Georisk: Assessment and Management of Risk for

Engineered Systems and Geohazards

ISSN: 1749-9518 (Print) 1749-9526 (Online) Journal homepage: https://www.tandfonline.com/loi/ngrk20

The story of statistics in geotechnical engineering

Kok-Kwang Phoon

To cite this article: Kok-Kwang Phoon (2020) The story of statistics in geotechnical engineering,
Georisk: Assessment and Management of Risk for Engineered Systems and Geohazards, 14:1,
3-25, DOI: 10.1080/17499518.2019.1700423

To link to this article: https://doi.org/10.1080/17499518.2019.1700423

Published online: 09 Dec 2019.

Submit your article to this journal

Article views: 1481

View related articles

View Crossmark data

Citing articles: 2 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=ngrk20
GEORISK
2020, VOL. 14, NO. 1, 3–25
https://doi.org/10.1080/17499518.2019.1700423

SPOTLIGHT ARTICLE

The story of statistics in geotechnical engineering

Kok-Kwang Phoon
Department of Civil and Environmental Engineering, National University of Singapore, Singapore, Singapore

ABSTRACT ARTICLE HISTORY

The story of statistics in geotechnical engineering can be traced to Lumb’s classical Canadian Received 8 October 2019
Geotechnical Journal paper on “The Variability of Natural Soils” published in 1966. In parallel, Accepted 30 November 2019
the story of risk management in geotechnical engineering has progressed from design by
KEYWORDS
prescriptive measures that do not require site-specific data, to more refined estimation of Big Indirect Data (BID);
site-specific response using limited data from site investigation as inputs to physical models, generic databases; MUSIC-X;
to quantitative risk assessment (QRA) requiring considerable data at regional/national scales. transformation models; site
In an era where data is recognised as the “new oil”, it makes sense for us to lean towards challenge; similarity index
decision making strategies that are more responsive to data, particularly if we have
zettabytes coming our way. In fact, we already have a lot of data, but the vast majority is
shelved after a project is completed (“dark data”). It does not make sense to reduce one
zettabyte to a few bytes describing a single cautious value. It does not make sense to
expect big data to be precise and to fit a particular favourite physical model as demanded
by the classical deterministic world view. This paper advocates the position that there is
value in data of any kind (good or not so good quality, or right or wrong fit to a physical
model) and the challenge is for the new generation of researchers to uncover this value by
hearing what data have to say for themselves, be it using probabilistic, machine learning, or
other data-driven methods including those informed by physics and human experience, and
to re-imagine the role of the geotechnical engineer in an immersive environment likely to be
imbued by machine intelligence.

1. Introduction
In response to a question on the “importance of stat-
Errors using inadequate data are much less than those istics as a tool in engineering applications” posed during
using no data at all. an interview before his retirement (Lam and Li 1986),
Charles Babbage
Professor Lumb opined that
Every story has a beginning. The idea that statistics can be
Traditionally engineering and civil engineering are very
used to quantify uncertainties in the properties of natural deterministic in their teaching and in the attitude of
soils (an intrinsic characteristic of site data) and this stat- their practitioners. When something goes wrong, it
istical approach can provide a rational basis for the selec- takes them by surprise. And yet all the things they are
tion of a suitably cautious design value may arguably be handling, the raw materials, the input and output, are
traced to Lumb’s classical Canadian Geotechnical Journal random processes. If that can be taken seriously the
method of design can be improved considerably. Instead
paper on “The Variability of Natural Soils” published in of the old fashioned safety factor, the probability of fail-
1966. One may view this paper as being ahead of its ure type of approach is more satisfactory and practically
time. The First International Conference on Applications far more useful.
of Statistics and Probability to Soil and Structural Engin-
eering was organised in Hong Kong in 1971. It was not He further added that
surprising that Professor Peter Lumb played a key role once you think of all these things as being random pro-
in launching this important conference series. The first cesses, it does clear up the engineer’s mind as well as
ICASP in Hong Kong was followed by Aachen (1975), improving his design. It makes him realize that he can-
Sydney (1979), Florence (1983), Vancouver (1987), not predict what is going to happen precisely. This is
what most engineers try to do. That is what they taught
Mexico City (1991), Paris (1995), Sydney (1999), in schools and in universities in general: Engineering is
San Francisco (2003), Tokyo (2007), Zurich (2011), Van- precise, it is a science. Yet, in reality, it is vague. It is not
couver (2015) and Seoul (2019). a science. It is more an art.

CONTACT Kok-Kwang Phoon [email protected] Department of Civil and Environmental Engineering, National University of Singapore, Block E1A,
#07-03, 1 Engineering Drive 2, Singapore 117576, Singapore
© 2019 Informa UK Limited, trading as Taylor & Francis Group
4 K.-K. PHOON

It was only in 1995 that a National Research Council and regulatory offices after completion of a project. They
report “Probabilistic Methods in Geotechnical Engineer- are stored primarily for compliance purposes, rather
ing” recommended that than shared and further analysed to provide more
insights and to support future decision making. In
probabilistic methods, while not a substitute for tra-
ditional deterministic design methods, do offer a sys- short, our data is mainly “dark”, a term defined in Gart-
tematic and quantitative way of accounting for ner’s IT glossary (Gartner 2019). This is not true even for
uncertainties encountered by geotechnical engineers, a smaller set of data published in the literature that has
and they are most effective when used to organize and been progressively compiled into generic soil/rock data-
quantify these uncertainties for engineering designs bases in recent years. However, this is frequently true for
and decisions.
one specific site. In fact, site-specific data are more chal-
A Recommended Practice DNV-RP-C207 (DNV 2012) lenging to deal with than simply “not enough” and
provides principles, guidance and recommendations for “uncertain”. We now understand that there are at least
use of statistical methods for analysis and representation seven rather than two attributes that define our data.
of soil data. The latest 4th edition of the international Phoon, Ching, and Wang (2019) refer to these attributes
standard “General Principles on Reliability for Struc- as “MUSIC-X”: Multivariate, Uncertain and Unique,
tures” (ISO2394:2015) includes a new Annex D dedi- Sparse, Incomplete, and potentially Corrupted with “X”
cated to the reliability of geotechnical structures. denoting the spatial/temporal dimension. The “unique”
Annex D recognises that geotechnical reliability-based and “potentially corrupted” (in the sense of data contain-
design should place site investigation and the interpret- ing outliers) attributes are very hard problems. It is
ation of site conditions/profile/data as the cornerstone important to emphasise here that the term “uncertainty”
of the methodology (Phoon et al. 2016). Despite these adopted in this paper refers to both imprecision in the
notable advances, it is accurate to say that data plays a knowledge of a particular physical parameter, say
supporting rather than a leading role in decision making undrained shear strength, and the deeper imprecision in
in practice. After all, data have not spoken for them- modeling this imprecise knowledge, say the mean and
selves. Phoon (2017) pointed out that data scarcity standard deviation of the random variable model or
(“curse of small sample size”) is more conspicuous in the scale of fluctuation of the random field model. This
geotechnical engineering than structural engineering. uncertainty of the uncertainty model, commonly
Decision making strategies have evolved to be effective referred to as statistical uncertainty, is notoriously
in such a data poor environment, thus making it even difficult to address for sparse data. This paper shows
harder to monetise data because these predominant that reasonable solutions are available even for our
strategies do not need much data – some such as design exceedingly modest site-specific data. There are also
by prescriptive measures requires almost none. One non-probabilistic solutions (Beer et al. 2013). This is out-
could say there is selection pressure against data in evol- side the scope of this paper.
utionary parlance. All sensible engineers know that generic correlation
This paper covers the estimation of useful statistics models must be used with caution. It is better to adopt
from the original classical univariate setting to a more quasi-local correlation models supported by site-specific
realistic incomplete multivariate setting encountered in data and data from “similar” sites possessing comparable
a typical site investigation programme. It traces the evol- geology. At present, the engineer relies almost entirely on
ution of uncertainties as an inconvenient feature out of his/her experience working on other sites to construct
step with a deterministic world view that could be miti- such quasi-local models. One research challenge is how
gated by adopting a judicious precedent-based cautious to do this algorithmically. This is called the “site chal-
stance, to something entirely undesirable that should lenge” (Phoon 2018). This will potentially extend the
be minimised, to a fact of reality that should be coped search for “similar” sites to anywhere anywhen, beyond
with explicitly using statistics, and to an asset that can regional/municipal databases that the engineer is fam-
be exploited using Bayesian machine learning. The iliar with over the duration of his practice and beyond
focus is not on mathematical gymnastics, but to demon- the deep isolation that human experience encases each
strate that data can produce valuable insights to support and every one of us in because it cannot be shared in
decision making in its own right, over and above its cur- full or with ease. Recent research shows that this site
rent value in physical modelling, ultimate/proof load challenge is tractable even under full MUSIC constraints.
testing, and monitoring. The common lament is that The existence of a quasi-local correlation model that
we do not have enough data to do this. This is not maintains an optimal balance between a generic database
true. We have a lot of data, but they are shelved in design (that may not be directly and completely applicable
 GEORISK: ASSESSMENT AND MANAGEMENT OF RISK FOR ENGINEERED SYSTEMS AND GEOHAZARDS 5

although extremely data rich) and a site database (that is assessment (QRA). One can manage risks cleverly with
fully applicable but extremely data poor) (cf. “Goldilocks experience alone. For example, Terzaghi and Peck
dilemma” in Phoon 2020) remains an open research (1967) cited Kidder-Parker Architects’ and Builders’
question at this point. Handbook (1931) in their Table 54.1 “Soil pressures
allowed by various building codes”. An engineer could
select an allowable bearing pressure based on the “char-
2. Role of data in design
acter of foundation bed” and the name of a city. Table 1
Simple calculations based on a range of variables are bet- of BS8004 (1986), “Presumed allowable bearing values
ter than elaborate ones based on limited input. Professor under static loading” provides a range of bearing values
Ralph Peck’s Legacy Website (Geoengineer 2019)
for each type of rock and soil. Qualitative information
All engineers make decisions in the face of uncertainty. such as “strong limestones and strong sandstones” and
After all, why would we need a factor of safety if we “schists and slates” for rocks and “firm clays” and “soft
have omniscient access to perfect knowledge and infor- clays and silts” for cohesive soils is sufficient. Section
mation? Einstein and Baecher (1982) put this across 2.5 of Eurocode 7 describes design by prescriptive
succinctly: measures (EN 1997−1:2004). These presumed/pre-
scribed values are conservative, but this is a sensible
In thinking about sources of uncertainty in engineering
approach to risk management in the absence of site-
geology, one is left with the fact that uncertainty is inevi-
table. One attempts to reduce it as much as possible, but specific data and calculation models.
it must ultimately be faced. It is a well-recognised part of In our current practice, building regulations typically
life for the engineer. The question is not whether to deal mandate minimum ground investigation at a specific site,
with uncertainty but how? for example, the number of boreholes should be the greater
Uncertainty, in its broadest sense, can range from known of (i) one borehole per 300 m2 or (ii) one borehole at every
unknowns where some knowledge and/or data exist for interval between 10 and 30 m, but no less than 3 boreholes
characterisation to unknown unknowns where alternate in a project site. It is possible to “predict” a reasonable site-
strategies such as robust or resilient design could be specific response (say bearing pressure) based on such lim-
more applicable. Phoon (2017) opined that in between ited information through the mediation of a physical
“white swans” (known unknowns) and “black swans” model and a healthy dose of engineering judgment. We
(unknown unknowns), there will be “grey swans” cover- need some site-specific data to use this approach. We
ing events that are foreseeable even in the absence of data also need the “right” kind of data, which is chained to the
or unforeseeable events that do not result in dispropor- input side of the model. What is “right” for one model
tionate consequences. In all likelihood, the factor of may not be “right” for another model. Lambe (1973)’s Ran-
safety is intended to cover only white and possibly kine Lecture explains the need to calibrate both data and
some grey swans. model together. In addition, a factor of safety and experi-
Casagrande (1965)’s classic paper and Terzaghi Lec- ence are still needed as noted by Burland (1987) in his
ture on “calculated risk” remains relevant to our practice. Nash Lecture. One may conclude that engineers are clever
One should not confuse managing risk in the broad sense in adopting risk management strategies that work with the
articulated by Casagrande with quantitative risk imperfect knowledge and the limited information they
have at hand. There is no doubt that we have been success-
ful. Failures are rare.
Table 1. Three-tier classification scheme of soil property So, what has changed? Some claimed that our internet
variability for reliability calibration (Source: Table 9.7, Phoon traffic has exceeded a zettabyte (1021 bytes or roughly the
and Kulhawy 2008). number of sand grains on all the beaches on the planet)
Geotechnical parameter Property variability COV (%)
throughput per year as of 2016. We may not have a zet-
Undrained shear strength Lowa 10–30
Mediumb 30–50
tabyte currently, but we have a lot of data. They are just
Highc 50–70 not directly useful such as not specific to the site of
Effective stress friction angle Lowa 5–10 interest or need to be transformed to fit the input
Mediumb 10–15
Highc 15–20 side of a physical model. Needless to say, they will
Horizontal stress coefficient Lowa 30–50 also be imperfect in the sense of being uncertain,
Mediumb 50–70
Highc 70–90 incomplete, and possibly even corrupted to some
a
Typical of good quality direct lab or field measurements. extent. Phoon, Ching, and Wang (2019) coined the
b
Typical of indirect correlations with good field data, except for the standard term Big Indirect Data (BID) to refer to any data that
penetration test (SPT).
c
Typical of indirect correlations with SPT field data and with strictly empirical are potentially useful but not directly applicable to
correlations. the decision at hand. All engineering decisions are
6 K.-K. PHOON

