State of the Art Paper 1

A framework for landslide risk assessment and management

R. Fell,
School of Civil and Environmental Engineering, The University of New South Wales, Sydney, Australia
K.K.S. Ho
Geotechnical Engineering Office, Civil Engineering and Development Department, Hong Kong Special
Administrative Region.
S. Lacasse
International Centre for Geohazards, Norwegian Geotechnical Institute, Oslo, Norway
E. Leroi
Urbater, Roquevaire, France
ABSTRACT: This paper provides a framework for landslide risk assessment and management. It outlines the
processes of hazard analysis, including characterization of the landslide (the danger); frequency analysis; the
risk estimation calculation; risk evaluation against risk tolerance criteria and value judgements. The paper
discusses the benefits and limitations of quantitative and qualitative risk management, and gives simplified
examples.
This paper also discusses the advantages, disad-
vantages and limitations of QRA for engineered
1 INTRODUCTION slopes and landslides. The other seven State of the
Art (SOA) papers in this Conference provide the
Landslides and engineered slopes have always in- details of the methods that can be used. The invited
volved some form of risk assessment and man- and submitted papers in this volume deal with spe-
agement. This was often done by the use of engi- cific applications, case studies, research and devel-
neering judgement by the Geotechnical Engineers opment.
or Engineering Geologists in consultation with
owners and regulators.
The more formal applications of risk assessment 2 TERMINOLOGY
and management principles, in a qualitative man-
ner, have been practised for landslide hazard zon- The International Society of Soil Mechanics and
ing for urban planning and highway slope man- Geotechnical Engineering (ISSMGE) Technical
agement since the 1970s. In the 1980s, and Committee on Risk Assessment and Management
particularly in the 1990s, these have been ex- (TC32) developed a Glossary of Terms for Risk
tended to quantitative methods, and to manage- Assessment, based on IUGS (1997), ICOLD
ment of individual slopes, pipeline routes, subma- (2003), and National Standards such as British
rine slopes and more global slope risk Standard BS 8444, Australia-New Zealand Stan-
management. dard AS/NZS 4360, and Canadian Standard
These developments are described by Varnes CAN/CSA Q 634-91. The Glossary is attached to
(1984), Whitman (1984), Einstein (1988, 1997), this volume and these terms are used throughout
Fell (1994), Leroi (1996), Wu, et al. (1996), Fell all the SOA papers.
and Hartford (1997), Nadim and Lacasse (1999) Readers are encouraged to use these terms so
Ho, et al. (2000) Kvalstad et al. (2001), Nadim et that there is consistency across the international
al. (2003), Nadim and Lacasse (2003, 2004), Hart- community. The most important terms and their
ford and Baecher, and (2004), and Lee and Jones definitions are:
(2004). Some guidelines have been developed Annual exceedance probability (AEP): The
(e.g. Australian Geomechanics Society, 2000). estimated probability that an event of specified
At this time there exists a generic framework magnitude will be exceeded in any year.
for the use of quantitative risk assessment (QRA) Consequence: In relation to risk analysis, the
for engineered slopes and landslides; including in- outcome or result of a hazard being realised.
dividual slopes, groups of slopes (such as cuts and Danger (Threat): The natural phenomenon that
fills on a length of highway), land use planning could lead to damage, described in terms of its ge-
and zoning for urban development and global or ometry, mechanical and other characteristics. The
regional landslide risk management. This paper de- danger can be an existing one (such as a creeping
scribes this framework. slope) or a potential one (such as a rockfall). The
characterisation of a danger or threat does not in- hazard, evaluation of the vulnerability of the ele-
clude any forecasting. ment(s) at risk, consequence identification, and
Elements at risk: Population, buildings and risk estimation. Consistent with the common dic-
engineering works, infrastructure, environmental tionary definition of analysis, viz. A detailed ex-
features and economic activities in the area af- amination of anything complex made in order to
fected by a hazard. understand its nature or to determine its essential
Frequency: A measure of likelihood expressed features, risk analysis involves the disaggregation
as the number of occurrences of an event in a given or decomposition of the system and sources of risk
time or in a given number of trials (see also likeli- into their fundamental parts.
hood and probability). Qualitative risk analysis: An analysis which
Hazard: Probability that a particular danger uses word form, descriptive or numeric rating
(threat) occurs within a given period of time. scales to describe the magnitude of potential con-
Individual risk to life: The increment of risk sequences and the likelihood that those conse-
imposed on a particular individual by the existence quences will occur.
of a hazard. This increment of risk is an addition to Quantitative risk analysis: An analysis based
the background risk to life, which the person on numerical values of the probability, vulnerabil-
would live with on a daily basis if the facility did ity and consequences, and resulting in a numerical
not exist. value of the risk.
Likelihood: Conditional probability of an out- Risk assessment: The process of making a de-
come given a set of data, assumptions and informa- cision recommendation on whether existing risks
tion. Also used as a qualitative description of prob- are tolerable and present risk control measures are
ability and frequency. adequate, and if not, whether alternative risk con-
Probability: A measure of the degree of cer- trol measures are justified or will be implemented.
tainty. This measure has a value between zero (im- Risk assessment incorporates the risk analysis and
possibility) and 1.0 (certainty). It is an estimate of risk evaluation phases.
the likelihood of the magnitude of the uncertain Risk control: The implementation and en-
quantity, or the likelihood of the occurrence of the forcement of actions to control risk, and the peri-
uncertain future event. odic re-evaluation of the effectiveness of these ac-
There are two main interpretations: tions.
i) Statistical frequency or fraction The Risk evaluation: The stage at which values and
outcome of a repetitive experiment of some kind judgement enter the decision process, explicitly or
like flipping coins. It includes also the idea of implicitly, by including consideration of the impor-
population variability. Such a number is called an tance of the estimated risks and the associated so-
objective or relative frequentist probability be- cial, environmental, and economic consequences,
cause it exists in the real world and is in principle in order to identify a range of alternatives for man-
measurable by doing the experiment. aging the risks.
Risk management: The systematic application
ii) Subjective probability (degree of belief)
of management policies, procedures and practices
Quantified measure of belief, judgement, or confi-
to the tasks of identifying, analysing, assessing,
dence in the likelihood of an outcome, obtained by mitigating and monitoring risk.
considering all available information honestly, Risk mitigation: A selective application of ap-
fairly, and with a minimum of bias. Subjective propriate techniques and management principles to
probability is affected by the state of understanding reduce either likelihood of an occurrence or its ad-
of a process, judgement regarding an evaluation, or verse consequences, or both.
the quality and quantity of information. It may Societal risk: The risk of widespread or large
change over time as the state of knowledge scale detriment from the realisation of a defined
changes. risk, the implication being that the consequence
Risk: Measure of the probability and severity of would be on such a scale as to provoke a
an adverse effect to life, health, property, or the socio/political response.
environment. Quantitatively, Risk = Hazard x Po- Temporal (spatial) probability: The probabil-
tential Worth of Loss. This can be also expressed ity that the element at risk is in the area affected by
as Probability of an adverse event times the con- the danger (threat) at the time of its occurrence.
sequences if the event occurs. Tolerable risk: A risk within a range that soci-
Risk analysis: the use of available information ety can live with so as to secure certain net bene-
to estimate the risk to individuals or populations, fits. It is a range of risk regarded as non-negligible
property or the environment, from hazards. Risk and needing to be kept under review and reduced
analyses generally contain the following steps: further if possible.
definition of scope, danger (threat) identification, Vulnerability: The degree of loss to a given
estimation of probability of occurrence to estimate element or set of elements within the area affected
by a hazard. It is expressed on a scale of 0 (no loss) ing used to prioritise the implementation of risk
to 1 (total loss). reduction measures.
Also, a set of conditions and processes resulting The risk management process in Figure 1 can be
from physical, social, economic, and environ- divided in phases. Five of these are illustrated by
mental factors, which increase the susceptibility of the darker shades in Figure 2. The graphics illus-
a community to the impact of hazards. trate that each new phase includes the previous
Other terms to describe landslide classification, one(s) and that the solution becomes more in-
features and geometry are detailed in Appendix A volved as one progresses through the different
of this volume. phases. The 5 phases together form an integrated
framework schematically illustrated in the graphics
in Figure 3.
3 THE RISK MANAGEMENT PROCESS

