Reliability of Slopes and Landslide Risk Assessment

RELIABILITY OF SLOPES AND LANDSLIDE RISK ASSESSMENT
G. R. Dodagoudar
Professor, Department of Civil Engineering
I.I.T. Madras, Chennai - 600 036, India
E-mail: goudar@iitm.ac.in
INTRODUCTION
The uncertainty is a fact of life in Civil Engineering. Recognition and assessment of

uncertainties in geotechnical engineering is of paramount importance. Conventional design
procedures adopted in geotechnical engineering often fall short of expectations in new
situations. In conventional geotechnical analyses, the properties of geomaterials are
assumed to be characterised by some fixed values that are determined from laboratory
testing of samples or in situ measurements. Analyses then proceed with these values in a
deterministic manner yielding a solution that is dependent on the quality of the
characteristic values that have been used. However, there is an increasing awareness that
geologic materials exhibit significant variability in their properties and distribution. Under
conditions of uncertainty, the design and planning of engineering systems involve risks
which in turn involve the probability and associated consequences. In this light,
probabilistic methods can be of great value to the designer since it is possible to express
many of the decision uncertainties in terms of numerical probabilities, thus allowing
quantification of judgement. Geotechnical engineers have increasingly recognized the
importance of probabilistic approaches for analysis and decision-making. In particular,
slope stability and landslide hazard analyses can benefit greatly from the adoption of
methods and techniques within a probabilistic framework.
The occurrence of disasters cannot be attributed to natural processes alone. While natural
factors are dominant features of hazards, disasters often occur due to increased
vulnerability of communities and infrastructure resulting from excessive urban
development, inadequate infrastructure, poor quality control, incomplete understanding of
hazards and of vulnerability and lack of action to mitigate risk. Engineered slopes and
landslides have always involved some form of risk assessment and management. This was
often done by the use of engineering judgement by the geotechnical engineers in
1
consultation with owners and regulators. With specific reference to a slope, the current
design practice relies on using calculated factor of safety (FS) values in the design process
to take care of uncertainties associated with soil parameters, site stratigraphy, ignorance
and potential consequences of slope failure. Under ideal conditions, a FS of at least 1.0
should ensure safe design, but because of all uncertainties, a higher value of the FS is
desirable for design recommendations. Based on past experience, the FS in the range of
1.3 to 1.5 is usually recommended as it represents a level of acceptable safety that
minimises failures. Unfortunately, the common approach of using such a single value of
the design FS does not effectively relate to the quality of site characterisation and cannot
reflect the uncertainty in its underlying parameters. The problem of defining a better and
more consistent measure of safety of slopes can best be attempted by probability theory,
which computes the standard deviation of FS, reliability index or probability of failure pf
of slope, that depend not only on the expected values of the uncertain parameters used in
the analysis but also on their inherent scatter.
There are several types of uncertainty and a recent review has been presented by Christian
(2004). There are many sources of uncertainty in probabilistic slope stability and landslide
risk analyses. These uncertainties include the lack of knowledge and inability to model
precisely the following features: Spatial uncertainties: (i) site stratigraphy and
variability, (ii) geologic origins and characteristics of subsurface materials, and (iii)
groundwater levels. Data uncertainties: (i) in-situ soil and/or rock characteristics, and (ii)
soil and/or rock properties. Additionally, it is impossible to predict with certainty, the
future environmental conditions such as external loading, earthquakes and rain-induced
porewater pressures. The amount of uncertainty varies from one analysis to another and
from one site to another. Natural variability of geotechnical parameters is the most
obvious source of uncertainty. Uncertain parameters that affect the stability of a slope are
treated as random variables in the probabilistic formulation and then the performance of
the slope is assessed in the probabilistic framework (Dodagoudar, 2001; Chowdhury et al.,
2010). Such an analysis can take into account uncertainties associated with the site
stratigraphy, soil parameters and even the method of analysis. The final result will be the
reliability index or the probability of failure, or alternatively, the probability of
unsatisfactory performance. This paper presents the available approaches for uncertainty
representation, methods for probabilistic reliability analysis and application of probability
theory to reliability analyses of finite and infinite earth slopes. A landslide risk assessment
2
carried out for a specific site in Western Ghats region is given at the end followed by
concluding remarks.
REPRESENTATION OF UNCERTAINTY
Uncertainty could simply be defined as the occurrence of events that are beyond our
control. Any realistic model of a real-world phenomenon must take into account the
possibility of randomness. The quantities we are interested in will not be predictable in
advance but, rather, will exhibit an inherent variation that should be taken into account by
the model. This is usually accomplished by allowing the model to be probabilistic in
nature. Such a model is referred to as a probability model. The uncertainty can be in the
problem definition, in the available information, in the alternative solution methodologies
and their results, or in the random nature of the solution outcomes. They relate mainly to
our lack of a perfect understanding with regard to the phenomena and processes involved,
and lack of a perfect knowledge of how to determine parameter values for processes that
are fairly well understood. The various approaches for representing uncertainty in the
context of different domains of applicability are presented by Klir (1994):
Classical Set Theory

Uncertainty is expressed by sets of mutually exclusive alternatives in situations where only
one alternative is desired. This includes diagnostic, predictive and retrodictive
uncertainties. Here, the uncertainty arises from the nonspecificity inherent in each set.
Large sets result in less specific predictions, retrodictions, etc., than smaller sets. Full
specificity is obtained only when one alternative is possible.
Probability Theory
Probabilistic analysis is the most widely used method for characterizing uncertainty in
physical systems, especially when estimates of the probability distributions of uncertain
parameters are available. Uncertainty is expressed in terms of a measure on subsets of a
universal set of events. The uncertainty measure is a function that assigns a number
between 0 and 1 to each subset of the universal set. This number, called probability of the
subset, expresses the likelihood that the desired unique alternative is in this subset. Here,
the uncertainty arises from the conflict among likelihood claims associated with the subsets
of the universal set, each consisting of exactly one alternative. Since these alternatives are
mutually exclusive, nonzero probabilities assigned to two or more events conflict with one
another since only one of them can be the desired one.
3
Fuzzy Set Theory
Fuzzy set theory is a method that facilitates uncertainty analysis of systems where
uncertainty arises due to vagueness or fuzziness rather than due to randomness alone.
Fuzzy sets, similar to classical sets, are capable of expressing nonspecificity. Vagueness is
different from nonspecificity in the sense that vagueness emerges from imprecision of
definitions, in particular definitions of linguistic terms. In fuzzy sets, the membership is not
a matter of affirmation or denial, but rather a matter of degree or belief (Ross, 2010).
Fuzzy Measure Theory

This theory considers a number of special classes of measures, each of which is
characterized by a special property. Some of the measures used in this theory are
plausibility and belief measures, and the classical probability measures. Fuzzy measure
theory and fuzzy set theory differ significantly. In the fuzzy set theory, the conditions for
the membership of an element into a set are vague, whereas in the fuzzy measure theory,
the conditions are precise, but the information about an element is insufficient to determine
whether it satisfies those conditions.
Rough Set Theory

A rough set is an imprecise representation of a crisp set in terms of two subsets, a lower
approximation and upper approximation. Further, the approximations could themselves be
imprecise or fuzzy.
In the present study, uncertainty representation for the reliability analysis of finite earth
slopes is attempted using probability theory.
UNCERTAINTY ANALYSIS
Probabilistic approach can describe uncertainty arising from stochastic disturbances,

