Probability Risk Paper As Published Discussions Updated Plots

Probability and Risk of Slope Failure
Francisco Silva, M.ASCE1; T. William Lambe, Hon.M.ASCE2; and W. Allen Marr, F.ASCE3
Abstract: This paper aims to demystify the use of risk assessment as a decision management tool and present a methodology that places
quantitative risk assessment within reach of every geotechnical engineer, even for routine engineering assignments. In particular, we
propose using quantification of expert judgment 共i.e., subjective probabilities兲 as a practical alternative for determining probability of
slope failure. The writers present a semiempirical relationship between factor of safety and annual probability of failure that permits
estimation of slope failure probabilities with relatively modest effort. The case study for a tailings dam shows that risk assessment based
on quantification of expert judgment provides a framework to arrive at rational management and engineering decisions related to dam
safety and other geotechnical problems. Using the semiempirical relationship presented here, practicing engineers can use this helpful tool
by applying their current skills.
DOI: 10.1061/共ASCE兲1090-0241共2008兲134:12共1691兲
CE Database subject headings: Decision making; Risk management; Soil erosion; Probability; Slope stability; Tailings; Dams, earth;
Dam safety.
Introduction done correctly. The engineer should focus his effort on obtaining
a representative geometry 共surface and subsurface兲 and correctly
This paper presents a risk assessment methodology the writers use defining pore pressures and strength.
in the practice of geotechnical engineering. This methodology has This paper moves beyond the determination of safety factor to:
evolved over the last 30 years in response to client’s needs for 1. More fully understand the significance of level of safety;
quantified risk assessments at affordable costs. Owners and 2. Describe a practical method of risk-based decision making
operators of industrial facilities understand the concept of risk. for situations involving slope failures; and
Risk concepts and risk management form part of many 3. Illustrate that geotechnical engineers have all the necessary
management programs. These owners and operators expect skills to perform these risk-based analyses through an
engineers to provide probability and risk data to help them make example.
informed business decisions. Although geotechnical engineers The engineering literature 共Morgenstern 1995; Vick 1994; and
have applied probability and risk concepts for several decades, too others兲 identifies three commonly accepted ways of estimating
frequently the numbers proved difficult to support or dependable event probabilities:
numbers required prohibitively costly investigations, evaluations, 1. Based on frequency of observations 共historical data兲;
or modeling. 2. Derived from probability theory 共mathematical modeling兲;
When considering slope stability problems in geotechnical and
engineering, an early step consists of correctly determining the 3. Quantification of expert judgment 共subjective probabilities兲.
level of safety of a slope. Numerous references exist to help The main objective of this paper is focused on demonstrating
engineers with this task 共Lambe and Silva 2003; Duncan and that quantification of expert judgment provides the engineer with
Wright 2005; Cornforth 2005兲. A correct determination of level of a practical method to determine probabilities for risk analyses. We
safety should properly handle the three geotechnical fundamentals combine historical and subjective probabilities to obtain a corre-
that control slope stability: geometry, pore pressures, and strength. lation between safety factor and failure probability suitable for
Lambe and Silva 共1992兲 have shown that nearly all the methods use in engineering practice.
commonly used to integrate these fundamentals into
determination of a safety factor provide similar answers when
Probability of Failure
1
Consulting Civil Engineer, GeoEngineering and Environment, 12
Baskin Rd., Lexington, MA 02421 共corresponding author兲. E-mail: When involved with a potentially unstable slope, engineers want
silva@alum.mit.edu to know whether or not the slope will fail. Since there are many
2
Consulting Geotechnical Engineer, 32550 Clay Gully Rd., Myakka uncertainties that affect this determination, the engineer has to
City, FL 34251-9457. settle for estimating the probability of whether the slope will fail.
3
President and CEO, Geocomp Corporation, 1145 Massachusetts One can estimate the probability of a failure 共or of any other
Ave., Boxborough, MA 01719.