Figures 1, 2 and 3 describe the overall risk man- 4 LANDSLIDE RISK ANALYSIS
agement process.
Hazard analysis involves characterising the 4.1 Scope definition
landslide (classification, size, velocity, mechanics,
To ensure that the risk analysis addresses the rele-
location, travel distance), and the corresponding
vant issues, satisfies the needs of those concerned,
frequency (annual probability) of occurrence.
and to avoid misunderstandings, it is important to
Risk analysis includes hazard analysis and con-
define the scope of the risk analysis:
sequence analyses. Consequence analysis includes
(a) Is the analysis for a single site (e.g. a road
identifying and quantifying the elements at risk
cutting, or a building); a number of sites, (e.g. all
(property, persons), their temporal spatial probabil-
the road cuttings on a length of road); hazard zon-
ity, their vulnerability either as conditional prob-
ing for land-use planning; or global risk assess-
ability of damage to conditional probability of
ment, where for example cut slopes on all roads
damage to property, or conditional probability of
in a local government area are being studied uni-
loss of life or injury.
versally to formulate policies and prioritise mitiga-
Risk assessment takes the output from risk
tion actions?
analysis and assesses these against values judge-
(b) The geographic limits. Note that to be com-
ments, and risk acceptance criteria.
plete, the effects of landsliding up slope of a site,
Risk management takes the output from the risk
not confined to the site may need to be considered;
assessment, and considers risk mitigation, includ-
and the impacts of the landsliding on sites
ing accepting the risk, reducing the likelihood, re-
downslope, e.g. of a road fill, may also need to be
ducing consequences e.g. by developing monitor-
part of the analysis.
ing, warning and evacuation plans or transferring
(c) Whether the analysis will be restricted to
risk (e.g. to insurance), develops a risk mitigation
property loss or damage, or it will also include as-
plan and possibly implements regulatory controls.
sessment of the potential for loss of life and injury.
It also includes monitoring of the risk outcomes,
(d) The extent of geotechnical engineering and
feedback and iteration when needed.
geological studies which will form the basis of the
The process is iterative within any one study,
analysis. These can control the overall standard of
and should be up-dated periodically as monitoring
the risk analysis.
results become available.
(e) The approach to be used to characterise the
Landslide risk management involves a number
landslides, and assess the frequency of landsliding,
of stakeholders including owners, occupiers, the
and their consequences.
affected public and regulatory authorities, as well
(f) Whether the analysis will be quantified or
as geotechnical professionals, and risk analysts.
qualitative.
It is an integral part of risk management that the
(g) How risk acceptance criteria will be deter-
estimated risks are compared to acceptance criteria
mined, by whom, and through what process? The
(either quantitative or qualitative). Geotechnical
extent to which the stakeholders (owners, public,
professionals are likely to be involved as the risk
regulator, risk analyst) will be involved.
analysts, and may help guide in the assessment and
(h) Operational (e.g. land access) and financial
decision process, but ultimately it is for owners,
constraints to the analysis.
regulators and governments to decide whether the
(i) Legal responsibilities of all parties.
calculated risks are acceptable or whether risk
(j) The nature of the end product of the risk
mitigation is required.
analysis report, maps, and how these will be
In some cases the absolute values of risk are not
communicated to the interested parties.
as important as the relative risks. This is often the
case for risk assessments for cuts and fills on
highways, where the risk assessment process is be-
 Scope
Definition

Hazard Analysis

Landslide (Danger)
Characterisation

RISK ANALYSIS
Analysis of Frequency

Consequence Analysis

RISK ASSESSMENT
Characterisation of
Consequence Scenarios

Analysis of Probability and

Severity of Consequence

RISK MANAGEMENT
Risk Estimation

Value Judgements and

Risk Tolerance Criteria

Risk Evaluation vs
Tolerance Criteria and Value
Judgements

Risk Mitigation
Options?

Risk Mitigation and

Control Plan

Implementation of risk
mitigation

Monitor, Review and Feedback

Figure 1 Flow chart for landslide risk management.

 (a) Landslide (danger) characterisation (b) Hazard analysis

Frequency
analysis

(c) Risk analysis (d) Risk evaluation

(e) Risk mitigation and control

Figure 2 Representation of 5 phases of the Risk Management Process

R I S K M A N A G E M E N T

RISK ASSESSMENT

RISK ANALYSIS

Political HAZARD ANALYSIS

Aspirations Social Elements at
demands risk LANDSLIDE (DANGER)
Other Frequency CHARACTERISATION
constraints Regulation Vulnerability analysis Mechanics, Location
Volume,Travel Distance
and Velocity
Budget Risk Temporal
acceptance Spatial
criteria probability

Consequences

Values
Judgement

Risk mitigation Monitor and

Control options & Control plan Review

Figure 3 Schematic representation of the integrated risk management process.