ultimately black and white, be it choosing the dimen- has limits in dealing with complex real world processes.
sions of a structure, time interval between maintenance, One should be mindful that real data emerge from such
or issuance of an evacuation notice, notwithstanding complex processes, not from idealised physical models.
the imperfect nature of our data, methods, and under- Another limitation is that a physical model cannot
standing of reality. The adjective “useful” is used in the “learn” on its own to be better as data and problem scen-
context of supporting such real world decisions. The arios evolve in real time.
generic soil/rock and load test databases presented in Actually, in the author’s opinion, our practice is more
the next section will be one type of BID. Monitoring accurately described as pure white box, because it does
data is another BID. not admit uncertainty explicitly. Although somewhat
It is timely to ask ourselves how existing strategies exaggerated, one could argue that we are philosophically
that are tailored to work effectively in a data poor aligned to Laplace’s Demon who famously said (Laplace,
environment can monetise BID. First, there is no mech- not the hypothetical demon) in his “A Philosophical
anism to update presumptive bearing pressures or fac- Essay on Probabilities”:
tors of safety using data. There is a formal mechanism
to do this for resistance factors that are calibrated from We ought then to regard the present state of the universe
statistics (e.g. Phoon, Kulhawy, and Grigoriu 2003; as the effect of its anterior state and as the cause of the
one which is to follow. Given for one instant an intelli-
Paikowsky et al. 2004; Fenton et al. 2016; Tang and
gence which could comprehend all the forces by which
Phoon 2018). There is no space to engage in a full discus- nature is animated and the respective situation of the
sion on reliability-based design, simplified or otherwise, beings who compose it – an intelligence sufficiently
but it is reasonable to lean towards mechanisms that are vast to submit these data to analysis – it would embrace
responsive to data and self-improve with data, particu- in the same formula the movements of the greatest
larly if we have zettabytes coming our way. Second, a bodies of the universe and those of the lightest atom;
for it, nothing would be uncertain and the future, as
physical model erects a significant computational barrier the past, would be present to its eyes.
between input and all other data. For example, monitor-
ing data constitutes the basis for our observational We do go to some lengths to collect good quality data
approach. However, it is assuredly “non-input” data. Sys- directly relevant to our physical models and if this is not
tem identification techniques are needed to back-calcu- possible, we apply empirical rules to be conservative. Do
late the equivalent input data before updated we collect only high quality data that are limited in quan-
predictions are possible. For a large 3D finite element tity or only lower quality data in larger quantity? Do we
model, forward calculations (inputs to outputs) can combine them? When shouldn’t they be combined? The
take days using current computers. Backward calcu- jury is still out on these important questions, but there is
lations (outputs to inputs) will take longer. It is difficult no hope to make progress without an explicit uncertainty
to leverage on our powerful physical models to glean model. Scott A. Barnhill left the following message on
deeper insights from monitoring data, particularly for Professor Ralph Peck’s legacy website: “Perhaps engin-
risk management of big projects in real time. eers trained in geology have an advantage. They are
System identification in its most general form can link more likely to accept mother nature as she exists, rather
inputs and outputs without the mediation of a physical than as created in the mind of the engineer” (Geoengi-
model (black box approach as opposed to the physics- neer 2019). This practical wisdom needs reinforcing if
based white box approach). The most common example we would like to engage emerging digital technologies
of system identification in geotechnical engineering is an with greater haste. The Institution of Civil Engineers
artificial neural network (Shahin, Jaksa, and Maier 2001; (UK) State of the Nation Report (2017) observed that
Jaksa, Maier, and Shahin 2008). Interestingly, machine- “the infrastructure sector has been slow to engage with
learning methods are also data-, rather than physics-dri- the uptake of new digital technologies compared with
ven, in part because they accommodate all data, whether other industries”.
they are good or not so good quality, or right or wrong fit Let me call the future of geotechnical engineering
to a physical model. In fact, some machine-learning (unimaginatively) as Geo 4.0. If Geo 4.0 needs to operate
methods have been successful even when the data used in an overwhelmingly data-rich cyber-physical environ-
have been judged “useless” by a human expert. Such a ment, it is reasonable to question if a pure white box
physics-free and judgment-neutral approach can be approach will continue to be a winning strategy. A
exceedingly powerful as demonstrated by Google’s black box or grey box (physics-informed data-driven)
AlphaGo project. It is interesting that our practice is approach may be more effective. The point is not to be
almost entirely dominated by a white box approach, philosophical, but to be pragmatic. The statistician
although it is known in many fields that this approach George Box once said: “All models are wrong, but
 GEORISK: ASSESSMENT AND MANAGEMENT OF RISK FOR ENGINEERED SYSTEMS AND GEOHAZARDS 7

some are useful”. This seems like a good adage to follow random variable model (measurements at different
when we explore how physics, data, and experience can depths are independent) to a random field model
be combined in even more clever ways to support (measurements at different depths are correlated). The
decision making. theoretical application of a random field in geotechnical
Peck (1980) adopted the intriguing question “Where engineering was popularised in an earlier paper by Van-
has all the judgment gone?” as the title for his fifth Laur- marcke (1977). Other seminal contributions were also
its Bjerrum memorial lecture. The honest answer nowa- made by pioneers such as Wilson Tang (Tang 1984;
days is no one knows. Philosophers are asking the same Lacasse, Liu, and Nadim 2017), Tien H Wu (Wu et al.
question as the power of artificial intelligence expands 1989, 1996; Baecher and Christian 2019), Harr (1987),
beyond what was previously thought to be reachable Lacasse and Nadim (1996), Gregory Baecher and John
only by the human mind. The Leverhulme Centre for Christian (Baecher 1987; Baecher and Christian 2003),
the Future of Intelligence (http://lcfi.ac.uk/) was estab- Herbert Einstein (Einstein and Baecher 1983; Einstein
lished for this reason. One can safely say that machine- et al. 1996), and many others. The author is unable to
human interactions will be transformed in unimaginable do even partial justice in this cursory overview of more
ways and our human minds will be enhanced (to put it than five decades of work in trying to coax data that
mildly) as part of this transformation. At a more mun- are 100% accurate (within measurement limits) to say
dane level, it is already possible for Bayesian methods something useful about the unknown state of the ground
to learn some limited aspects of expert judgment, between measured locations. It goes without saying that
which will vastly expand the sharing of this “digitized these ground truths can only be approximate (no free
experience” beyond what we can do with conventional lunch!) and an explicit uncertainty model is again
“on the job” training at the individual level (Vick 2002). necessary.
The next two sections will briefly discuss: (1) generic Over the years, the terms random field and spatial
databases (BID) to provide an overview of the attributes variability have become synonymous, although the two
of geotechnical data and (2) preliminary research to concepts are distinct. The former is a mathematical
address the “site challenge” under the realistic data con- model. The latter is a description of a geological reality.
straints of MUSIC as an example of what Bayesian There is no guarantee that a random field model, particu-
machine learning could do. larly the common second-order (stationary) version fully
described by an autocorrelation function, is an adequate
representation of this reality for all geologic settings.
3. Generic databases
The predominance of a second-order field model in the
Numbers have an important story to tell. They rely on literature arises in part from the difficulty of characterising
you to give them a clear and convincing voice. higher-order fields with limited data. Theoretical higher-
Stephen Few
order fields do exist (Shields and Kim 2017). Phoon,
Lumb (1966)’s classic paper on “The Variability of Natu- Ching, and Wang (2019) pointed to several other limit-
ral Soils” showed that the variations in the properties of ations of the widely used second-order field model.
four typical Hong Kong soils (a soft marine clay, an allu- While observing that this model is a closer match to reality
vial sandy clay, a residual silty sand, and a residual clayey compared to the independent and identically distributed
silt) about a mean trend can be characterised as random (i.i.d.) model, the authors observed that
variables following distributions such as normal, lognor-
mal, and bi-normal distributions. The ensuing body of it cannot be applied in its most general non-stationary
form because we do not have sufficient site investigation
work on soil properties was published in diverse venues
data for statistical characterization. The current practice
such as ICASP, ASCE symposiums (e.g. Characterisation is to assume a trend function can be removed from the
of soil properties: bridge between theory and practice, data and the residuals are second-order stationary
Atlanta, Georgia, 1984; Uncertainty in the geologic within a typical site. The reason for this assumption is
environment: from theory to practice, Madison, Wiscon- that pairs of measurements regardless of where they
sin, 1996), and reports (e.g. Filippas, Kulhawy, and Gri- are measured can be used to estimate the autocorrela-
tion function. Needless to say, there is no trend, no
goriu 1988; Orchant, Kulhawy, and Trautmann 1988; stationary residuals, and no autocorrelation function
Spry, Kulhawy, and Grigoriu 1988; Kulhawy, Birgisson, in reality. These concepts exist purely within the station-
and Grigoriu 1992; National Research Council 1995; ary random field model.
Vanmarcke and Fenton 2003), before culminating in
The authors further highlighted that
an extensive compilation of univariate statistics by
Phoon & Kulhawy (1999a, 1999b) that further general- trend removal can be difficult (Ching, Wu, and Phoon
ised the characterisation of natural variations from a 2016; Ching et al. 2017; Ching and Phoon 2017).
8 K.-K. PHOON

Estimation of random field parameters is also computa- which gives us a fair chance of swaying more business-
tionally challenging (Tian et al. 2016; Wang H. et al. oriented clients to accept site investigation as an invest-
2018; Xiao et al. 2018). Fine details of the autocorrela- ment rather than a cost (Ching, Phoon, and Yu 2014).
tion function such as sample path “smoothness” are
important (Ching and Phoon 2019a). Characterization The practical value of characterising spatial correlations
of site stratigraphy is a major missing feature of past in the form of a SOF is that the COV of a spatial average
random field studies until quite recently. (Wang, can be reduced, thus permitting higher resistance factors
Huang, and Cao 2013; Ching et al. 2015; Li et al. 2016; to be used for problems governed by spatial averages.
Qi et al. 2016; Wang X. et al. 2016; Wang H. et al. Another practical value is that soil properties at
2017; Wang X. et al. 2018; Cao et al. 2019; Wang
unsampled locations can be interpolated more accurately
H. et al. 2019; Wang X. et al. 2019; Wang, Hu, and
Zhao 2019) using kriging or general regression (Yuen, Ortiz, and
Huang 2016; Yuen and Ortiz 2016, 2018) when spatial
correlations are available. This brief review is not
The difficulties have nothing to do with theory. They intended to be up-to-date or comprehensive on what
have everything to do with statistical characterisation we know about spatial variability, its value, and its
using actual data. Two statistics are needed to describe impact on design. Research has advanced considerably
a second-order (stationary) random field model, namely beyond Vanmarcke’s classic paper in 1977. The inter-
the coefficient of variation (COV) and the scale of fluctu- ested reader can refer to the Joint TC205/TC304 Work-
ation (SOF). The COV is needed to describe the scatter ing Group Report (2017) on “Discussion of statistical/
about the mean trend in the basic random variable reliability methods for Eurocodes”, which has been
model, basically a characteristic amplitude of the fluctu- made available at the ISSMGE TC304 website: http://
ations. The SOF is an additional statistics that roughly 140.112.12.21/issmge/tc304.htm.
describe the distance over which the measurements are The characterisation of geotechnical data has become
strongly correlated in a random field model (DeGroot even more realistic with the compilation of multivariate
and Baecher 1993; Jaksa 1995; DeGroot 1996). It is a databases over the past decade. Momentum is gathering
characteristic wavelength of the fluctuations. If the SOF worldwide to screen, organise, and share our valuable
is much larger than the mobilised volume of soil, the data to hasten the pace of our digital transformation
basic random variable model can be adopted. It is accu- such as Project 304 dB (TC304 2019). Ching, Li, and
rate to say that measurements sampled at a depth inter- Phoon (2016) provided a useful overview of generic
val larger than the SOF can be modelled as independent multivariate databases on soil/rock properties. Table 2
random variables. Or to put this in another way, infor- shows an updated summary of these databases, labelled
mation on spatial variability cannot be captured by a as (geo-material type)/(number of parameters of inter-
sparse sampling grid. From this practical perspective, est)/(number of data points). For example, the CLAY/
the random field model merely allows measurements 10/7490 database consists of 7490 records from 251
sampled at any depth interval, including near continuous studies carried out in 30 countries. Each record contains
cone penetration test soundings, to be modelled with ten clay parameters measured at roughly the same depth,
greater realism. although some may be missing. The CLAY/10/7490
The most complete compilation of COVs for both database is global in coverage. In contrast, the SH-
soils and rocks to date is given by Phoon et al. (2016). CLAY/11/4051 municipal database covers 50 sites in
An updated compilation for the SOF is currently in pro- Shanghai (Zhang et al. 2019). Another source of infor-
gress (Cami, Javankhoshdel, and Phoon 2020). The prac- mation frequently collected comes from pile load tests.
tical value of characterising natural variations in the The performance databases for other geotechnical struc-
form of a COV is that resistance factors can be calibrated tures (in addition to piles) are available, but less com-
more realistically based on our knowledge of soil par- monly reported in the literature. A comprehensive
ameters, such as the three-tier classification scheme of survey of these databases was recently carried out by
soil property variability shown in Table 1. A similar Phoon and Tang (2019). Table 3 includes further
scheme appears in the 2014 edition of the Canadian updates. The following geotechnical structures are cov-
Highway Bridge Design Code (CAN/CSAS614:2014) ered: (1) shallow and deep foundations, (2) offshore
that presents different resistance factors depending on spudcans, (3) mechanically stabilised earth and soil
the “degree of understanding” (low, typical, high) (Fen- nail walls, (4) pipes and anchors (plate, helical, and shor-
ton et al. 2016) and others (e.g. Paikowsky et al. 2004; ing), (5) slopes and base heave, (6) cantilever walls, and
Bathurst, Javankhoshdel, and Allen 2017). For the first (7) braced excavations. Details are given elsewhere
time, we can establish a defensible link between site (Phoon and Tang 2019). Case studies are even more
investigation efforts and the economy of the design, informative, but no systematic compilation has been
Table 2. Summary of some soil/rock property databases (updated from Phoon and Ching 2017).
Range of parameters
Database Reference Parameters of interest # Data points # Sites/studies OCR PI St
′ ′
CLAY/5/345 Ching and Phoon (2012) u , s p, s v
LI, su, sre 345 37 sites 1–4 – Sensitive to quick clays
CLAY/6/535 Ching, Phoon, and Yu (2014) su /s′ v , OCR, (qt − sv )/s′ v , (qt − u2 )/s′ v , (u2 − u0 )/s′ v , Bq 535 40 sites 1–6 Low to very high plasticity Insensitive to quick
clays
CLAY/7/6310 Ching and Phoon (2013, su from 7 different test procedures 6310 164 studies 1–10 Low to very high plasticity Insensitive to quick
2015) clays

