See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/266601998

Bearing capacity of clay: Influence of spatial variation of soil properties and

reliability based approach

Chapter · January 2011

CITATIONS READS

0 357

1 author:

Dr. Amit Srivastava

Lovely Professional University
91 PUBLICATIONS 990 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Quaternary Binders and Quaternary concrete View project

All content following this page was uploaded by Dr. Amit Srivastava on 11 October 2014.

The user has requested enhancement of the downloaded file.

In: Clay: Types, Properties and Uses ISBN: 978-1-61324-449-4
Editors: Justin P. Humphrey and Daniel E. Boyd ©2011 Nova Science Publishers, Inc.

Chapter 10

BEARING CAPACITY OF CLAY – INFLUENCE OF

SPATIAL VARIATION OF SOIL PROPERTIES AND
RELIABILITY BASED APPROACH

Amit Srivastava
Assistant Professor, Department of Civil Engineering, Jaypee University of Engineering
& Technology, Guna, MP, India. Email: amit.srivastava@jiet.ac.in

ABSTRACT
In the conventional design, the allowable bearing pressure on a shallow footing
resting on clayey deposit is estimated by dividing its ultimate bearing capacity by a factor
of safety. The selection of appropriate value of factor of safety depends on various factors
that involve social and economic considerations, type of structure and its importance,
type of soil, size of footing, and type and amount of load on the footing. Based on past
experiences and engineering judgments, it is established that a minimum factor of safety
of 3.0 is desirable. The approach is straight forward but does not handle variability in an
appropriate manner.
It should be noted that soil is a natural material and therefore geotechnical data are
always random and variable in space. Different sources of uncertainties include: inherent
variability, model transformation uncertainty, testing errors. In conventional procedure, it
is assumed that the selection of appropriate value of factor of safety takes into
consideration all the variability in the geotechnical parameters.
In the recent years, there have been considerable advances in the characterization of
soil variability and its application in geotechnical designs. It is realized that the use of
Reliability Based Design (RBD) is necessary to consider all sources of uncertainty in the
analysis and incorporate them in geotechnical design. The approach provides
mathematical basis to handle variability and gives rationality in the decision making. In
the present chapter, the influence of spatially variable soil parameters on the reliability-
based geotechnical deign is studied by considering the case of bearing capacity analysis
of a shallow strip footing resting on a clayey soil. For numerical modeling and analysis,
commercially available finite difference numerical code FLAC 5.0 is used.
2
Amit Srivastava

INTRODUCTION
The design of shallow foundation requires the following two criteria to be satisfied for
the successful performance, i.e., (i) it should have sufficient factor of safety with respect to
shear failure, and (ii) the settlements should be within tolerable limits. Except for the cases,
where the width of the footing is few meters wide, the settlement criterion governs the
estimation of allowable pressure for footings and piers resting on sands/gravels. In these
cases, almost all the settlements take place at the end of the construction and hence do not
pose any problem to the structures founded on such soils subsequently. In case of clays, both
the shear failure and settlements are the governing criteria (Bowles 1996). There are well
established theories for the estimation of ultimate bearing capacity of the clayey soil from
shear failure criterion. The settlement analysis requires the following two aspects to be
investigated: (i) An estimate of the net “final” settlement at the end of the consolidation and
(ii) time for settlement. The settlement calculations in clays are based on the classical
Terzaghi’s consolidation theory (Terzaghi and Peck 1948). With experience and from
economic considerations, it is established that a factor of safety of not more than 3.0 will be
appropriate for the design of shallow foundation resting on clayey soils in case where the
shear failure is the governing criterion (Terzaghi and Peck 1948, Bowles 1996). In this
context, the guidelines based on measurements of final settlements assume considerable
importance. Skempton (1951) quantified the selection of appropriate value of factor of safety
for the footings resting on clay based upon theoretical and field observations and concluded
that a factor of safety of at least 3.0 is desirable in estimating the allowable bearing capacity.
In cases where the foundation design is controlled by settlement considerations, a factor of
safety greater than 3.0 is desirable to restrict the settlement to a magnitude compatible with
the structural requirements. Skempton (1951) has also addressed the variation of soil
properties in an indirect manner and suggested the factors of safety. Following section
provides brief discussion and literature work on the ultimate bearing capacity of a shallow
foundation resting on a clayey soil based upon shear failure and settlement consideration.

ULTIMATE BEARING CAPACITY OF CLAYEY SOIL

Shear Failure Criterion

The net ultimate bearing capacity (qnu) of a shallow foundation on a clayey soil can be
estimated with the help of the following equation (Bowles, 1996).

qnu  cNC (1)

Where c is the undrained cohesion (for a saturated clayey soil for which undrained shear
strength parameter u = 0) and Nc is the bearing capacity factor. Applying a factor of safety
(FS), the allowable bearing capacity (qa) of the foundation soil can be calculated as below:

qnu cN c
qa   (2)
FS FS
 3
Bearing Capacity of Clay – Influence of Spatial Variation of Soil Properties and Reliability…
It can be seen that the net ultimate bearing capacity of shallow foundation on clayey soil
depends on two factors c and Nc and a global factor of safety (FS) is applied to cNc to obtain
the allowable bearing capacity of the foundation soil.
According to Skempton (1951), if the shear strength of the soil does not vary by more
than 50% of the average value for a depth of 2/3B below the footing, the average value of
“c” can be used for the calculation of ultimate bearing capacity of the clayey soil.
For the estimation of bearing capacity factor (Nc), theoretical solutions are available in
the literature. The most popular of them is the Prandtl (1920) solution in which the analysis of
the bearing capacity of the strip footing on the surface (D = 0) showed value of bearing
capacity factor (Nc) equal to 2+ = 5.14. In the analysis, the failure mechanism assumes that
the footing pushes in front of itself a wedge of clay, which, in turn, pushes the adjacent
material sideways and upwards.
For the footings placed at a considerable depth, Meyerhof (1950) modified the Prandtl
(1920) solution in which the slip surface is assumed to curve back on the sides of the
foundation. The corresponding value of Nc for strip footing is 8.3. The solution provides an
upper limit to the bearing capacity factor (Nc) value because of the greater length of assumed
shear surface.
Ishlinsky (1944) obtained a rigorous solution for Nc for circular footings, with a smooth
base placed on the surface of the clay. The solution provided a numerical value of Nc = 5.68.
For a rough footing, Meyerhof (1950) used an approximate analysis and for the same case, a
value of Nc equal to 6.2 was evaluated.
For circular footings located at a considerable depth beneath the surface there are three
theoretical solutions available in the literature. (i) Meyerhof (1950) calculated a value of Nc
equal to 9.3, which is certainly an upper limit. (ii) Gibson (1950) combined the approach
originated by Bishop et al., (1945) for metals with large strain theory by Swainger (1947) and
found the solution for bearing capacity factor for clays. The failure mechanism assumed that
the penetration of footing at ultimate failure is equivalent to expanding a spherical cavity in
the clay of diameter equal to the diameter of the footing (B). A plastic zone develops with a
B3 E
radius of (where E is the secant modulus of clays at a stress equal to ½ the yield
2 c
value) beyond which the clay remains in elastic state. The following expression for Nc was
proposed by Gibson (1950);

