Dale, S. G. (1992). Comparison of interpretations
of a major landslide at an earthfill dam site in Papua New
Guinea. Quarterly Journal of Engineering Geology and
Hydrogeology, 25, 313–330.
Fookes, P. G., Dale, S. G and Land, J. M. (1991). Some observations
on a comparative aerial-photography interpretation of a
landslipped area. Quarterly Journal of Engineering Geology
and Hydrogeology, 24, 249–265.
Griffiths, J. S., Hutchinson, J. N., Brunsden, D., Petley, D. J. and
Fookes, P. G. (2004). The reactivation of a landslide during the
construction of the Ok Ma tailings dam, Papua New Guinea.
Quarterly Journal of Engineering Geology and
Hydrogeology, 37, 173–186.
Hearn, G. J. (2002). Engineering geomorphology for road design
in unstable mountainous areas: lessons learnt after 25 years in
Nepal. Quarterly Journal of Engineering Geology and
Hydrogeology, 35, 143–154.
Ibsen, M-L. and Brunsden, D. (1996). The nature, use and problems
of historical archives for the temporal occurrence of
landslides, with specific reference to the south coast of
Britain, Ventnor, Isle of Wight. Geomorphology, 15, 241–258.
Keefer, D. K. (1994). The importance of earthquake-induced
landslides to long-term slope erosion and slope-failure hazards
in seismically active regions. Geomorphology, 10, 265–284.
Korup, O., McSaveney, M. J. and Davies, T. R. H. (2004).
Sediment generation and delivery from large historic landslides
in the Southern Alps, New Zealand. Geomorphology,
61, 189–207.
Lang, A. (1999). Classic and new dating methods for assessing
the temporal occurrence of mass movements.
Geomorphology, 30, 33–52.
Lin, J-C, Petley, D. N., Jen, C-H. and Hsu, M-L. (2006). Slope
movements in a dynamic environment: A case study of Tachia
river, Central Taiwan. Quaternary International, 147,
103–112.
Lynett, P. J., Borrero J. C., Liu, P. L.-F. and Synolakis, C. E.
(2003). Field survey and numerical simulations: a review of
the 1998 Papua New Guinea earthquake and tsunami. Pure
and Applied Geophysics, 160, 2119–2146.
Malamud, B. D., Turcotte, D. L., Guzzetti, F. and
Reichenbach, P. (2004). Landslide inventories and their statistical
properties. Earth Surface Processes and Landforms, 29,
687–711.
Mitchell, W. A., McSaveney, M., Zondervan, A. et al. (2007).
The Keylong Serai rock avalanche, NW Indian Himalaya:
geomorphology and palaeoseismic implications. Landslides,
4, 245–254.
Petley, D. N., Dunning, S. A. and Rosser, N. J. (2005a). The
analysis of global landslide risk through the creation of a
database of worldwide landslide fatalities. In O. Hungr,
R. Fell, R. Couture and E. Eberhardt, (eds.), Landslide Risk
Management, Amsterdam: A. T. Balkema, pp. 367–374.
David Petley72
Downloaded from Cambridge Books Online by IP 147.251.65.55 on Thu Mar 17 10:25:08 GMT 2016.
http://dx.doi.org/10.1017/CBO9780511807527.006
Cambridge Books Online © Cambridge University Press, 2016
Petley, D. N., Mantovani, F., Bulmer, M. H. K. and Zannoni, F.
(2005b). The interpretation of landslide monitoring data for
movement forecasting. Geomorphology, 66, 133–147.
Petley, D. N., Hearn, G. J., Hart, A. et al. (2007). Trends in
landslide occurrence in Nepal. Natural Hazards, 43, 23–44.
Tagliavini, F., Mantovani, M., Marcato, G., Pasuto, A. and
Silvano, S. (2007). Validation of landslide hazard assessment
by means of GPS monitoring technique: a case study in the
Dolomites (Eastern Alps, Italy). Natural Hazards and Earth
System Sciences, 7, 185–193.
Varnes D. J. (1978). Slope movement types and processes. In
R. L. Schuster and R. J. Krizek (eds.), Landslides, Analysis
and Control. Transportation Research Board Sp. Rep. No.
176, National Academy of Sciences, pp. 11–33.
73Landslide hazards
Downloaded from Cambridge Books Online by IP 147.251.65.55 on Thu Mar 17 10:25:08 GMT 2016.
http://dx.doi.org/10.1017/CBO9780511807527.006
Cambridge Books Online © Cambridge University Press, 2016
Downloaded from Cambridge Books Online by IP 147.251.65.55 on Thu Mar 17 10:25:08 GMT 2016.
http://dx.doi.org/10.1017/CBO9780511807527.006
Cambridge Books Online © Cambridge University Press, 2016