variability conditions and risk considerations. In the analysis and design of geotechnical
engineering systems, uncertainties arising in geotechnical and hydrologic aspects are of
particular interest. There are two types of uncertainties: objective uncertainties associated
with any random process or deducible from statistical samples and subjective
uncertainties for which no quantitative factual information is available. Knowledge of
statistical information describing the uncertainties is essential in reliability analysis.
Therefore, uncertainty analysis is a prerequisite for reliability analysis. Answers provided
by the reliability analysis can be used for making decisions or choosing actions. The
reliability analysis requires the clearly defined goals for the investigation. The main
4
objective of uncertainty analysis is to identify the statistical properties of a model output as
a function of uncertain input parameters. Furthermore, it offers the designer an insight
regarding the contribution of each uncertain input parameter to the overall uncertainty of
the model output.
Probability Distributions
Uncertain parameters that affect the stability of an engineering system are treated as
random variables in the probabilistic formulation. Such a parameter was presumed to have
a single, unique value for the deterministic analysis, but in the uncertainty analysis it will
have a range of values as defined by its probability density function (PDF). For practical
purposes, continuous distribution functions are usually adopted for reliability analysis. The
following are the two density functions most widely used models for the distribution of
random variables in geotechnical reliability analysis.
Normal distribution
The best known and most widely used probability distribution is the Gaussian distribution
or the normal distribution. Random variables with different means and variances can be
modelled by normal probability density functions with appropriate choices of the center
and width of the curve. The expected value of random variable X, E[X] = determines the
center of the probability density function and the variance V[X] = 2 determines the width,
where is the mean and is the standard deviation. Each has the characteristic symmetric
bell-shaped curve, but centers and dispersions differ. A normal random variable having the
mean and standard deviation has the PDF:
1 1 x 2
f X ( x) exp for < x < (1)
2 2
Skewness coefficient of the normal random variable is zero. The normal distribution
sometimes provides a viable alternative to approximate the probability of a random
variable, which is not normally distributed.
Lognormal distribution
If a random variable has a Lognormal distribution, then its natural logarithm has a normal
distribution. The PDF of a lognormal variable is given by
5
1 1 ln x 2
f X ( x) exp X
for 0 x < (2)
2 X x
2 X
where X and X are the two parameters of the lognormal distribution. The choice of a
Lognormal distribution for FS is often considered to be preferable to that of a Normal
distribution since there is no implication of negative values of FS and also because the tail
of the distribution is less dominating.
Propagation of Uncertainties
Probability concepts are important for both site-specific and regional studies concerning
slopes. Both qualitative and quantitative assessments of landslide susceptibility and hazard
require a good grasp of probability concepts. For natural slopes as well as for built-up or
engineered slopes such as embankments, excavations and dams, the uncertainties in
geotechnical properties are usually the most important. Such uncertainties have both
spatial and systematic components. A third source of uncertainty is model error. The
methods applicable for uncertainty analysis involve different levels of mathematical
complexity and data requirements. The choice of appropriate technique depends on the
nature of the problem at hand including the availability of information, model complexity,
and type and accuracy of results desired. A central problem in engineering uncertainty
analysis is that: Given a function of random variables, Y = g(X) where X = (X1, X2, . . ., Xn),
and the probability distribution of X, what are the probability moments of Y? When g is a
simple function, the answer can be obtained by direct integration as follows (Ang and
Tang, 1984; 2007; Melchers, 1999; Ditlevsen and Madsen, 2007)
E[Y] = g( x) f X ( x) dx (3)
2 [Y] g( x) E[ y]
2
f X (x) dx (4)
and so on. For more complex functions these integrations can be evaluated by numerical
quadratures. The direct approach encounters major difficulties when g is complex or not
known explicitly. The direct approach also requires knowledge of the distribution of X. In
probabilistic approach, the uncertainties associated with model inputs are described by
probability distributions, and the objective is to estimate the output probability
distributions. This process comprises of two stages:
6
Probability Encoding of Inputs
This process involves the determination of probability distribution of the input parameters,
and incorporation of random variations due to both natural variability and errors. This is
accomplished by using either statistical estimation techniques or expert judgements.
Statistical estimation techniques involve estimating probability distributions from available
data or by collection of a large number of representative samples. In cases where limited
data are available, an expert judgement provides the information about the input probability
distribution. For example, a uniform distribution is selected if only a range of possible
values for an input is available, but no information about which values are more likely to
occur is available. Similarly, normal distribution is typically used to describe unbiased
measurement errors.
Propagation of Uncertainty through Models

In the presence of uncertainty, it is not simple to satisfy the basic design requirements. The
main steps involved in the uncertainty analysis of a model are: (a) estimation of
uncertainties in model inputs and parameters (characterisation of input uncertainties), (b)
estimation of the uncertainty in model outputs resulting from the uncertainty in model
parameters (uncertainty propagation), (c) characterisation of uncertainties associated with
different model structures and model formulations (characterisation of model uncertainty),
and (d) characterisation of the uncertainties in model predictions resulting from
uncertainties in the evaluation data.
In many engineering applications, the available data base is not broad enough to define
moments of Xi beyond the first and second moments, i.e. mean and variance. Since the
propagation of the first two moments is also a simpler task than the exact solution, the use
of the means and variances and their propagation through functional relationships have
received much attention. Methods related to this class of problems are called first-order
second moment methods. The common methods that are used by geotechnical engineers for
uncertainty analysis are: Monte Carlo simulation technique, Taylor series method, and
point estimate method.
The Monte Carlo simulation technique enables the determination as well as the estimation
of statistical moments. It involves the generation of random numbers and values of Y
associated with a set of random values of the basic random variables. After many values of
7
Y have been calculated in this way, the probability distribution of Y is generated and the
statistical parameters are calculated. The probability of failure can then be estimated from
the generated distribution.
The partial derivative or Taylor series expansion method is a first-order second moment
approximation given the mean and standard deviation of each of the basic random
variables, the mean and standard deviation of the function is then calculated. With the
appropriate distribution model for Y, a probability of failure is evaluated.
Point Estimate Method (PEM) is an approximate numerical integration approach. The PEM
can be considered as a variation of the partial derivative method and yet embodies a distinct
concept. The concept is the use of point estimates, i.e. point-by-point evaluations of the
functional Y = g(X) for particular values of X. Point estimates become the only means to
evaluate the functional relationship when g(X) is not known in closed form, e.g. an
algorithm or a computer code. If Y is the expression for FS, the expected value of variable
FS is found by adding several terms for example, four terms if the basic variables are only
two (cohesion c and friction angle ). These terms constitute the magnitudes of FS
calculated at values of the variables, one standard deviation on either side of the means; the
correlation between parameters is a multiplying factor to the terms. This method is very
convenient to use and has proved to be successful. It does not require the calculation of
derivatives of FS and can therefore be used in any geotechnical and especially slope
stability problems regardless of how complex the expression for FS is.
METHODS OF RELIABILITY ANALYSIS
One of the objectives of geotechnical engineering design is to ensure the safety and
performance of an engineering system for a given period of time and for specified loading
conditions and functions. Reliability assessment methods have received a great deal of
attention by researchers since the early days of geotechnical reliability. Since 1970, when
Wu and Kraft published their work on Safety analysis of slopes (Wu and Kraft, 1970),
many methods have been developed and suggested for reliability assessment of slopes
(Alonso, 1976; Tang et al., 1976; Vanmarcke, 1977; Whitman, 1984; Bergado and
Anderson, 1985; Li and Lumb, 1987; Chowdhury and Xu, 1993; Low et al., 1998;
Dodagoudar, 2001; Baecher and Christian, 2003; Chowdhury et al., 2010).
8
Duncan (2000, 2001) presented very useful guidance for adoption of probabilistic analysis
in geotechnical engineering practice. A spreadsheet approach has been suggested by Low
(1996). An approach for probabilistic analysis based on the finite element method (FEM)
has been outlined by Xu and Low (2006). Christian (2004) has drawn attention to
conceptual and practical issues concerning several geotechnical applications. Babu and
Murthy (2005) carried out reliability analysis of unsaturated soil slopes. The reliability of
braced excavations in clay against basal heave has been analyzed by Goh et al. (2008).
Probabilistic approach assumes a limiting equilibrium condition as with the deterministic
methods, but takes into account the statistical distribution of the uncertain input parameters.
Reliability calculations provide a means of evaluating the combined effects of
uncertainties, and a means of distinguishing between conditions where uncertainties are
particularly high or low (Duncan, 2000). In spite of the fact that it has potential value,
reliability theory has not been much used in routine geotechnical practice. One of the most
important applications of probability concepts in engineering reliability analysis is in the
evaluation of the safety and reliability of engineering systems, and the formulation of
associated design criteria. The reliability analysis consists of two main steps: (1)
identification and analysis of uncertainties of each of the contributing factors, and (2)
combining the uncertainties of the random factors to determine the overall reliability of the
system. The reliability of the system may be expressed as a probability of failure or
probability of safety. Two methods of probabilistic reliability analyses are considered in the
present study (Dodagoudar, 2001): (1) First-Order Second Moment (FOSM) Method, and
(2) Rosenblueths Point Estimate Method (RPEM).
First-Order Second Moment Method

The basic concept of second moment method is to set all random quantities in terms of
their two moments. These operational procedures were introduced for codification purposes
by Cornell (1969) within the context of the so-called first-order second moment (FOSM)
method. In FOSM method, the information on the distribution of random variables is
ignored. Joint probabilistic behaviour is expressed by the correlation coefficient; the
moments of higher order are not used. The FOSM method is also referred to as the Mean
value First-Order Second Moment (MFOSM) method in the literature. The FOSM method
estimates the uncertainty of the factor of safety of a slope against instability as a function of
the variances of the random input variables, such as soil friction angle, soil cohesion, soil
9
unit weight and pore pressure ratio, etc. It uses Taylor series expansion to estimate the local
uncertainty of the factor of safety.
Reliability Evaluation of Slope

For reliability evaluation of slopes, the functional Y is taken as the FS expression, i.e. FS =
g(X1, X2, . . ., Xn), in which X is an n-dimensional vector of random input parameters and
g( ) represents a functional relationship for the FS. In the context of the present study, g( )
is the factor of safety formulation using the Bishops simplified method of slope stability
analysis (Bishop, 1955). The random input vector X consists of the soil variables and pore
pressure ratio. Depending on the number of variables involved in the FS expression, the
mean and variance of the FS are calculated using the FOSM method. The uncertainty of the
safety factor (variance of FS), V[FS] depends not only on the uncertainty of individual
random input parameters as measured by V[Xi], but also on the associated sensitivity
coefficients (derivatives of the function with respect to random input parameters). In this
study, for numerical approximation of the partial derivatives, the Taylor series finite
difference (TSFD) estimation procedure is used to evaluate the variance of the FS for
reliability analysis of slopes.
Rosenblueths Point Estimate Method