4 E 
Nc   loge  1  1 (3)
3 c 

The range of E/c for majority of undisturbed clays lies in the range of 50 to 200 that
provides a corresponding Nc value between 7.6 to 9.4 calculated using eq (3) with an average
value of Nc = 8.5. (iii) Wilson (1950) provided a value of Nc equal to 8.0 for the bearing
capacity of clay loaded at depth by a rigid circular plate by finding the foundation pressure
necessary to bring about the merging of the two plastic zones originating from the edges of
the footing. It can be seen that the available theoretical solutions for Nc although involve
different approaches, yet they lead to values of Nc in the range of 10% covered by Mott-
Gibson analysis for clays.
4
Amit Srivastava
Skempton (1951) proposed the following expression for the estimation of bearing
capacity factor (Nc), on the basis of theory, laboratory tests and field observations:

For strip footings:

 Df 
N c  51  0.2  ; with a maximum limiting value of 7.5 (4a)
 B 

For square and circular footings:

 Df 
N c  61  0.2  ; with a maximum limiting value of 9.0 (4b)
 B 

For rectangular footings:

 Df  B D
N c  5.01  0.2 1  0.2  ; For f  2.5 (4c)
 B  L B

 B Df
N c  7.51  0.2  ;for  2.5 (4d)
 L B

Table [1] summarizes the available field evidence on the ultimate bearing capacity of
clays that provides a satisfactory confirmation to the suggested values of Nc by Skempton
(1951).

Settlement Analysis

For the shallow foundation resting on a clayey soil, the final settlement and rate of
settlement can be calculated from Terzaghi’s theory of one dimensional consolidation
(Terzaghi and Peck 1948). Considering the estimation of net final settlement of the clayey
soil under an applied net pressure (qn), the following Eq. 5 can be used for the foundation
resting directly on a thick layer of clayey soil:

S n  mv qn BI f (5)

where mv is the coefficient of volumetric compressibility at a depth beneath the

foundation, as measured in oedometer tests on undisturbed samples and determined over the
range of overburden pressure from (po) to (po+z). po is the initial effective overburden
pressure at depth z and z is the increment of vertical pressure due to application of net
foundation pressure (qn). If = influence factor at a given depth z the value of which can be
obtained from published literature (Terzaghi 1943).
 5

Table 1. Comparison of field data with the suggested Nc value for ultimate bearing capacity of clays (adopted from Skempton 1951).

Dimensions Avg. shear Value of Nc

Net
Avg. (S) strength of
foundation
settlement at clay beneath Proposed
References S/B % pressure at Actual
B (m) L (m) D (m) failure the (Skempton
failure (qnf) (qnf/c)
(cm) foundation 1951)
kPa
(kPa)
Odenstad (1948) 0.0 (lower) 5.88 5.4
7 kPa (Vane)
Cadling and 0.40 2.0 12.7 3 41.20
6.4 (compr.)
Odenstad (1950) 0.3 (upper) 6.44 6.5
91.0
(with side 5.58
friction)
Skempton (1942) 2.44 2.74 1.68 25.4 10
110.0 15.32
(no side 7.18 7.2
friction)
Morgan (1944)
2.44 2.44 15.24 27.94 12 182.0 21.0 8.67 9.0
Skempton (1950)
1.83 (in clay) 7.4
Wilson (1950) 2.44 2.44 35.56 15 277.0 34.47 8.0
6.0 (total) 8.6
Nixon (1949) 7.62 7.62 0 - - 80.0 12.93 6.18 6.2
6
Amit Srivastava
Skempton (1951) indicated that the above conventional approach is adequate for
estimating the final settlement of foundations on deep beds of clays and it can be written in
the following form to obtain a relationship between final settlement and factor of safety
against ultimate failure.

qn qnu
Sn  B mv c I f (6)
qnu c

Defining the modulus of compressibility Kv = 1/mv, (qun/c)If  5.0, the net final settlement
is given by the following expression.

5 qn 5 1
Sn  B B (7)
K v / c q nu K v / c FS

Figure 1. Final settlements of the foundations in saturated clays (comparison of theoretical and recorded
settlement values as reported in Skempton 1951).
 7
Bearing Capacity of Clay – Influence of Spatial Variation of Soil Properties and Reliability…

For over-consolidated clays Kv/c lies in the range from 70 to 200 while for normally-
consolidated clays the range is approximately from 25 to 80 (Skempton 1951). Eq. 7 has been
plotted in Fig. 1 for several typical values of Kv/c and it is compared with the field
observation results (Skempton 1951).
It can be observed that for any given clay the settlement is approximately proportional to
the width of the footing for a given factor of safety. Conversely, the allowable net foundation
pressure on given clay will decrease in direct proportion to the foundation width, if it is
required to restrict the settlement to some specified magnitude. A close observation of Fig. 1
leads to the following inferences:
1) For a limiting value of final settlement and breadth of the footing, the normally-
consolidated soil required higher factor of safety than over-consolidated soil.
2) For a limiting allowable settlement, a larger width of the footing means a higher
factor of safety from settlement considerations and hence the design is governed
by settlement considerations.
3) The stability criterion is relevant only for small footings on over-consolidated
clays. For other cases the design is governed by settlement considerations.
From the above discussion on the bearing capacity evaluation of clayey soil based on
shear failure and settlement criteria, it can be concluded that a factor of safety of at least 3.0 is
desirable. In circumstances, where the design of foundation is controlled by the settlement
considerations, a factor of safety much greater than 3.0 could be used in order to restrict the
settlements within the tolerable limits. The analysis conducted in early 1950s provided a basis
for the selection of appropriate value of factor of safety in the conventional design of shallow
foundations resting on a clayey soil and since then it has been used routinely in the
conventional designs.

ISSUES OF VARIABILITY AND PROBABILISTIC APPROACH

A question that often arises, in practice, is to know “how safe is safe?” or to what extent
the factors of safety that are routinely used to address the question of safety and economy are
adequate. These factors of safety represent the combined influence of total variability and
deviations of analytical formulations based on simplified assumptions. To take care of
different sources of uncertainties involved in the estimation of input strength and stiffness
parameters, a selection of appropriate value of factor of safety comes from past experiences
and good engineering judgment. It is well understood that the approach is simple and
straightforward but does take into account the variability in an appropriate manner (Duncan,
2000; Schweiger et al, 2001; Baecher and Christian, 2003). In the recent years, it has been
established that reliability analysis is a useful tool to understand the role of variability and
studies have been conducted on the reliability analysis of shallow foundations (Sivakumar
Babu et. al, 2006; Babu and Srivastava, 2007).
In a probabilistic approach, the performance of the structure is expressed in a
probabilistic framework, i.e., either probability of failure (pf) or in terms of safety index
known as reliability index (). The development of reliability based design procedures is
receiving considerable attention and guidelines on the targeted reliability indices have been
suggested in literature. USACE (1997) made specific recommendation on target probability
8
Amit Srivastava
of failure (pf) and reliability indices () in geotechnical and infrastructure projects. The
suggested guidelines are given in Figure 2, which indicate that a reliability index () value of
at least 4.0 is considered to indicate good performance of the system and 3.0 for the above
average performance.