GEORISK: ASSESSMENT AND MANAGEMENT OF RISK FOR ENGINEERED SYSTEMS AND GEOHAZARDS
CLAY/10/7490 Ching and Phoon (2014) LL, PI, LI, s′ v /Pa , St, Bq, s′ p /Pa , su /s′ v , (qt − sv )/s′ v , 7490 251 studies 1–10 Low to very high plasticity Insensitive to quick
(qt − u2 )/s′ v clays
CLAY/9/249 D’Ignazio et al. (2019) s′ v /Pa , σv/Pa, s′ v /Pa , qt/Pa, u2/Pa, u0/Pa, PI, wn, St 249 18 sites 1–10 Low to very high plasticity Insensitive to quick
clays
FG/KSAT/4/1358 Feng and Vardanega (2019) e, LL, wn/LL, −ln(ksat) 1358 33 studies – Low to very high plasticity –
′ ′
FI-CLAY/7/216a D’Ignazio et al. (2016) u , s v , s p , wn, LL, PL, St
sFV 216 24 sites 1– Low to very high plasticity Insensitive to quick
7.5 clays
JS-Clay/5/124b Liu et al. (2016) Mr, qc, fs, wn, γd 124 16 Soft to stiff clayey soils and silty clay soils with high variability of
the strength and stiffness characteristicsMr = 12.54–95.82 MPa,
qc = 0.22–3.93 MPa, fs = 0.03–0.14 MPa, wn (%) = 6.91–78.11,
γd = 10.47–19.92 kN/m3
RFG/TXCU-278 Beesley and Vardanega su /s′ v , γ50 CIU, OCR, γ50 CKU 278 21 studies 1–32 Low to medium-high
(2019) plasticity
′
SE-CLAY/4/499 c
Hov et al. (2019) sFV DSS
u , su , s p , LL 499 Sweden s′ p = 13–505 kPa; sFC u = 5–101 kPa; 10–20
sDSS
u = 6–53 kPa; LL = 22–145%
SH-CLAY/11/ Zhang et al. (2019) LL, PI, LI, e, K0, s′ v /Pa Su /s′ v(UCST) , St(UCST), Su /s′ v(VST) , St(VST), 4051 50 sites Normal consolidated to slightly over-consolidated clay; Very soft
4051 ps /s′ v (Shanghai) clay (LI = 0.49–2.19) with slight to medium plasticity (PI = 10.4–
26.5) and with medium to high sensitivity (St = 2.7–7.8)
SAND/7/2794 Ching et al. (2017) D50, Cu, Dr, s′ v /Pa , fʹ, qt1, (N1)60 2794 176 studies 1–15 D50 = 0.1–40 mm, Cu = 1–1000 + Dr = −0.1–117%
ROCK/9/4069 Ching et al. (2018) n, γ, RL, Sh, σbt, Is50, Vp, σc, E 4069 184 studies γ = 15–35 kN/m3, n = 0.01–55%σc = 0.7–380 MPa, E = 0.03–
120 GPa
a
F-CLAY renamed as FI-CLAY to follow internet domain for Finland (FI).
b
J-CLAY renamed as JS-CLAY to follow phonetics abbreviation of Jiangsu (JS).
c
SE-CLAY/4/499 based on S. Larsson (personal communications, 2019).
Notes: LL = liquid limit; PL = plastic limit; PI = plasticity index; LI = liquidity index; wn = natural water content; e = void ratio; ksat=saturated hydraulic conductivity; Mr = resilient modulus; qc = cone tip resistance; fs = sleeve
friction; γd = dry density; D50 = median grain size; Cu = coefficient of uniformity; Dr = relative density; σv = vertical total stress; s′ v = vertical effective stress; s′ p = preconsolidation stress; su = undrained shear strength; sFV
u =
′ ′
undrained shear strength from field vane; sre u = remoulded su; fʹ = effective friction angle; St = sensitivity; OCR = overconsolidation ratio, (qt − sv )/s v = normalised cone tip resistance; (qt − u2 )/s v = effective cone tip
resistance; u0 = hydrostatic pore pressure; (u2 − u0 )/s′ v = normalised excess pore pressure; Bq = pore pressure ratio = (u2-u0)/(qt-σv); Pa = atmospheric pressure = 101.3 kPa; qt1 = (qt/Pa) × CN (CN is the correction factor
for overburden stress); (N1)60 = N60×CN (N60 is the N value corrected for the energy ratio); n = porosity; γ = unit weight; R = Schmidt hammer hardness (RL = L-type Schmidt hammer hardness); Sh = Shore scleroscope hardness;
σbt = Brazilian tensile strength; Is = point load strength index (Is50 = Is for diameter 50 mm); Vp = P-wave velocity; σc = uniaxial compressive strength; E = Young’s modulus; γ50 CIU = shear strain to mobilise 0.5su under iso-
tropically-consolidated undrained conditions; γ50 CKU = shear strain to mobilise 0.5(su – τ0); τ0 = initial shear stress; ps = cone tip resistance from CPT which is unique in China without the measurement of pore pressure; SDSS u =
undrained shear strength from direct simple shear test.

9
10 K.-K. PHOON

Table 3. Summary of performance databases for some geotechnical structures (updated from Table 1; Phoon and Tang 2019).
Geotechnical structures Database/reference Data source Test type Geomaterial N
Shallow foundations UML-GTR ShalFound07 (Paikowsky et al. 2010) Global Laboratory/field Cohesionless 549
UML-GTR RockFound07 (Paikowsky et al. 2010) Global Field Rock 122
Akbas (2007) Global Field Cohesionless 400
Samtani and Allen (2018) USA/Europe Field Cohesionless 80
SpreadFound/1026 (Tang et al. 2019) Worldwide Prototype Various 1026
Offshore spudcans Tang and Phoon (2019a) – Centrifuge Clay with sand 159
Drilled shafts (vertical load) Ng et al. (2001) Hong Kong Field Rock/saprolite 38
AbdelSalam, Baligh, and El-Naggar (2015) Egypt Field Various 318
Asem, Long, and Gardoni (2018) Global Field Soft rock 190
DSHAFT (Garder et al. 2012) Iowa, USA Field Various 38
Motamed, Elfass, and Stanton (2016) Las Vegas Valley Field Caliche 41
Stark et al. (2017) Illinois, USA Field Weak rock 155
TxDOT (Moghaddam et al. 2018) Texas Field Various 27
Tang, Phoon, and Chen (2019) Global Field Various 320
Drilled shafts (lateral load) EPRI (Chen and Kulhawy 1994) Global Field Clay/sand 88
Chen and Lee (2010) Global Field Clay/sand 99
Chen, Lin, and Kulhawy (2011) Global Field Clay/sand 40
Marcos and Chen (2013) Global Field Gravel 24
Augered cast-in-place piles Reddy and Stuedlein (2017) USA Field Cohesionless 112
McVay et al. (2016) Florida, USA Field Various 78
Driven piles AAU-NGI (Augustesen 2006) Global Field Various 420
Zhang et al. (2006) Hong Kong Field (static/dynamic) Weathered granite 1514
Long et al. (2009) Wisconsin, USA Field (dynamic) Various 316
PILOT (Roling, Sritharan, and Suleiman 2011) Iowa, USA Field Various 275
PSU (Smith et al. 2011) Global Field Various 322
Long and Anderson (2014) Illinois, USA Field (dynamic) Various 111
ZJU-ICL (Yang et al. 2016) Global Field Sand 117
Long (2016) Wisconsin, USA Field (static/dynamic) IGM 215
Lehane et al. (2017) Global Field Various 120
Adhikari et al. (2018) Wyoming, USA Field Soft rock 25
TxDOT (Moghaddam et al. 2018) Texas Field Various 33
Tang and Phoon (2018a, 2018b, 2019b) Global Field Various 783
Helical piles Tang and Phoon (2018c, 2019c) Canada/USA Field Various 1010
Driven cast-in-situ piles Long (2013) Wisconsin, USA Field Various 182
Flynn (2014) United Kingdom Field Sand 116
Pile foundations FHWA DFTLD (Abu-Hejleh et al. 2015) Mainly in USA Field Various 1567
Dithinde et al. (2011) South Africa Field Various 174
IFSTTAR (Burlon et al. 2014) France Field Various 174
Niazi (2014) Global Field Various 330
Galbraith, Farrell, and Byrne (2014) Ireland Field Various 175
AUT-CPT (Moshfeghi and Eslami 2018) Global Field Various 466
WBPLT (Chen et al. 2014) Global Field Various 613
LADOTD (Rauser and Tsai 2016) Louisiana, USA Field (static/dynamic) Various 1465
Nanazawa et al. (2019) Japan Field Various 441
Micropiles Almeida and Liu (2018) Canada Field Ontario soils 47
Foundations EPRI (Kulhawy et al. 1983) USA Field Various 804
Mechanically stabilised earth walls Huang and Bathurst (2009) – Laboratory Cohesionless 318
Miyata and Bathurst (2012a) Japan Laboratory/in situ Cohesionless 652
Miyata and Bathurst (2012b) Japan Laboratory Various 503
Miyata, Bathurst, and Allen (2014) Japan Laboratory N/A 362
Miyata and Bathurst (2015) Japan Field Various 520
Miyata and Bathurst (2019) Global In situ Cohesionless 113
Allen and Bathurst (2018) – Field Various 378
Miyata, Yu, and Bathurst (2018) – In situ/laboratory Various 202
Wood et al. (2012a, 2012b) Texas, USA Laboratory Cohesionless 650
Soil nail walls Lazarte (2011) – Field – 166
Cheung and Shum (2012) Hong Kong Field CDG/CDV 913
Lin, Bathurst, and Liu (2017) Global In situ – 123
Liu et al. (2018) – In situ – 95
Yuan et al. (2019) China In situ Various 144
Multi-anchor walls Miyata, Bathurst, and Konami (2011) Japan In situ Various 28
Slopes Travis, Schmeeckle, and Sebert (2011) Global Field Various 157
Bahsan et al. (2014) – Field Clay 43
Excavations (base heave) Wu, Ou, and Ching (2014) Global In situ Cohesive 24
Pipes White, Cheuk, and Bolton (2008) – Small/full-scale Sand 61
Stuyts, Cathie, and Powell (2016) – Small/full-scale Sand 108
Ismail, Najjar, and Sadek (2018) – Small scale/centrifuge Sand 143
Plate anchors White, Cheuk, and Bolton (2008) – Small/full-scale Sand 54
Plate anchors Stuyts, Cathie, and Powell (2016) – Small/full-scale Sand 192
Helical anchors Tang and Phoon (2016) – Laboratory Cohesive 78
Field 25

(Continued )
 GEORISK: ASSESSMENT AND MANAGEMENT OF RISK FOR ENGINEERED SYSTEMS AND GEOHAZARDS 11

Table 3. Continued.
Geotechnical structures Database/reference Data source Test type Geomaterial N
Shoring anchors Chahbaz, Sadek, and Najjar (2019) Beirut Field Clay/marl/ 70
limestone
Cantilever wall Phoon et al. (2009) – Centrifuge Sand 20
Excavation (stability) Marsland (1953) – Small-scale Loose/dense sand 23
Large-scale 10
Excavation (wall displacement) Long (2001) Global Field Various 296
Moormann (2004) Global Field Soft soil 530
Wang, Xu, and Wang (2010) Shanghai Field Soft soil 300
Wu, Ching, and Ou (2013) Taipei Field Soft clay 22
Notes: CDG = completely decomposed granite; CDV = completely decomposed volcanic; IGM = intermediate geomaterial; N = number of load tests; NUS =
National University of Singapore; UWA = University of Western Australia; ZJU = Zhejiang University; ICL = Imperial College London.

carried out perhaps with the exception of liquefaction sites in an acceptable way. The million dollar question
(Andrus, Stokoe, and Chung 1999; Cetin et al. 2004; (literally, if one were to consider how many mandatory
Moss et al. 2006; Idriss and Boulanger 2010; Ku et al. site investigations are carried out worldwide in any
2012; Juang, Ching, and Luo 2013; Kayen et al. 2013). time period) is whether we can get more value from
We have a lot of data, but our data is mainly “dark” site data beyond establishing generic transformation
because it is typically not exploited to provide insights models.
or to support decision making once the project that pro-
duced the data is completed.
For soil/rock properties, the most basic design decision 4. Value of data
in current geotechnical practice is to estimate their values Data is not information, information is not knowledge,
from other test results, typically field test results. Phoon knowledge is not understanding, understanding is not
and Kulhawy (1999a) identified at least three sources of wisdom.
Clifford Stoll
uncertainties in a comprehensive statistical study of a
broad range of laboratory and field test data: (1) spatial Geotechnical data are often referred to as “uncertain” in
variability, (2) measurement errors (including statistical the qualitative sense of “I don’t know”, rather than with
uncertainty due to limited measurements), and (3) trans- a mathematical formalism in mind. In fact, it is accu-
formation uncertainty. The third source of uncertainty rate to say that the majority harbours the sentiment
arising from transforming a soil/rock parameter such as that formalism such as statistics is not possible, because
the overconsolidation ratio to a design parameter such of data scarcity. One may venture to guess that the slow
as the normalised undrained shear strength can be signifi- progress in geotechnical reliability-based design or
cant as shown in the data scatter in Figure 1. other formal risk-informed design methodology is par-
Although it is widely known that local transformation tially impeded by the lack of an accurate understanding
models (dashed lines in Figure 1) are preferred to those of the attributes of geotechnical data beyond broad gen-
calibrated from a generic database such as CLAY/10/ eralities such as “uncertain” and “scarce”. In fact, many
7490, there are no methods to quantify this “site effect” practitioners do not appreciate the power of statistics in
for data routinely collected in a typical project (in con- its unusual ability to quantify even the uncertainties in
trast to data specially collected for a research study). the models it posits. In short, statistics can quantify the
Nonetheless, the variety of dashed lines, each referring precision limits of its own models consistently based on
to a local transformation model, clearly shows that this the available data at hand. The National Research
site effect is important. The need for building regulations Council (1995) made this point rather clearly: “The
to mandate a site investigation in every project is a rec- lack of a large data set does not preclude the use of
ognition that every site is unique to some degree. Clearly, probability theory. Probability theory can be used to
geotechnical data are “uncertain” and “unique” to some evaluate the uncertainties involved in working with
extent. The former characteristic is more familiar and meager information”. Nonetheless, this deep insight
better studied. has been lost, as practitioners continue to express
Building regulations do not permit site investigation some reservations under the misconception that there
efforts to vary as a function of how much is known at is insufficient data to characterise a probability model,
neighbouring/comparable sites, even if the sites were to such as its parameters (mean, COV) and its shape (nor-
be adjacent to the site of interest, possibly because mal, lognormal, beta, etc.). Many miss the point that
there are no statistical methods that can manage uncer- even ignorance can be approximately quantified. This
tain, sparse, and somewhat unique data from different is exceedingly powerful, because it allows an engineer
12 K.-K. PHOON