Point estimate method (PEM) was presented by Rosenblueth (1975; 1981) and proposed by
Alfaro and Harr (1981) for the solution of geotechnical problems. The multivariate form of
Rosenblueths method for uncorrelated variables has been applied widely in geotechnical
reliability studies (Wolff, 1996; Dodagoudar, 2001). Rosenblueth introduced the PEM by
seeking expected value of functional Y, E[ Y r ] for real number r, given the first three
moments of X. The computation of E[ Y r ], however, must satisfy certain conditions, which
are specified in the following for a function of a single random variable, Y = g(X):
(1) E[ Y r ] should be expressed as linear combinations of powers of the point estimates:
g(x+), g(x), . . . where x+, x , . . ., etc. are specific values of X and g(x+), g(x), . . .
, etc. are evaluations of the function g(X) at these values of X.
(2) E[ Y r ] should be exact when Y = X, for r = 1, 2, 3.
The basic idea of Rosenblueths PEM (RPEM) is to approximate the original probability
density function (PDF) or probability mass function (PMF) of the random variable X by
assuming that the entire probability mass of X is concentrated at two points x and x+. The
method was initially limited to the analysis of three correlated variables. Its extension to
10
any number of correlated or independent random variables was given by Rosenblueth
(1981). For problems involving N random variables, they are located at the vertices of an
N-dimensional hypercube, away from the central or mean values, the two points for
each variable are computed and permutated to produce a total of 2N possible points of
evaluation in the parameter space. The rth moment of Y = g(X) = g(X1, X2, . . . , XN) about
the origin can be approximated as
E[ Y r ] = p (1,2 ,...,N ) y (r1,2 ,...,N ) (5)
in which the subscript i is a sign indicator that can only be + or for representing the
random variable Xi having the value of xi = i + i or xi = i i, respectively;
p( 1 , 2 ,..., N ) is determined by
N N 1 N ij
p ( 1, 2 ,..., N ) p i , i ( i ) ( j ) N (6)
i 1 i 1 j i 1 2
where ij is the correlation coefficient between random variables Xi and Xj. The number of
terms in the summation of Equation (5) is 2N which corresponds to the total number of
possible combinations of + and for all N random variables. For each term of the
summation in Equation (5), the model has to be evaluated once at the corresponding point
in the parameter space. Equation (5) has been used to evaluate the expected value and
standard deviation of the performance function Y. In the case of slope stability analysis, a
factor of safety expression is taken as the performance function. The expected value and
standard deviation of the FS are used to evaluate the reliability index or probability of
failure.
The reliability index obtained by the RPEM is independent of the type of distribution or
correlations among the basic variables which enables sensitivity studies with respect to
these parameters without much computational effort. The versatility of the RPEM is that, it
can be used even when the functional relationships are not given as an explicit equation
(Christian and Baecher, 1999). In the present study, the RPEM has also been used to
evaluate the reliability of finite earth slopes.
Reliability of Slopes
Reported failures of slopes having computed safety factors larger than unity indicate that
no matter what number is found by analytical means there is always some chance of failure
11
(Chowdhury, 1978; Chowdhury et al., 2010). The conventional deterministic approach is
concerned with the calculation of the magnitude of the FS under given conditions. Within a
probabilistic framework the FS under given conditions is regarded as a random variable
following some probability distribution. In this section, the usual model of the safety factor
will be examined to determine its reliability, or more particularly its complement, the
probability of failure.
The factor of safety FS is a function of several random variables and constants, i.e.
FS = f (c, , , u, H, , . . . . ) (7)
in which c is the cohesion, is the angle of internal friction, is the unit weight, u is the
porewater pressure, H is the slope height and is the inclination of slope.
Failure of a slope is defined as the event whereby FS receives a value smaller than or equal
to unity. Thus, if F denotes failure, one has
F = [FS 1] (8)
If f(FSi) represents the PDF of the factor of safety FSi along a potential failure surface Si,
then the probability of failure of a slope along Si is
1
P[FSi ] f ( FSi ) d ( FSi ) (9)

in which P[FSi] denotes the conditional probability of failure given that the latter occurs
along Si. If there are N possible failure surfaces, each with probability of occurrence P[Si],
i = 1, 2, . . ., N, then, from the total probability theorem one has that the probability of
failure pf of the slope is equal to
N
pf = P[F] = P[FSi ] P[Si ] (10)
i 1
N
with P[S ] 1i
i 1
Among the possible failure surfaces, the critical failure surface is one that corresponds to
the minimum value of the conventional factor of safety (FS = FSmin.). If the critical failure
surface is denoted by Sc and the probability of its occurrence by P[Sc], then Equation (10)
may be written as (A-Grivas, 1981)
12
N 1
pf = P[FSc] P[Sc] + P[FS ] P[S ]
i i (11)
i 1
The system probability of failure of the slope will be determined using Equation (11). This
probability of failure has traditionally been seen as the next evolutionary stage beyond the
factor of safety. If the pf is large, say > 10-3, as it is for most slopes with critical stability,
the probability distribution is not too important, because the pf is not too sensitive to
variations in the form of the PDF. However, for very small risk, << 10-3, the pf can be very
different from one distribution to other (Ang, 1972). The next step is the calculation of the
system probability of success ps more commonly called as the system reliability. The
system reliability of the slope is given by
ps 1 p f (12)
It should be noted that pf and ps are functions of the reliability index and of the probability
distribution of the factor of safety.
Reliability Index and Failure Probability

Uncertainty modelling and analysis in geotechnical engineering started with the
employment of safety factors using deterministic analysis. Then it was followed by
probabilistic analysis with reliability based safety factors. The conventional definition of
factor of safety FS = C/D where C is the capacity (resistance) and D is the demand
(loading), and this is only a function of the variability in so far as C and D are chosen by
the engineer to represent the conservative values. The ratio of mean values is the Central
Factor of Safety (CFS) and it generally exceeds the conventional factor of safety. If the
maximum demand (Dmax) exceeds the minimum capacity (Cmin), the distributions
overlap, and there is nonzero probability of failure.
The difference between the two random variables (i.e. capacity and demand functions) C
and D is the safety margin S and failure will occur if the safety margin is less than zero.
The safety margin is expressed as (Harr, 1979)
S=CD (13)
The safety margin is itself a random variable and hence the probability of failure pf is the
probability that the safety margin will be less than zero. Mathematically,
13

pf = P [ C < D ], i.e., pf = PC ( C ) p D ( D ) dD (14)

where Pc(C) is the cumulative distribution function of the capacity C, and pD(D) is the
probability density function of the demand D.
Many different terms are used to describe the reliability of an engineering system. The
commonly used term probability of failure is always associated with a particular
performance criterion. For practical structures and performance criteria, it is difficult to
compute the probability of failure precisely. Therefore, a first-order estimate is frequently
used in probabilistic design specifications (Leporati, 1979). This first-order estimate
employs a measure known as the reliability index or safety index .
Reliability index
In most applications of probability theory to the analysis of slope stability, the uncertainties
have been stated in the form of reliability index. The number of standard deviations that the
mean value of the safety margin ( S ) is beyond S = 0 is called the reliability index :
S
(15)
[S]
A number of approaches have been proposed to calculate , including the invariant solution
by Hasofer and Lind (1974) and the simpler FOSM method (Madsen et al., 1986). The
latter approach can adopt different performance functions and limit states, including FS 1
= 0, ln FS = 0, etc. Taking the performance function and limit state as FS 1 = 0, the
reliability index based on the FOSM method is
E[FS] 1
(16)
[FS]
where E[FS] is the expected value of the FS and [FS] is the standard deviation of the FS.
The FOSM method does not require the probability distribution of FS to be assumed. In
some cases, a common approach is to assume that FS is lognormally distributed. In this
case the reliability index is given by
E[ln FS]
(17)
ln [ FS]
where E[ln FS] and ln [FS] are given by (Benjamin and Cornell 1970)
14
E[ln FS] = ln (E[FS]) ln [ FS]
2
(18)
2
and
[ FS] 2
ln [FS] = ln 1
(19)
E[ FS]
The variability of [FS] is also valid when adopting the lognormal distribution of FS, since
ln[FS] is directly related to the variability in [FS]. The Bishops simplified method is
used in the present work as a basis for evaluating the above two statistical moments of the
safety factor in the FOSM method and RPEM.
According to Equation (16 or 17), the reliability index of a slope depends not only on a
measure of the central tendency of FS but also on a measure of its dispersion. It can also be
considered as a way of normalising the factor of safety with respect to its standard
deviation. Therefore, whether or not the probability distribution of FS is known, the
reliability index expresses uncertainty in the stability of the slope (Christian et al., 1994).
For example, consider two cases, in both of which FS is assumed to be normally distributed
about the mean value. It is clear from Equation (16) that the case with a E[FS] of 1.5 and
[FS] of 0.5 has a higher overall probability of failure than does the case with a E[FS] of
1.2 and a smaller [FS] of 0.1. The rational approach is to develop a design to achieve a
consistent reliability index rather than relying on a value of FS alone.
Failure probability
Uncertainties in basic parameters are quantified and uncertainty associated with FS is
calculated, which is then incorporated in the reliability index defined above. The definition
does not require that the probability distribution of FS be known, but there is a tacit
assumption that FS is distributed in some way which can be described meaningfully by an
expected value and standard deviation (Christian et al., 1994). When the shape of the
probability distribution of the factor of safety is known, the reliability index can be related
to the probability of failure.
It is known that the sum or difference of two normal variates is also a normal variate. If the
FS is normally distributed, the probability of failure pf and reliability index are related by
pf = 1.0 () = () (20)
where () is the cumulative distribution function (CDF) of the normal distribution.