 4.2.2 Frequency analysis
The frequency of landsliding can be expressed in
4.2 Hazard analysis terms of (IUGS 1997):
Hazard analysis is the process of identification - The number of landslides of a certain charac-
and characterisation of the potential landslides to- teristic that may occur in a study area per year.
gether with evaluation of their corresponding fre- - The probability of a particular slope experi-
quency of occurrence. encing landsliding in a given period, e.g. a
year.
4.2.1 Landslide (Danger) characterisation - The driving forces exceeding the resistant
Landslide (danger) characterisation requires an un- forces in probability or reliability terms, with
derstanding of the slope processes and the relation- the frequency of occurrence being determined
ship of those processes to geomorphology, geol- by considering the annual probability of the
ogy, hydrogeology, failure and slide mechanics, critical pore water pressures being exceeded in
climate and vegetation. From this understanding it the analysis.
will be possible to: - This should be done for each type of landslide
- Classify the types of potential landsliding: the which has been identified and characterised as
classification system as proposed by Varnes affecting the analysis.
(1984) or modified by Cruden & Varnes (1996) There are several ways of calculating frequency
forms suitable systems. A site may be affected (IUGS 1997):
by more than one type of landslide hazard e.g., (1) Historic data within the area of study, or ar-
slow rotational earth slides on the site, and very eas with similar characteristics, e.g. geology, geo-
rapid rockfall and debris flows from above the morphology.
site. (2) Empirical methods based on correlations in
Assess the physical extent of each potential accordance with slope instability ranking systems.
landslide,including the location, areal extent (3) Use of geomorphological evidence (coupled
and volume involved. with historical data), or based on expert judgement.
Assess the likely initiating event(s), the physical (4) Relationship to the frequency and intensity
characteristics of the materials involved, such as of the triggering event, e.g. rainfall, earthquake.
shear strength, pore pressures; and the slide me- (5) Direct assessment based on expert judge-
chanics. The latter is critical to understanding ment, which may be undertaken with reference to a
the pre and post failure behaviour of the land- conceptual model, e.g. use of a fault tree method-
slide. ology.
Estimate the resulting anticipated travel dis- (6) Modelling the primary variable, e.g. piezo-
tance, travel path, depth and velocity of move- metric pressures versus the triggering event, cou-
ment if failure occurs, taking account of the pled with varying levels of knowledge of geometry
slide mechanics, and estimating the probability and shear strength.
that the land slide will affect the area in which (7) Application of probabilistic methods, taking
the element at risk is located (PT:L) into account the uncertainty in slope geometry,
Identify possible pre-failure warning signs shear strength, failure mechanism, and piezometric
which may be monitored. pressures. This may be done either in a reliability
A list of possible landslides (dangers) should be framework, or taking into account the frequency of
developed. Consideration must be given to hazards failure (for example by considering pore pressures
located off site as well as within the site as it is on a frequency basis).
possible for landslides both upslope and (8) Combinations of the above methods.
downslope to affect the elements at risk. It is vital In practice it may be appropriate and advisable
that the full range of hazards (e.g. from small, high to use more than one method for the analysis.
frequency events to large, low frequency events) Details of the methods and their applicability
be properly characterised and considered in the are given in SOA Paper 2 in this volume. It is im-
risk analysis. Often the risk is dominated by the portant to express the probability of sliding in fre-
smaller, more frequent landslides. The effects of quency (per annum) terms, because quantitative
proposed development in an area should also be risk acceptance criteria for loss of life are usually
considered, as these effects may alter the nature expressed in per annum terms. Financial analysis
and frequency of potential hazards. of damage also usually requires frequency as an
It is important that geotechnical professionals input.
with training and experience in landsliding and The authors have a preference for estimating
slope processes are involved in this stage of the frequencies quantitatively. This gives a uniformity
analysis because the omission or under/over esti- of outcomes in quantified terms (rather than using
mation of the effects of different landslides often ill-defined subjective terms such as likely, unlikely
can control the outcomes of the analysis. etc.), allows risk to be compared with quantitative
acceptance criteria, and allows comparison with 4.3.1 Elements at risk
risks from other hazards with which the parties in- The elements at risk include the population, build-
volved may be able to associate. However it is rec- ings, engineering works, infrastructure, vehicles,
ognised that many practitioners are not familiar environmental features and economic activities
with quantifying landslide frequencies, and it is which are in the area affected by the hazard. In
important there are sanity checks on the results practical terms, this usually means on the land-
against historical performance data, and for more slide, and/or in the area onto which the landslide
important analyses, reviews by persons who are may travel if it occurs. It may also include property
experienced in landslide risk analysis. immediately adjacent to or upslope of the land-
For most hazard analyses, the estimation of fre- slide, if the property or its value would be affected
quency based on historical data, geomorphological by landsliding and infrastructure which may in-
evidence, relationship to trigger event frequencies clude powerlines, water supply, sewage, drainage,
etc. are typically more reliable than the apparently roads, communication facilities. The population at
more rigorous and detailed probabilistic analyses risk includes persons who live, work, or travel
because of the many uncertainties involved and through the area affected by the hazard.
data constraints. Also, some of the causes or con- It would be usual to categorise vehicles into
tributory factors to slope instability may not be cars, trucks and buses, because of the different
amenable to conventional limit equilibrium analy- number of persons likely to be in the vehicles.
sis, e.g. effects of topography on surface water The elements at risk are likely to be dependent
flows. on the nature of the landslide hazard e.g. for a
This is particularly true for smaller slopes, and boulder fall, or debris flow at a given site.
for landslides on natural hillsides, where it is very
difficult to estimate pore water pressures, and 4.3.2 Probability of landslide reaching the ele-
where small variations in strengths, and geometry ment at risk (PT:L)
and geological anomalies have large effects on the The probability of the landslide reaching the ele-
outcomes. There is also seldom sufficient data to ment at risk depends on the relative location of the
properly model such factors as auto-correlation of element at risk and the landslide source, together
parameters, so reliance is often placed on pub- with the path the landslide is likely to travel below
lished generalised information which may not be the source. It is a conditional probability between 0
applicable to the site under consideration. and 1.
(a) For buildings which are located on the
source landslide PT:L = 1.
4.3 Consequence analysis (b) For buildings or persons located below the
Consequence analysis involves: source landslide and in the path of the resulting
(a) Identifying and quantifying the elements at travel of the landslide, PT:L is calculated taking ac-
risk including property and persons. count of the travel distance of the landslide, the lo-
(b) Assessing temporal spatial probabilities for cation of the source landslide, and the element at
the elements at risk (PS:T). risk.
(c) Assessing vulnerability of the elements at (c) For vehicles or persons in vehicles, or per-
risk, in terms of property damage (Vprop:T) and loss sons walking in the area below the source landslide
of life/injury (VD:T) as appropriate. in the path of the resulting travel (runout) of the
This has to be done for each of the landslide landslide, PT:L is calculated taking account of the
hazards. travel distance of the landslide, and the path to be
The consequences may not be limited to prop- followed by the vehicle or person. Whether the ve-
erty damage and loss of life/injury. Other conse- hicle or person is in the path at the time of the
quences may include loss of reputation of the landslide is taken account through the temporal
owner and geotechnical engineers, consequential spatial probability (PS:T).
costs (e.g. a road is closed for some time affecting The methods for estimation of travel distance
businesses along the road), litigation from those in- are described in SOA 4 of this volume. This in-
jured or the relatives of those killed, potential volves some uncertainty which should be taken de-
criminal charges for those involved, political re- terminded.
percussions, adverse social and environmental ef-
fects. Most of these may not be readily quantifi- 4.3.3 Temporal spatial probability (PS:T)
able, but may need to be systematically considered, The temporal spatial probability is the probability
in consultation with owners and factored into the that the element at risk is in the area affected by
decision-making process as appropriate, at least for the hazard at the time of its occurrence. It is a con-
comprehensive risk analysis studies. ditional probability, and is between 0 and 1.
(a) For buildings on or in the path of the land-
slide, the temporal spatial probability is 1.
 (b) For a single vehicle which passes below a Landslides which move slowly (particularly
single landslide, it is the proportion of time in a those with a nearly planar, horizontal surface of
year when it will be in the path of the landslide. rupture) may cause little damage, other than to
(c) For all the vehicles which pass below a sin- structures which are on the boundaries of the land-
gle landslide, it is the proportion of time in a year slide and hence experience differential displace-
when a vehicle will be in the path of the landslide. ment.
Where there are a number of potential landslides in The rate of movement is less important for
any year, e.g. rockfalls, the calculation is some- structures than it is for loss of life, except in so far
what more complicated as described in SOA 5 in as it affects the time rate of damage, i.e. buildings
this volume. on a slow moving slide (which moves intermit-
(d) For persons in a building, it is the proportion tently every year) can be expected to have a lower
of time in a year which the persons occupy the vulnerability than those on a fast moving one.
building (0 to 1.0). This is likely to be different for Factors which most affect the vulnerability of
each person. persons include:
For persons in vehicles, the temporal spatial The velocity of landsliding. Persons are more
probability will be as for (b) and (c). However it likely to be killed by a rapid landslide than slow
may vary for say one person in a car, and four per- regardless of the landslide volume.
sons in a car. Landslide volume persons are more likely to
The range of credible consequence scenarios be buried or crushed by large landslides than
will need to be considered in societal risk calcula- small.
tions. Details of how to calculate temporal spatial Whether the person(s) are in the open, or in a
probability are given in SOA 5 of this volume. vehicle or building (ie. a function of the degree
For some situations it will be necessary to build of protection the person(s) has from the land-
into the calculation of temporal spatial probability, slide impact).
whether the person(s) at risk may have sufficient If they are in a building, whether the building
warning to evacuate from the area affected by the collapses upon impact by the landslide, and the
hazard. Persons on a landslide are more likely to nature of the collapse.
observe the initiation of movement and move off Persons who are buried by a landsliding mass
the slide than those who are below a slide falling have a high vulnerability. Death is more likely to
or flowing onto them. result from asphyxia than from crushing or impact.
Each case should take into account the nature of SOA 5 in this volume gives detailed information
the landslide including its volume, and velocity, on the assessment of vulnerability.
monitoring results, warning signs, evacuation sys-
tems, the elements at risk, and the mobility of the
persons. 4.4 Risk estimation
4.4.1 Risk calculation
4.3.4 Vulnerability (Vprop:T and VD:T) The risk can be presented in a number of ways:
Vulnerability is the degree of loss (or damage) to a (a) The annual risk (expected value) in which
given element, or set of elements, within the area the probability of occurrence of the danger is mul-
affected by the hazard. It is a conditional probabil- tiplied by the consequences summed over all the
ity, given the landslide occurs and the element at hazards. This is expressed as $x damage per an-
risk is on or in the path of the landslide. For prop- num; or potential loss of lives per annum.
erty, it is expressed on a scale of 0 (no loss or (b) Frequency consequence (f N) pairs for
damage) to 1 (total loss or damage) for property. example for property, the annual probability of
For persons it is usually the probability (be- minor ($x) damage; medium ($y) damage and ma-
tween 0 and 1) that given the person is on or in the jor ($z) damage; and for risk to life, the annual
path of the landslide, the person is killed. It may probability of loss of 1 life, 5 lives, 100 lives etc.