Figure 1. Correlation between normalised undrained shear strength (su /s′ v ) and overconsolidation ratio (OCR) (Ching and Phoon
2019b).

to weigh the cost of making a decision against the cost data. The screening for extremes or outliers and spatial
of collecting more information to reduce imprecision in variability are clearly important, but these aspects are
some aspects of the problem. This paper does not cover not covered in this paper. Table 4 is a site-specific
sensitivity analysis, but its usefulness to decision mak- example of an actual MUSIC database from Taipei.
ing is clear. A naïve one-at-a-time deterministic sensi- With the exception of the mobilised undrained shear
tivity analysis can produce misleading results for a strength [su(mob)], each column contains the results
number of reasons, but one simple reason would be from a different and independent test. Good practice
the inability to take care of dependencies in determinis- requires a suite of tests to be conducted for cross-vali-
tic analysis. Dependency is a feature of all multivariate dation, identification of layer boundaries, estimation of
real world data (Ching, Li, and Phoon 2016). It is most design properties, and others. Hence, geotechnical data
commonly captured by a correlation coefficient in basic are intrinsically “multivariate” in nature. There is an
statistics (Ching, Phoon, and Li 2016). obvious tradeoff between conducting different tests in
Phoon (2018) suggested that the attributes of geo- different locations and conducting different tests in
technical data can be succinctly described as MUSIC: the same location. The former strategy collects more
Multivariate, Uncertain and Unique, Sparse, and information on the spatial variability of the site. The
InComplete. It is useful to clarify in passing that latter strategy collects information on the cross-corre-
“scarce” refers to a small number of measurements, lations among all tests. In practice, it is common to
while “sparse” refers to a small number of measure- adopt an intermediate strategy that involves conducting
ments widely distributed in space. Given that all site different test combinations at different depths and
data are situated in space (and sometimes, in time), locations. The greyed out cells in Table 4 denote absent
the term “sparse” is more descriptive as it is unlikely measurements. Hence, geotechnical data are typically
for measurements to be taken at one corner of a site. “incomplete”. A data table without missing entries is
Phoon, Ching, and Wang (2019) further suggested an exception rather than the norm in geotechnical
that MUSIC can be re-interpreted to cover extremes: engineering. It is useful to observe that the greyed out
Multivariate, Uncertain and Unique, Sparse, Incom- cells are not randomly distributed. They occur more
plete, and potentially Corrupted. Ching and Phoon frequently in columns where measurements are more
(2019b) subsequently extended MUSIC to MUSIC-X, costly and the percentage of absent measurements can
where “X” denotes the spatial/temporal context of the be very high in these columns.
 GEORISK: ASSESSMENT AND MANAGEMENT OF RISK FOR ENGINEERED SYSTEMS AND GEOHAZARDS 13

Table 4. Site investigation data for a silty clay layer at a Taipei site (Ou and Liao 1987).
Test results
′
Depth (m) 2
su (kN/m ) su(mob) (kN/m ) 2
LL PI LI s v /Pa s′ p /Pa su (mob)/s′ v qt1
12.8 UU 55.2 46.9 30.1 9.1 1.20 1.26 1.71 0.37 3.35
14.8 VST 50.7 52.9 32.8 12.8 1.43 1.43 0.36 3.34
16.1 UU 61.9 51.7 36.4 14.5 1.24 1.54 0.33 3.15
17.8 UU 54.2 42.8 41.9 18.9 0.90 1.68 1.79 0.25 2.74
18.3 VST 59.5 59.3 1.72 0.34 2.76
20.2 UU 73.1 60.5 38.1 17.3 0.70 1.88 0.32 2.73
22.7 VST 63.3 64.4 37.0 16.0 0.58 2.08 0.31 2.97
24.0 UU 82.2 67.5 38.0 16.2 0.75 2.19 2.19 0.30 2.80
26.6 UU 98.1 82.1 34.8 13.8 0.80 2.41 0.34 3.92
Note: LL = liquid limit; PI = plasticity index; LI = liquidity index; s′ v = vertical effective stress; s′ p = preconsolidation stress; Pa = atmospheric pressure = 101.3 kPa;
qt = (corrected) cone tip resistance; qt1 = (qt − sv )/s′ v ; su = undrained shear strength; su(mob) = mobilised su values (Mesri and Huvaj 2007).

Each row (record) in a MUSIC database such as Table beyond the original boundary conditions. ‘Local’ cali-
4 refers to data collected at the same depth from different brations, where available, are to be preferred over the
tests conducted in close proximity. There are only n = 9 broad, generalized correlations. (Kulhawy and Mayne
1990)
rows in Table 4. Hence, geotechnical data are “sparse”.
However, this is not true for a generic database such as Notwithstanding this sensible caveat, the engineer is typi-
CLAY/10/7490. The CLAY/10/7490 database consists cally left with no recourse but to use these generalised cor-
of n = 7490 records from 30 countries for ten clay par- relations in the absence of “local” versions. Hence, BID is
ameters. In short, site-specific data can be sparse, but already routinely used in practice in the form of Figure 1.
generic data are not sparse, although they may not be One could surmise that it has some real value. The
complete and directly applicable to a specific site. To research challenge is to distil more value out of BID.
counter the prevalent sentiment that there is no big
data in geotechnical engineering, Phoon, Ching, and
Wang (2019) explicitly refer to any data that are poten- 4.1. Bayesian machine learning
tially useful but not directly applicable to the decision at Can we address some attributes of MUSIC-X based on
hand as Big Indirect Data (BID). A generic database will the meagre information we have at hand? This is argu-
be one type of BID. A compilation of case studies can be ably the central question that practitioners are most
regarded as another type of BID. interested to know. The short answer is yes. Recently,
Although site effects are well known, they are mainly Ching and Phoon (2019c) proposed a novel Bayesian
characterised in research studies through a testing machine learning method to do this, namely to construct
programme that is more detailed than what is routinely a site-specific distribution function for a MUSIC data-
carried out in practice and for rather distinctive geo- base such as that shown in Table 4. Each database is
materials. Kulhawy and Mayne (1990) pointed out that effectively a table with m columns representing soil par-
“comprehensive characterization of the soil at a particular ameters (Y1, Y2, … , Ym) and n rows representing
site would require an elaborate and costly testing measurements at different depths. The observed data
programme, well beyond the scope of most project bud- are denoted by Y o and unobserved data denoted by
gets”. To the knowledge of the author, no one has quan- Y u. Because soil parameters can be highly non-normal,
tified site effects numerically based on more routine data Ching and Phoon (2015) adopted an analytical trans-
such as those shown in Table 4 commonly collected at a formation based on the Johnson distribution to convert
project level. In practice, site effects are broadly appreciated (Y1, … , Ym) to approximately normal data. The approxi-
based on geology, soil mechanics, and experiences at com- mately normal data are denoted by x = (X1, … , Xm)T,
parable sites, rather than characterised quantitatively where “T” refers to vector/matrix transpose. A key
through a detailed multivariate analysis of the site data assumption made in Ching and Phoon (2019c) is that
that meets MUSIC constraints. The typical caveat included x at a certain depth follows the multivariate normal
in design guides would include a general statement such as probability density function (PDF):
caution must always be exercised when using broad, 1 m
generalized correlations of index parameters or in-situ −
−
f (x|ms , Cs ) = |Cs | 2 (2p) 2
test results with soil properties. The source, extent, limit-

ations of each correlation should be examined carefully 1 T −1
before use to ensure that extrapolation is not being done × exp − (x − ms ) Cs (x − ms ) (1)
2
14 K.-K. PHOON

The multivariate normal PDF has mean vector = μs and 9 site-specific data points are available, it can be seen that
covariance matrix = Cs; the subscript “s” is to highlight the site-specific PDF (solid grey markers for MUSIC-X
that μs and Cs are “site-specific”. Because site-specific and open black markers for MUSIC) is less scattered
data are sparse (small n), it is technically challenging to and arguably more informative for this particular site
estimate μs and Cs using conventional methods such as than the generic PDF (blue markers). The MUSIC and
matching moments or maximising likelihood. It is also MUSIC-X PDFs are similar, because the sampling inter-
very challenging to estimate the statistical uncertainties vals in Table 4 are large and spatial variability is thus not
associated with μs and Cs, which are significant for a well captured. Figure 2(d) further shows that Su /s′ v is
fairly typical record size of around 10. Ching and less correlated to qt1 at the Taipei site than the generic
Phoon (2014) pointed out that it is impossible to guaran- version. The generic correlations in CLAY/10/7490 are
tee Cs to be positive definite when the multivariate data is 0.91, −0.57, −0.50, and 0.73 for the transformation
incomplete. To address these critical limitations, Ching models shown in Figure 2(a–d), respectively, in standard
and Phoon (2019c) developed a novel Gibbs sampler normal space (Ching and Phoon 2014). To the author’s
(GS) to overcome this long standing challenge. The key knowledge, this learning algorithm (MUSIC or
idea is to treat μs, Cs, and x u (transformed from Y u) as MUSIC-X) is the first of its kind. Even when used by
unknown random quantities and to sequentially sample itself, the site-specific PDF can guide the engineers to
one random quantity at a time from distributions con- select conservative design values more appropriate for
ditioned on the rest of the quantities and the observed a particular site by using the approximate lower bounds
data x o (transformed from Y o) using GS. Simulation is of the solid grey markers (rather than generic blue mar-
practical because these conditional probabilities are kers) shown in Figure 2. If these values were ascertained
available in closed-form for suitably chosen conjugate to be overly conservative, more measurements could be
priors. While Bayesian methods are known to be very taken and the favourite question “how many measure-
powerful, there is no acknowledgment in the literature ments are enough” can be addressed quantitatively by
that these methods are very complex and computation- the reduction in the scatter of the solid grey markers
ally intensive. Excessive emphasis on what works in prin- that will improve the lower bounds. This MUSIC or
ciple rather than what works in practice will not attract MUSIC-X PDF is basically a quantification of site
more users. This GS has been generalised to MUSIC-X “uniqueness” from a data-informed perspective and
recently (Ching and Phoon 2019b). there is clear value to do this. Once site uniqueness can
Consider properties at a new depth (xnew) that does be captured numerically, it opens all kinds of interesting
not appear in the training data previously used in the research avenues to combine site-specific data with gen-
GS. Based on the total probability theorem, the con- eric data from other sites in the entire world, not merely
ditional multivariate PDF f(xnew|X o) is a mixture of in the restricted region that an engineer practices in. The
multivariate normal PDFs: similarity index approach outlined below is one such
example.

new |X ) = f (x new |m s , Cs )f (m s , Cs |X 0 )dm s dCs
0
f (x
  4.2. Similarity index approach
1 
T
≈ N(x new |m s.t , Cs,t ) The next natural question is how to combine a site-
T − tb t=tb +1 specific PDF with a generic PDF in a more discriminate
(2) way that accounts for site “uniqueness”. Ching and
Phoon (2019d) developed a similarity index (S) based
where (μs,t, Cs,t) are the GS samples at time step = t; tb is on f(xnew|X o) to identify records from a generic database
the end of the burning-period; and T is the total number that are “similar” to those from a specific site. A second
of GS time steps or samples. The simulation of a site- example of a MUSIC database from Onsøy, Norway is
specific probability distribution appears very compli- given in Table 5. This set of site-specific data is shown
cated to the average engineer, but it can support a critical as red solid triangles in Figure 3 against a background
design decision on how to choose soil/rock properties at of generic data from CLAY/10/7490 shown as grey
a particular site by “learning” from site-specific data solid circles. Figure 3 also presents data from another
alone. An example based on actual data from a Taipei site in Norway (Drammen) from Lacasse and Lunne
site (Table 4) is shown in Figure 2. Although Table 4 (1982). The Drammen and Onsøy sites are roughly
contains 9 records, note that s′ p /Pa is only measured 50 km apart with comparable geologic origins (Lacasse
at 3 depths. It is not surprising that the statistical uncer- et al. 1981; Lacasse and Lunne 1982). The data are ident-
tainties in Figure 2(b) is large. For Figure 2(c or d) where ified as “similar” (S > 1) (black solid circles) or
 GEORISK: ASSESSMENT AND MANAGEMENT OF RISK FOR ENGINEERED SYSTEMS AND GEOHAZARDS 15

Figure 2. Transformation models based on generic data (CLAY/10/7490) and site-specific data (MUSIC and MUSIC-X simulations) for a
Taipei site (Table 4) (J. Ching, personal communications, 2019).
Note: solid and dashed lines are the median and 95% confidence interval for the generic data.