15
The reliability index represents an alternative means of presenting pf on a more convenient
scale as shown in Table 1. The reliability indexes for most geotechnical engineering
problems lie between 1 and 4, corresponding to probability of failure ranging from about
15.87 percent to 0.0032 percent. Note that pf decreases as increases, but the change is not
linear. A proper understanding of these two terms and their interrelationship is essential
because they play a fundamental role in geotechnical reliability analysis. The probability of
failure of a slope for static slide corresponds to a FS, which has some uncertainty
associated with it. The application of reliability analysis to slopes suggests that reliability
index offers a rational alternative to the deterministic factor of safety approach. Published
results describing the application of reliability theory indicate for a heterogeneous slope
that has not been placed with engineering control, standard deviation [FS] is at least 0.25
(Christian et al., 1994; Wolff, 1996; Dodagoudar, 2001; Chowdhury et al., 2010 ).
Table 1 Reliability Index (as normal variate) and Probability of Failure
Reliability index Probability of failure

[ ] [pf = (-)]
1.0 0.1587
1.5 0.06681
2.0 0.02275
2.5 0.6210 x 10-2
3.0 0.1350 x 10-2
3.5 0.2326 x 10-3
4.0 0.3162 x 10-4
5.0 0.2859 x 10-6
RELIABILITY ANALYSIS: FINITE EARTH SLOPES
Both the FOSM method and RPEM are used to evaluate the reliability of safety of finite
earth slopes. Bishops simplified method (Bishop, 1955) is used to evaluate the factors of
safety needed in the FOSM method and RPEM. Reliability indexes and failure probabilities
are evaluated assuming FS to be a normally distributed random variable. One numerical
example and one case study of slope failure are used to demonstrate the applicability of the
16
probability theory in reliability analysis of finite earth slopes. A methodology for
probabilistic reliability analysis of slope is given in Figure 1.
Numerical Example: Homogeneous Slope: A slope section shown in Figure 2 is

considered for the reliability analysis. The given data are: Height of slope (H) = 8 m,
Inclination () = 45 degrees and the hard rock surface is 5 m below the bottom of the slope.
The other data are given in Table 2. The coefficients of variation (COVs) for c, and are
taken from published literature (Harr, 1995). The mean value of pore pressure ratio ru is
taken as 0.35 with a coefficient of variation, COV(ru) = 0.50.
Reliability analysis: Results
The reliability analysis of the above numerical example problem has been carried out using
both the FOSM method and RPEM. Results obtained using both the methods are presented
below. The Bishops simplified method is used to determine the FS values needed to be
used in both these methods. A central factor of safety (CFS) of the slope using mean values
of the uncertain input parameters is 1.252.
The geometric parameters corresponding to the slip circle (radius and coordinates of
centre) for the mean values of the variables are: R = 10.936 m, X = 4.40 m, and Y =
16.00 m.
When ru = 0.0, the CFS = 1.710, R = 13.046 m, X = 4.00 m and Y = 18.00 m. This
shows how the FS is sensitive to the changes in the pore pressure ratio.
The variation in unit weight is small (i.e., COV() = 3.0 %), so its contribution to the E[FS]
and standard deviation of FS is neglected. Soil strength parameters c and and pore
pressure ratio ru are considered as input random variables. The expected values and
standard deviations of the FS are evaluated by using both the FOSM method and RPEM.
The E[FS] obtained by the FOSM method with first-order approximation is the CFS itself.
As the FOSM method uses first-order approximations, the E[FS] values are same for
correlated as well as for uncorrelated random variables and only the standard deviations
vary.
17
Y
8m
Soil
c, and
0
45
5m
5m
X
Rock surface
Fig. 2 Slope Section Used for the Numerical Example
Table 2 Soil Data Used for the Numerical Example
Soil parameter Mean value COV (%)

Cohesion, c (kN/m2) 18.0 22.22
Friction angle, (degrees) 30.0 10.0
Unit weight, (kN/m3) 19.5 3.0
Pore pressure ratio, ru 0.35 50.0
18
Specification of the slope
geometry Other non-random
parameters
Specification of the probability distribution of the

uncertain soil properties and pore pressure ratio
Determination of critical slip surface and its

associated FS using limit equilibrium method
Analysis using FOSM method Analysis using RPEM
Calculation of the partial derivative Evaluate the FS expression for all

of FS with respect to each of the combinations of the N uncertain
uncertain input parameters input parameters, i.e. 2N times
Calculation of the expected value Calculation of the expected value

and variance of FS using FOSM and variance of FS using RPEM
method
Computation of reliability index and

probability of failure
Fig. 1 Methodology for Probabilistic Reliability Analysis
The results of the reliability analysis, using Taylor series expansion with second-order
mean and first-order variance, are given in Table 3. Different correlation coefficients are
considered between the strength parameters c and , the results of which are presented in
Table 3. Using the values of E[FS] and [FS] the overall reliability indexes are evaluated
using the FOSM method. Equation (16) is used to estimate the reliability index and is
taken as an alternative risk measure to the usual factor of safety. The parameter measures
the difference between best estimate of the FS from its failure value (i.e. FS = 1.0) in units
of [FS]. This measure is more complete than FS alone and allows more consistency to be
brought to the selection of a design or target FS. It can be seen from the results that both
19
the E[FS] and [FS] increase as the correlation coefficient becomes positive and the values
of decrease. The results of the reliability analysis using the RPEM are given in Table 4.
The [FS] is considered as the parameter representing uncertainty in the FS. As can be
seen from the results of both the methods there is not much difference in the values of
[FS]. Therefore, it is decided to use the RPEM for further study on the reliability analysis
of slopes. In the following section, a parametric sensitivity-reliability analysis of the
example slope problem is carried out using the RPEM.
Table 3 Results of the Reliability Analysis by Taylor Series Expansion of FS (with

second-order mean and first-order variance)
Correlation between variables, c and

Values of (c, ) = 0.50 (c, ) = 0.0 (c, ) = 0.25
E[FS] 1.215 1.217 1.218
[FS] 0.281 0.298 0.306
Overall
reliability index () 0.765 0.728 0.712
pf (FS as a normal
variate) pf = 0.2221 pf = 0.23325 pf = 0.2381
FS 1.0
Table 4 Results of the Reliability Analysis by RPEM
Correlation between variables, c and

Values of (c, ) = 0.50 (c, ) = 0.0 (c, ) = 0.25
E[FS] 1.235 1.238 1.240
[FS] 0.282 0.301 0.309
Overall
reliability index () 0.832 0.792 0.775
pf (FS as a normal
variate) pf = 0.20233 pf = 0.21456 pf = 0.21867
FS 1.0
20
Parametric study
Parametric sensitivity analysis is an important aspect of modern reliability analysis. The
main aim of carrying out reliability analysis is calculating the sensitivity of the estimated
failure probability to changes in distribution parameters. These sensitivities are useful
because they quantify the importance of distribution parameters such as means, standard
deviations and correlations. This quantification helps us in assessing the validity of the
reliability estimates (in light of the assumptions made in the choice of the parameter
values) and helps us define the roles of the random variables in subsequent analyses (e.g.
an unimportant random variable can be treated as a constant). Since the differences in the
values of standard deviation are very small for both the FOSM method and RPEM, in this
section a parametric study has been carried out using RPEM only.
The cohesion c, angle of internal friction and pore pressure ratio ru are considered as the
only random input variables in the analysis. The parametric sensitivity analysis has carried
out by considering the variation in cohesion only. The total number of COV(c) considered
for the analysis is seven. A single coefficient of variation is used for the angle of internal
friction and pore pressure ratio. Thus, COV(c) = 0.10, 0.15, 0.20, 0.25, 0.30, 0.35 and 0.40,
COV() = 0.10 and COV(ru) = 0.50. Four cases of mean values of ru are considered:
Case 1: E[c] = 18 kN/m2, E[] = 300 and E[ru] = 0.10;

Case 2: E[c] = 18 kN/m2, E[] = 300 and E[ru] = 0.20;
Case 3: E[c] = 18 kN/m2, E[] = 300 and E[ru] = 0.30; and
Case 4: E[c] = 18 kN/m2, E[] = 300 and E[ru] = 0.35.
The variation in the values of E[FS] with COV(c) for each of the cases is given in Tables 5
8 for uncorrelated as well as for correlated strength parameters c and . The correlation
coefficient between c and is taken as 0.50 and +0.25. The mean values of factors of
safety (i.e., CFS values) corresponding to these cases are: 1.581, 1.444, 1.317 and 1.252
respectively. The plots of variation of standard deviation of FS with COV(c) and variation
of overall failure probability with COV(c) for the above four cases are presented in Figures
3 10.
It is seen from Table 5 that, as the correlation between strength parameters is more towards
positive, a slight increase in the E[FS] is observed. The same trend can also be seen from
Tables 6 8. It is also seen from the tables that the E[FS] is slightly decreased as the value
21
of COV(c) is increased. For the same value of COV(c), with the increase in the value of
E[ru], the decrease in the value of E[FS] is more. This shows how the factor of safety of a
slope is sensitive to the changes in the values of pore pressure ratio.
Table 5 Sensitivity of the Estimated Expected Factor of Safety with COV(c) [Case 1:
E[c] = 18 kN/m2, E[] = 300, E[ru] = 0.10]
Expected value of FS (E[FS])

COV(c) (c, ) = 0.50 (c, ) = 0.0 (c, ) = +0.25
0.10 1.580 1.581 1.581
0.15 1.575 1.577 1.579
0.20 1.573 1.577 1.579
0.25 1.571 1.576 1.578
0.30 1.569 1.574 1.576
0.35 1.567 1.572 1.574
0.40 1.562 1.567 1.569
E[c] = 18 kN/m2, E[] = 300, E[ru] = 0.20]

COV(c) (c, ) = 0.50 (c, ) = 0.0 (c, ) = +0.25
0.10 1.444 1.446 1.447

0.15 1.442 1.444 1.446
0.20 1.440 1.444 1.445
0.25 1.436 1.439 1.441
0.30 1.433 1.438 1.440
0.35 1.431 1.435 1.437
0.40 1.426 1.430 1.432
22
E[c] = 18 kN/m2, E[] = 300, E[ru] = 0.30]