Figure 2. USACE (1997) guidelines for reliability index () and corresponding pf.

Methods of reliability analysis such as first order reliability method (FORM), second-
order reliability method (SORM), Point Estimate Method (PEM), and Monte Carlo simulation
(MCS) are available in literature (Harr 1987, Baecher and Christian 2003).
Normally, in the probabilistic analysis, the performance function g() is defined as C-D;
where C is the capacity and D is the demand. g() > 0 will be in the safe state while g() < 0
will be in the unsafe state. In case of bearing capacity analysis of shallow foundation, from
shear failure consideration, C will be taken as ultimate bearing capacity of the foundation soil
and D will be the applied pressure on the footing. On the other hand, from the settlement
considerations C will be the allowable settlement and D will be the calculated settlement. For
uncorrelated normally distributed C and D, reliability index () can be calculated using the
following equation:

C   D
 (8)
 C2   D2

On the other hand, for uncorrelated log-normally distributed C and D, reliability index
() can be calculated as below:
 9
Bearing Capacity of Clay – Influence of Spatial Variation of Soil Properties and Reliability…
  1  CoV 2 
ln  C D

  D 1  CoVC 
2

 (9)
 
ln 1  CoVC2 1  CoVD2 
Where C, D are the mean values of C and D; C and D are the standard deviation in the
estimation of C and D. CoV is the coefficient of variation which is obtained by dividing the
standard deviation () by the mean ().

QUANTITATIVE ASSESSMENT OF SOIL VARIABILITY

There are three major sources of uncertainty associated with geotechnical engineering
practice (Phoon and Kulhawy, 1999a & 1999b), viz., (a) the natural heterogeneity or inherent
variability (the physical phenomenon contributing to the variability), (b) measurement error
(due to equipment, procedural-operator, and random testing errors), and (c) model
transformation uncertainty (due to approximation present in empirical, semi-empirical or
theoretical models to relate measured quantities to design parameters).
Quantitative assessment of soil uncertainty modeling requires use of statistics, as well as
probabilistic modeling to process data from laboratory or in situ measurements. In the
probabilistic analysis, the input parameters are modeled as continuous random variables
defined by their probability density functions (pdf) and the parameters of distributions.
Normally, in geotechnical practice, the soil parameters are either modeled as normally
distributed or log-normally distributed continuous random variables (Baecher and Christian,
2003).

Random Variables

A random variable is a function defined on a sample space that assigns a probability or

likelihood to each possible event within the sample space, either assuming a specific value
(discrete random variable) or being within a range of values (continuous random variable).
Commonly employed models for discrete random variables include the Binomial and Poisson
distributions, and for continuous random variables include the Normal, Lognormal, Beta and
Uniform distributions. Random variables are discussed in terms of probability distributions,
moments, etc. (Ang and Tang, 1975; Benjamin and Cornell, 1970; Harr, 1987; Baecher and
Christian, 2003).
It should be noted that the selection of any probability distribution to characterize a
random variable is essentially an assumption, made because certain distributions facilitate
computations. It cannot in general be proved that a random variable fits a certain distribution,
although the goodness of fit between a data set and one or more candidate distributions can be
assessed by some standard statistical tests, such as the Chi-squared and Kolmogorv-Smirnov
tests, found in most statistical texts (Haldar and Mahadevan, 2000; Baecher and Christian,
2003).
10
Amit Srivastava
Moments of Random Variables

Moments of random variables are required in calculating the reliability index () or
probability of failure (pf) by first-order second-moment methods. The first moment about the
origin is the mean or expected value and the second central moment is the variance. (Central
moments are calculated with respect to the mean). The square root of the variance is the
standard deviation, and the ratio of the standard deviation () to the mean value () is the
coefficient of variation (CoV). Procedures for the calculation of moments are discussed in
standards text books (Baecher and Christian, 2003).
Any quantitative geotechnical variability relies on sets of measured data, which are often
limited in size and hence, it is referred to sample statistics. The uncertainty in the measured
data is expressed in terms of sample mean () and variance (2) evaluated from the following
expression:

Sample Mean

1 n
   xi (10)
n i 1

Variance (2)

It is a measure of dispersion of data about the mean value. The square root of variance is
defined as standard deviation ().

n
2 
1
 xi   2 (11)
(n  1) i 1

The coefficient of variation (CoV), which is obtained by dividing the sample standard
deviation () by the sample mean (), is commonly used in quantifying the geotechnical
uncertainty analysis because of the advantages of being dimensionless as well as providing a
meaningful measure of relative dispersion of data around the sample mean. The parameters of
the normal and log-normal probability distribution function (pdf) are directly related to the
unbiased estimates of statistical moments, i.e., sample mean () and variance (2) of the
measured data set.

Table 2. CoV% for the selected geotechnical parameters (Duncan, 2000).

Property CoV% range

Dry unit weight (d) 2 – 13
Undrained shear strength (cu) 6 – 80
Effective Friction angle () 7 – 20
Elastic Modulus (Es) 15 – 70
Coefficient of permeability (k) 68 - 90
 11
Bearing Capacity of Clay – Influence of Spatial Variation of Soil Properties and Reliability…
Consideration of these uncertainties in the input soil parameters and its impact on the
performance of a geotechnical system are studied using the reliability-based design
procedures. Reliability analysis focuses on the most important aspect of performance, i.e.,
probability of failure (pf).
Where site-specific data are not available to estimate parameters of random variables,
uncertainty can be characterized by assuming that the coefficient of variation of a parameter
is similar in magnitude to that observed at other sites. Typical values of coefficients of
variation for soil properties have been compiled and reported by Harr, 1987; Lacasse and
Nadim, 1996; Kulhawy and Trautman, 1996; Phoon and Kulhawy, 1999a & 1999b; Duncan,
2000; Uzielli et al., 2007. Typical values of range of coefficient of variation (CoV%) for
selected geotechnical parameters are provided in Table 2.

Independent and Correlated Random Variables

Independent random variables are those for which the likelihood of the random variable
assuming a specific value does not depend on the value of any other variable. Where the
value of a random variable depends on the value of another random variable, the two are said
to be correlated. Some examples of random variables that may be correlated are:
(i) Unit weight and friction angle of sand.
(ii) Preconsolidation pressure and undrained strength of clay.
(iii) The c and  parameters in a consolidated-undrained strength envelope.
Where random variables are correlated, their probability distributions form a joint
distribution, and one additional moment, the covariance, is necessary to model the parameters
when using second-moment methods. An alternative way to express the interdependence is
with the correlation coefficient, which relates the covariance to the variances of the two
variables. Calculation of correlation coefficients is further discussed in standard texts
(Baecher and Christian, 2003). An investigation into the values of the correlation coefficients
between soil parameters is reported in Uzielli et al. (2007). The effect of parameter
correlation is to increase or decrease the total uncertainty, depending on whether correlation is
positive or negative. Although parameter correlation can be shown to significantly affect the
results of probabilistic analysis, independence of random variables is often assumed in
probabilistic analysis. This may be done for two reasons, both for computational simplicity
and the fact that data are often insufficient to make reliable estimates of the required
correlation coefficients.