also include the probability of injury. (c) Cumulative frequency consequence plots
Factors that most affect vulnerability of prop- (F N plots), for example a plot of the annual
erty include: probability of N or more lives being lost (see sec-
The volume of the landslide in relation to the tion 5.2 and Figure 4).
element at risk It is often useful to calculate all three. The an-
The position of the element at risk, e.g. on the nual risk for property can be calculated from:
landslide, or immediately downslope
R(prop) = P(L) x P(T:L) x P(S:T) x V(prop:S) x E (1)
The magnitude of landslide displacement, and
relative displacements within the landslide (for where
elements located on the landslide) R(prop) is the annual loss of property value
The velocity of landslide movement. P(L) is the frequency of the landsliding
 P(T:L) is the probability of the landslide reach- From de Morgan's rule, the estimated upper
ing the element at risk bound conditional probability is
P(S:T) is the temporal spatial probability of the
PUB = 1 (1 P1)(1 P2) (1 Pn) (5)
element at risk
V(prop:S) is the vulnerability of the element at risk where
to the landslide event PUB = estimated upper bound conditional prob-
E is the element at risk (e.g. the value or ability
net present value of the property) P1 to Pn = the estimate of several individual
The annual probability that a particular person hazard conditional probabilities.
may lose his/her life can be calculated from: This calculation should be done before applying
the annual probability of the common causative
P(LOL) = P(L) x P(T:L) x P(S:T) x V(D:T) (2)
event. If all the conditional probabilities P1 to Pn
where are small (<0.01), equation 5 yields the same
(P(LOL) is the annual probability that the person value, within acceptable accuracy, as obtained by
will be killed adding all the estimated conditional probabilities.
V(D:T) is the vulnerability of the person to the (ii) The lower bound
landslide event The lower bound estimate is the maximum in-
and P(L), P(T:L) and P(S:T) are as defined above dividual conditional probability.
To estimate annual loss of life risk, equation (3)
is expanded to be as for equation (2) with E being 4.4.2 Uncertainty and sensitivity analysis
the number of persons at risk. The inputs into the risk estimation are not precise,
There are a number of situations where the risks usually involving a large contribution from engi-
from a number of landslide hazards have to be neering judgement, or uncertainty in input parame-
summed to give the total risk. These include: ters (e.g. for formal probabilistic analysis) (La-
Where the element at risk is exposed to a num- casse et al. 2003; 2004). Uncertainty describes any
ber of types of landsliding e.g. boulder fall, de- situation without certainty, whether or described
bris flows, and translational sliding by a probability distribution. Uncertainty is caused
Where the landsliding may be triggered by by natural variation and/or incomplete knowledge
more than one phenomena e.g. rainfall, earth- (lack of understanding or insufficient data). In the
quake, human activity. context of structural safety, uncertainty can be ei-
Where the element at risk is exposed to a num- ther aleatory (inherent variability in natural proper-
ber of different sizes of landslide of the same ties and events) or epistemic (incomplete knowl-
classification e.g. debris flows of 50m3, 5,000 edge of parameters and the relationships between
m3 and 100,000m3 volume. input and output values).
Where the element at risk is exposed to a num- Often for landslide risk assessments, it is not
ber of slopes on which landsliding can occur practical to model uncertainties formally e.g. by
e.g. a vehicle driving along a road in which assigning probability distributions to each input
there are 20 cut slopes each of which is a poten- and using Monte Carlo type analysis (e.g. Morgan
tial source of boulder falls. and Henrion, 1990). However, it is possible to do
In these cases, equations (1) and (2) should be sensitivity analysis by considering the effects of
written as: different assumed values for the inputs. It should
n
be recognised that the use of upper or lower limits
R(prop) = ( P( L ) x P( T:L ) x P(S:T ) x V( prop:S) x E) (3) of input variables in order to estimate upper and
1 lower bound results gives extremely low likelihood
and values, and that the analysis may be almost mean-
ingless.
n
P( LOL) = (P
1
( L) x P( T:L ) x P(S:T ) x VD:T ) (4)
4.4.3 Qualitative risk estimation
Qualitative risk analysis uses descriptors to de-
where n is the number of landslide hazards. scribe the frequency of landsliding and the conse-
This assumes that the hazards are independent quences. This may comprise tools such as risk rat-
of each other, which may often not be correct. If ing systems, risk scoring schemes, and risk ranking
one or more of the hazards may result from the matries (e.g. Stewart, et al. 2002). These can serve
same causative event e.g. a single rain event, or a useful role in landslide risk management in pro-
earthquake, then the probabilities should be esti- viding a relative comparison of risks of different
mated using the theory of uni-modal bounds as fol- sites and prioritisation of follow-up actions in ad-
lows: dressing the risk portfolio posed by a large number
(i) The upper bound of sites. In some cases, a hybrid approach may be
adopted whereby qualitative risk analysis can fa-
cilitate a first-pass screening of the more domi- Qualitative Risk Analysis Matrix Classes of Risk to Prop-
nant hazards in a given site so that attention can be erty
Likelihood Consequences to property
focused on the more deserving areas or hazards,
Catastrophic Major MediumMinor Insignificant
which can be evaluated in detail using quantitative Almost VH VH H H M
methods. Qualitative risk assessment may also be certain
used, coupled with engineering judgement, to ex- Likely VH H H M L-M
amine whether a given landslide hazard is posing a Possible H H M L-M VL-L
significant risk to life (e.g. a precariously perched Unlikely M-H M L-M VL-L VL
boulder above a busy highway with signs of dis- Rare M-L L-M VL-L VL VL
tress) and the need for prompt risk reduction Not VL VL VL VL VL
credible
measures (e.g. boulder removal) in order to safe-
Legend: VH very high risk
guard public safety, without the need for elaborate
H high risk
quantitative analysis. In general, qualitative risk
M moderate risk
assessment must be undertaken critically and pref-
L low risk
erably subject to expert review to avoid spurious
VL very low risk
outcomes and for it to be value-adding.
Table 1 gives an example adapted from AGS
(2000). In this case, the likelihood incorporates
Combining likelihood with consequence results
the frequency of landsliding, the probability of the
in a risk matrix divided into 5 classes from very
landslide reaching the element at risk, and tempo-
low risk (VL) to very high risk (VH).
ral spatial probability. The consequences incorpo-
Other schemes may be developed by the geo-
rate the vulnerability and the value of the element
technical risk analyst in consultation with the own-
at risk.
ers or other stakeholders where appropriate, to best
suit a given problem.
Table 1. Example of qualitative terminology for use in as- Qualitative risk assessment is subject to limita-
sessing risk to property adapted from AGS (2000) tions, which include potentially imprecise and sub-
Qualitative Measures of Likelihood of landsliding jective description of the likelihood term, for ex-
Level Descriptor Description ample adverse or could occur and hence are
A Almost certain The event is expected to occur liable to result in wide differences in the estimated
B Likely The event will probably occur un- risks, together with lack of risk acceptance criteria
der adverse conditions against which the qualitatively assessed risks can
C Possible The event could occur under ad-
verse conditions be evaluated.
D Unlikely The event could occur under very AGS (2000) recommended that schemes such as
adverse circumstances that shown in Table 1 are only applicable to con-
D Rare The event is conceivable but only sideration of risks to property. Extreme care must
under exceptional circumstances be exercised where qualitative risk assessment ap-
E Not credible The event is inconceivable or fan-
ciful
proaches are used for estimating risk of loss of life
and decision-making on site-specific basis, espe-
Qualitative Measures of Consequences to Property cially for marginal cases, because of the associated
Level Descriptor Description shortcomings
1 Catastrophic Structure completely destroyed or
large scale damage requiring major
engineering works for stabilisation. 5 LANDSLIDE RISK ASSESSMENT
2 Major Extensive damage to most of struc-
ture, or extending beyond site
boundaries requiring significant sta- 5.1 Risk assessment process
bilisation works.
3 Medium Moderate damage to some of struc-
Risk assessment involves taking the outputs from
ture, or significant part of site requir- the risk analysis and comparing them against val-
ing large stabilisation works. ues judgements and risk tolerance criteria to de-
4 Minor Limited damage to part of structure, termine if the risks are low enough to be tolerable.
or part of site requiring some rein- The process is one of making judgements, tak-
statement/stabilisation works. ing account of political, legal, environmental, regu-
5 Insignificant Little damage
latory and societal factors. The decision is usually
the responsibility of the owner and regulator,
sometimes consulting with the affected public or
stakeholders. Non-technical clients may seek guid-
ance from the risk analyst on whether to accept the
risk, but from a legal viewpoint it is important that
the owner and regulator make the final decision.
 Assessment of the risk may involve considera- other risks to which a person is exposed in eve-
tion of values such as: ryday life.
(a) Property or financial loss The incremental risk from a hazard should,
Annualised risk cost wherever reasonably practicable, be reduced,
Financial capability i.e. The As Low As Reasonably Practicable
Impact on corporate reputations (ALARP) principle should apply.
Insurance available If the possible loss of life from a landslide inci-
For railways and roads; accidents per million dent is high, the likelihood that the incident
tonnes of freight hauled, frequency of accidents might actually occur should be low. This ac-
Indirect costs e.g. loss of road access counts for societys particular intolerance to in-
When mitigation measures are being consid- cidents that cause many simultaneous casual-
ered, cost benefit ratio. ties, and is embodied in societal tolerable risk
criteria.
(b) Loss of life Persons in society will tolerate higher risks than
Individual risk to life. they regard as acceptable, when they are unable
Societal risk e.g. as a frequency versus number to control or reduce the risk because of financial
of deaths (known as f N) or cumulative fre- or other limitations.
quency versus number of deaths (known as F Higher risks are likely to be tolerated for exist-
N) criteria. ing slopes than for planned projects, and for
Annualised potential loss of life workers in industries with hazardous slopes,
When mitigation measures are being consid- e.g. mines, than for society as a whole.
ered, cost per statistical life saved. These principles are common with other dan-
gers such as Potentially Hazardous Industries
(PHI) and dams. (IUGS 1997) considered that
5.2 Risk acceptance criteria there are other principles that are applicable to risk
It is important to recognise the difference between from slopes and landslides:
acceptable and tolerable risks: Tolerable risks are higher for landslides on
Acceptable risk: A risk which everyone im- natural hillsides than those from engineered
pacted is prepared to accept. Action to further re- slopes.
duce such risk is usually not required unless rea- Once a natural slope has been placed under
sonably practicable measures are available at low monitoring, or risk mitigation measures have
cost in terms of money, time and effort. been executed, the tolerable risks approach
Tolerable risk: A risk within a range that soci- those of engineered slopes.
ety can live with so as to secure certain net bene- Tolerable risks may vary from country to coun-
fits. It is a range of risk regarded as non-negligible, try, as well as within a country, depending on
and needing to be kept under review and reduced historic exposure to landslide hazard, and the
further if possible. system of ownership and control of slopes and
Factors that affect an individual's attitude to ac- natural landslides hazards.
ceptable or tolerable risk will include (adapted There are no universally established individual
from AGS 2000): or societal risk acceptance criteria for loss of life
Resources available to reduce the risk. due to landslides. Guidance on what has been ac-
Whether there is a real choice, e.g. can the per- cepted in various countries is given in SOA 6 in
son afford to vacate a house despite the high this volume.
risk? The following are some examples:
The individuals commitment to the property (i) Individual risk
and its value relative to the individuals income. AGS (2000) suggested that, based on criteria
Age and character of the individual. adopted for Potentially Hazardous Industries, Aus-
Exposure the individual has experienced in the tralian National Committee on Large Dams (AN-
past, especially with regards to risk associated COLD 1994, which were also adopted in AN-
with landslides. COLD 2003); and the review in Fell and Hartford
Availability of insurance. (1997) the tolerable risk criteria shown in Table 2
Regulatory or policy requirements. might reasonably be concluded to apply to engi-
Whether the risk analysis is perceived to be re- neered slopes. They suggested that acceptable
liable. risks are usually considered to be one order of
There are some common general principles that magnitude smaller than these tolerable risks.
can be applied when considering tolerable risk to It should be noted the AGS (2000) guidelines