COV(c) (c, ) = 0.50 (c, ) = 0.0 (c, ) = +0.25
0.10 1.308 1.310 1.311

0.15 1.305 1.308 1.309
0.20 1.303 1.306 1.308
0.25 1.302 1.305 1.307
0.30 1.299 1.303 1.305
0.35 1.297 1.300 1.302
0.40 1.295 1.299 1.300
E[c] = 18 kN/m2, E[] = 300, E[ru] = 0.35]

COV(c) (c, ) = 0.50 (c, ) = 0.0 (c, ) = +0.25
0.10 1.239 1.240 1.241
0.15 1.238 1.239 1.241
0.20 1.237 1.240 1.240
0.25 1.234 1.238 1.240
0.30 1.234 1.235 1.237
0.35 1.232 1.234 1.235
0.40 1.231 1.232 1.232
It is seen from Figures 3, 5, 7 and 9 that the variation of [FS] with COV(c) is almost
linear. The [FS] increases as the value of COV(c) increases. As the correlation between
strength parameters is more towards positive there is more increase in the value of [FS].
As the value of COV(c) increases from 0.15 to 0.40, much increase in the value of [FS] is
observed. Because of this, one can observe much higher values of overall failure
probabilities associated with the slope (Figures 4, 6, 8 and 10). The relation between
overall failure probability and COV(c) is somewhat complex. Therefore, for smaller values
23
of the failure probability, a log-scale is used along y-axis whereas for greater values, a
linear scale is used. It can be seen from the above figures that the values of failure
probabilities are increased as the correlation is changing from negative to positive.
It can be concluded that the positive correlation between the strength parameters is
responsible for the increase in the values of failure probabilities. For the probabilistic
design of slopes, one has to consider the positive correlation between the strength
parameters as well as the possible variations in their values. The probabilistic design
criteria for static cases have to be based on the target probability of failure with a minimum
reliability index of 1.4 on any of the potential failure surfaces, with the specified minimum
FS of 1.35.
In the next section, a case study of slope failure reported in the literature is considered to
validate the methods of probabilistic reliability analysis. Both the FOSM method and
RPEM are used for the reliability evaluation.
0.40
(c',') = 0.25
(c',') = 0.0
0.35
Standard Deviation of FS, [FS]
(c',') = -0.5
0.30
0.25
0.20
0.15
0.10
0.10 0.15 0.20 0.25 0.30 0.35 0.40

Fig. 3 Variation of [FS] with COV(c): Case 1: E[c] = 18 kPa, E[] = 300, E[ru] =
COV(c')
0.10.
24
(c',') = 0.25
(c',') = 0.0
(c',') =-0.5
Overall Failure Probability, pf 0.01
1E-3
1E-4
0.10 0.15 0.20 0.25 0.30 0.35 0.40

COV(c')
Fig. 4 Variation of Overall pf with COV(c): Case 1: E[c] = 18 kPa, E[] = 300, E[ru] =
0.10.
0.40
(c',') = 0.25
(c',') = 0.0
0.35
(c',') = -0.5
0.30
0.25
0.20
0.15
0.10 0.15 0.20 0.25 0.30 0.35 0.40

COV(c')
0.20
25
0.1 (c',') = 0.25
(c',') = 0.0
Overall Failure Probability, p f (c',') = -0.5
0.01
1E-3
0.10
Fig. 6 Variation of Overall 0.15
pf with 0.20
COV(c ): Case
0.25
] = 18
0.30
2: E[c 0.35
] = 300, E[ru] =
0.40
kPa, E[
COV(c')
0.20
0.40 (c',') = 0.25

(c',') = 0.0
0.38
(c',') =-0.50
0.36
0.34
0.32
0.30
0.28
0.26
0.24
0.22
0.20
0.10 0.15 0.20 0.25 0.30 0.35 0.40
COV(c')
0.30.
26
0.24
0.22
(c',') = 0.25
0.2 (c',') = 0.0
0.18 (c',') = -0.5
Ovearall Failure Probability, p f
0.16
0.14
0.12
0.1
0.08
0.06
0.10 0.15 0.20 0.25 0.30 0.35 0.40
COV(c')
Fig. 8 Variation of Overall pf with COV(c): Case 3: E[c] = 18 kPa, E[] = 300, E[ru] =
0.30.
0.42
(c',') = 0.25
0.40 (c',') = 0.0
(c',') = -0.5
0.38
0.36
0.34
0.32
0.30
0.28
0.26
0.24
0.10 0.15 0.20 0.25 0.30 0.35 0.40

COV(c')
0.35
27
0.30
(c',') = 0.25
0.28 (c',') = 0.0
(c',') = -0.5
Overall Failure Probability, p f
0.26
0.24
0.22
0.20
0.18
0.16
0.10 0.15 0.20 0.25 0.30 0.35 0.40
): Case 4: E[c] = 18 kPa, E[] = 300, E[ru]
Fig. 10 Variation of Overall pf with COV(cCOV(c')
= 0.35
CASE STUDY: SELSET LANDSLIDE
A case study of the Selset landslide reported in Skempton and Brown (1961) has been
considered. The Selset landslide was a rotational slide within a deposit of heavily
overconsolidated boulder clay. The clay was remarkably uniform, without fissures or joints
and the long-term conditions existed. The landslide took place in the south slope of the
River Lune Valley. As the slip was within a deposit of nonfissured overconsolidated
boulder clay, no significant variation of mean soil properties was observed within a depth
of 18.3 m. The soil profile could therefore be modelled as a statistically homogeneous
medium. An average cross section through the slope was given by Skempton and Brown
(1961), with an average height of 12.8 m and an average inclination of 280. The failure
width was about 55 m and the exact location of the slip surface was not determined.
However, Skempton and Brown (1961) examined various possible slip circles and the most
critical slip surface was identified. A mean value of 0.45 was suggested by the authors as a
suitable value for the pore pressure ratio ru. A judgemental value of 10 % is used here for
the COV of ru. A summary of test results on the soil is given in Skempton and Brown
28
(1961), from which the following input data parameters are derived and are presented in
Table 9.
Table 9 Soil Parameters for Selset Landslide
Parameter Mean value COV (%)

c 8.6 kN/m2 30.0
32.00 7.0
21.8 kN/m3 0.70
ru 0.45 10.0
Results
A two-dimensional probabilistic analysis of the Selset landslide has been carried out here.
The CFS of the slope is 0.958 with geometrical parameters of the slip circle: X = 8.050 m,
Y = 33.250 m and R = 27.776 m for the mean values as given above. The CFS = 1.723
when ru = 0.0, but the critical slip surface was almost the same. The FOSM method and
RPEM are used to carry out the probabilistic stability analysis. The uncertain parameters c,
and ru are considered as random input variables. The variability of is small (i.e.
COV() = 0.7 %), so its variance contribution to FS is neglected. The results of the analysis
are presented in Table 10.
As can be seen from the table, the E[FS] and [FS] are almost equal for both the FOSM
method and RPEM. Little is known about the statistical properties, especially the
correlation structure of the strength parameters as well as porewater pressure. Therefore,
here the variables are considered as uncorrelated and the probability of failure of a slope is
evaluated considering the reliability index as a normal variable. The difference in failure
probability values is marginal for both the methods. For the actual slope height of 12.8 m,
the failure probability of the slope is high (> 0.5). This is to be expected as the failure of the
slope had indeed occurred. This can also be seen from the values of CFS and E[FS].
Whenever the probability of failure is more than 0.5, the reliability index is less than zero.
For practical purposes, in any case, the value of should not be less than 1.4 with the
corresponding static factor of safety of 1.35 at least on any of the critical slip surfaces.
29
Table 10 Results of the Reliability Analysis: Selset Landslide
Values of FOSM method RPEM

E[FS] 0.958 0.957
[FS] 0.131 0.132
Overall
reliability index () 0.321 0.328
pf (FS as a normal pf = 0.62575 pf = 0.62770