SPATIAL VARIATION AND AUTO-CORRELATION FUNCTION

It is well understood that second moment statistics, i.e., mean () and variance (2), alone
are insufficient to describe the spatial variation of soil properties, which vary in the 2- or 3-
dimensional space, whether measured in the laboratory or in situ. Structured explanations of
the statistical techniques used for the investigation of spatial variability are provided by
Priestly (1981); Baecher and Christian (2003). While the former provides an exhaustive
insight into the mathematical framework of time series analysis, the latter focuses specifically
on the application of such techniques to geotechnical engineering.
12
Amit Srivastava

Trend, t(z)

Deviation from trend,

L (characteristics
(z)
length or Zone of
influence)
Variation of soil
property, (z)

Figure 3. Statistical description of soil spatial variability in two dimensional space.

The spatial variability of soil property is the variation of soil properties from one point to
another in space (Schweiger and Peschl, 2005). A spatial variation of soil deposit in any
direction can, in principle, be characterized in detail if a sufficiently large number of
measurements are taken. This, however, is impossible in practice, and, therefore use of
statistical techniques for investigating the spatial variability of soil properties is needed.
Fig. 3 shows a typical spatial variation of soil properties [(z)] in a 2-dimensional space
characterized by (i) the vertical scale of fluctuation () or auto-correlation distance (ro), (ii)
trend function [t(z)], and (iii) deviation from the trend [w(z)], which constitute important
parameters for site characterization and reliability based design.
By decomposition the in situ soil properties are modeled as spatially variable parameters
in space, which consists of two components: (i) a deterministic trend and, (ii) the variations
off the trend. Considering a one dimensional case in which the soil properties vary with depth
(most frequently encountered in practice and in the literature), at any depth (z), decomposition
is expressed by the following relationship:

 z   t z    z  (12)

where, (z) is the measured geotechnical property, t(z) is the trend function (It should be
noted that the presence of trend in soil properties even for extremely homogeneous soil is
unavoidable due to factors such as overburden stress and stress history and hence, it is present
in most cases for geotechnical parameters); and (z) is the set of residuals off the trend which
includes different sources of associated uncertainties. Normally, variation off the trend, i.e.,
(z) is treated as a weak stationary process (Chatfield, 2004).
Trend removal by least squares regression has been utilized by several researchers in the
geotechnical literature (Jaksa et al., 1997; Wu, 2003). The main reason for performing data
transformation (e.g. by decomposition) is to obtain stationary residuals. This is desirable
because virtually all classical statistics are based on the assumption of stationarity. Among the
classical statistical tests (e.g. statistical runs test, Spearman’s rank coefficient, Kendall’s 
test), which have been used in the geotechnical literature, Kendall’s  test has been identified
as the most powerful tool for the geotechnical applications (Jaksa, 1995). Kendall’s  test
 13
Bearing Capacity of Clay – Influence of Spatial Variation of Soil Properties and Reliability…
(e.g. Daniel, 1990) is a non-parametric test for statistical independence (see Appendix I). As
statistical independence implies stationarity (where as the converse is not true), Kendall’s 
test can be used for assessing the latter (Uzielli et al., 2007).
A fundamental assumption inherent to decomposition and trend fitting is that the
residuals of decompositions are spatially uncorrelated, i.e., their fluctuation is completely
random. Due to high level of complexity involved in the selection of trend function and
associated uncertainty in the trend parameters; Uzielli et al. (2007) stated that it is better to
identify a simple trend function and model the set of residuals as a random variable that can
be effectively investigated using appropriate and well established statistical techniques. In
second-moment perspective, the part of variability that remains after trend removal
corresponds to the variance of the residuals and the pattern that remains in the values of soil
properties after the removal of a deterministic trend can be referred to as correlation structure
or, equivalently, auto-correlation because it refers to correlation between elements of the
same set (Vanmarcke, 1983).
The auto-correlation function describes the variation of the strength of spatial correlation
as a function of the spatial separation distance (i.e., lag distance, r) between two spatial
locations at which data are available. Since, the data sets are, in practice, always limited in
size, it is referred to as sample auto-correlation, which is an approximation of the real auto-
correlation function. The sample auto-correlation is calculated from the available set of data
that is deemed representative of the stochastic process. For a finite number of data sets,
sample auto-covariance [Cx(r)] and sample auto-correlation coefficient [x(r)] are calculated
using the following Eqs 13 & 14 (Box and Jenkins, 1970):

1
C x [r ] 
(n  1)
 ( xi   )(xir   ) (13)

C x (r )
 x (r )  (14)
C x (0)

where, xi is the value of soil property x at location i, and Cx(r) represents the auto-
covariance of the data at separation (or lag) distance r, and n is the number of all pairs of data
having a lag distance r. To identify how a given soil properties varies spatially, one needs to
fit one or more auto-correlation models to the sample auto-correlation calculated by Eq. 13 &
14. Various auto-correlation models (i.e., single exponential, cosine exponential, second-
order Markov, squared exponential) have been employed in geotechnical literature (DeGroot
and Baecher, 1993; Jaksa, 1995; Lacasse and Nadim, 1996; Fenton 1999a&b; Phoon et al.,
2003). Uzielli et al. (2005) provided a detailed discussion on various theoretical auto-
correlation models available in the literature and also proposed a method (MSBR approach)
for fitting the theoretical auto-correlation model using stress – normalized cone penetration
testing parameters.
With the aim of setting criteria on the choice of the correlation structure of the residuals,
Spry et al. (1988) opined that the auto-correlation structure should be identified by fitting
empirical auto-correlation models to the sample auto-correlation coefficients exceeding
Bartlett limits (IB) i.e. IB = ±1.96/ N . If the sample auto-correlation structure of the
14
Amit Srivastava
detrended data indicates that more than 95% of the sample auto-correlation falls between the
bound  1.96/ N (1.96 is the 0.975 quantile of the standard normal distribution), then the
variation off the trend in the data is considered to be uncorrelated and it can be taken as white
noise or an i.i.d. (independently and identically distributed) process (Priestley, 1981; Fenton,
1999a). The auto-correlation function has been widely used for investigating spatial
variability in the context of geotechnical engineering (Baecher and Christian, 2003; Jaksa et
al., 1997; Phoon and Kulhawy, 1999a and 1999b; Sivakumar Babu et al., 2006).
For the spatial variability modeling, a parameter, i.e., an auto-correlation distance (ro) is
defined as “the distance within which the soil property exhibits relatively strong correlation”.
To obtain the numeric value of auto-correlation distance ( or r0), it is taken as the distance at
which x(r) decays to 1/e. In a physical sense, it is the same as the scale of fluctuation (),
although the methodologies of obtaining scale of fluctuation () and auto-correlation distance
(ro) are different (Vanmarcke, 1977). A numerical relationship is established between scale of
fluctuation () and auto-correlation distance (ro) for the type of auto-correlation function used
in the analysis (Vanmarcke, 1983) as indicated in Table 3.