loss of life criteria (IUGS 1997): do not represent a regulatory position. ANCOLD
The incremental risk from a hazard to an indi- (2003) deleted reference to the average of persons
vidual should not be significant compared to
at risk, taking account only of the person most at established fact that people, including engineers,
risk. have a lot of trouble understanding small probabili-
ties and that in recent years, the f-N and F-N dia-
grams have proven to be useful tools for describing
Table 2. AGS (2000) suggested tolerable risk criteria the meaning of probabilities and risks in the con-
Situation Suggested tolerable risk for loss of life text of other risks with which society is familiar.
Existing engi- 10-4/annum person most at risk He points out that computed absolute probabilities
neered slopes 10-5/annum average of persons at risk
New engineered 10-5/annum person most at risk may not include all contributions; an effective ap-
slopes 10-6/annum average of the persons at risk proach is to compare probabilities of different op-
tions or alternatives. Probabilistic methodologies
also provide insight into the relative contributions
of different parameters to the uncertainty of the re-
(ii) Societal risk sult and thus give guidance for where further in-
The application of societal risk to life criteria is vestigations will be most fruitful.
to reflect the reality that society is less tolerant of Whether such quantitative criteria as the exam-
events in which a large number of lives are lost in ples given are acceptable in principle will depend
a single event, than of the same number of lives are on the country and legal system in which the land-
lost in a large number of separate events. Examples sliding is being considered. In some societies, e.g.
are public concern to the loss of large numbers of Australia, Hong Kong, and the United Kingdom,
lives in airlines crashes, compared to the many the use of such criteria for Potentially Hazardous
more lives lost in small aircraft accidents. Industries, and to a lesser extent dams and land-
The use of cumulative F-N curves to reflect this slides is gaining acceptance. In others, such as
is not universal. An example which has been tri- France, the legal framework currently precludes
alled on an interim basis to assist landslide risk the use at least in absolute terms. This is discussed
management of natural hillside hazards is shown in further in SOA6.
Figure 4. As pointed out in IUGS (1997), those who use
QRA for slopes and landslides should keep the fol-
lowing in mind when analysing, assessing and
managing risk:
(a) Estimates of risk are inevitably approximate,
and should not be considered as absolute values.
This is best understood by allowing for the uncer-
tainty in the input parameters, and in reporting the
risk analysis outcomes.
(b) Tolerable risk criteria are themselves not ab-
solute boundaries. Society shows a wide range of
tolerance to risk, and the risk criteria are only a
mathematical expression of the assessment of gen-
eral societal opinion.
(c) It is often useful to use several measures of
tolerable risk criteria, e.g. f-N pairs, individual and
societal risk, and measures such as cost to save a
life and maximum justifiable cost if risk mitigation
is being considered.
(d) It must be recognised that QRA is only one
input to the decision process. Owners, society and
regulators will also consider political, social and
legal issues in their assessments and may consult
the public affected by the hazard.
(e) The risk can change with time because of
natural processes and development. For example:
Depletion of debris from slopes can lead to a
Figure 4 Interim societal risk tolerance criteria (Geotechni-
reduction in risk with time
cal Engineering Office, 1998). Removal of vegetation by natural processes,
e.g. fire or human intervention, can lead to an
increase in risk
Christian (2004) also discusses the use of F-N Construction of roads on a slope may increase
criteria. He suggests that using the output of prob- the probability of landsliding and/or the ele-
abilistic analyses is hindered by the well- ments at risk, and hence the risk.
 (f) Extreme events should be considered as part for low occupancy use such as sports field or pas-
of the spectrum of events. This is relevant to the sive recreation. In some cases mitigation measures
triggering events (landslides, earthquake) the size as outlined above may be appropriate.
of the landslide and the consequences. Sometimes Apart from the consideration of risk mitigation
it is the smaller, more frequent, landslides that con- using engineering measures, landslide risk man-
tribute most to risk, not the low frequency very agement also consists of the use of soft (or non-
large event. engineering) options, such as public education
campaigns, public information services, etc. to ad-
dress the issue of risk tolerance by the general pub-
6 LANDSLIDE RISK MANAGEMENT lic or the stakeholders and avoid unduly high ex-
pectations of the level of safety that can be
6.1 Risk management process achieved in practice. Risk tolerance is related, in
The outcomes of the Risk Assessment will be ei- part, to the perception and understanding of land-
ther: slide risk. Risk communication to lay people forms
(a) The risks are tolerable, or even acceptable a key element of the landslide risk management
and no mitigation options need be considered. process in facilitating a better understanding of the
or nature and reality of landslide risk, and promoting
(b) The risks are intolerable, and risk mitigation the build-up of trust in, and credibility of, the risk
options need to be considered. analyst. Geotechnical professionals involved in
The risk management process is iterative, re- landslide risk assessment and risk management
quiring consideration of the risk mitigation options have an important role to play in risk communica-
and the results of the implementation of the mitiga- tion, which is best done using languages and
tion measures and of the monitoring. means that can be easily comprehensible by the
Examples of options for mitigation of risks for a general public.
slope or group of slopes would include:
Reduce the frequency of landsliding by stabi-
lization measures such as groundwater drainage, 7 THE BENEFITS AND LIMITATIONS OF
slope modification, anchors; or by scaling loose LANDSLIDE RISK MANAGEMENT
rocks,
Reduce the probability of the landslide reaching Some of the benefits of the use of quantitative risk
the element at risk e.g. for rockfalls, construct assessment in landslide risk management include:
rock catch fences; for debris flows construct (a) It encourages a rational, systematic approach
catch dams; to assessing the safety of natural and engineered
Reduce the temporal spatial probability of the slopes, by requiring an assessment of the charac-
element at risk e.g. by installing monitoring and teristics of the landslides, their travel distance and
warning systems so persons can evacuate; relo- velocity, frequency of sliding, the elements at risk,
cation of buildings to be further from the land- their temporal spatial probability and vulnerability.
slide; (b) It can be applied to situations which are not
Other risk management options may include: amenable to conventional deterministic analysis
Avoid the risk e.g. abandon the project, seek- e.g. rockfalls, small landslides in cut slopes, shal-
ing an alternative site or form of development low landslides and resulting debris flows on steep
such that the risk will be tolerable natural slopes.
Transfer the risk, by requiring another authority (c) It can be applied to land-use planning, with
to accept the risk, or to compensate for the risk specific loss of life acceptance criteria used to de-
such as by insurance (for property) termine the zoning where building is acceptable.
Postpone the decision if there is sufficient un- (d) It allows comparison of risks across an
certainty, awaiting the outcomes of further in- owners portfolio of slopes e.g. cut slopes on
vestigations, assessment of mitigation options, highways, and thereby allows prioritisation of re-
and monitoring. This would usually only be a medial works, and potentially setting of risk-based
temporary measure. standards for acceptable designs.
Finally a risk mitigation plan will be decided (e) Some local and regional government plan-
upon. There may be elements of control in this ners are familiar with risk management principles,
plan i.e. regulations imposed by local or other and welcome landslide risk management being
governments. presented in terms they can relate to other hazards.
For hazard analysis for land use planning, the (f) The process requires consideration of risks
emphasis may be on limiting building development for all levels of loading, rather than relying on ex-
to those areas where risks are assessed as likely to treme event loadings. Often failure paths will be
be acceptable, and using the higher hazard areas identified in the analysis which have been over-
looked.
 (g) It focuses attention or what happens if the though inevitably people will refer to what it is be-
slope fails, including the possibility of the slide ing done in societies with similar legal and social
travelling rapidly onto buildings below, causing values.
damage and loss of life. (g) Some over rely on the results of risk as-
(h) It focuses attention on liabilities and respon- sessments and do not understand the uncertainty
sibilities if the parties involved. in the probabilities calculated. This is for the ana-
(i) It provides a framework to put uncertainties lyst to understand, and convey in the reporting
and engineering judgement into a system. This re- process and when communicating with the public.
sults in an enhanced awareness of the need to con- (h) The authors experience is that many ex-
sider uncertainties, and insight on what can go perienced practitioners are reluctant to use quanti-
wrong, and their potential consequences, together tative approaches to estimating landslide frequen-
with how the uncertainties and risks can be best cies, because of their lack of experience in doing
managed this. This needs to be addressed by systematic, on
(j) It provides an open and transparent process the site training and review by experienced profes-
on the nature and key contributors of landslide risk sionals.
and the corresponding uncertainty for discussion (i) There is still a lack of general acceptance of
with the regulators, owners, stakeholders, etc. the method by the profession. It should be recog-
(k) It allows systematic consideration of risk nised that QRA is an engineering tool that may be
mitigation options and cost benefit ratios, consis- used for an appropriate problem or to supplement
tent with the As Low As Reasonably Practical other conventional tools for landslide risk man-
(ALARP) principles, thus encouraging optimisa- agement
tion and enhancing cost benefit.
Some of the challenges and perceived limita-
tions include (adapted from IUGS 1997): 8 EXAMPLES OF LANDSLIDE RISK
(a) The potential uncertainty in estimating fre- ASSESSMENT
quencies, travel distance and vulnerability. How-
ever these uncertainties can be modelled in the Figures 5, 6 and 7 give examples of certain ele-
analysis, or sensitivity studies done to get a feel for ments of landslides risk assessment. These are
their influence. simplified to illustrate the basic principles in-
(b) The variety of approaches, and the need for volved. Note that for convenience it has been as-
expert judgement to assess frequency of landslid- sumed that the tolerable risk criteria in Table 2 and
ing in many cases. This requires those doing the Figure 4 apply to the cases considered. Other ex-
analysis to be trained, and calibrated. Baynes et amples can be found in Lee and Jones (2004), La-
al (2002) give a good example of how this can be casse (1998), Ho et al (2000), and Fell and Hart-
achieved. ford (1997).
(c) Revisiting an assessment can lead to a sig-
nificant change in the assessed risk due to in-
creased data, or development of more advanced 9 CONCLUDING REMARKS
methods. This however is common to a conven-
tional deterministic approach. (a) The risk management framework presented
(d) Poor estimates of risk because significant in this paper has been successfully used in land-
hazards have been overlooked. This is a problem slide risk assessment and management for engi-
whichever approach is used, and can only be over- neered and natural slopes. The framework may be
come by using well trained and experienced geo- adapted to suit a variety of problems, with due re-
technical professionals to do the analyses. gard to the nature of the issues involved.
(e) Results of an assessment are seldom verifi- (b) Recent developments have included more
able. A possible approach to overcome this is to widespread use of quantitative methods; more re-
use systematic peer review by individuals or for fined hazard and risk zoning which often involves
larger projects, panels. The first author has seen use of digital technologies; improved rainfall-
how successful this can be in risk assessment for landslide incidence correlation models; and im-
dams. For slopes, where budgets are often smaller, proved methods for assessing travel distances and
peer review while still essential, is more likely to travel paths.
be done on a sample of the slopes being assessed, (c) While the emphasis in this paper is on quan-
but it still should be done. titative methods, current practice also involves the
(f) Acceptable and tolerable loss of life criteria use of risk-based qualitative methods in many ap-
for slopes and landslides are not well established. plications, including management of landslide
This is an issue which has to be overcome at the risks for roads and railways, and in land use plan-
country, state or local government level. It will not ning. These are valuable in that the landslide proc-
be practical to establish universal guidelines, al- esses are systematically studied, and can lead to
 FIGURE 5 EXAMPLE I LANDSLIDING IN ROAD FILL