event occurring兲 using one of the three methods listed in the
Note. Discussion open until May 1, 2009. Separate discussions must
be submitted for individual papers. The manuscript for this paper was
previous section. All three methods have wide acceptance within
submitted for review and possible publication on July 26, 2005; approved the profession. Benjamin and Cornell 共1970兲 state: “The sources
on January 2, 2008. This paper is part of the Journal of Geotechnical of these probabilities may include observed frequencies, deduc-
and Geoenvironmental Engineering, Vol. 134, No. 12, December 1, tions from mathematical models, and in addition, measures of an
2008. ©ASCE, ISSN 1090-0241/2008/12-1691–1699/$25.00. engineer’s subjective degree of belief regarding the possible states
JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2008 / 1691
Downloaded 08 Dec 2008 to 18.51.1.222. Redistribution subject to ASCE license or copyright; see http://pubs.asce.org/copyright
Fig. 1. Factor of safety versus annual probability of failure
of nature.” The difficulty and expense associated with determin- standard construction or operation. Temporary facilities and
ing probability of failure by rigorous mathematical means has those with low failure consequences often fall into this cat-
held back the widespread adoption of risk-based analyses in geo- egory; and
technical engineering. During the last 2 decades, we have con- 4. Category IV—facilities with little or no engineering.
cluded that quantified expert judgment places at our disposal an The family of curves in Fig. 1 and the associated Table 1 with
underutilized tool for making better engineering and management the four levels of engineering reflect the generally accepted con-
decisions. cept that: “A larger factor of safety does not necessarily imply a
Fig. 1 shows relationships between factor of safety and annual smaller risk, because its effect can be negated by the presence of
probability of failure based on actual engineering projects and larger uncertainties in the design environment” 共Kulhawy and
developed through quantified expert judgment. This plot, an up- Phoon 1996; D’Andrea and Sangrey 1982, Tavares and Serafim
dated version of the one originally presented by Lambe 共1985兲 1983; and Christian et al. 1994兲. The curves in Fig. 1 provide a
and Baecher and Christian 共2003兲, is the most important figure in practical grouping related to the standard deviation of the safety
this paper. Fig. 1 classifies earth structures into the four categories factors calculated with four different levels of engineering.
described in Table 1. The category of the earth structure, based on Fig. 1 shows data from over 75 projects spanning over 4 de-
the level of engineering, ranges from best 共Category I兲 to poor cades used to develop the relationships. The projects included
共Category IV兲. We establish the level of engineering by examin- zoned and homogeneous earth dams, tailings dams, natural and
ing the practices followed for design, investigation, testing, analy- cut slopes, and some earth retaining structures. The writers used
ses and documentation, construction, and operation and the iterative process described below to arrive at the probability
monitoring. The four categories correspond to the following types of failure determinations.
of facilities: Two frequently mentioned data points served as reference
1. Category I—facilities designed, built, and operated with points for the curves in Fig. 1:
state-of-the-practice engineering. Generally these facilities 1. 共1.5, 0.0001兲—Baecher et al. 共1980兲, Whitman 共1984兲, and
have high failure consequences; Christian et al. 共1992兲 based on historical performance of
2. Category II—facilities designed, built, and operated using earth dams designed and constructed with conservative engi-
standard engineering practice. Many ordinary facilities fall neering practice; and
into this category; 2. 共1.0, 0.5兲—Vick 共1994兲 based on the theoretical fact that a
3. Category III—facilities without site-specific design and sub- normally distributed uncertainty on factor of safety gives a
1692 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2008
Table 1. Earth Structure Categories and Characteristics
Operation and
Design Construction monitoring
Level of Analyses and

engineering Investigation Testing documentation
I (Best) • Evaluate design and • Run lab tests on undisturbed • Determine FS using effective stress • Full time supervision • Complete performance program
performance of nearby structures specimens at field conditions parameters based on measured data by qualified engineer including comparison between
共geometry, strength, pore pressure兲 for predicted and measured
site performance 共e.g., pore pressure,
strength, deformations兲
Facilities with • Analyze historic aerial • Run strength test along field • Consider field stress path in stability • Construction control • No malfunctions 共slides, cracks,
high failure photographs effective and total stress paths determination tests by qualified artesian heads兲
consequences engineers and technicians
• Locate all nonuniformities 共soft, • Run index field tests 共e.g., • Prepare flow net for instrumented • No errors or omissions • Continuous maintenance by trained
wet, loose, high, or low field vane, cone penetrometer兲 to sections crews
permeability zones兲 detect all soft, wet, loose, high, or
low permeability zones
• Determine site geologic history • Calibrate equipment and • Predict pore pressure and other relevant • Construction report