“dissimilar” (S < 1) (black open circles) using this simi- “similarity” to the data at one site, and (3) perform a
larity index. The practical benefit of doing this is that weighted regression on a combined dataset containing
we can replace a generic transformation model such as the site-specific data and the generic database. The
Figure 1 by a quasi-site-specific version that is based equivalent generic sample size (Neq) is the sum of the
on the site-specific data and appropriately weighted gen- weights produced by the generic records. Figure 4 pre-
eric data. Ching and Phoon (2019d) recommended the sents the construction of quasi-site-specific transform-
following procedure to do this: (1) assume the weight ation models using different number of records from
of a site-specific data point is 1, (2) assign weights to Table 5: (a) all 9 records; (b) 6 records (depths = 1.9,
records in a generic database as a function of their 3.5, 5.2, 9.5, 10.8, and 13.4 m); (c) 2 records (depths =

Table 5. Site investigation data for a marine clay layer in Onsøy (Norway) (Lacasse and Lunne 1982).
Site-specific data Y
Index Depth (m) LL PI LI s′ v /Pa s′ p /Pa su (mob)/s′ v St Bq qt1 qtu OCR
1 1 56.2 20 1.54 0.06 0.85 2.03 6 0.16 29.11 25.57 13.99
2 1.9 50.2 18.1 1.82 0.12 0.6 0.91 14 0.24 17.69 14.58 5.2
3 3.5 59.9 30.5 0.93 0.22 0.48 0.48 15 0.3 10.52 8.41 2.26
4 5.2 56.8 22.9 1.07 0.32 0.45 0.37 7 0.35 7.7 6.11 1.42
5 7.6 66.3 31.5 0.87 0.47 0.54 0.24 14 0.47 5.89 4.25 1.17
6 9.5 65.1 29.6 0.97 0.58 0.25 12 0.41 6.19 4.74
7 10.8 74.4 36.1 0.81 0.65 0.84 0.25 9 0.46 5.93 4.31 1.28
8 13.4 71.4 35.8 0.87 0.81 1.05 0.24 0.47 5.95 4.24 1.29
9 16.3 72.7 34.7 0.76 0.99 0.99 0.24 0.55 6.13 3.88 1
Notes: LL = liquid limit; PI = plasticity index; LI = liquidity index; s′ v = vertical effective stress; s′ p = preconsolidation stress; Pa = atmospheric pressure =
101.3 kPa; qt = (corrected) cone tip resistance; qt1 = (qt − sv )/s′ v ; su = undrained shear strength; su(mob) = mobilised su values (Mesri and Huvaj 2007); St
= sensitivity; qt = (corrected) cone tip resistance; u2 = pore pressure behind cone; Bq = pore pressure ratio = (u2-u0)/(qt-σv); u0 = hydrostatic pore pressure;
qt1 = (qt − sv )/s′ v ; qtu = (qt − u2 )/s′ v .
16 K.-K. PHOON

Figure 3. Automatic detection of records from a generic database CLAY/10/7490 that are “similar” to those from a specific site in Onsøy,
Norway (Ching and Phoon 2019d).

1.9 and 13.4 m); and (d) no site-specific data is available. algorithms that can learn from both data and their inter-
Note that s′ p /Pa is not measured at a depth = 9.5 m. actions with the engineers.
Hence, the number of site records for this specific trans-
formation model is Ns = 8 and 5 for Figure 4(a and b),
respectively. It could be seen that the quasi-site-specific 5. Let data speak for themselves
transformation model is more customised to Onsøy
when there are 8 site records but revert back to the gen- Hand (2014) said:
eric form when there are only 2 site records. Note that In general, when building statistical models, we must
the black solid circles are meant for reference only; the not forget that the aim is to understand something
full set of generic records (grey solid circles) with appro- about the real world. Or predict, choose an action,
priate weights is used for regression. This similarity make a decision, summarize evidence, and so on, but
index approach treats the records in a generic database always about the real world, not an abstract mathemat-
ical world: our models are not the reality.
as independent, although a natural grouping based on
site exists. Data records measured within a site tend to Let’s take a step back and ask ourselves why we need a
be more similar with each other than those measured model in the first place. One answer is that we do not
in other sites. Ching, Wu, and Phoon (2020) proposed have sufficient data to make a decision without mediation
a hierarchical Bayesian model to capture this additional by a model. The simplest statistical model is to assume
site information commonly made available in generic data are independent and identically distributed (i.i.d.).
databases. No quantitative information related to the Limited data are needed to characterise this model, but
site location such as GPS location, nearest city, region, it clearly deviates from a reality that exhibits spatial varia-
country and others is needed. Only qualitative knowl- bility. The classical random field model tries to do better,
edge that a group of data records are measured within but it requires more data for statistical characterisation.
the same project site is needed. In the author’s opinion, More recent multiple point methods in geostatistics can
the above research is preliminary, because many struc- consider more than two-point autocorrelation (or
tural elements of a generic database such as groups second-order) information (Mariethoz and Caers
have not been studied and the algorithms are not in a 2015), but they require richer data in the form of
full learning mode, including learning from expert judg- training images. Phoon, Ching, and Wang (2019) opined
ment. However, they are founded on Bayesian theory that a “taxonomy of methods based on the type/amount
and they will set the scene for more powerful algorithms of data available could help guide future development in
to emerge in the near future. It is quite likely that some data-driven algorithms and strengthen a virtuous cycle of
aspects of our engineering experience would be “digi- data collection hardware developing hand in hand with
tized” in the future in the sense of being captured by algorithms”. Wang and Zhao (2016, 2017) explored a
 GEORISK: ASSESSMENT AND MANAGEMENT OF RISK FOR ENGINEERED SYSTEMS AND GEOHAZARDS 17

Figure 4. Construction of quasi-site-specific transformation models by combining different amount of Onsøy data with appropriately
weighted records in CLAY/10/7490 (Ching and Phoon 2019b).
Note: median (solid line) and 95% confidence interval (dashed lines).

new sampling paradigm in digital signal processing de-trending (Wang Y et al. 2019). In addition, the BCS-
called compressive sampling (or sensing, CS) that can KL generator may be readily extended to simulate
reconstruct a near replica of the original signal from a cross-correlated bivariate RFSs (Zhao and Wang 2018).
small number of measurements. Wang, Zhao, and An important open question is whether the basis func-
Phoon (2018) subsequently developed a Bayesian Com- tions in BCS (which are prescribed) can reproduce the
pressive Sampling-Karhunen-Loève (BCS-KL) expan- finer details of some sample paths such as those produced
sion version that can generate random field samples by the Whittle-Matérn autocorrelation function correctly
(RFSs) directly from sparse measurements. This BCS- when there are sufficient measurements (convergence). A
KL generator is shown to be capable of dealing with second related question is whether it can retain its key
much more general non-Gaussian and non-stationary practical advantage of representing such “rough” sample
RFSs, including RFSs with unknown non-stationary paths using sparse measurements (rate of convergence).
auto-covariance structure without explicit estimation of It will be interesting to further explore the possibility of
the autocorrelation function (Montoya-Noguera et al. decoupling BCS from the KL expansion, which contains
2019) and RFSs with unknown trend function without a fairly strong multivariate Gaussianity assumption.
18 K.-K. PHOON

Notwithstanding the above, it is clear that more recent decisions to account for uncertainties explicitly. Geo
non-classical methods can handle much more realistic 4.0 will do a lot more, but we (the entire geotechnical
data up to bivariate vector fields that are potentially engineering community) need to engage in re-imagining
non-stationary and non-Gaussian, without making the future of our profession with greater boldness. So,
strong demands on data such as sample sizes and com- where did engineering judgment go? To quote Professor
pleteness that cannot be fulfilled in practice. Classical Ralph Peck, who is widely recognised for his practical
models have reigned supreme as we have always assumed engineering wisdom: “Theory and calculation are not
one can collect sufficient and appropriate data for charac- substitute for judgement, but are the basis for sounder
terisation. Many models were developed without a judgments” (NGI 2019). Professor Peck may not have
characterisation method in mind and in fact, many imagined digitalisation, but he would agree that better
models exist without a satisfactory characterisation calculations grounded on actual data will make our
method in place even years after they were introduced. decisions even better. In the much longer term, no one
One cannot help but ask if this supposedly minor foot- really knows the role of humans in an immersive
note pertaining to data collection and empirical charac- cyber-physical reality.
terisation thereof is really minor. It will be fruitful to In fact, why don’t we initiate an “AlphaGeo”
ask ourselves the inverse question: what data we have project to see how far we can monetise our data and
and what models can we develop to make full use of to sharpen the role of engineering judgment? One sus-
the data at hand, warts and all? In some sense, these pects (with good reasons) that the story of statistics
non-classical models that can function under very gen- will unfold in exciting and unexpected ways in the near
eral conditions are allowing our data to speak for future. It is not far-fetched to wonder if data may be
themselves. the only reality.

6. Concluding thoughts Acknowledgements

The moral of this story is that the value of geotechnical This paper is an update of the 10th Lumb Lecture, delivered at
data is significantly under-appreciated and not fully the University of Hong Kong, 6 December 2018. The author
exploited for decision making. Our data is “dark” in would like to thank Professor Limin Zhang, Editor in Chief
the sense that it is stored primarily for compliance pur- of Georisk, for his encouragement to prepare this paper. The
author is also grateful to the Department of Civil Engineering,
poses, rather than shared and actively mined for insights The University of Hong Kong and the Geotechnical Division,
that can inform future decision making. The world is The Hong Kong Institution of Engineers for their kind invita-
being revolutionised by new and powerful ways of col- tion to deliver this lecture. In particular, the generous hospital-
lecting, sharing, analysing, and monetising data. Clearly, ity extended by Professor Zhongqi Quentin Yue, Honorary
there is a pressing need for the geotechnical engineering Professor Chack Fan Lee, and Dr Victor Li is deeply appreci-
ated. The author also thanked Dr Victor Li for sharing the
community to engage in this digital transformation.
article: “Excerpts from interview with Professor Peter
There is no doubt that reasoned judgment is further Lumb”, Hong Kong Statistical Society Newsletter, Vol. 9,
enhanced when it is guided by relevant data and analyti- Issue 1, 1986. This paper was drafted during the author’s sab-
cal tools that make the most sensible use of data, be it batical at the Institute for Risk and Reliability, Leibniz Univer-
using physics, statistics, machine learning, or some com- sity, which was funded by the Alexander von Humboldt
binations thereof. One Geo 4.0 approach could be to Foundation. Last but not least, the author is deeply indebted
to Prof Jianye Ching, National Taiwan University, for sharing
develop a clever physics-informed and data-driven his many deep insights and research in Bayesian learning and
“grey box” algorithm to shortlist “similar” sites from for preparing all the figures, to Dr Chong Tang for his exten-
BID for the engineer to further refine based on his/her sive editorial assistance, and to the following colleagues for
experience and for the algorithm to “learn” and thus their invaluable comments: Zijun Cao, Marco D’Ignazio (for
become even more discriminating in the future. In this updating CLAY/9/249), Sina Javankhoshdel, C. Hsein Juang,
Leena Korkiala-Tanttu, Tim Länsivaara, Stefan Larsson (for
way, our collective experiences could be partially “digi-
updating SE-CLAY/4/499), Andy Yat-fai Leung, Monica Löf-
tized” as well. For the first time, we are starting to realise man, Sukumar Pathmanandavel, Anders Prästings, Mengfen
we can combine experience and data in even more clever Shen (for updating liquefaction databases), Yu Wang (for dis-
ways. This paper describes some preliminary research on cussions on Bayesian compressive sampling), and Dongming
Bayesian learning that grew out of a simple idea studied Zhang (for updating SH-CLAY/11/4051).
by pioneers such as Professor Peter Lumb since the
sixties. The idea is that uncertainties and even limited
knowledge of these uncertainties can be quantified Disclosure statement
numerically using statistics, thus allowing design No potential conflict of interest was reported by the author.
 GEORISK: ASSESSMENT AND MANAGEMENT OF RISK FOR ENGINEERED SYSTEMS AND GEOHAZARDS 19