Danger -
landslide in roadfill
Road
13
? Range of travel
20 distance angle

? Elements at risk -
? house and persons

0 5 10m Seepage
Scale

1. Scope definition
Calculate the risk to persons living in the house below a road as shown in the figure. Assess the tolerability of
this risk against the tolerable risk criteria shown in Table 1 and Figure 4.

2. Risk analysis

(i) Danger (Landslide) characterisation

The road was built 50 years ago, by cut and fill with a bulldozer. There was no proper compaction of the fill.
The site is underlain by granitic rocks, and the fill is derived from residual soils and completely weathered
granite which classifies as a silty sand. A thorough search of records has indicated that over the length of this
road, which is all in similar topography, geology and climatic conditions to this fill, there have been 4 land-
slides in a total of 60 fills.
Based on the geometry of the fill, and the landslides which have occurred, it is assessed that the likely vol-
ume of the slide is about 1000m3. Because of the loose, saturated nature of the fill it is anticipated that there
may be a large loss of undrained shear strength on sliding (static liquefaction) and the movement after fail-
ure is likely to be rapid.
Using empirical methods, it is estimated that the travel distance angle will be between 13o and 20o. Based
on this estimate, and the geometry of the slope, it is estimated that the probability of the landslide reaching the
element at risk (the house and its occupants) PT:L = 0.4.

(ii) Frequency analysis

Assuming this fill is similar to the other 60 fills on the road and that the 50 years of the roads performance
road is representative of the future, the frequency of sliding of the fill is:
4
PL = = 1.33 x 10 3 / annum
60 x 50

(iii) Consequence analysis

(a) Temporal spatial probability (P(S:T)) of the persons
Four persons live in the house. One of those persons is in the house 20 hours per day, 7 days per week; while
the other three are in the house 12 hours per day, 2 days per week.
For the person most at risk: For the other three persons:
20 12 2
P(S:T ) = = 0.83 P(S:T ) = x = 0.14
24 24 7

12 2
P(S:T ) = x = 0.14 assuming no warning.
24 7
(b) Vulnerability (of the persons (V(D:T))
Based on the volume of landsliding, its likely velocity when it hits the house, it is estimated that the vul-
nerability of the persons to being killed if they are in the house when the landslides hits is 0.4.
FIGURE 5 continued

(iv) Risk estimation

The annual probability of the person most at risk losing his/her life is
P( LOL ) = P( L ) x P( T:L ) x P(S:T ) x V( D:T )
= (1.33 x 10 3 ) x (0.4) x (0.83) x (0.4) / annum
= 1.7 x 10 4 / annum
The annual probability of four persons being in the house where it is hit by the slide (assuming the time
they spend in the house overlap)
= (1.33 x 10 3 ) x (0.4) x (0.14)
= 0.74 x 10 4 / annum
Since their vulnerability is 0.4, so 1.6 persons (say 1 to 2) would be killed.

3. Risk assessment
(i) Risk evaluation
(a) Individual Risk
From Table 2, the tolerable individual risk for an existing slope is 1 x 10-4/annum; so for the individual
most at risk, with P(LOL) = 1.7 x 10-4, the risk is just in the intolerable range.
(b) Societal Risk
From Figure 4 reproduced below, the societal risk is below the limit of tolerability line, but in the ALARP
region.

(ii) Comment
At this time, possible risk mitigation options would be considered, and the risks re-calculated. The ALARP
principle might be used along with values judgements to determine a risk mitigation and/or monitoring plan,
or to consider doing more geotechnical investigations to get an improved more accurate assessment of the
risk.
 FIGURE 6 EXAMPLE II ROCKFALLS FROM CUTTINGS ON A HIGHWAY

Danger -
rockfall in road cuttings

Elements at risk -
vehicles and their occupants

Lane Lane
N S

Highway

0 5 10m

Scale

1. Scope definition
Calculate the risk to persons travelling on the highway as shown in the figure. Assess the tolerability of this
risk against the tolerable risk criteria shown in Table 1 and Figure 4. Only consider direct impact falls.

2. Risk analysis

(i) Danger (landslide) characterisation

The road to a ski resort is privately owned and was built 10 years ago. The 50 cuts in the road were con-
structed at relatively steep slopes, and without treatment to control weathering, erosion and shallow instability
leading to rockfalls.
A thorough search of the maintenance records and observations of boulder impacts on the road surface in-
dicated that for the average cutting on the road, there have been 2 rockfalls per annum, with boulders ranging
in size from 0.5m dia to 1m dia. The cuttings are in similar topography, geology and climatic conditions.
Based on the recorded boulder impacts on the road surface, and the use of rockfall simulation programs, it is
assessed that 60% of rocks falling from the slope will impact on Lane N which is closest to the cut, and 10%
on Lane S.

(ii) Frequency analysis

The average frequency of rockfalls for each cutting is 2 per annum. There are a total of 50 cuts along the
road, giving a total of 100 rockfalls per annum or 0.27/day, the average frequency of rockfalls (NR) onto lane,
N = 0.6 x 0.27 = 0.16/day, and on Lane S, = 0.1 x 0.27 = 0.027/day.