sensors prior to testing program performance parameters 共e.g., stress, clearly documents
deformation, flow rates兲 for instrumented construction activities
section
• Determine subsoil profile using • Have design report clearly document
continuous sampling parameters and analyses used for design
• Obtain undisturbed samples for • No errors or omissions
lab testing of foundation soils
• Determine field pore pressures • Peer review
共0.20兲共0.20兲共0.20兲共0.20兲共0.20兲
II (Above • Evaluate design and performance • Run standard lab tests on • Determine FS using effective stress • Part-time supervision • Periodic inspection by qualified
average) of nearby structures undisturbed specimens parameters and pore pressures by qualified engineer engineer
Ordinary facilities • Exploration program tailored to • Measure pore pressure in strength • Adjust for significant differences No errors or omissions • No uncorrected malfunctions
project conditions by qualified tests between field stress paths and stress path
engineer implied in analysis that could affect
design
• Evaluate differences between • Selected field measurements
laboratory test conditions and field
conditions
• Routine maintenance
共0.40兲共0.40兲共0.40兲共0.40兲共0.40兲
III (Average) • Evaluate performance of nearby • Index tests on samples from site • Rational analyses using parameters • Informal construction • Annual inspection by qualified
structures inferred from index tests supervision engineer
Unimportant or • Estimate subsoil profile from • No field measurements
temporary facilities existing data and borings
with low failure • Maintenance limited to emergency
consequences repairs
共0.60兲共0.60兲共0.60兲共0.60兲共0.60兲
IV (Poor) • No field investigation • No laboratory tests on samples • Approximate analyses using assumed • No construction • Occasional inspection by
obtained at the site parameters supervision by non-qualified person
qualified engineer
• No construction control tests. • No field measurements
Little or no 共0.80兲共0.80兲共0.80兲共0.80兲共0.80兲
engineering
probability of failure of 0.5 at a factor of safety of 1.
The failure base frequency of 10−4 for an annual probability of
failure for dam slopes provides an anchor based on historic prob-
abilities. We used this historic value for the curve corresponding Curves for
to above average facilities 共Category II兲 because earthen dams lognormal
considered by Baecher et al. 共1980兲 were typically designed and distribution
operated with good conservative engineering practice. The Cat-
egory III curve uses the 50% probability of failure for a computed
safety factor equal to 1.0 suggested by Vick 共1994兲 and others.
Vick 共1994兲 states that “The first-year failure probability of 0.5
for the base case was consistent with initially expressed judg-
ments that failure was essentially already in progress.”
The writers then selected engineering projects with well
known design, construction, and operation characteristics from
their practice. Stability analyses helped perform deterministic as-
Fig. 2. FS versus p共f兲—comparison with lognormal distribution and
sessments of the degree of safety in terms of safety factor.
data from Whitman 共1984兲
Through an elicitation process they then obtained failure prob-
abilities by comparing each case with the two historic data points
above 共1.5 : 10−4 and 1.0 : 0.5兲 and with each other. When asso-
ciates of Professor Lambe participated in the determination of 1. The general shape of the four curves agrees with theoretical
probabilities, all reached general consensus on the appropriate probability distributions; and
values. Using this process, they successively iterated through the 2. The magnitudes of the numbers used agree well with theo-
projects to attain consistency in the failure probabilities. The flat- retical work and field observations.
tening of the curves near factor of safety 共FS兲 = 2 reflects the The main difference among Fig. 1 and the work depicted in
diminishing returns obtained from overbuilding a constructed fa- Fig. 2 consists of the intentional flattening of the Fig. 1 curves for
cility. Below FS= 1, the curves approach p共f兲 = 1 asymptotically. higher safety factors. The more rigorously derived relationships
Fig. 1 depicts a semiempirical relationship between factor of yield ever-diminishing failure probabilities for increasing safety
safety and annual probability of failure. The final shape of the factors and have different probabilities of failure at a factor of
curves reflects the fitting to the data points for actual projects in safety of one. We purposely flattened the curves in Fig. 1 to
reflect our belief that increasing the safety factor well beyond the
each of the four categories. The failure probabilities reflect their
typical values used for earth structures provides little benefit with
judgment of the relative degree of safety of the various earth
respect to the corresponding probability of failure. Discontinui-
structures. They sought a methodology to codify judgments based
ties, weak zones, wet zones, high or low permeability zones, and
on geotechnical engineering fundamentals, not just make unrea-
other features that can elude a geotechnical investigation control
soned guesses. Roberds 共1990兲 and Baecher and Christian 共2003兲
the level of safety for grossly overdesigned facilities. Terzaghi
discuss in detail methodologies to obtain subjective probabilities
共1929兲 presents a treatise of the importance of these “minor geo-
through elicitation of expert judgment. The writers generally fol-
logical defects” on the performance of earth structures. The
lowed these concepts.