variate)
FS 1.0
Discussion
The uncertainties in the random input variables have been incorporated into the
probabilistic reliability analysis in order to arrive at quantitative information concerning the
reliability index and probability of failure for the finite earth slopes. The Bishops
simplified method is used as a basis for evaluating statistical moments of the safety factor
in both the FOSM method and RPEM. The other methods of stability analysis can also be
used to get the statistical moments. The methods of reliability analysis are independent of
the slope stability model. These reliability methods can also be used along with the finite
element method. The difference in the values of standard deviation, which is used as the
uncertainty in the FS, estimated by both the FOSM method and RPEM is very small. The
RPEM has been used to carry out the sensitivity analysis for the numerical example. The
sensitivity analysis for reliability of slopes is carried out with respect to COV(c). A case
study of Selset landslide has been used to demonstrate the applicability of the probabilistic
reliability analysis. The results obtained by both the FOSM method and RPEM are in good
agreement. As the RPEM does not involve the evaluation of derivatives, it can be used in
practice for the reliability-based analysis and design of earth slopes. Based on the analysis
presented in this study, some design guidelines are evolved for the safety of slopes in the
form of minimum reliability index, expected minimum FS and standard deviation of FS.
30
PARMACHI LANDSLIDE, WESTERN GHATS, INDIA: RISK ASSESSMENT
One of the forms of natural disasters given much attention by researchers all over the world
during the International Decade for Natural Disaster Reduction (1990-2000) is the
occurrence of landslides. In hilly terrains of India, including Himalayan mountains and
Western Ghats, landslides have been a major and widely spread natural disaster that strike
life and property almost perennially and occupy a position of major concern. These
landslides, year after year, bring about untold misery to human settlements apart from
causing devastating damages to transportation and communication network.
This part of the paper describes: (i) rainfall-induced landslide evaluation using Digital
Elevation Model (DEM) and Geographic Information System (GIS), (ii) fractal modelling
of mass movement including critical failure surface determination using shortest path
concept of GIS and (iii) evaluation of risk due to landslide. The development of GIS which
has revolutionized the way in which a variety of spatial data can be handled, checked,
analyzed, manipulated, synthesized and updated. The GIS has proved to be a powerful tool
for working with large bodies of data, deriving suitable spatial and temporal parameters
and performance indicators and for displaying information in a suitable form such as
maps. Moreover, GIS-based maps can be updated, altered and enhanced with the
availability of additional data or better quality data or as a consequence of improved
knowledge about any aspect of a project. The methods used here demonstrate the
usefulness of the GIS based technique in identifying potential slide regions in a long
stretch of a hill slope and the corresponding risk, which could serve as supports for
landslide hazard management.
Description of the Study Area
A large landslide occurred at Parmachi (Lat. 18 8' N and Long. 73 36' E) in the Western
Ghats on June 28, 1994 resulting in the death of about 15 persons and destruction of a few
houses. In the study area the detailed geological and geotechnical investigations were
carried out. Geological investigations have provided important information on the slide
geometry and structure. The lithological map shows that the outcrops in the area composed
of the following: i) soils in the study area are residual soils, and ii) the soil cover varies
from 0.5 metres to about even 10 metres. Prior to the occurrence of the landslide there was
a heavy downpour in the area for three days touching a value of 240 mm/day (0.01m/hr)
during the twenty-four hours just prior to the landslide. The landslide involved movement
31
of about 30,000 m3 of soil and rock over a distance of approximately 300 m. The elevations
at the location vary from about 240 m to 500 m. The slopes vary from 400 to 600. The
stability evaluation of Parmachi landslide (Dodagoudar, 2001) is considered for the
rainfall-induced slope failure analysis. Based on this study, the criterion for threshold
rainfall intensity-duration is suggested for the Parmachi region of Western Ghats, India.
Rain-Induced Landslide
Landslides seldom occur abruptly and it is doubtful if shear failure occurs simultaneously
along the entire slip surface except perhaps in very small slides. Moreover, many large
slides are a succession of small individual slides. First a small slide occurs and as a
consequence of the movement of this mass, other adjacent masses become unstable. This
process may then be repeated so that landslide grows during its movement (Chowdhury et
al., 2010). Major contributing factors include the growth of urbanization, lack of adequate
planning for development of hilly areas and marginal lands and inadequate management of
hazards. There is an increasing recognition of the need for assessment of landslide
susceptibility and hazard and for landslide risk management.
Landslides have been frequently found to occur following periods of heavy rainfall which
tends to raise the groundwater level. This results in an increased weight of a potentially
sliding mass, an increase in the porewater pressure and a consequent reduction in the shear
strength on a potential sliding surface. Occurrence of landslides along long stretches of hill
roads is a common sight during heavy monsoon. This paper attempts to analyse the
stability of a long stretch of a hill slope using DEM and GIS concepts. The case study of a
landslide, which occurred on June 28, 1994 in the village of Parmachi in the Western
Ghats is considered. The landslide had occurred as a result of a heavy downpour of about
240 mm/day (0.1 m/hr) of rainfall. The methodology adopted here involves building a
DEM first. For this, the study area of about 500 m x 400 m extent has been divided into
grids of 16 m x 16 m size. The hill where the landslide occurred has elevations varying
from about 240 m to 500 m. The slopes vary from 400 to 600. The soil is residual and the
cover varies from 0.5 m to as much as 10 m.
Relatively simple methods, such as infinite slope analyses have been used successfully
for regional landslide hazard studies within the GIS framework. Stability analysis has
been carried out for different intensities and durations of rainfall (Dodagoudar, 2001)
using the infinite slope method. This method has been used since it allows superposition of
32
thematic information, especially for subsequent risk evaluation. A factor of safety value is
obtained for each grid using Equation (21):
FS = (21)
where cd = effective soil cohesion (kN/m2), sat = saturated unit weight of the soil (kN/m3),
1 = unit weight of the soil (kN/m3), H = thickness of soil cover (m), h = groundwater table
height (m), z = depth of the failure surface from the top (m), d = angle of friction (degrees)
and = slope inclination (degrees).
One failure along a slope may be followed by another although the time interval between
such successive failures can vary greatly. Moreover, one or both of two successive failures
may involve either limited movement or catastrophic failure. In the latter case, whole or
part of the volume of material involved in the slope failure may move away from the
exposed slope face. A fluctuation of the groundwater table and the effect of vegetation in
the form of root cohesion ct and tree weight T are incorporated in the stability analysis.
The thematic information required for this is obtained from different sources. Remote
sensing products are useful in obtaining information about vegetation and landuse and the
variable soil conditions. The result (Figure 11) clearly shows that a rainfall of 240 mm/day
(0.01 m/hr) leads to a large patch of contiguous grids failing, indicating failure of the
slope, i.e. landslide.
Runout and Deposition of the Landslide Mass
The runout of a typical debris flow during a landslide depends on momentum transfer. In a
landslide the sliding mass continuously decreases as deposition progresses. Here a fractal
deposition model is evolved and used.
Deposition Pattern with Distance
The deposited mass Md and distance travelled along the paths have a definite relationship.
Two non-dimensional terms are calculated: (i) the term Mp/Mi, where Mi is the initial
displaced mass and Mp is mass participating in movement (= 1 - Md) at a location at
distance s and (ii) the term (sf - s)/sf. A power law relationship for deposition pattern is
derived from known deposition profiles of a number of landslides (Quadri, 2000).
In order to compute the Mi, the critical slip surface is located by the concept of shortest
path, which is widely used in GIS network problems. The critical slip surface will be
33
analogous to the shortest path', i.e. the path with the least weighted distance. The least
distance, in this case, would mean the least factor of safety.
Fig. 11 Simulation of Parmachi Landslide
Minimization of the well-known expression for the overall factor of safety (Chowdhury et
al., 2010) gives the critical slip surface:
FS = (22)
where FL is the local factor of safety and is given by the term inside the brackets, and
are the normal and shear stresses prevalent along a plane with as the inclination with the
horizontal; c, and u are the cohesion, angle of shearing resistance and pore pressure; b is
the width of slices; and L is the total length of the slip surface. The term inside the
summation sign in Equation (22) is a weighted sum of local factors of safety, the
weightage for each slice being the base length, b sec, of the slice concerned.
The following relations are now derived in terms of the fractal dimension (D) of the terrain
from the fundamental principles of mechanics [Quadri, 2000]:
34
For runout distance:
(23)
For depositional profile:
(24)
A case study of the Madison Canyon Landslide, which occurred on August 17, 1959 in the
mountain terrain of south western Montana on Madison river in North America is used for
validation. The displaced mass moved down about 2200 m and travelled about 600 m,
crossed the river and rode up the opposite bank. The computed deposition profile is
compared with the observed ones in Table 11. After having validated the fractal model,
the deposition profile of the Parmachi landslide is computed and presented in Table 12.
This agrees well with the observed values. The volume of displaced soil and rock was
approximately 30,000 m3 and the run-out was about 300 m.
Table 11 Deposition Profile
35
Table 12 Deposition Profile for Parmachi Landslide
Landslide Risk Evaluation and Assessment
More formal applications of risk assessment and management principles, in a qualitative