(iii) Consequence analysis

(a) Temporal spatial probability (P(S:T)) of vehicles

The probability of a vehicle occupying the length of road onto which the rock falls is given by
NV L 1
P(S:T ) =
24 1000 VV
where NV = average number of vehicles/day
FIGURE 6 continued

L = average length of vehicle (metres)

VV = velocity of vehicle (km/hour)
For each lane, the average number of vehicles per day over the year is 2000, the average length of the ve-
hicles is 6 metres, and they are travelling at 60 km/hr, ignoring the width of the boulder:

For each lane For a particular vehicle travelling once each day in one direc-
tion
2000 6 1 1 6 1
P(S:T ) = P(S:T ) =
24 1000 60 24 1000 60
= 0.0083 = 0.0000042

(b) Vulnerability of the persons in the vehicles V(D:T)

Based on published information and judgement, it is estimated that the vulnerability of persons in vehicles
in lane N is 0.3 and in lane S, 0.15.
(iv) Risk estimation
The annual probability of the person most at risk losing his/her life by driving along the road is:
(a) For lane N (b) For lane S
P( LOL ) = P( S ) x VD:T = (1 (1 P( S :T ) ) N R ) x VD:T
P( LOL ) = (1 (1 0.0000084) 0.027 x 0.15
= (1 (1 0.0000042) 0.16
) x 0.3
7 = 0.3 x 10 7 / annum
= 2.0 x 10 / annum
The total probability of death for the person most at risk is 2.3 x 10-7/annum. For a person who only travels
on the road once per year in each direction, P(LOL) = 6.3 x 10-10/annum (2.3 x 10-7/365). The total annual risk
assuming each of the 2000 vehicles/day carries an average of 3 persons is 2000 x 365 x 3 x 6.3 x 10-10/annum
= 0.0014 persons/annum. The F-N plot has not been determined in this case.

3. Risk assessment
(i) Risk evaluation
(a) Individual risk
From Table 1, the tolerable individual risk for existing slopes is 1 x 10-4/annum. So for the individual most
at risk, with P(LOL) = 2.3 x 10-7/annum, the risks are within the tolerable limit. For an individual who drives on
the road only once per year, the risk is 6.3 x 10-10/annum, which would be acceptable. The societal risk limit
of tolerability for one life lost is 10-3/annum (see Figure 4). The estimated probability of one or more lives lost
is about 5 x 10-4/annum, near the tolerable limit.
(ii) Comment
(a) It is considered reasonable to sum the risks for all the road cuttings because the road is the responsibil-
ity of one organization.
(b) At this time, risk mitigation options would be considered. These could include engineering option to
reduce the frequency of rockfalls (rock-bolting, shotcreting, scaling of loose rocks in a regulated manner); re-
ducing the probability the rocks will fall onto the road (e.g. mesh protection over the slope, catch drain); or
reducing the probability of vehicles being below a rockfall when it occurs (e.g. closing the road in periods of
heavy rain if it could be demonstrated that is when most rockfalls occurred).
(c) See SOA Paper 5 for the equations for estimating risk.
 FIGURE 7 EXAMPLE III LANDSLIDING OF MINE WASTE DUMP

Mine waste
Water table

Danger -
Concrete landslide in mine waste
culvert

SECTION A-A
Cracked
culvert
Str
ea
m

A
Culvert Overflow
water

20
0

18
0
16
0

14
0
12
0
200

180

160

140

120

100

10
0

A
Str
ea
m

0 100 200m
d
Scale Roa Elements at risk -
houses and their occupants

Houses

PLAN

1. Scope definition
Calculate the risk to persons living in the houses and travelling on the road below the mine waste dump. As-
sess the tolerability of these risks against individual and societal tolerable risk criteria.

2. Risk analysis

(i) Danger (landslide) characterisation

The mine waste is silty sandy gravel and gravelly silty sand coarse reject from a coal washing. It was de-
posited over 50 years by end tipping. Geotechnical site investigations, hydrological and engineering analyses
have shown that:
(a) The waste is loose, and the lower part is saturated.
(b) The waste is likely to liquefy and flow liquefaction occurs for earthquakes loadings larger than 10-3
AEP
(c) The culvert through the waste dump exceeds its capacity and runs full for floods greater than 0.1 AEP.
For floods larger than this water flows over the sides of the waste dump and leaks onto the waste material
through cracks in the culvert, increasing the pore pressures in the waste.
FIGURE 7 continued

(d) The factor of safety of the dump under static loading is about 1.2 for water table levels which are
reached annually.
(e) If the dump slides even under static loading, it is likely to flow because of its loose, saturated granular
nature. The probability of this occurring given sliding occurs and the resultant debris flow reaching the houses
is 0.5 based on post liquefaction shear strengths, and empirical methods for estimating travel distance.
(f) The volume of the anticipated landslide and resulting debris flow is about 100,000m3 and the debris
flows are likely to be travelling at a high velocity when they reaches the road and houses.

(ii) Frequency analysis

The potential failure modes are:
(a) Culvert runs full, water leaks, saturates downstream toe, causes slide.
(b) As for (a), but a smaller slide, blocks/shears culvert, causes slide.
(c) Culvert collapses, flow saturates downstream toe, causes slide.
(d) A bigger flood, causes the culvert overflow, saturates fill, causes slide.
(e) As for (d), but scour of flowing water at toe of fill initiates slide.
(f) Rainfall infiltration, remobilizes slide.
(g) Earthquake causes liquefaction.
Based on the hydrology of the catchment, the hydraulics of the culvert, stability analyses and engineering
judgement, it is estimated that the frequency of landsliding of the waste for modes (a) to (f) is 0.01/annum.
Based on an analysis of liquefaction using a Youd et al (2001) approach, and post liquefaction stability
analysis, it is estimated that the frequency of landsliding for mode G is 0.005/annum.
Hence the total P(L) = 0.015/annum.

(iii) Consequence analysis

(a) Temporal spatial probability (P(S:T)) of the persons in the houses, and on the road
A survey of occupancy of the houses shows that the person most at risk in one of the houses is in the house
on average 18 hours/day, 365 days per year, so P(S:T) = 0.75.
Each house is occupied by a further 4 persons, for 10 hours/day, 325 days/year. Assuming they are all in
the houses at the same time. So:
10 325
P(S:T ) for 16 persons = x
24 365
= 0.36
Vehicles on the road travel at an average velocity of 30 km/hour as they pass by the 100 metres of road po-
tentially affected by the debris flow. So for each time the vehicle drives along the road,
100
P(S:T ) =
30,000 x 365 x 24
= 3.8 x 10 7
If a vehicle travels along the road 250 times a year (such as the school bus)
P(S:T) = 250 x 3.8 x 10-7 = 9.5 x 10-5
The critical vehicles for risk assessment are buses which travel 250 days/year.

(b) Vulnerability of persons (V(D:T)

Bases on the likely high velocity of sliding and large volume, it is estimated that the vulnerability of per-
sons in the houses is 0.9, and in a bus, 0.8.

(iv) Risk estimation

The annual probability of the person most at risk losing his or her life is
PLOL = p( L ) x P(T :L ) x P( S :T ) x V( D:T )
PLOL = (0.015) x (0.5) x (0.75) x 0.9/annum
= 5 x 10 3 / annum
FIGURE 7 continued

If all four houses are hit by the landslide, 0.9 x 16 or say 14 of the 16 persons would be killed. The annual
probability that this would happen is:
= 0.015 x 0.5 x 0.36/annum
= 2.7 x 10- 3 /annum
If a bus with 40 persons on it is hit by the landslide, 0.8 x 40 = 32 persons would be killed. The annual
probability this would happen is:
= 0.015 x 0.5 x 9.5 x 10-5 /annum
= 7.1 x 10- 7 /annum
So if loss of life of persons in other vehicles on the road is ignored, the cumulative F-N pair are:
One or more lives F = 5 x 10-3 + 2.7 x 10 3 + 7.1 x 10 7 = 7.7 x 10 3 / annum
15 or more lives = 2.7 x 10 3 + 7.1 x 10 7 = 2.7 x 10 3 / annum
33 lives F = 7.1 x 10 7 / annum
3. Risk assessment

(i) Risk evaluation

(a) Individual risk.
The risk for the person most at risk is 5 x 10-3/annum which is well in excess of the tolerable individual
risk in Table 1.
(b) Societal risk
The three points on the F-N curve are shown below. It can be seen that the risks are well in excess of the
tolerable for 1 and 15 lives, but in the ALARP range for 33 lives lost in a bus.