changes in probability of failure near the factor of safety of one
An earlier version of Fig. 1 presented by Lambe at the First come from the consideration that engineers inherently apply more
Terzaghi Oration in 1985 and by Silva at the Peck Symposium conservatism in their choice of assumptions and selection of pa-
presentation of Lambe et al. 共1987兲 differs from Fig. 1 for low rameters as the importance of the structure and the consequences
values of safety factor. Over the years, work with earth structures of failure increase. This produces a bias in their analyses such that
on the verge of failure allowed us to refine the curves for safety their computed safety factors at low values does not represent the
factors near one. actual expected safety factor value, but something higher. Hence
In order to use Fig. 1 for future cases, the reader should deter- the probability of failure would actually be lower due to this
mine the safety factors, defined as shear strength along the sliding built-in conservatism. From the comparison presented in Fig. 2
surface divided by shear stress along the same surface, in a man- and similar comparisons with the work by Meyerhof 共1970兲,
ner consistent with the analyses used for its development. Lambe Alonzo 共1976兲, and Santamarina 共1987兲, we conclude that Fig. 1
and Silva 共1992兲 show that a wide variety of slope stability meth- makes good engineering sense and agrees with the considerations
ods give the same and correct value of shear stress. Thus any of of probability theory. We were pleased to note that Fig. 1 agrees
the commonly used method of stability analysis should suffice. with our judgment that increasing the level of engineering on
The strength determination corresponds to the best estimate of earth dams from above average 共Category II兲 to best 共Category I兲
the strength acting in the field and not necessarily the average and using a FS of 1.5 should lower the probability of failure by
strength or a “conservative” value of strength. Lambe and Silva about 100 times 共from 10−4 to 10−6兲.
共2003兲 present our approach for strength determination, but a sta- To estimate the annual probability of failure using Fig. 1, the
bility analysis where the engineer models the strength acting in engineer first determines the category for the earth structure under
the field should prove compatible. Pragmatically, this means consideration using Table 1 as a guide. With a compatible calcu-
using effective stresses for strength determination or considering lation of factor of safety 共see recommendations above兲, the engi-
the stress path to failure if using strength in terms of total stresses. neer looks vertically to the appropriate curve, Categories I–IV,
Fig. 2 compares Fig. 1 with theoretical curves computed using and reads the corresponding annual probability of failure horizon-
a lognormal distribution, probabilities of failure for earth dams by tally. Other engineers have also presented the concept of relating
Whitman 共1984兲, and an estimate of the industry experience for factor of safety to probability of failure 共Wu and Kraft 1967兲.
stability of mine pit slopes. This comparison shows that: In our practice we interpolate between curves by assigning
Table 2. Evidence of Subjective Probability Acceptance among Geotechnical Engineering Profession
Statement Reference
“Judgment is essential in evaluating and representing uncertainty.” Benjamin and Cornell 共1970兲
“. . . the probability measure may be assigned by the engineer in a wholly subjective Benjamin and Cornell 共1970兲
manner.”
“Subjectively derived probability distributions can represent the opinions of individuals Roberds 共1990兲, p. 183
or of groups.”
“Subjective probability assessment by solicitation of expert opinion is not new, and its Vick 共1994兲, p. 56
legitimacy is generally accepted.”
“Fig. 11 presents an attempt to organize the various data streams that enter into a QRA Morgenstern 共1995兲, p. 120
in different ways. The first distinction is between objective and subjective data.
Objective data arise from observation while subjective data arise from belief, usually
based on expert opinion. Objective data might be regarded as inherently superior to
subjective data, but this is not necessarily true.”
“. . . the probability of an event might be simply a subjective measure of the degree of Gilbert and McGrath 1996
belief an engineer has in a judgment or prediction.”
“Historical investigations and engineering judgment have established that the annual Christian and Urzua 共1998兲, p. 1140
probability of slides in the relevant portion of the limestone slopes of Montes de
Caneja near San Juan, Puerto Rico, is approximately 0.1.” “The annual probability of
static failure is accepted here as given and is not the subject of further discussion.”
“Determine the annual probability of failure on the basis of historical data, observation, Christian and Urzua 共1998兲, p. 1142
judgment, or other input.”
“What we in the dam business, and indeed also in many aspects of hazardous Stewart 共2000兲
industries outside the dam business, must employ to define uncertainty by means of
probabilities are subjective processes. You may discover that ‘subjective probability’ is
the predominant approach quoted in the literature for risk based dam assessments over
the past decade or so.”