ORCID Alternative Approaches in a Geotechnical Engineering

Context.” Structural Safety 41: 1–10.
Kok-Kwang Phoon http://orcid.org/0000-0003-2577-8639 Beesley, M. E., and P. J. Vardanega. 2019. “Parameter
Variability of Undrained Shear Strength and Strain Using
References a Database of Reconstituted Soil Tests.” Canadian
Geotechnical Journal, in press.
AbdelSalam, S., F. Baligh, and H. M. El-Naggar. 2015. “A BS8004. 1986. Code of Practice for Foundations. London:
Database to Ensure Reliability of Bored Pile Design in British Standards Institution.
Egypt.” Proceedings of the Institution of Civil Engineers – Burland, J. B. 1987. “The Teaching of Soil Mechanics – a
Geotechnical Engineering 168 (2): 131–143. Personal View.” Proceedings, 9th European Conference on
Abu-Hejleh, N., M. Abu-Farsakh, M. Suleiman, and C. Tsai. 2015. Soil Mechanics and Foundation Engineering, Dublin, 3
“Development and Use of High-Quality Databases of Deep vols., 1427–1447.
Foundation Load Tests.” Transportation Research Record: Burlon, S., R. Frank, F. Baguelin, J. Habert, and S. Legrand.
Journal of the Transportation Research Board 2511: 27–36. 2014. “Model Factor for the Bearing Capacity of Piles
Adhikari, P., Y. Gebreslasie, K. Ng, T. Sullivan, and S. Wulff. from Pressuremeter Test Results: Eurocode 7 Approach.”
2018. “Static and Dynamic Analysis of Driven Piles in Soft Géotechnique 64 (7): 513–525.
Rocks Considering LRFD Using a Recently Developed Cami, B., S. Javankhoshdel, and K. K. Phoon. 2020. “A Review
Electronic Database.” In Installation, Testing, and Analysis of Scale of Fluctuation for Spatially Varying Soils:
of Deep Foundations (GSP 294), 83–92. Reston, VA: ASCE. Estimation Methods and Values.” ASCE-ASME Journal of
Akbas, S. 2007. “Deterministic and Probabilistic Assessment of Risk and Uncertainty in Engineering Systems, Part A: Civil
Settlements of Shallow Foundations in Cohesionless Soils.” Engineering, under review.
Ph.D. thesis, Cornell University Cao, Z., S. Zheng, D. Q. Li, and K. K. Phoon. 2019. “Bayesian
Allen, T., and R. Bathurst. 2018. “Application of the Simplified Identification of Soil Stratigraphy Based on Soil Behavior
Stiffness Method to Design of Reinforced Soil Walls.” Type Index.” Canadian Geotechnical Journal 56 (4): 570–586.
Journal of Geotechnical and Geoenvironmental Engineering Casagrande, A. 1965. “Role of the ‘Calculated Risk’ in
144 (5): 04018024. Earthwork and Foundation Engineering.” Journal of Soil
Almeida, A. P. R. P., and J. Liu. 2018. “Statistical Evaluation of Mechanics and Foundations Division 91 (SM4): 1–40.
Design Methods for Micropiles in Ontario Soils.” DFI Cetin, K. O., R. B. Seed, A. Der Kiureghian, K. Tokimatsu, L. F.
Journal - The Journal of the Deep Foundation Institute 12 Harder, R. E. Kayen, and R. E. S. Moss. 2004. “Standard
(3): 133–146. Penetration Test-based Probabilistic and Deterministic
Andrus, R. D., K. H. Stokoe, and R. M. Chung. 1999. Draft Assessment of Seismic Soil Liquefaction Potential.”
Guidelines for Evaluating Liquefaction Resistance Using Journal of Geotechnical and Geoenvironmental Engineering
Shear Wave Velocity Measurements and Simplified 130 (12): 1314–1340.
Procedures. Gaithersburg, MD: US Department of Chahbaz, R., S. Sadek, and S. Najjar. 2019. “Uncertainty
Commerce, Technology Administration, National Institute Quantification of the Bond Stress – Displacement
of Standards and Technology. Relationship of Shoring Anchors in Different Geologic
Asem, P., J. Long, and P. Gardoni. 2018. “Probabilistic Model Units.” Georisk: Assessment and Management of Risk for
and LRFD Resistance Factors for the Tip Resistance of Engineered Systems and Geohazards 13 (4): 276–283.
Drilled Shafts in Soft Sedimentary Rock Based on Axial Chen, Y. J., and F. H. Kulhawy. 1994. Case History Evaluation
Load Tests.” In Innovations in Geotechnical Engineering: of Behavior of Drilled Shafts under Axial and Lateral
Honoring Jean-Louis Briaud (GSP 299), 1–46. Reston, VA: Loading. Report EPRI TR-104601. Palo Alto, CA: Electric
ASCE. Power Research Institute (EPRI)
Augustesen, A. 2006. “The Effects of Time on Soil Behaviour Chen, Y. J., and Y. H. Lee. 2010. “Evaluation of Lateral
and Pile Capacity.” Ph.D. thesis, Aalborg University Interpretation Criteria for Drilled Shaft Capacity.” Journal
Baecher, G. B. 1987. Statistical Analysis of Geotechnical Data. of Geotechnical and Geoenvironmental Engineering 136
Report No. GL-87-1, U.S. Vicksburg: Army Engineer (8): 1124–1136.
Waterways Experiment Station. Chen, Y. J., M. R. Liao, S. S. Lin, J. K. Huang, and M. C. M.
Baecher, G. B., and J. T. Christian. 2003. Reliability and .Marcos. 2014. “Development of an Integrated Web-Based
Statistics in Geotechnical Engineering. London and System with a Pile Load Test Database and pre-Analyzed
New York: John Wiley and Sons. Data.” Geomechanics and Engineering 7 (1): 37–53.
Baecher, G. B., and J. T. Christian. 2019. “TH Wu and the Chen, Y. J., S. W. Lin, and F. H. Kulhawy. 2011. “Evaluation of
Origins of Geotechnical Reliability.” Georisk 13 (4): 242–246. Lateral Interpretation Criteria for Rigid Drilled Shafts.”
Bahsan, E., H. J. Liao, J. Y. Ching, and S. W. Lee. 2014. Canadian Geotechnical Journal 48 (4): 634–643.
“Statistics for the Calculated Safety Factors of Undrained Cheung, R. W. M., and K. W. Shum. 2012. Review of the
Failure Slopes.” Engineering Geology 172: 85–94. Approach for Estimation of Pullout Resistance of Soil Nails.
Bathurst, R. J., S. Javankhoshdel, and T. M. Allen. 2017. “LRFD GEO Report No. 264. Geotechnical Engineering Office,
Calibration of Simple Soil-structure Limit States Considering Civil Engineering and Development Department, Hong
Method Bias and Design Parameter Variability.” Journal of Kong
Geotechnical and Geoenvironmental Engineering 143 (9): Ching, J., D. Q. Li, and K. K. Phoon. 2016. “Statistical
04017053. Characterization of Multivariate Geotechnical Data.” In
Beer, M., Y. Zhang, S. T. Quek, and K. K. Phoon. 2013. Reliability of Geotechnical Structures in ISO2394, edited by
“Reliability Analysis with Scarce Information: Comparing K. K. Phoon and J. V. Retief, 89–126. Balkema: CRC Press.
20 K.-K. PHOON

Ching, J., K. H. Li, K. K. Phoon, and M. C. Weng. 2018. Environment: From Theory to Practice (GSP 58), edited by
“Generic transformation models for some intact rock prop- C. D. Shackelford, P. P. Nelson, and M. J. S. Roth, 210–
erties.” Canadian Geotechnical Journal 55 (12): 1702–1741. 238. Reston, VA: ASCE.
Ching, J., and K. K. Phoon. 2012. “Modeling Parameters of DeGroot, D. J., and G. B. Baecher. 1993. “Estimating
Structured Clays as a Multivariate Normal Distribution.” Autocovariance of In-situ Soil Properties.” Journal of
Canadian Geotechnical Journal 49 (5): 522–545. Geotechnical Engineering 119 (1): 147–166.
Ching, J., and K. K. Phoon. 2013. “Multivariate Distribution for D’Ignazio, M., T. Lunne, K. H. Andersen, S. L. Yang, B. Di Buò,
Undrained Shear Strengths Under Various Test Procedures.” and T. T. Länsivaara. 2019. “Estimation of Preconsolidation
Canadian Geotechnical Journal 50 (9): 907–923. Stress of Clays from Piezocone by Means of High-quality
Ching, J., and K. K. Phoon. 2014. “Correlations Among Some Calibration Data.” AIMS Geosciences 5 (2): 104–116.
Clay Parameters – the Multivariate Distribution.” Canadian D’Ignazio, M., K. K. Phoon, S. A. Tan, and T. T. Länsivaara.
Geotechnical Journal 51 (6): 686–704. 2016. “Correlations for Undrained Shear Strength of
Ching, J., and K. K. Phoon. 2015. “Constructing Multivariate Finnish Soft Clays.” Canadian Geotechnical Journal 53
Distribution for Soil Parameters.” Chap. 1 in Risk and (10): 1628–1645.
Reliability in Geotechnical Engineering, 3–76. Balkema: Dithinde, M., K. K. Phoon, M. Wet, and J. Retief. 2011.
CRC Press. “Characterization of Model Uncertainty in the Static Pile
Ching, J., and K. K. Phoon. 2017. “Characterizing Uncertain Design Formula.” Journal of Geotechnical and
Site-specific Trend Function by Sparse Bayesian Learning.” Geoenvironmental Engineering 137 (1): 70–85.
Journal of Engineering Mechanics 143 (7): 04017028. DNV. 2012. “Statistical Representation of Soil Data.”
Ching, J., and K. K. Phoon. 2019a. “Impact of Auto-correlation Recommended Practice DNVGL-RP-C207, Det Norske
Function Model on the Probability of Failure.” Journal of Veritas (DNV), Oslo, Norway.
Engineering Mechanics 145 (1): 04018123. Einstein, H. H., and G. B. Baecher. 1982. “Probabilistic and
Ching, J., and K. K. Phoon. 2019b. “Constructing a Site- Statistical Methods in Engineering Geology I. Problem
specific Multivariate Probability Distribution Using Statement and Introduction to Solution.” In
Sparse, Incomplete, and Spatially Variable (MUSIC-X) Ingenieurgeologie und Geomechanik als Grundlagen des
Data.” Journal of Engineering Mechanics, under review. Felsbaues/Engineering Geology and Geomechanics as
Ching, J., and K. K. Phoon. 2019c. “Constructing Site-specific Fundamentals of Rock Engineering, edited by L. Muller,
Multivariate Probability Distribution Model Using Bayesian 47–61. Vienna: Springer.
Machine Learning.” Journal of Engineering Mechanics 145 Einstein, H. H., and G. B. Baecher. 1983. “Probabilistic and
(1): 04018126. Statistical Methods in Engineering Geology. Specific
Ching, J., and K. K. Phoon. 2019d. “Measuring Similarity Methods and Examples Part I: Exploration.” Rock
between Site-specific Data and Records from Other Sites.” Mechanics and Rock Engineering 16 (1): 39–72.
ASCE-ASME Journal of Risk and Uncertainty in Einstein, H. H., V. B. Halabe, J.-P. Dudt, and F. Descoeudres.
Engineering Systems, Part A: Civil Engineering, in press. 1996. “Geologic Uncertainties in Tunnelling.” In
Ching, J., K. K. Phoon, J. L. Beck, and Y. Huang. 2017. Uncertainty in the Geologic Environment: From Theory to
“Identifiability of Geotechnical Site-specific Trend Practice (GSP 58), edited by C. D. Shackelford, P. P.
Functions.” ASCE-ASME Journal of Risk and Uncertainty Nelson, and M. J. S. Roth, 239–253. Reston, VA: ASCE.
in Engineering Systems, Part A: Civil Engineering 3 (4): EN 1997−1:2004. Eurocode 7: Geotechnical Design — Part 1:
04017021. General Rules. Brussels, Belgium: European Committee for
Ching, J., K. K. Phoon, and D. Q. Li. 2016. “Robust Estimation Standardization (CEN).
of Correlation Coefficients Among Soil Parameters Under Feng, S., and P. J. Vardanega. 2019. “Correlation of the
the Multivariate Normal Framework.” Structural Safety Hydraulic Conductivity of Fine-grained Soils with Water
63: 21–32. Content Ratio Using a Database.” Environmental
Ching, J., K. K. Phoon, and J. W. Yu. 2014. “Linking Site Geotechnics 6 (5): 253–268.
Investigation Efforts to Final Design Savings with Fenton, G. A., F. Naghibi, D. Dundas, R. J. Bathurst, and D. V.
Simplified Reliability-based Design Methods.” Journal of Griffiths. 2016. “Reliability-based Geotechnical Design in
Geotechnical and Geoenvironmental Engineering 140 (3): the 2014 Canadian Highway Bridge Design Code.”
04013032. Canadian Geotechnical Journal 53 (2): 236–251.
Ching, J., J. S. Wang, C. H. Juang, and C. S. Ku. 2015. “Cone Filippas, O. B., F. H. Kulhawy, and M. D. Grigoriu. 1988.
Penetration Test (CPT)-based Stratigraphic Profiling Reliability-based Foundation Design for Transmission Line
Using the Wavelet Transform Modulus Maxima Method.” Structures: Uncertainties in Soil Property Measurement.
Canadian Geotechnical Journal 52 (12): 1993–2007. Report EL-5507(3). Palo Alto: Electric Power Research
Ching, J., S. S. Wu, and K. K. Phoon. 2016. “Statistical Institute.
Characterization of Random Field Parameters Using Flynn, K. 2014. “Experimental Investigations of Driven Cast
Frequentist and Bayesian Approaches.” Canadian in-Situ Piles.” Ph.D. thesis, National University of Ireland,
Geotechnical Journal 53 (2): 285–298. Galway
Ching, J., S. Wu, and K. K. Phoon. 2020. “Constructing Quasi- Galbraith, A., E. Farrell, and J. Byrne. 2014. “Uncertainty in
site-specific Multivariate Probability Distribution Using Pile Resistance from Static Load Tests Database.”
Hierarchical Bayesian Model.” Journal of Engineering Proceedings of the Institution of Civil Engineers -
Mechanics, ASCE, under review. Geotechnical Engineering 167 (5): 431–446.
DeGroot, D. J. 1996. “Analyzing Spatial Variability of in Situ Garder, J., K. Ng, S. Sritharan, and M. Roling. 2012. An
Soil Properties.” In Uncertainty in the Geologic Electronic Database for Drilled SHAft Foundation Testing
 GEORISK: ASSESSMENT AND MANAGEMENT OF RISK FOR ENGINEERED SYSTEMS AND GEOHAZARDS 21