manner, have been practiced for landslide hazard zoning for urban planning and highway
slope management since the 1970s. In the 1980s, and particularly in the 1990s, these
have been extended to quantitative methods, and to management of individual slopes,
pipeline routes, and more global slope risk management. In the present case, instability
due to excessive rainfall is considered. A storm of June 28, 1994, which caused more
extensive damage is considered for the risk evaluation. Probabilistic method is used for
computing the factor of safety so as to account for the variability in the soil parameters
involved in the formulation and for the different rainfall intensities and duration. The
resulting mass movement is also calculated. A common definition of risk is:
Risk = Probability of occurrence of the catastrophic event (i.e., critical rainfall) x
Probability of failure of slope Consequences of failure.
It is considered that risk assessment methods for landslides and slopes have been
developed to a level that they are applicable in practical terms and form a useful tool to
complement engineering judgement. The level of analysis possible will vary from project
to project and may increase as further data becomes available. Readers are encouraged to
use these terms so that there is consistency across the international community (Fell and
Hartford, 1997).
Annual Exceedance Probability (AEP): The estimated probability that an event of
specified magnitude will be exceeded in any year.
Hazard: Probability that a particular danger (threat) occurs within a given period of time.
36
Vulnerability: The degree of loss to a given element or set of elements within the area
affected by a hazard.
Landslide Susceptibility: The classification, and volume (or area) of landslides which
exist or potentially may occur in an area or may travel or retrogress onto it. It may also
include a description of the velocity and intensity of the existing or potential landsliding.
Elements at risk: Population, buildings and engineering works, infrastructure,
environmental features and economic activities in the area affected by a hazard.
Individual risk to life: The increment of risk imposed on a particular individual by the
existence of a hazard. This increment of risk is an addition to the background risk to life,
which the person would live with on a daily basis if the facility did not exist.
Risk: Measure of the probability and severity of an adverse effect to life, health, property,
or the environment. Quantitatively, Risk = Hazard Potential worth of loss Or Risk =
Probability of an adverse event times the consequences if the event occurs.
Quantitative risk analysis: An analysis based on numerical values of the probability,
vulnerability and consequences, and resulting in a numerical value of the risk.
Risk assessment: The process of making a decision recommendation on whether existing
risks are tolerable and present risk control measures are adequate, and if not, whether
alternative risk control measures are justified or will be implemented. Risk assessment
incorporates the risk analysis and risk evaluation phases.
Risk control: The implementation and enforcement of actions to control risk, and the
periodic re-evaluation of the effectiveness of these actions.
Risk evaluation: The stage at which values and judgement enter the decision process,
explicitly or implicitly, by including consideration of the importance of the estimated risks
and the associated social, environmental, and economic consequences, in order to identify
a range of alternatives for managing the risks.
Risk management: The systematic application of management policies, procedures and
practices to the tasks of identifying, analysing, assessing, mitigating and monitoring risk.
Risk mitigation: A selective application of appropriate techniques and management
principles to reduce either likelihood of an occurrence or its adverse consequences, or
both.
Figure 12 shows the flowchart of the overall landslide risk management strategy adopted by
the geotechnical engineers. Risk analysis includes hazard analysis and consequence
analyses. Consequence analysis includes identifying and quantifying the elements at risk
37
(property, persons), their temporal spatial probability, their vulnerability either as
conditional probability of damage to conditional probability of damage to property, or
conditional probability of loss of life or injury. Risk assessment takes the output from risk
analysis and assesses these against values judgements, and risk acceptance criteria. In this
study, for the present case, the probability of failure is calculated based on the consideration
that the critical surface corresponds to the minimum value of the conventional factor of
safety. If the latter is denoted by Sc (i.e. Sc is the critical surface for which FS is minimum)
and the probability of its occurrence by P[Sc], then the probability of failure pf is written as
N 1
pf = P[FSc] P[Sc] + P[FS ] P[S ]
i i (25)
i 1
Here, the expected value and standard deviation of FS, which are required for the
computation of pf are obtained using the Rosenblueth's Point Estimate Method. The
computations can be further improved if fuzzy uncertainties are also considered in the
analysis [Dodagoudar, 2001].
The results obtained for constant rainfall intensities and varying duration have demonstrated
how gradually the slope reaches failure state when duration of rainfall increases. Based on
the results of infinite slope analysis, the critical rainfall is taken as 240 mm/day (0.01 m/hr)
for 24 hours for the Parmachi area of Western Ghats. However, it was seen that increase in
duration is less critical than increase in intensity of rainfall. For the critical slope section,
mass movement analysis using a momentum transfer law shows a run out distance of 300 m
and a duration of 18 s. The computed thickness of deposit was 0.89 m. When superimposed
over the digital terrain model (DTM), this indicates that the human settlement in the area lies
within the run out distance.
Table 13 shows the probabilities of all likely events for a critical rainfall of 240 mm/day
(0.01 m/hr).The alternative with the least expected cost is chosen if the expected value is
the criterion for decision. Thus the evaluation of risk could help in deciding and planning
suitable remedial or preventive measures. However, the recent developments include the
use of computer intelligence and knowledge-based methods which take into consideration
the location and distribution of existing landslides in a region, in addition to the relevant
influencing factors. Therefore, the tools and techniques for regional landslide hazard
studies are thus improving significantly.
38
Fig. 12 Schematic of Landslide Risk Management
39
Table 13 Decision Table
CONCLUDING REMARKS
The engineering design consists of proportioning the elements of a system so that it