Table 3. Theoritical autocorrelation function, and corresponding autocorrelation distance

and scale of fluctuation

RANDOM FIELD MODELING PROCEDURE

The modelling of spatial correlation using autocorrelation function or semivariogram
does not allow full representation of the effects of soil variability on the performance and
reliability of spatially random geotechnical systems. Important effects such as the spatial
averaging effect and the existence of a critical spatial correlation require additional operations
and parameters (Uzielli et al. 2005).
The spatial variability of the soil properties can be expressed concisely by means of
random field theory. In most general terms, a random field is essentially a random (or
stochastic) process consisting of an indexed, i.e., ordered according to one or more reference
dimensions, set of random variables (Vanmarcke, 1983). The description of a random field in
 15
Bearing Capacity of Clay – Influence of Spatial Variation of Soil Properties and Reliability…
the second-moment sense through a mean, standard deviation, a scale of fluctuation and a
spatial correlation function is useful to characterise a spatially variable soil property.
Moreover, it allows the spatial averaging effect to be investigated, which significantly
affects quantitative variability assessment. However, if the results of random field modelling
are to be used in reliability-based engineering applications, some possible limitations in this
approach should be recognised. For instance, it could be suspected that, if spatial variability
of soil properties is included in an engineering model, stresses and/or displacements which
would not appear in the homogeneous case (i.e. in which variability is not addressed) could
be present.
Hence, one of the most important benefits of random field modelling is the capacity to
simulate data series. By using sets of random field simulations and implementing the
variability in non-linear finite element meshes, the Monte Carlo technique can be used to
predict reliability of geotechnical systems with spatially variable properties.
A number of studies have focused, in recent years, on the combined utilisation of random
fields, nonlinear finite element analysis and Monte Carlo simulation for investigating the
behaviour and reliability of geotechnical systems when the variability of soil properties which
are relevant to the main presumable failure mechanisms is considered. A detailed description
about the approach for the spatially variable modeling of soil property is discussed by Haldar
and Sivakumar Babu (2008) and Haldar (2008) and it is presented as follows:
The property variations of the in situ soil represented by the mean value, coefficient of
variation and scale of fluctuation influence the likely parameters for design. In a random field
modeling, assumption of log-normal distribution is appropriate as the soil properties are non-
negative and the distribution also has a simple relationship with normal distribution. For
example, if the soil property k is considered as random variable and assumed to be a log-
normally distributed value represented by parameters mean (k,), standard deviation (k), and
correlation distance (z), a log-normally distributed random field is given by;

xi   expln k ~
k ~ x    ln k ~ x 
x   Gi ~ (15)

x is the spatial position at which k is desired. G ~

where ~ x  is a normally distributed
random field with zero mean with unit variance. The values of ln k and  ln k are determined
using log-normal distribution transformations given by,

 2 
 ln2 k  ln1  k2  = ln1  COVk2  (16)
 k 

1
 ln k  ln  k   ln2 k (17)
2

The correlation function is considered as exponentially decaying as given by the

following equation:
16
Amit Srivastava

 2 
 k    exp   (18)
 z 

where,   ~
x1  ~
x2 is the absolute distance between the two point. The correlation matrix
is decomposed into the product of a lower triangular and its transpose by Cholesky
decomposition,

L  LT   k (19)

Given the matrix L, correlated standard normal random field is obtained as follows;

i
Gi   Lij Z j , i = 1, 2, 3, . n (20)
j 1

where, Zj is the sequence of independent standard normal random variables. For a

parametric study, typical values of COV (%) for various input soil parameters are taken from
published work as summarized in Table [2].

In the present study, the results are presented assuming that the soil properties have
isotropic correlation structure; therefore the correlation distance is the same in both horizontal
and vertical directions. The correlation matrix is generated considering equation (18). The
value of the lag distance () is considered to be the center to center distance of the
consecutive grids.

Figure 4: Discretization of finite difference grid in the random filed modeling

Figure 4. Discretization of finite difference grid in the random filed modeling.
 17
Bearing Capacity of Clay – Influence of Spatial Variation of Soil Properties and Reliability…

Fig. 4 explains the evaluation of correlation matrix considering the discretization of finite
difference grid. For example, if the center to center distance between grids 1 and 2 id dx, the
correlation between these two grids can be calculated by putting the value  = dx in equation
(18). Similarly, correlation between grid 1 with 3, 4, 5 can be established by placing  = 2 ×

dx, and 4 × dx and between grid 1 with 31, 32, 33 are dy, dx 2  dy 2 and 2dx 2  dy 2 ,
so on. Therefore values in the first row of the correlation matrix are the correlation
coefficients between grid 1 and other grids, and leads to 900 values in a row (if suppose the
number of grids are 30 × 30). Hence, considering all the grids, the size of the correlation
matrix is 900 × 900. Once the correlation matrix is established, it is decomposed into lower
and upper triangular matrices using Cholesky decomposition technique (Press et al, 2002).
The correlated standard normal random field is obtained by generating a sequence of
independent standard normal random variables (with zero mean and unit standard deviation)
and decomposed correlation matrix by equation (19) knowing the correlation distance (z),
whereas CoVk utilized to determine the standard deviation of the soil properties using
equation (16).
Realizations of log- normally distributed soil properties at each grid location are obtained
by transformation presented in equation (15) for a specified mean, and standard deviation of
the given soil property (k). The total computation process is conducted by developing a
subroutine in ‘FISH’ code in FLAC. Monte Carlo simulation approach is used in the
generation of sample functions of 2D Log-Normal random field.
Fig. 5 shows comparison of the combination of different realization of random variation
of uncorrelated and correlated log-normally distributed cohesion parameter (mean = 5.0 kPa,
CoV = 30%, 80%, and correlation distance = 0.0m, 5.0m).
JOB TITLE : cohesion parameter (*10^1)

JOB TITLE : cohesion parameter (*10^1)

FLAC (Version 5.00)
FLAC (Version 5.00)
1.000
LEGEND
1.000
LEGEND
21-Dec-09 9:17
step 0 21-Dec-09 9:20
-1.667E+00 <x< 1.167E+01
step 0 0.800
-1.667E+00 <y< -1.667E+00
1.167E+01 <x< 1.167E+01 0.800
-1.667E+00 <y< 1.167E+01
Grid plot
Grid plot
0.600
0 2E 0
0.600
cohesion 0 2E 0
2.00E+03 cohesion
3.00E+03 2.00E+03
4.00E+03 3.00E+03 0.400

5.00E+03 4.00E+03 0.400

6.00E+03 5.00E+03
7.00E+03 6.00E+03
8.00E+03 7.00E+03 0.200
9.00E+03 8.00E+03
1.00E+04 0.200
9.00E+03
1.00E+04
Contour interval= 1.00E+03
Contour interval= 1.00E+03 0.000

0.000

Amit Srivastava
IISc, Bangalore Amit Srivastava
0.000 0.200 0.400 0.600 0.800 1.000

Figure 5a. Mean value of cohesion = 5 kPa (values in figure are in Pa), CoV = 30%, Correlation dist =
IISc, Bangalore
0.000
(*10^1)
0.200 0.400
(*10^1)
0.600 0.800 1.000

0.0 m. JOB TITLE : cohesion parameter (*10^1)