(DSHAFT). Report No. InTrans Project 10-366, Iowa Structures: Transformation Models for In-Situ Tests.
Department of Transportation Report EL-5507(4). Palo Alto: Electric Power Research
Gartner. 2019. “IT Glossary.” Gartner. Accessed October 3, Institute.
2019. https://www.gartner.com/it-glossary/dark-data. Kulhawy, F. H., and P. W. Mayne. 1990. Manual on Estimating
Geoengineer. 2019. “Professor Ralph Peck’s Legacy Website.” Soil Properties for Foundation Design. Report EL-6800. Palo
Geoengineer.org. Accessed October 3, 2019. https://peck. Alto, CA: Electric Power Research Institute.
geoengineer.org/resources/words-of-wisdom. Kulhawy, F. H., T. D. O’Rourke, J. P. Stewart, and J. F. Beech.
Hand, D. J. 2014. “Wonderful Examples, but Let’s not Close 1983. Transmission Line Structure Foundations for Uplift-
Our Eyes.” Statistical Science 29: 98–100. Compression Loading: Load Test Summaries. Appendix to
Harr, M. E. 1987. Reliability-based Design in Civil Engineering. EPRI Final Report EL-2870. Report No. EL-3160-LD. Palo
New York: McGraw-Hill. Alto, CA: Electric Power Research Institute (EPRI)
Hov, S., A. Prästings, E. Persson, and S. Larsson. 2019. “On Lacasse, S., M. Jamiolkowski, R. Lancellotta, and T. Lunne.
Empirical Correlations for Normalised Shear Strengths 1981. “In Situ Characteristics of Two Norwegian Clays.”
from Fall Cone and Direct Simple Shear Tests in Soft Proceedings, 10th International Conference on Soil
Swedish Clays.” Geotechnical and Geological Engineering, Mechanics and Foundation Engineering, Stockholm, 2
under review. vols., 507–511.
Huang, B. Q., and R. J. Bathurst. 2009. “Evaluation of Soil- Lacasse, S., Z. Liu, and F. Nadim. 2017. “Probabilistic
Geogrid Pullout Models Using a Statistical Approach.” Characterization of Soil Properties—Recognition of
Geotechnical Testing Journal 32 (6): 489–504. Wilson Tang’s Contribution to Geotechnical Practice.” In
Idriss, I. M., and R. W. Boulanger. 2010. SPT-based Geotechnical Safety and Reliability: Honoring Wilson
Liquefaction Triggering Procedures. Davis, CA: Center for H. Tang (GSP 286), edited by C. Hsein Juang, R. B.
Geotechnical Modeling, Department of Civil and Gilbert, L. Zhang, J. Zhang, and L. Zhang, 2–26. Reston,
Environmental Engineering, University of California at VA: ASCE.
Davis. Report No. UCD/CGM-10/02. Lacasse, S., and T. Lunne. 1982. “Penetration Tests in Two
Ismail, S., S. S. Najjar, and S. Sadek. 2018. “Reliability Analysis Norwegian Clays.” Proceedings, Second European
of Buried Offshore Pipelines in Sand Subjected to Upheaval Symposium on Penetration Testing, Amsterdam, 661–670.
Buckling.” In Proceedings of the Offshore Technology Lacasse, S., and F. Nadim. 1996. “Uncertainties in
Conference (OTC). Houston, TX: American Petroleum Characterising Soil Properties.” In Uncertainty in the
Institute. OTC-28882-MS. Geologic Environment: From Theory to Practice (GSP 58),
ISO2394:1973/1986/1998/2015. General Principles on Reliability edited by C. D. Shackelford, P. P. Nelson, and M. J. S.
for Structures. Geneva, Switzerland: International Roth, 49–75. Reston, VA: ASCE.
Organization for Standardization. Lam, K., and W. K. Li. 1986. “Excerpts from Interview with
Jaksa, M. B. 1995. “The Influence of Spatial Variability on the Professor Peter Lumb.” Hong Kong Statistical Society
Geotechnical Design Properties of a Stiff, Overconsolidated Newsletter 9 (1): 2–5.
Clay.” Ph.D. Dissertation, University of Adelaide. Lambe, T. W. 1973. “Predictions in Soil Engineering.”
Jaksa, M. B., H. R. Maier, and M. A. Shahin. 2008. “Future Géotechnique 23 (2): 151–202.
Challenges for Artificial Neural Network Modelling in Lazarte, C. A. 2011. Proposed Specifications for LRFD Soil-
Geotechnical Engineering.” Proceedings, 12th International Nailing Design and Construction. NCHRP Report 701.
Conference of International Association for Computer Washington, DC: Transportation Research Board
Methods and Advances in Geomechanics (IACMAG), Lehane, B. M., J. K. Kim, P. Carotenuto, F. Nadim, S. Lacasse,
1710–1719, Goa, India, October 1–6. R. J. Jardine, and B. F. J. Van Dijk. 2017. “Characteristics of
Joint TC205/TC304 Working Group Report. 2017. “Discussion Unified Databases for Driven Piles.” In 8th International
of Statistical/Reliability Methods for Eurocodes. International Conference of Offshore Site Investigation and
Society for Soil Mechanics and Geotechnical Engineering.” Geomechanics. Vol. 1, 162–191. London: Society for
http://140.112.12.21/issmge/tc304.htm. Underwater Technology.
Juang, C. H., J. Ching, and Z. Luo. 2013. “Assessing SPT-based Li, Z., X. Wang, H. Wang, and R. Y. Liang. 2016. “Quantifying
Probabilistic Models for Liquefaction Potential Evaluation: Stratigraphic Uncertainties by Stochastic Simulation
A 10-Year Update.” Georisk: Assessment and Management Techniques Based on Markov Random Field.” Engineering
of Risk for Engineered Systems and Geohazards 7 (3): 137– Geology 201: 106–122.
150. Lin, P. Y., R. Bathurst, and J. Y. Liu. 2017. “Statistical
Kayen, R., R. E. S. Moss, E. M. Thompson, R. B. Seed, K. O. Evaluation of the FHWA Simplified Method and
Cetin, A. D. Kiureghian, Y. Tanaka, and K. Tokimatsu. Modifications for Predicting Soil Nail Loads.” Journal of
2013. “Shear-wave Velocity–Based Probabilistic and Geotechnical and Geoenvironmental Engineering 143 (3):
Deterministic Assessment of Seismic Soil Liquefaction 04016107.
Potential.” Journal of Geotechnical and Geoenvironmental Liu, H. F., L. S. Tang, P. Y. Lin, and G. X. Mei. 2018. “Accuracy
Engineering 139 (3): 407–419. Assessment of Default and Modified Federal Highway
Ku, C. S., C. H. Juang, C. W. Chang, and J. Ching. 2012. Administration (FHWA) Simplified Models for
“Probabilistic Version of the Robertson and Wride Estimation of Facing Tensile Forces of Soil Nail Walls.”
Method for Liquefaction Evaluation: Development and Canadian Geotechnical Journal 55 (8): 1104–1115.
Application.” Canadian Geotechnical Journal 49 (1): 27–44. Liu, S., H. Zou, G. Cai, B. V. Bheemasetti, A. J. Puppala, and J.
Kulhawy, F. H., B. Birgisson, and M. D. Grigoriu. 1992. Lin. 2016. “Multivariate Correlation among Resilient
Reliability-based Foundation Design for Transmission Line Modulus and Cone Penetration Test Parameters of
22 K.-K. PHOON

Cohesive Subgrade Soils.” Engineering Geology 209: 128– Journal of Geotechnical and Geoenvironmental Engineering
142. 144 (2): 04017106.
Long, M. 2001. “Database for Retaining Wall and Ground Moghaddam, R. B., P. W. Jayawickrama, W. D. Lawson, J. G.
Movements due to Deep Excavations.” Journal of Surles, and H. Seo. 2018. “Texas Cone Penetrometer
Geotechnical and Geoenvironmental Engineering 127 (3): Foundation Design Method: Qualitative and Quantitative
203–224. Assessment.” DFI Journal – The Journal of the Deep
Long, J. 2013. Improving Agreement between Static Method Foundations Institute 12 (2): 69–80.
and Dynamic Formula for Driven Cast-in-Place Piles in Montoya-Noguera, S., T. Zhao, Y. Hu, Y. Wang, and K. K.
Wisconsin. Report No. 0092-10-09, Wisconsin Phoon. 2019. “Simulation of Non-stationary Non-
Department of Transportation Gaussian Random Fields from Sparse Measurements
Long, J. 2016. Static Pile Load Tests on Driven Piles into Using Bayesian Compressive Sampling and Karhunen-
Intermediate-geo Materials. Report No. WHRP 0092-12- Loève Expansion.” Structural Safety 79: 66–79.
08, Wisconsin Department of Transportation Moormann, C. 2004. “Analysis of Wall and Ground
Long, J., and A. Anderson. 2014. Improved Design for Driven Movements due to Deep Excavations in Soft Soil Based on
Piles Based on a Pile Load Test Program in Illinois: phase a new Worldwide Database.” Soils and Foundations 44
2. Report No. FHWA-ICT-14-019, Illinois Department of (1): 87–98.
Transportation Moshfeghi, S., and A. Eslami. 2018. “Study on Pile Ultimate
Long, J., J. Hendrix, and D. Jaromin. 2009. Comparison of Five Capacity Criteria and CPT-Based Direct Methods.”
Different Methods for Determining Pile Bearing Capacities. International Journal of Geotechnical Engineering 12 (1):
Report No. WisDOT 0092-07-04, Wisconsin Department 28–39.
of Transportation. Moss, R. E. S., R. B. Seed, R. E. Kayen, J. P. Stewart, A. Der
Lumb, P. 1966. “Variability of Natural Soils.” Canadian Kiureghian, and K. O. Cetin. 2006. “CPT-based
Geotechnical Journal 3 (2): 74–97. Probabilistic and Deterministic Assessment of in Situ
Marcos, M. C. M., and Y. J. Chen. 2013. “Evaluation of Lateral Seismic Soil Liquefaction Potential.” Journal of
Interpretation Criteria for Drilled Shaft Capacity in Geotechnical and Geoenvironmental Engineering 132 (8):
Gravels.” Geotechnical and Geological Engineering 31 (5): 1032–1051.
1411–1420. Motamed, R., S. Elfass, and K. Stanton. 2016. LRFD Resistance
Mariethoz, G., and J. Caers. 2015. Multiple-point Geostatistics - Factor Calibration for Axially Loaded Drilled Shafts in the
Stochastic Modeling with Training Images. West Sussex: Las Vegas Valley. Report No. 515-13-803, Nevada
John Wiley & Sons. Department of Transportation.
Marsland, A. 1953. “Model Experiments to Study the Influence Nanazawa, T., T. Kouno, G. Sakashita, and K. Oshiro. 2019.
of Seepage on the Stability of a Sheeted Excavation in Sand.” “Development of Partial Factor Design Method on
Géotechnique 3 (6): 223–241. Bearing Capacity of Pile Foundations for Japanese
McVay, M., S. Wasman, L. Huang, and S. Crawford. 2016. Specifications for Highway Bridges.” Georisk: Assessment
Load and Resistance Factor Design (LRFD) Resistance and Management of Risk for Engineered Systems and
Factors for Auger Cast in Place Piles. Tallahassee, FL: Geohazards 13 (3): 166–175.
Florida Department of Transportation. National Research Council. 1995. Probabilistic Methods in
Mesri, G., and N. Huvaj. 2007. “Shear Strength Mobilized in Geotechnical Engineering. Washington, DC: National
Undrained Failure of Soft Clay and Silt Deposits.” In Academies Press.
Advances in Measurement and Modeling of Soil Behavior Ng, C. W. W., T. L. Y. Yau, J. H. M. Li, and Tang. 2001. “New
(GSP 173), 1–22. Reston, VA: ASCE. Failure Load Criterion for Large Diameter Bored Piles in
Miyata, Y., and R. J. Bathurst. 2012a. “Measured and Predicted Weathered Geomaterials.” Journal of Geotechnical and
Loads in Steel Strip Reinforced c-Φ Soil Walls in Japan.” Geoenvironmental Engineering 127 (6): 488–498.
Soils and Foundations 52 (1): 1–17. NGI. 2019. “Peck Library.” Norwegian Geotechnical Institute.
Miyata, Y., and R. J. Bathurst. 2012b. “Analysis and Calibration Accessed October 3, 2019. https://www.ngi.no/eng/
of Default Steel Strip Pullout Models Used in Japan.” Soils Publications-and-library/NGI-s-Historical-Libraries/Peck-
and Foundations 52 (3): 481–497. Library.
Miyata, Y., and R. J. Bathurst. 2015. “Reliability Analysis of Niazi, F. S. 2014. “Static Axial Pile Foundation Response Using
Geogrid Installation Damage Test Data in Japan.” Soils Seismic Piezocone Data.” Ph.D. thesis, Georgia Institute
and Foundations 55 (2): 393–403. ofTechnology
Miyata, Y., and R. J. Bathurst. 2019. “Statistical Assessment of Orchant, C. J., F. H. Kulhawy, and C. H. Trautmann. 1988.
Load Model Accuracy for Steel Grid-Reinforced Soil Walls.” Reliability-based Foundation Design for Transmission Line
Acta Geotechnica 14 (1): 57–70. Structures: Critical Evaluation of In-situ Test Methods.
Miyata, Y., R. J. Bathurst, and T. M. Allen. 2014. “Reliability Report EL-5507(2). Palo Alto: Electric Power Research
Analysis of Geogrid Creep Data in Japan.” Soils and Institute.
Foundations 54 (4): 608–620. Ou, C. Y., and J. T. Liao. 1987. Geotechnical Engineering
Miyata, Y., R. J. Bathurst, and T. Konami. 2011. “Evaluation of Research Report. GT96008. Taipei: National Taiwan
Two Anchor Plate Capacity Models for MAW Systems.” University of Science and Technology.
Soils and Foundations 51 (5): 885–895. Paikowsky, S. G., B. Birgisson, M. McVay, T. Nguyen, C. Kuo,
Miyata, Y., Y. Yu, and R. J. Bathurst. 2018. “Calibration of Soil- G. B. Baecher, B. Ayyub, et al. 2004. Load and Resistance
Steel Grid Pullout Models Using a Statistical Approach.” Factors Design for Deep Foundations. NCHRP Report 507.
 GEORISK: ASSESSMENT AND MANAGEMENT OF RISK FOR ENGINEERED SYSTEMS AND GEOHAZARDS 23