satisfies various criteria of performance. The evaluation of reliability and safety of a given
or proposed system is one of the purposes of slope reliability analysis and is necessary in
the development of proper criteria for design. The need to treat uncertainties in the design
or assessment of slope stability today induces a lot of concern. The problem of defining a
better and more consistent measure of safety of slopes can best be attempted by probability
theory. Under conditions of uncertainty, the probability theory is useful in taking into
account the variability and lack of definitions in the parameters used for FS calculation and
thus it helps in objective selection of FS for the design of slopes. The reliability index and
probability of failure are the appropriate measures of safety of slope and provide more
information than a single fixed FS value as obtained from the deterministic approaches.
The Rosenblueths point estimate method and First-order second moment method have
been compared and found to give comparable results for uncertainties of FS. As the RPEM
does not involve derivatives it is preferred in this study. Based on the results of the
analyses, it is noted that for practical purposes, the reliability index should not be less than
1.4, with the corresponding static factor of safety of at least 1.35 on any of the critical slip
surfaces. Uncertainties due to imperfect knowledge should be provided for in adopting the
minimum factors of safety for design. The probability of failure provides more information
than the factor of safety obtained from the conventional methods. The probability values
are useful in planning mitigation measures and to reduce economic and social losses due to
slope failures and landslides.
40
It is of prime importance to note that risk studies involve a lot more work such as
identifying and classifying assets or elements at risk and estimating the potential for
destruction or damage to those assets. Thus risk studies are not made routinely for
individual slopes or for geotechnical projects in which slope analysis is a component.
Therefore, it is important that reliability analysis for a slope within a probabilistic
framework be carried out as a complete exercise in itself. The proposed approach is useful
in evaluating the occurrence of landslides along long stretches of hill slopes and in urban
areas. The concept of GIS can be used in three different contexts in a landslide problem.
Firstly, the shortest path method is useful in determining the critical slip surface. The
DEM and raster GIS based approach is useful in determining the regions of the slope,
which are likely to slide. The information about landuse is useful in risk evaluation. The
computation of risk can be improved further by considering fuzzy uncertainties as well in
the formulation. The interpretation of the risk probabilities, if done with care and insight,
could serve as instrument for anticipating and planning mitigation measures.
The risk assessment approach presented in this paper has been successfully used in
landslide risk assessment and management for engineered and natural slopes. The
approach may be adapted to suit a variety of problems, with due regard to the nature of the
issues involved. Recent developments have included more widespread use of quantitative
methods; more refined hazard and risk zoning which often involve use of digital
technologies such as remote sensing, GIS; improved rainfall-landslide incidence
correlation models; and improved methods for assessing travel distances and travel paths.
The intensity of critical rainfall of 240 mm/day, computed in the present study agrees well
with the reported rainfall intensity in the study area of Parmachi at the time of the
landslide. The runout distance and the thickness of deposit also agree with the observed
values at the site. The duration of runout, however, could not be verified due to lack of
reliable eye-witness accounts. The probability based decision analysis is a simple and
convenient tool, which gives a fair idea of the overall risk that is faced by an area exposed
to landslide proneness. Mitigation measures can be planned on the basis of the magnitudes
of the overall probabilities. It is to be noted that the need for proper geotechnical inputs to
the risk analysis, particularly with respect to the hazard identification and quantification is
a must and mandatory. The risk assessment is not a substitute for good geotechnical
engineering knowledge and judgement.
41
ACKNOWLEDGEMENT
Author wishes to acknowledge the support given by Mr. Adarsh S. Chatra and Ms. B.
Divya Priya, Research Scholars, Department of Civil Engineering, IIT Madras, Chennai,
for their help in the preparation of the manuscript.
REFERENCES
A-Grivas, D. (1981). How Reliable are Present Methods of Slope Failure Prediction,
Proc. 10th Int. Conf. on Soil Mech. and Found. Engg., Stockholm. A. A. Balkema,
Rotterdam, 3: pp 427-430.
Alfaro, L. D. and Harr, M. E. (1981). Reliability of Soil Slopes, Transportation Research
Record, No. 89, Washington D C: Nat. Research Council, pp. 78-82.
Alonso, E. E. (1976). Risk Analysis of Slopes and its Application to Slopes in Canadian
Sensitive Clays, Geotechnique, 26, pp. 453-472.
Ang, A. H. S. (1972). Probabilistic Bases of Safety, Performance and Design, Proc. of
the Special Conf. on Safety and Reliability of Metal Structures, ASCE, Pittsburgh, PA,
pp. 321-345.
Ang, A. H. S. and Tang, W. H. (1984). Probability Concepts in Engineering Planning and
Design II: Decision, Risk And Reliability, John Wiley and Sons, New York.
Ang, A. H.-S. and Tang, W. H. (2007). Probability Concepts in Engineering: Emphasis on
Applications to Civil and Environmental Engineering, John Wiley and Sons, New
York.
Babu, G. L. S. and Murthy, D. S. N. (2005). Reliability Analysis of Unsaturated Soil
Slopes, Journal of Geotechnical and Geoenvironmental Engineering, ASCE, 131, pp.
14231428.
Baecher, G. B. and Christian, J. T. (2003). Reliability and Statistics in Geotechnical
Engineering, John Wiley and Sons, New York.
Benjamin, J. R. and Cornell, C. A. (1970). Probability, Statistics and Decision for Civil
Engineers, McGraw-Hill, New York.
Bergado, D. T. and Anderson, L. R. (1985). Stochastic Analysis of Pore Pressure
Uncertainty for the Probabilistic Assessment of the Safety of Earth Slopes, Soils and
Foundations, 25, pp. 87-105.
42
Bishop, A. W. (1955). The Use of the Slip Circle in the Stability Analysis of Slopes,
Geotechnique, 5, pp. 7-17.
Chowdhury, R. N. (1978). Slope Analysis, Developments in Geotechnical Engineering, No.
22, Elsevier, New York.
Chowdhury, R. N. and Xu, D. W. (1993). Rational Polynomial Technique in Slope-
Reliability Analysis, Journal of Geotechnical Engineering, ASCE, 119, pp. 1910-
1927.
Chowdhury, R. N., Flentje, P. and Bhattacharya, G. (2010). Geotechnical Slope Analysis,
CRC Press, Boca Raton.
Christian, J. T. (2004). Geotechnical Engineering Reliability How Well Do We Know
What We Are Doing?, The 39th Terzaghi Lecture, Journal of Geotechnical and
Geoenvironmental Engineering, ASCE, 130, pp. 9851003.
Christian, J. T., Ladd, C. C. and Baecher, G. B. (1994). Reliability Applied to Slope
Stability Analysis, Journal of Geotechnical Engineering, ASCE, 120, pp. 2180-2207.
Christian, J. T. and Baecher, G. B. (1999). Point-estimate Method as Numerical
Quadrature, Journal of Geotechnical and Geoenvironmental Engineering, ASCE, 125,
pp. 779-786.
Cornell, C. A. (1969). A Probability-based Structural Code, Journal of the American
Concrete Institute, 66, pp. 974-985.
Ditlevsen, O. and Madsen, H. O. (2007). Structural Reliability Methods, Department of
Mechanical Engineering, Technical University of Denmark, Lyngby, Web edition.
Dodagoudar, G. R. (2001). Probabilistic and Fuzzy Reliability Analysis of Earth Slopes,
Ph. D. Thesis, Indian Institute of Technology Bombay, Mumbai, India.
Duncan, J. M. (2000). Factors of Safety and Reliability in Geotechnical Engineering,
Journal of Geotechnical and Geoenvironmental Engineering, ASCE, 126, pp. 307-316.
Duncan, J. M. (2001). Closure to the discussion on Factors of Safety and Reliability in
Geotechnical Engineering by J. M. Duncan, Journal of Geotechnical and
Geoenvironmental Engineering, ASCE, 127, pp. 717721 (Discussion pp. 700717).
Fell, R. and Hartford, D. (1997). Landslide Risk Management, Landslide Risk
Assessment, D. M. Cruden and R. Fell (Eds.), Balkema, Rotterdam, pp. 51-110.
Goh, A. T. C., Kulhawy, F. H. and Wong, K. S. (2008). Reliability Assessment of Basal-
Heave Stability for Braced Excavations in Clay, Journal of Geotechnical and
Geoenvironmental Engineering, ASCE, 134, pp. 145153.
43
Harr, M. E. (1979). Probability of Failure in Geotechnical Engineering, Proc. of the 3rd
Int. Conf. Application of Statistics and Probability in Soil and Struct. Engg., Sydney, 2,
pp. 680-685.
Harr, M. E. (1995). Accounting for Variability (Reliability), Civil Engineering
Handbook, W. F. Chen (Ed.), CRC Press, Boca Raton, pp. 683-704.
Hasofer, A. M. and Lind, N. C. (1974). Exact and Invariant Second-moment Code
Format, Journal of Engineering Mechanics, ASCE, 100, pp. 111-121.
Klir, G. J. (1994). The Many Facets of Uncertainty, Uncertainty Modelling and Analysis:
Theory and Applications, B. M. Ayyub and M. M. Gupta (Eds.), Elsevier, New York,
pp. 3-19.
Leporati, E. (1979). The assessment of Structural Safety: A Comparative Statistical Study
of the Evolution and Use of Level 3, Level 2, and Level 1, Research Studies Press,
Oregon.
Li, K. S. and Lumb, P. (1987). Probabilistic Design of Slopes, Canadian Geotechnical
Journal, 24, pp. 520-535.
Low, B. K. (1996). Practical Probabilistic Approach Using Spreadsheets, Geotechnical
Special Publication No.: 58, Uncertainty in the Geologic Environment: From Theory to
Practice, ASCE, 2, pp. 12841302, Madison, Wis. USA.
Low, B. K., Gilbert, R. B. and Wright, S. G. (1998). Slope Reliability Analysis using
Generalised Method of Slices, Journal of Geotechnical and Geoenvironmental
Engineering, ASCE, 124, pp. 350-362.
Madsen, H. O., Krenk, S. and Lind, N. C. (1986). Methods of Structural Safety, Prentice-
Hall, Englewood Cliffs, New Jersey.
Melchers, R. E. (1999). Structural Reliability Analysis and Prediction, John Wiley and
Sons, New York.
Quadri, S. S. (2000). Centrifuge Studies on Rainfall Induced Slope Instability and
Modelling of Mass Movement Using Fractals and DEM, Ph. D. Thesis, Indian Institute
of Technology Bombay, Mumbai, India.
Rosenblueth, E. (1975). Point Estimates for Probability Moments, Proc. Nat. Acad. of
Sciences, USA, 72, pp. 3812-3814.
Rosenblueth, E. (1981). Two-point Estimates in Probabilities, Applied Mathematical
Modelling, 5, pp. 329-335.
44
Ross, T. J. (2010). Fuzzy Logic with Engineering Applications, John Wiley and Sons, New
York.
Skempton, A. W. and Brown, J. D. (1961). A Landslide in Boulder Clay at Selset,
Geotechnique, 11, pp. 280-293.
Tang, W. H., Yucemen, M. S. and Ang, A. H. S. (1976). Probability-based Short Term
Design of Soil Slopes, Canadian Geotechnical Journal, 13, pp. 201-215.
Vanmarcke, E. H. (1977). Reliability of Earth Slopes, Journal of Geotechnical
Engineering, ASCE, 103, pp. 1247-1266.
Whitman, R. V. (1984). Evaluating Calculated Risk in Geotechnical Engineering,
Journal of Geotechnical of Engineering, ASCE, 110, pp. 143-188.
Wolff, T. F. (1996). Probabilistic Slope Stability in Theory and Practice, Uncertainty in
the Geologic Environment: From Theory to Practice ASCE Geotechnical Special
Publication No. 58, ASCE, New York.
Wu, T. H. and Kraft, L. M. (1970). Safety Analysis of Slopes, Journal of the Soil
Mechanics and Foundation Division, ASCE, 96, pp. 609-630.
Xu, B. and Low, B. K. (2006). Probabilistic Stability Analyses of Embankments Based on
Finite Element Method, Journal of Geotechnical and Geoenvironmental Engineering,
ASCE, 132, pp. 14441454.
45

Reliability of Slopes and Landslide Risk Assessment

Încărcat de

Informații document

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Reliability of Slopes and Landslide Risk Assessment

Încărcat de

Drepturi de autor:

Formate disponibile

RELIABILITY OF SLOPES AND LANDSLIDE RISK ASSESSMENT

The uncertainty is a fact of life in Civil Engineering. Recognition and assessment of

Classical Set Theory

Fuzzy Measure Theory

Rough Set Theory

Probabilistic approach can describe uncertainty arising from stochastic disturbances,

Propagation of Uncertainty through Models

METHODS OF RELIABILITY ANALYSIS

First-Order Second Moment Method

Reliability Evaluation of Slope

Rosenblueths Point Estimate Method

E[ Y r ] = p (1,2 ,...,N ) y (r1,2 ,...,N ) (5)

Reliability Index and Failure Probability

where () is the cumulative distribution function (CDF) of the normal distribution.

Table 1 Reliability Index (as normal variate) and Probability of Failure

Reliability index Probability of failure

RELIABILITY ANALYSIS: FINITE EARTH SLOPES

Numerical Example: Homogeneous Slope: A slope section shown in Figure 2 is

Reliability analysis: Results

Fig. 2 Slope Section Used for the Numerical Example

Table 2 Soil Data Used for the Numerical Example

Soil parameter Mean value COV (%)

Specification of the probability distribution of the

Determination of critical slip surface and its

Analysis using FOSM method Analysis using RPEM

Calculation of the partial derivative Evaluate the FS expression for all

Calculation of the expected value Calculation of the expected value

Computation of reliability index and

Fig. 1 Methodology for Probabilistic Reliability Analysis

Table 3 Results of the Reliability Analysis by Taylor Series Expansion of FS (with

Correlation between variables, c and

Table 4 Results of the Reliability Analysis by RPEM

Correlation between variables, c and

Case 1: E[c] = 18 kN/m2, E[] = 300 and E[ru] = 0.10;

Expected value of FS (E[FS])

Expected value of FS (E[FS])

0.10 1.444 1.446 1.447

Expected value of FS (E[FS])

0.10 1.308 1.310 1.311

Expected value of FS (E[FS])

0.10 0.15 0.20 0.25 0.30 0.35 0.40

Overall Failure Probability, pf 0.01

0.10 0.15 0.20 0.25 0.30 0.35 0.40

0.10 0.15 0.20 0.25 0.30 0.35 0.40

0.40 (c',') = 0.25

0.10 0.15 0.20 0.25 0.30 0.35 0.40

CASE STUDY: SELSET LANDSLIDE

Table 9 Soil Parameters for Selset Landslide

Parameter Mean value COV (%)

Values of FOSM method RPEM

pf (FS as a normal pf = 0.62575 pf = 0.62770

Description of the Study Area

Runout and Deposition of the Landslide Mass

Fig. 11 Simulation of Parmachi Landslide

For depositional profile:

Table 11 Deposition Profile

Landslide Risk Evaluation and Assessment

More formal applications of risk assessment and management principles, in a qualitative

The engineering design consists of proportioning the elements of a system so that it

S-ar putea să vă placă și