FLAC (Version 5.00)

JOB TITLE : cohesion parameter 1.000 (*10^1)

LEGEND
FLAC (Version 5.00)
21-Dec-09 9:27
step 0 1.000
-1.667E+00 <x< 1.167E+01 LEGEND 0.800
-1.667E+00 <y< 1.167E+01
21-Dec-09 9:27
Grid plot step 0
-1.667E+00 <x< 1.167E+01 0.800
0.600
0 -1.667E+00
2E 0 <y< 1.167E+01
cohesion
Grid plot
2.00E+03
3.00E+03 0 2E 0 0.600
4.00E+03 0.400

5.00E+03 cohesion
6.00E+03 2.00E+03
7.00E+03 3.00E+03
8.00E+03 4.00E+03 0.400
0.200
5.00E+03
Contour interval= 6.00E+03
1.00E+03
7.00E+03
8.00E+03
0.200
0.000
Contour interval= 1.00E+03

Figure 5b: Mean value of cohesion = 5 kPa (values in figure are in

Amit Srivastava
0.000
IISc, Bangalore
Pa), CoV = 30%, Correlation dist = 2.5 m
0.000 0.200 0.400 0.600 0.800 1.000
(*10^1)

Figure 5b. Mean value of cohesion = 5 kPa (values in figure are in Pa), CoV = 30%, Correlation dist =
Amit Srivastava
IISc, Bangalore
0.000 0.200 0.400 0.600 0.800 1.000

2.5 m. (*10^1)
18
JOB TITLE : cohesion parameter
Amit Srivastava
JOB TITLE : cohesion parameter
FLAC (Version 5.00)
(*10^1)

(*10^1)

FLAC (Version 5.00)

1.000
LEGEND
1.000
LEGEND
21-Dec-09 9:32
step 0 21-Dec-09 9:32
-1.667E+00 <x<step1.167E+01
0 0.800
-1.667E+00 <y< 1.167E+01
-1.667E+00 <x< 1.167E+01 0.800
-1.667E+00 <y< 1.167E+01
Grid plot
Grid plot
0.600
0 2E 0
0.600
0 2E 0
cohesion
0.00E+00
cohesion
2.50E+03 0.00E+00
5.00E+03 2.50E+03 0.400

7.50E+03 5.00E+03 0.400

1.00E+04 7.50E+03
1.25E+04 1.00E+04
1.50E+04 1.25E+04
0.200
1.75E+04 1.50E+04
0.200
2.00E+04 1.75E+04
2.25E+04 2.00E+04
2.25E+04
Contour interval= 2.50E+03 0.000
Contour interval= 2.50E+03 0.000

Amit Srivastava
IISc, Bangalore
Amit Srivastava
0.000 0.200 0.400 0.600 0.800 1.000
IISc, Bangalore (*10^1)

Figure 5c. Mean value of cohesion = 5 kPa (values in figure are in Pa), CoV = 80%, Correlation dist =
0.000 0.200 0.400 0.600 0.800 1.000
(*10^1)

0.0 m. JOB TITLE : cohesion parameter

JOB TITLE : cohesion parameter
(*10^1)
(*10^1)
FLAC (Version 5.00)
FLAC (Version 5.00)
1.000
LEGEND
1.000
LEGEND
21-Dec-09 9:37
step 0 21-Dec-09 9:37
-1.667E+00 <x< 1.167E+01
step 0 0.800
-1.667E+00 <y< 1.167E+01
-1.667E+00 <x< 1.167E+01 0.800
-1.667E+00 <y< 1.167E+01
Grid plot
Grid plot
0.600
0 2E 0
0.600
0 2E 0
cohesion
0.00E+00 cohesion
2.50E+03 0.00E+00
5.00E+03 2.50E+03 0.400

7.50E+03 5.00E+03 0.400

1.00E+04 7.50E+03
1.25E+04 1.00E+04
1.50E+04 1.25E+04
0.200
1.75E+04 1.50E+04
0.200
2.00E+04 1.75E+04
2.25E+04 2.00E+04
2.25E+04
Contour interval= 2.50E+03 0.000
Contour interval= 2.50E+03 0.000

Amit Srivastava
IISc, Bangalore Amit Srivastava
0.000 0.200 0.400 0.600 0.800 1.000
IISc, Bangalore (*10^1)

Figure 5d. Mean value of cohesion = 5 kPa (values in figure are in Pa), CoV = 80%, Correlation dist =
0.000 0.200 0.400 0.600 0.800 1.000
(*10^1)

2.5 m.

With this brief discussion on the methodology for incorporating spatial variation of soil
strength and stiffness parameters into the numerical modeling procedures, the following
section presents influence of spatially variable soil parameters on the performance assessment
of geotechnical systems considering the case of “bearing capacity of shallow strip footing
resting on a clayey soil”.

BEARING CAPACITY OF SHALLOW STRIP FOUNDATION RESTING ON

SPATIALLY VARIABLE CLAYEY SOIL
The influence of spatially variable soil parameters on the bearing capacity and
deformation pattern of a shallow foundation resting on a clayey soil deposit is investigated
using the numerical analysis. The influence of spatial variation of undrained cohesion (cu)
parameter on the bearing capacity of clayey soil is studied in relation to the well known
Prandtl solution. The Prandtl solution provides a bearing capacity factor (qu/cu = Nc) equal to
5.14 for the uniformly constant value of cohesion parameter under undrained loading
conditions (Taylor, 1948).
Prior to the bearing capacity analysis of clayey soil for the spatially variable cohesion
parameter, the validation of the numerical model is performed assuming uniformly constant
value of cohesion parameter. Figure 5 shows the finite difference numerical model having
size 50 m × 20 m with uniformly constant values of input soil parameters i.e. density ( ),
cohesion (c), elastic modulus (Es), and Poisson’s ratio (). The mean values of these input soil
parameters are taken from Table 4.
 19
Bearing Capacity of Clay – Influence of Spatial Variation of Soil Properties and Reliability…
Table 4. Mean values and coefficient of variations of input soil parameters (adopted
from Srivastava and Babu 2009).