(ii) Comment
At this point, possible risk mitigation options would be considered, and the risks recalculated. The mitiga-
tion options could include reducing the probability of sliding by repairing the cracks in the culvert, controlling
water which overflows when the culvert capacity is exceeded; removing and replacing the outer waste well
compacted so it will not flow if it fails; adding a stabilizing berm; installing a warning system so persons in
the houses can be evacuated and the road blocked to traffic when movement is detected in the waste.
uniform classification of hazards and risks, which ANCOLD 2003. Guidelines on risk assessment. Australian
can be understood by those responsible for risk National Committee on Large Dams.
AS/NZS 4360:1999. Australian/New Zealand Standard, Risk
management. Qualitative approaches are better if Management Standards Australia, Standards New Zea-
they are underpinned by quantitative studies par- land.
ticularly where loss of life is an issue. RTA(2001) Australian Geomechanics Society 2000. Landslide risk
is an example of this for risk management of land- management concepts and guidelines. Australian Ge-
sliding affecting highways. Other examples in- omechanics Society, Sub-Committee on Landslide Risk
clude the design event approach for assessing miti- Management, Australian Geomechanics, 35: 49-92.
Baynes, F.J., Lee, I.K. and Stewart, I.E. 2002. A study of the
gation measures for natural hillside landslide accuracy and precision of some landslide risk analyses.
hazards (Ho, 2004). Australian Geomechanics, Vol.37, No.2, 149-156.
(d) Adoption of quantitative methods is likely to Canadian Standards Association 1991. Risk analysis re-
assist in risk communication in many cases be- quirements and guidelines. CAN/CSA-Q634-91, Cana-
cause regulators, politicians and managers of larger dian Standards Association, Toronto, Canada, 42.
Christian, J.T. 2004. Geotechnical Engineering reliability:
organizations are often familiar with quantifying how well do we know what we are doing? Journal Geo-
risks within other parts of their responsibilities. technical and Geoenvironmental Engineering, Vol.130,
Quantifying landslide risks allows these people to No.10, 985-1003.
assess them in perspective with those from other Cruden, D.M. & Varnes, D.J. 1996. Landslide types and
hazards. In some cases the use of quantitative risk processes: in Landslides Investigation and Mitigation.
assessment is stipulated by the regulator or owner. In K. Turner, R.L. Schuster (eds), Transportation Re-
search Board Special Report 247, National Research
(e) While the nature of the problem and avail- Council, Washington, D.C.
able methods for many studies always involve Einstein, H.H. 1988. Special lecture, landslide risk assess-
some degree of uncertainty in the risk estimates, ment. Proc. 5th Int. Symp. On Landslides, Lausanne, Swit-
this is not to say they should not be estimated, pro- zerland. A.A. Balkema, Rotterdam, The Netherlands, 2:
vided the limitations are acknowledged. Decisions 1075-1090.
Einstein, H.H. 1997. Landslide risk systematic approaches
have to be made despite the uncertainties, and it is to assessment and management in Landslide Risk As-
better to have an approximate estimate of the risks, sessment, Cruden and Fell (eds), Balkema, Rotterdam,
than none at all. The level of sophistication to be 25-50.
adopted in risk estimation for a particular problem Fell, R. & Hartford, D. (1997). Landslide risk management;
only needs to be sufficient to facilitate an informed in Landslide Risk Aassessment, Cruden and Fell (eds),
decision. Balkema, Rotterdam, 51-110.
Fell, R. 1994. Landslide risk assessment and acceptable risk,
(f) There is often an overemphasis on the risk Canadian Geotechnical Journal, 31: 261-272.
analysis, and not enough attention put on the risk Geotechnical Engineering Office 1998. Landslides and Boul-
assessment and management. It is important that der Falls from Natural Terrain: Interim Risk Guidelines.
Geotechnical Professionals involve themselves in GEO Report No.75, Geotechnical Engineering Office,
the assessment and management process because The Government of the Hong Kong Special Administra-
tive Region.
they often have the best understanding of the na- Hartford, D.N.D. & Baecher, G.S. 2004. Risk and uncer-
ture of the hazard and the risk. However the final tainty in Dam Safety. Thomas Telford, London.
decisions on tolerable risks lie with owners, regu- Ho, K.K.S. 2004. Recent advances in geotechnology for
lators and politicians. slope stabilization and landslide mitigation perspective
(g) The authors cannot over-emphasise the need from Hong Kong, in Landslides: evaluation and stabiliza-
for proper geotechnical inputs to the risk analysis, tion. Lacerda, Ehrlich, Fontoura and Sayo (eds). Taylor
and Francis Group, London, 1507-1560.
particularly with respect to the hazard identifica- Ho, K., Leroi, E. & Roberds, B. 2000. Quantitative risk as-
tion and quantification. Risk assessment is not a sessment. Application, myths and future directions.
substitute for good geotechnical engineering GeoEng 2000, Technomic Publishing, 269-312.
knowledge and judgement. It enhances it by add- ICOLD 2003. Risk assessment in dam safety management,
ing insight. Draft ICOLD Bulletin. International Commission on
Large Dams, Paris.
IUGS 1997. Quantitative risk assessment for slopes and land-
slides the State of the Art. IUGS Working Group on
10 ACKNOWLEDGEMENTS Landsldies, Committee on Risk Assessment, in Landslide
risk assessment, Cruden and Fell (eds), Balkema, Rotter-
This paper is published with the permission of the Head of dam, 3-12.
the Geotechnical Engineering Office and the Director of Kvalstad, T.J., F. Nadim, and C.B. Harbitz 2001. Deepwater
Civil Engineering and Development, Government of the geohazards: Geotechnical concerns and solutions. Off-
Hong Kong Special Administrative Region. shore Technology Conference, OTC 01, Houston, Texas,
30 April 3 May, Paper OTC 12958., 11 p
Lacasse, S. 1998. Risk and Reliability in Geotechnical Engi-
neering. State-of-the-Art paper. Fourth International Con-
11 REFERENCES ference on Case Histories in Geotechnical Engineering,
St-Louis, Missouri, USA. March 1998. (NGI Report
ANCOLD 1994. Guidelines on risk assessment. Australian 594000-4.)
Committee on Large Dams.
Lacasse, S. 2004. Risk Assessment for Geotechnical Solu-
tions Offshore. Keynote Paper. OMAE2004-51144. Proc.
of OMAE 2004 -23rd International Conference on Off-
shore Mechanics and Arctic Engineering. Vancouver,
Canada. June 2004, 16 p.
Lacasse, S., Nadim, F. and Heg, K. 2003. Risk Assessment
in Soil and Rock Engineering. PanAm Conference,
SARA, June 2003, pp. 2743-2750
Lacasse, S., Nadim, F., Heg, K. and Gregersen, O. 2004.
Risk Assessment in Geotechnical Engineering: The Im-
portance of Engineering Judgement. The Skempton Con-
ference, Proc. London UK. Vol.2, pp 856-867.
Lee, E.M. and Jones, D.K.C. 2004. Landslide Risk Assess-
ment. Thomas Tilford Publishing, London.
Leroi, E. 1996. Landslide hazard risk maps at different
scales: objectives, tools and developments. In Landslides
Proc. Int. Symp. On Landslides, Trondheim, 17-21 June,
K. Senneset (ed), 35-52.
Morgan, M.G. and Henrion, M. 1990. Uncertainty a guide
to dealing with uncertainty in quantitative risk and policy
analysis. Cambridge University Press.
Nadim, F. 2002. Probabilistic methods for geohazard prob-
lems: State-of-the-Art. Probabilistics in GeoTechnics:
Technical and Economic Risk Estimation, Graz, Austria,
September 15 -19, pp. 333-350
Nadim, F. and Lacasse, S. 2004. Mapping of landslide hazard
and risk along the pipeline route. GeoPipe 2004. Terrain
and Geohazard Challenges facing Onshore Oil and Gas
Pipelines. Proc. (loose leaf binder), London, June 2004.
Nadim, F. and S. Lacasse (1999). Probabilistic Slope Stabil-
ity Evaluation. Proc. "Geotechnical Risk Management",
Hong Kong Institution of Engineers, Hong Kong, 177-
186.
Nadim, F. and S. Lacasse 2003. Review of probabilistic
methods for quantification and mapping of geohazards.
Geohazards 2003, Edmonton, June 2003, pp. 279-286
Nadim, F., D. Krunic, and P. Jeanjean 2003. Probabilistic
slope stability analyses of the Sigsbee Escarpment. Procs,
OTC 15203, Offshore Technology Conference 03, Hous-
ton, Texas, May 2003, OTC paper 15203, 8 p.
RTA 2001. Guide to slope risk analysis version 3.1. Roads
and Traffic Authority, Sydney.
Stewart, I.E., Baynes, F.J. and Lee, I.K. 2002. The RTA
guide to slope risk analysis version 3.1. Australian Ge-
omechanics, Volume 37, No.2, 115-149.
Varnes, D.J. & The International Association of Engineering
Geology Commission on Landslides and Other Mass
Movements 1984. Landslide hazard zonation: A review of
principles and practice. Natural Hazards, 3: 63. Paris,
France. UNESCO.
Whitman, R.V. 1984. Evaluating calculated risk in geotech-
nical engineering, J. Geotech. Eng., ASCE, Vol.110,
No.2, 145-188.
Wu, T.H., Tang, W.H. and Einstein, H.H. (1996). Landslide
hazard and risk assessment; in landslides investigations
and mitigation, Transportation Research Board Special
Report 247, national Research Council Washington DC.
Youd, T.L., Idriss, I.M., Andrus, R.D., Arango, I., Castro, G.,
Christian J.T., Dobry, R., Finn, W.D.L., Harder, L.F., Hynes,
M.E., Ishihara, K., Koester, J.P., Liao, S.S.C., Marcuson,
W.F., Martin, G.R., Mitchell, J.K., Moriwaki, Y., Power,
MS., Robertson, P.K., Seed, R.B., and Stokoe, K.H. 2001.
Liquefaction resistance of soils: summary report from the
1996 NCEER and 1998 NCEER/NSF Workshops on evalua-
tion of liquefaction resistance of soils. J. Geotechnical and
Geoenvironmental Engineering, ASCE, Vol. 127, No. 10,
817-834.