Washington, DC: Transportation Research Board of the Reddy, S., and A. Stuedlein. 2017. “Ultimate Limit State
National Academies. Reliability-Based Design of Augered Cast-in-Place Piles
Paikowsky, S., M. Canniff, K. Lesny, A. Kisse, S. Amatya, and Considering Lower-Bound Capacities.” Canadian
R. Muganga. 2010. “LRFD Design and Construction of Geotechnical Journal 54 (12): 1693–1703.
Shallow Foundations for Highway Bridge Structures.” Roling, M., S. Sritharan, and M. Suleiman. 2011. Development
NCHRP Report 651. Washington, DC: Transportation of LRFD Procedures for Bridge Pile Foundations in Iowa.
Research Board. Vol. 1: An electronic Database for Pile Load Tests (PILOT).
Peck, R. B. 1980. “Where has All the Judgment Gone?” Report No. IHRB Project TR-573, Iowa Department of
Canadian Geotechnical Journal 17 (4): 584–590. Transportation
Phoon, K. K. 2017. “Role of Reliability Calculations in Samtani, N. C., and T. C. Allen. 2018. “Implementation Report
Geotechnical Design.” Georisk 11 (1): 4–21. – Expanded Database for Service Limit State Calibration of
Phoon, K. K. 2018. “Editorial for Special Collection on Immediate Settlement of Bridge Foundation on Soil. Report
Probabilistic Site Characterization.” ASCE-ASME Journal No. FHWA-HIF-18-008. Washington, DC: Federal
of Risk and Uncertainty in Engineering Systems, Part A: Highway Administration (FHWA).
Civil Engineering 4 (4): 02018002. Shahin, M. A., M. B. Jaksa, and H. R. Maier. 2001. “Artificial
Phoon, K. K. 2020. “The Goldilocks Dilemma - too Little or Neural Network Applications in Geotechnical
too Much Data.” GeoStrata 24 (1), in press. Engineering.” Australian Geomechanics 36 (1): 49–62.
Phoon, K. K., and J. Ching. 2017. “Better Correlations for Shields, M. D., and H. Kim. 2017. “Simulation of Higher-
Geotechnical Engineering.” A Decade of Geotechnical order Stochastic Processes by Spectral Representation.”
Advances, Geotechnical Society of Singapore (GeoSS), Probabilistic Engineering Mechanics 47: 1–15.
73–102. Smith, T., A. Banas, M. Gummer, and J. Jin. 2011.
Phoon, K. K., J. Ching, and Y. Wang. 2019. “Managing Risk in Recalibration of the GRLWEAP LRFD Resistance Factor
Geotechnical Engineering – from Data to Digitalization.” for Oregon DOT. Report No. FHWA-OR-RD-11-08,
Proceedings, 7th International Symposium on Oregon Department of Transportation
Geotechnical Safety and Risk (ISGSR 2019), Taipei, Spry, M. J., F. H. Kulhawy, and M. D. Grigoriu. 1988.
Taiwan, in press. Reliability-based Foundation Design for Transmission Line
Phoon, K. K., and F. H. Kulhawy. 1999a. “Characterization of Structures: Geotechnical Site Characterization Strategy.
Geotechnical Variability.” Canadian Geotechnical Journal Report EL-5507(1). Palo Alto: Electric Power Research
36 (4): 612–624. Institute.
Phoon, K. K., and F. H. Kulhawy. 1999b. “Evaluation of Stark, T., J. Long, A. Baghdady, and A. Osouli. 2017. Modified
Geotechnical Property Variability.” Canadian Geotechnical Standard Penetration Test-based Drilled Shaft Design
Journal 36 (4): 625–639. Method for Weak Rocks (phase 2 study). Report No.
Phoon, K. K., and F. H. Kulhawy. 2008. “Serviceability Limit FHWA-ICT-17-018, Illinois Department of Transportation
State Reliability-based Design.” In Reliability-based Design State of the Nation Report. 2017. “Digital Transformation.”
in Geotechnical Engineering: Computations and Institution of Civil Engineers. https://www.ice.org.uk/
Applications, 344–383. New York: Taylor & Francis. news-and-insight/policy/state-of-the-nation-2017-digital-
Phoon, K. K., F. H. Kulhawy, and M. D. Grigoriu. 2003. transformation.
“Multiple Resistance Factor Design for Shallow Stuyts, B., D. Cathie, and T. Powell. 2016. “Model Uncertainty
Transmission Line Structure Foundations.” Journal of in Uplift Resistance Calculations for Sandy Backfills.”
Geotechnical and Geoenvironmental Engineering 129 (9): Canadian Geotechnical Journal 53 (11): 1831–1840.
807–818. Tang, W. H. 1984. “Principles of Probabilistic Characterization
Phoon, K. K., S. L. Liu, and Y. K. Chow. 2009. of Soil Properties.” In Probabilistic Characterization of Soil
“Characterization of Model Uncertainties for Cantilever Properties: Bridge Between Theory and Practice, edited by
Walls in Sand.” Journal of GeoEngineering 4 (3): 75–85. D. S. Bowles and H.-Y. Ko, 74–89. Atlanta, GA: ASCE.
Phoon, K. K., W. A. Prakoso, Y. Wang, and J. Ching. 2016. Tang, C., and K. K. Phoon. 2016. “Model Uncertainty of
“Uncertainty Representation of Geotechnical Design Cylindrical Shear Method for Calculating the Uplift
Parameters.” Chap. 3 in Reliability of Geotechnical Capacity of Helical Anchors in Clay.” Engineering Geology
Structures in ISO2394, 49–87. Balkema: CRC Press. 207: 14–23.
Phoon, K. K., J. V. Retief, J. Ching, M. Dithinde, T. Tang, C., and K. K. Phoon. 2018a. “Evaluation of Model
Schweckendiek, Y. Wang, and L. M. Zhang. 2016. “Some Uncertainties in Reliability-Based Design of Steel H-Piles
Observations on ISO2394:2015 Annex D (Reliability of in Axial Compression.” Canadian Geotechnical Journal 55
Geotechnical Structures).” Structural Safety 62: 24–33. (11): 1513–1532.
Phoon, K. K., and C. Tang. 2019. “Characterization Tang, C., and K. K. Phoon. 2018b. “Statistics of Model Factors
of Geotechnical Model Uncertainty.” Georisk 13 (2): in Reliability-Based Design of Axially Loaded Driven Piles
101–130. in Sand.” Canadian Geotechnical Journal 55 (11): 1592–
Qi, X. H., D. Q. Li, K. K. Phoon, Z. Cao, and X. S. Tang. 2016. 1610.
“Simulation of Geologic Uncertainty Using Coupled Tang, C., and K. K. Phoon. 2018c. “Statistics of Model Factors
Markov Chain.” Engineering Geology 207: 129–140. and Consideration in Reliability-Based Design of Axially
Rauser, J., and C. Tsai. 2016. “Beneficial use of the Louisiana Loaded Helical Piles.” Journal of Geotechnical and
Foundation Load Test Database.” In Transportation Geoenvironmental Engineering 144 (8): 04018050.
Research Board 95th Annual Meeting, 1–17. Washington, Tang, C., and K. K. Phoon. 2018. “Statistics of Model Factors
DC: Transportation Research Board. and Consideration in Reliability-Based Design of Axially
24 K.-K. PHOON

Loaded Helical Piles.” Journal of Geotechnical and Wang, X., H. Wang, R. Y. Liang, H. Zhu, and H. Di. 2018. “A
Geoenvironmental Engineering 144 (8): 04018050. Hidden Markov Random Field Model Based Approach for
Tang, C., and K. K. Phoon. 2019a. “Evaluation of Stress Probabilistic Site Characterization Using Multiple Cone
Dependent Methods for the Punch-Through Capacity of Penetration Test Data.” Structural Safety 70: 128–138.
Foundations in Clay with Sand.” ASCE-ASME Journal of Wang, H., X. Wang, J. F. Wellmann, and R. Y. Liang. 2018.
Risk and Uncertainty in Engineering Systems, Part A: Civil “Bayesian Stochastic Soil Modeling Framework Using
Engineering 5 (3): 04019008. Gaussian Markov Random Fields.” ASCE-ASME Journal
Tang, C., and K. K. Phoon. 2019b. “Characterization of Model of Risk and Uncertainty in Engineering Systems, Part A:
Uncertainty in Predicting Axial Resistance of Piles Driven Civil Engineering 4 (2): 04018014.
In to Clay.” Canadian Geotechnical Journal 56 (8): 1098– Wang, H., X. Wang, J. F. Wellmann, and R. Y. Liang. 2019. “A
1118. Bayesian Unsupervised Learning Approach for Identifying
Tang, C., and K. K. Phoon. 2019c. “Statistical Evaluation of Soil Stratification Using Cone Penetration Data.”
Model Factors in Reliability Calibration of High Canadian Geotechnical Journal 56 (8): 1184–1205.
Displacement Helical Piles Under Axial Loading.” Wang, H., J. F. Wellmann, Z. Li, X. Wang, and R. Y. Liang.
Canadian Geotechnical Journal. doi:10.1139/cgj-2018-0754. 2017. “A Segmentation Approach for Stochastic
Tang, C., K. K. Phoon, and Y. J. Chen. 2019. “Statistical Geological Modeling Using Hidden Markov Random
Analyses of Model Factors in Reliability-Based Limit State Fields.” Mathematical Geosciences 49 (2): 145–177.
Design of Drilled Shafts under Axial Loading.” Journal of Wang, J., Z. Xu, and W. Wang. 2010. “Wall and Ground
Geotechnical and Geoenvironmental Engineering 145 (9): Movements due to Deep Excavations in Shanghai Soft
04019042. Soils.” Journal of Geotechnical and Geoenvironmental
Tang, C., K. K. Phoon, D. Q. Li, and S. O. Akbas. 2019. Engineering 136 (7): 985–994.
“Expanded Database Assessment of Design Methods for Wang, Y., and T. Zhao. 2016. “Interpretation of Soil Property
Spread Foundations Under Axial Uplift and Compression.” Profile from Limited Measurement Data: A Compressive
under preparation. Sampling Perspective.” Canadian Geotechnical Journal 53
TC304. 2019. “304dB – TC304 Databases.” TC304 (9): 1547–1559.
(Engineering Practice of Risk Assessment and Wang, Y., and T. Zhao. 2017. “Statistical Interpretation of Soil
Management), International Society for Soil Mechanics Property Profiles from Sparse Data Using Bayesian
and Geotechnical Engineering. Accessed October 3, 2019. Compressive Sampling.” Géotechnique 67 (6): 523–536.
http://140.112.12.21/issmge/tc304.htm. Wang, Y., T. Zhao, Y. Hu, and K. K. Phoon. 2019. “Simulation
Terzaghi, K., and R. B. Peck. 1967. Soil Mechanics in of Random Fields with Trend from Sparse Measurements
Engineering Practice. New York: John Wiley & Sons Inc. without De-trending.” Journal of Engineering Mechanics
Tian, M., D. Q. Li, Z. Cao, K. K. Phoon, and Y. Wang. 2016. 145 (2): 04018130.
“Bayesian Identification of Random Field Model Using Wang, Y., T. Zhao, and K. K. Phoon. 2018. “Direct Simulation
Indirect Test Data.” Engineering Geology 210: 197–211. of Random Field Samples from Sparsely Measured
Travis, Q., M. Schmeeckle, and D. Sebert. 2011. “Meta-analysis Geotechnical Data with Consideration of Uncertainty in
of 301 Slope Failure Calculations. I: Database Description.” Interpretation.” Canadian Geotechnical Journal 55 (6):
Journal of Geotechnical and Geoenvironmental Engineering 862–880.
137 (5): 453–470. White, D., C. Cheuk, and M. Bolton. 2008. “The Uplift
Vanmarcke, E. H. 1977. “Probabilistic Modeling of Soil Resistance of Pipes and Plate Anchors Buried in Sand.”
Profiles.” Journal of the Geotechnical Engineering Division Géotechnique 58 (10): 771–779.
103 (11): 1227–1246. Wood, T., P. Jayawickrama, J. Surles, and W. Lawson. 2012a.
Vanmarcke, E. H., and G. A. Fenton. 2003. Probabilistic Site “Pullout Resistance of MSE Reinforcements in Backfills
Characterization at the National Geotechnical Typically Used in Texas: Vol. 2, Test Reports for MSE
Experimentation Sites (GSP 121). Reston, VA: ASCE. Reinforcements in Type B (sandy) Backfill.” Research
Vick, S. G. 2002. Degrees of Belief: Subjective Probability and Report: FHWA/TX-13/0-6493-3, Vol. 2, Texas
Engineering Judgment. Reston, VA: ASCE. Department of Transportation
Wang, Y., Y. Hu, and T. Zhao. 2019. “CPT-based Subsurface Wood, T., P. Jayawickrama, J. Surles, and W. Lawson. 2012b.
Soil Classification and Zonation in a 2D Vertical Cross-sec- “Pullout Resistance of MSE Reinforcements in Backfills
tion Using Bayesian Compressive Sampling.” Canadian Typically Used in Texas: Vol. 3, Test Reports for MSE
Geotechnical Journal. doi:10.1139/cgj-2019-0131. Reinforcements in Type A (gravelly) Backfill.” Research
Wang, Y., K. Huang, and Z. Cao. 2013. “Probabilistic Report: FHWA/TX-13/0-6493-3, Vol. 3, Texas
Identification of Underground Soil Stratification Using Department of Transportation
Cone Penetration Tests.” Canadian Geotechnical Journal Wu, T. H., M. A. Abdel-Latif, M. A. Nuhfer, and B. B.
50 (7): 766–776. Curry. 1996. Uncertainty in the Geologic Environment:
Wang, X., Z. Li, H. Wang, Q. Rong, and R. Y. Liang. 2016. from Theory to Practice (GSP 58), 76–90. Reston, VA:
“Probabilistic Analysis of Shield-driven Tunnel in ASCE.
Multiple Strata Considering Stratigraphic Uncertainty.” Wu, S. H., J. Y. Ching, and C. Y. Ou. 2013. “Predicting Wall
Structural Safety 62: 88–100. Displacements for Excavations with Cross Walls in Soft
Wang, X., H. Wang, R. Y. Liang, and Y. Liu. 2019. “A Semi- Clay.” Journal of Geotechnical and Geoenvironmental
supervised Clustering-based Approach for Stratification Engineering 139 (6): 914–924.
Identification Using Borehole and Cone Penetration Test Wu, S. H., C. Y. Ou, and J. Y. Ching. 2014. “Calibration of
Data.” Engineering Geology 248: 102–116. Model Uncertainties in Base Heave Stability for Wide
 GEORISK: ASSESSMENT AND MANAGEMENT OF RISK FOR ENGINEERED SYSTEMS AND GEOHAZARDS 25

Excavations in Clay.” Soils and Foundations 54 (6): 1159– Yuen, K. V., and G. A. Ortiz. 2018. “Multi-resolution Bayesian
1174. Nonparametric General Regression for Structural Model
Wu, T. H., W. H. Tang, D. A. Sangrey, and G. B. Baecher. Updating.” Structural Control and Health Monitoring 25
1989. “Reliability of Offshore Foundations-state of the Art.” (2): e2077.
Journal of Geotechnical Engineering 115 (2): 157–178. Yuen, K. V., G. A. Ortiz, and K. Huang. 2016. “Novel
Xiao, T., D. Q. Li, Z. Cao, and L. M. Zhang. 2018. “CPT- Nonparametric Modeling of Seismic Attenuation and
based Probabilistic Characterization of Three-dimensional Directivity Relationship.” Computer Methods in Applied
Spatial Variability Using MLE.” Journal of Geotechnical and Mechanics and Engineering 311: 537–555.
Geoenvironmental Engineering 144 (5): 04018023. Zhang, L. M., L. M. P. Shek, H. W. Pang, and C. F. Pang. 2006.
Yang, Z., R. Jardine, W. Guo, and F. Chow. 2016. A “Knowledge-based Design and Construction of Driven
Comprehensive Database of Tests on Axially Loaded Piles Piles.” Proceedings of the Institution of Civil Engineers -
Driven in Sand. London: Academic Press. Geotechnical Engineering 159 (3): 177–185.
Yuan, J., P. Y. Lin, R. Huang, and Y. Que. 2019. “Statistical Zhang, D. M., Y. Zhou, K. K. Phoon, and H. W. Huang. 2019.
Evaluation and Calibration of Two Methods for “Multivariate Probability Distribution of Shanghai Clay
Predicting Nail Loads of Soil Nail Walls in China.” Properties.” Engineering Geology, under review.
Computers and Geotechnics 108: 269–279. Zhao, T., and Y. Wang. 2018. “Simulation of Cross-correlated
Yuen, K. V., and G. A. Ortiz. 2016. “Bayesian Nonparametric Random Field Samples from Sparse Measurements Using
General Regression.” International Journal for Uncertainty Bayesian Compressive Sensing.” Mechanical Systems and
Quantification 6 (3): 195–213. Signal Processing 112: 384–400.