Properties Mean CoV%

* 3
Density (kN/m ) 16.0 -
Cohesion (kPa) 7.0 39%
Friction angle () 27 11%
Elastic Modulus (MPa) 5.5 34%
*
Poisson’s ratio () 0.35 -
*considered as deterministic parameters

The vertical load is applied at the surface. The maximum shear strain increment contours,
as shown in Fig. 6, indicate identical displacement of footing about the center line, which in
turn leads to the equal settlement of the footing placed on a soil medium having uniformly
constant values of input soil parameter.
For the given case, from Fig. 7 it can be noted that the results of the numerical analysis
provide a bearing capacity factor value close to the Prandtl solution (e.g. qu/cu = Nc = 5.14). It
validates the numerical procedure and geometric extent of the numerical model.
It is well understood that the actual soil profile will never have a uniformly constant
value of soil properties, instead it is always spatially variable and the ultimate bearing
capacity expressed in terms of bearing capacity factor (Nc) is likely to be different. In order to
investigate this aspect, the numerical modeling of spatially variable soil profile is performed
using the statistical parameters of the input soil parameters.
JOB TITLE : prandtle solution (*10^1)

FLAC (Version 5.00)

Vertical load
LEGEND 3.000

Equal shear strain

29-Apr-09 13:40
step 5165 increment about the
-2.778E+00 <x< 5.278E+01
0.5 m width
center line of the
-1.778E+01 <y< 3.778E+01 X YYY X
X footing X
2.000

X X
Grid plot X X
X X
X X
0 1E 1 X X
X X
X X
cohesion X X
X X
7.000E+03 X 20 m X
1.000

X X
Max. shear strain increment X X
Contour interval= 2.50E-02 X X
X 50 m X
Minimum: 0.00E+00 X X
X X
Maximum: 2.00E-01 X X
X X
Fixed Gridpoints B BBBBBBBBBBBBBBBBBBBBBBBB B BBBBBBBBBBBBBBBBBBBBBBBBB 0.000
X X-direction
Y Y-direction
B Both directions
Figure 6: Uniformly
Figure constant
6. Uniformly value
constant value of cohesion
of cohesion (Pa) parameter
(Pa) parameter.

-1.000

IISc, Bangalore
Amit Srivastava
0.500 1.500 2.500 3.500 4.500
(*10^1)
 20
Amit Srivastava

Figure 7. Load vs. displacement (mm) curve indicating numerical solution approaches to Prandtl
solution.

JOB TITLE : Bearing capacity of spatial variable soil (*10^1)

FLAC (Version 5.00) Vertical load

Non-constant distribution
LEGEND of shear strain increment 2.250

contours
11-Jan-09 19:19
0.5 m width
step 6468
8.300E+00 <x< 4.170E+01 1.750
-6.700E+00 <y< 2.670E+01

cohesion
2.50E+03
1.250
5.00E+03
7.50E+03
1.00E+04
1.25E+04
1.50E+04 SPATIALLY VARIABLE CLAYEY SOIL 0.750

Contour interval= 2.50E+03

Max. shear strain increment
Contour interval= 5.00E-02
Minimum: 0.00E+00 0.250

Maximum: 4.00E-01

-0.250

Figure 8. Spatially variable cohesion (Pa) parameter; indicating non-constant distribution of maximum
IISc, Bangalore
shear strain increment and differential settlement.
Amit Srivastava
1.000 1.500 2.000 2.500 3.000 3.500 4.000
(*10^1)
 21
Bearing Capacity of Clay – Influence of Spatial Variation of Soil Properties and Reliability…
Fig. 8 shows a typical realization of spatially variable cohesion parameter (Pa) with
inherent linear trend in the data and the maximum shear strain increment rate for the applied
vertical load. It can be noted that there is a non-constant shear strain rate increment about the
center line of the footing leading to the differential settlement of the footing. Since the soil
properties are modeled as uncorrelated normally distributed random process, therefore each
realization presents a completely different picture, which in turn leads to different bearing
capacity factor for each realization. Hence, the probabilistic assessment of bearing capacity of
spatially variable soil profile requires large number of Monte Carlo simulation runs.
It should be noted that in the probabilistic analysis, the number of Monte Carlo
simulation cycles influences the accuracy of the results. Since the increase in number of
simulations also increases the computational efforts, a compromise between accuracy and
computational time is achieved. From the literature (Haldar and Mahadevan, 2000; Baecher
and Christian, 2003), it is suggested that for an acceptable error of 5.0 % and  = 0.05 i.e.
95% confidence level in the estimated mean and variance of the output response, a minimum
sample size of 1536 is suggested. Therefore, in the present analysis, 2000 Monte Carlo
simulations are run for estimating the statistical parameters of the bearing capacity factor (Nc)
for the spatially variable soil profile.
12.00

10.00

8.00
qu/cu

6.00
5.14
4.00

2.00

0.00
0 20 40 60 80 100
Displacement (mm)

Figure 9. Load-displacement curves corresponding to 2000 Monte Carlo simulations for bearing
capacity factor (Nc) of spatially variable soil.

Fig. 9 provides results of 2000 numerical simulations and with these results; the mean
value of bearing capacity factor (Nc) is obtained as 5.03 with a CoV of 24.0 %. In the
conventional approach, the bearing capacity factor (Nc) value for uniformly constant soil
property is taken as 5.14 (Prandtl solution). It can be observed that the mean value of bearing
capacity factor for spatially variable undrained cohesion parameter is smaller than the
corresponding value taken for uniformly constant undrained cohesion parameter.
22
Amit Srivastava
In the conventional approach it is suggested that the applied load on the footing should be
such that it should provide a factor of safety of at least 3.0 with respect to bearing capacity
failure. Using the probabilistic approach, it is demonstrated that the approach allows the
selection of appropriate value of factor of safety in a rational way.
For the reliability index () calculations if the applied load on the footing is taken as
5.14/FS i.e. demand (D), which is assumed as deterministic parameter with no variance
(  D2 = 0), and the mean value of bearing capacity factor for the spatially variable soil deposit
is taken as the capacity i.e. C = Nc = 5.03, with coefficient of variation (CoV) equal to 24%;
the reliability index () values for different values of factor of safety (i.e. 2.0, 2.5, 3.0, 3.5,
4.0, 4.5, and 5.0) can be calculated. The results of the reliability index () calculations are
summarized in Table 5.

Table 5. Effect choice of factor of safety (FS) and corresponding reliability indices ().

FS  (Normally Dist.)  (Log-normal Dist.)

2.0 1.80 2.26
2.5 2.29 3.22
3.0 2.61 4.00
3.5 2.85 4.67
4.0 3.02 5.24
4.5 3.16 5.74
5.0 3.26 6.20
It can be noted that for a given factor of safety, the reliability index () values for log-
normally distributed Nc are higher than the corresponding values obtained for normally
distributed Nc. Although an assumption of normal distribution in the Nc factor leads to
conservatism in the design procedure, it allows consideration of its negative values, which is
not realistic (a bearing capacity factor, Nc, can never be negative). Therefore, results obtained
from log-normally distributed Nc factor are more realistic and acceptable.
From Table 5 it can also be noted that for the existing spatial variation of in situ soil
parameters a factor of safety of 3.0 provides a reliability index value of 4.46 (> 4.0), which
ensures good performance level and on the other hand a factor of safety of 2.5 indicates the
performance above average with a reliability index () value in the range of 3.0 to 4.0. Hence,
a choice of factor of safety can be based upon the performance level of the foundation system.
With these results it can be noted that the results of the probabilistic analysis provide a
justification in the choice of factor of safety to be used in the deign procedures.

CONCLUDING REMARKS
The present chapter discusses the incorporation of spatial variability of the soil
parameters in the numerical modeling and its influence on the bearing capacity of shallow
foundation resting on clayey soil. The results of the reliability analysis in the light of
conventional approach demonstrated that the former approach enables a rational choice of
design parameters based on the predetermined performance level of the structure in terms of
reliability index values. It is also concluded that the probabilistic analysis, when used in
conjunction with conventional approaches, enables a rational choice of performance limits
such as factors of safety and helps in the decision making process.
 23
Bearing Capacity of Clay – Influence of Spatial Variation of Soil Properties and Reliability…

REFERENCES
Ang, A.H.-S., and Tang, W. H. (1975). Probability Concepts in Engineering Planning. and
Design, Vol. 1, Basic Principles, John Wiley, New York.
Baecher, G.B., Christian, J.T., 2003. Reliability and statistics in geotechnical Engineering.
Chi Chester, John Wiley Publications.
Benjamin, J.R., Cornell, J.R., 1970. Probability, Statistics, and Decisions for Civil Engineers:
McGraw- Hill. New York. USA.
Bishop, R. F., HILL, R., and Mott, N. F. (1945): The theory on indentation and hardness test.
Proc. Physics society. 57:147.
Bowles, J. E. (1996): Foundation analysis and design, 5th edition, McGraw-Hill Book Co.,
New York.
Box, G. E. P., Jenkins, G. M. (1970). Time series analysis: forecasting and control. San
Francisco: Holden-Day.
Chatfield, C. (2004). The Analysis of Time Series: An Introduction. Sixth edition, Boca
Raton: Chapman & Hall/CRC Press.
Daniel, W.W. (1990). Applied nonparametric statistics. 2nd edition, Boston: PWS-Kent.
DeGroot , D. J., Baecher, G. B. (1993). Estimating auto-covariance of in situ soil properties.
Journal of Geotechnical Engineering, ASCE , Vol. 119(1), 147-166.
Duncan, J.M. 2000. Factors of safety and reliability in geotechnical engineering. Journal
Geotechnical and Geoenvironmental Engineering, ASCE, 126(4): 307-316.
Fenton, G. A. (1999a). Estimation for stochastic models. Journal of Geotechnical and
Geoenvironmental Engineering, ASCE, Vol. 125(6), 470-485.
Fenton, G. A. (1999b). Random field modeling of CPT data. Journal of Geotechnical and
Geoenvironmental Engineering ASCE, Vol. 125(6), 486-498.
Gibson, R. E. (1950): Discussion on paper by Guthlac Wilson (1950), Journal of institution of
Civil Engineers, 34:382.
Haldar, A., Mahadevan, S., 2000. Probability, reliability and statistical methods in
engineering design. John Wiley & Sons, NY.
Haldar, S. (2008). Reliability based design methods of pile foundations under static and
seismic loads. Ph.D. thesis, Indian Institute of Science, Bangalore, India.
Haldar, S., and Sivakumar Babu, G. L. (2008). Effect of soil spatial variability on the
response of laterally loaded pile in undrained clay. Computers and Geotechnics, 35,
537–547.
Harr, M.E., 1987. Reliability-based design in civil engineering. McGraw-Hill, NY.
Ishlinsky, A. J. (1944): The Axial Symmetrical Problem in Plasticity and the Brinell Test,
Journal of Applied Mathematics and mechanics, U.S.S.R., 8:201.
Jaksa, M. B. (1995). The influence of spatial variability on the geotechnical design properties
of a stiff, overconsolidated clay. Ph.D. thesis, University of Adelaide.
Jaksa, M. B., Brooker, P. I., and Kaggwa, W. S. (1997). Inaccuracies associated with
estimating random measurement errors. Journal of Geotechnical and Geoenvironmental
Engineering, ASCE, Vol. 123(5), 393–401.
Kulhawy, F. H., and Trautmann, C. H. (1996). Estimation of In-situ Test Variability,
Uncertainty in the Geologic Environment: From Theory to Practice (GSP 58), Ed. CD
Shackelford, PP Nelson & MJS Roth, ASCE, New York, Aug 1996, 269-286.
 24
Amit Srivastava
Lacasse, S. and Nadim, F. (1996). Uncertainties in characterizing soil properties. In
Shackleford, C. D., Nelson, P. P., and Roth, M. J. S. (Eds.), Uncertainty in the Geologic
Environment (GSP 58), ASCE, New York, 49-75.
Meyerhof, G. G. (1950): A general theory of bearing capacity, Building Research Station,
Note No. c.143.
Phoon, K.K., Kulhawy, F.H., 1999a. Characterization of geotechnical variability. Canadian
Geotechnical Journal 36, 612-624.
Phoon, K.K., Kulhawy, F.H., 1999b. Evaluation of geotechnical property variability.
Canadian Geotechnical Journal 36, 625-639.
Prandtl, L. (1920): Uber die Harte plastischer Korper. Nach. Gesell. Wiss Gottingen. Math-
Phys. Kl. P. 74.
Press, W. H., Teukolsky, S. A., Vetterling, W. T., Flannery, B. P., 2002. Numerical recipes in
C++: the art of scientific computing. Second edition, Cambridge University Press.
Priestley, M. B. (1981). Spectral analysis and time series. I: Univariate series, NY, Academic
Press.
Schweiger, H.F., Thurner, R., Pottler, R., 2001. Reliability analysis in geotechnics with
deterministic finite elements. International Journal of Geomechanics 1(4), 389–413.
Sivakumar Babu, G. L. and Amit Srivastava (2007): Reliability analysis of allowable
pressure on shallow foundation using response surface method, Computers and
Geotechnics, 34(3), pp.187-194.
Sivakumar Babu, G. L., Amit Srivastava, and Murthy, D.S.N. (2006): Reliability analysis of
bearing capacity of shallow foundation resting on cohesive soil, Canadian Geotechnical
Journal, 43, pp. 217-223.
Skempton, A. W. (1951): The bearing capacity of clays, Building research congress, London,
1, pp. 180-189.
Spry, M. J., Kulhawy, F. H., Grigoriu, M. D. (1988). Reliability–based foundation design for
transmission line structures: geotechnical site characterization strategy. Report EL –
5507(1). Palo Alto: Electric Power Research Institute.
Swainger, K. H. (1947): stress-strain compatibility in greatly deformed engineering metals,
Phil. Mag., 38:422.
Terzaghi, K. and Peck, R. B. (1948): Soil mechanics in engineering practice, John Wiley.
USACE, 1997. Risk-based analysis in Geotechnical Engineering for Support of Planning
Studies, Engineering and Design. US Army Corps of Engineers, Department of Army,
Washington, DC, 1997. 20314-100.
Uzielli, M., Lacasse, S., Nadim, F., and Phoon, K. K. (2007). Soil variability analysis for
geotechnical practice. Characterization and engineering properties of natural soils – Tan,
Phoon, Hight& Leroueil (eds), Taylor & Francis group.
Vanmarcke, E. H. (1977). Probabilistic modeling of soil profiles. Journal of the Geotechnical
Engineering Division, ASCE, Vol. 103, 1227–1246.
Vanmarcke, E. H. (1983). Random fields: analysis and synthesis. MIT Press, Cambridge.
Wilson, G. (1950): The bearing capacity of screw piles and screwcrete cylinders, Journal of
institution of Civil Engineers, 34:4.
Wu, T. H. (2003). Variations in clay deposits of Chicago. In G.A. Fenton & E.H. Vanmarcke
(eds.), Probabilistic Site Characterization at the National Geotechnical Experimentation
Sites, Geotechnical Special Publication, Vol. 121, 13–28.

View publication stats