“Subjective probability means that an individual asserts his or her degree of belief in a Stewart 共2000兲
particular proposition, measured by a number between 0 and 1.”
partial values for the different characteristics. For example, if a Quantification of Expert Judgment: Accepted
particular structure meets most of the Category I criteria but only Method to Determine Probabilities
benefitted from part time supervision by a qualified engineer dur-
ing construction 共a Category II attribute兲, we use the weighting Geotechnical engineers frequently approach the determination of
number in parentheses in Table 1 to compute the interpolated failure probabilities with trepidation, but it does not have to be
value as shown below: that way. The statements cited in Table 2 show that subjective
probabilities, such as quantified expert judgment, have been ac-
Design cepted for decades by practitioners and academics alike.
Investigation 0.2
Testing 0.2 Risk Assessment Methodology
Analysis and documentation 0.2
Construction 0.4 Definitions
Operation ⫹0.2
1. Hazard—condition, event, or activity that may present some
degree of risk;
Interpolated Category 1.2
2. Risk—potential for realization of some unwanted conse-
quence arising from a hazard. Risk= f 共annual probability
We then use a linearly interpolated curve located 20% of the of failure, consequences of failure兲. Risk always has two
distance between Curves I and II, or read the value sought components:
interpolating linearly between Curves I and II for the safety factor • The likelihood or probability of an event with unwanted
value under consideration. consequences occurring 共e.g., failure兲, and
Estimate
Event
Likelihood
Scenario Estimate Is It Yes
Identify Hazards Risks Adequately
(types of events) ( t d Safe? Continue
No Decision
Estimate Process
Consequences
of Each Event
Modify System
Fig. 3. Risk assessment methodology 共Bendixen 2003, personal

communication, with permission兲
• The magnitude or severity of the consequences if the event

occurs.
3. Average expected loss= annual probability of failure ⴱ cost of
failure
• Loss of human life typically considered separately from
cost. Fig. 4. FS versus p共f兲 for internal erosion
Overview Risk Assessment—Example from Engineering

Practice
Numerous industries routinely use a variety of risk assessment
tools and techniques. These approaches offer significant benefits The following section presents an example from engineering
in risk decision making, even with limited data. Simple tools and practice that applies the concepts and tools discussed above to
techniques can help organize thoughts and existing data. Risk help decide how to best use limited funds to increase the safety of
assessment helps the engineer to understand uncertainties in the tailings dams. Fig. 6 shows the plan view of the dams and reser-
project. Perhaps most importantly, risk assessment methods pro- voirs used for retention and decanting industrial fluids from a
vide a logical process of identifying hazards, evaluating the seri- mining operation.
ousness of each hazard, and assessing the effectiveness of risk The Main Dam, a nonengineered tailings pile, leaked at sev-
reduction measures. Fig. 3 shows the risk assessment process eral locations and showed evidence of numerous other malfunc-
schematically. tions 共e.g., large settlements, sinkholes, presence of old timbers,
One can perform risk assessments with various degrees of de- and very soft soils兲. Dam No. 2, designed by an experienced dam
tail or accuracy, from qualitative to quantitative. Engineers deter- engineering firm and built under their supervision, suffered from
mine failure probabilities for constructed facilities mainly as input low effective stresses near the toe since the first filling. In fact,
for risk-based decisions. As engineers, we should aim to back our
decisions with numbers, not words. Thus a practical method to
perform quantified risks assessments for routine projects can have
a positive impact on the safety of the facilities we design and help
build.
In performing a risk assessment the engineer must ask: failure
probability with respect to what? Fig. 1 relates safety factor to
annual probability of failure for slope failure. Other failure modes
require their own relationships. Fig. 4 shows a relationship similar
to that in Fig. 1 but applicable to safety with respect to internal
erosion. Fig. 4 is based on fewer data points and has not been as
thoroughly used as Fig. 1.
Tolerable Level of Risk
In any risk assessment, the engineer eventually has to face the

issue of tolerable risk, a term that we now prefer to use instead of
acceptable risk. Baecher and Christian 共2003兲 present a summary
of this important issue. In our practice, we have found that clients
can best confront this difficult issue when the level of risk is
compared to other ordinary activities. Fig. 5 presents a plot, de-
veloped for a project in Japan, which has proved useful for these
comparisons. Other risk management tools 共Morgan 1997兲 sug-
gest more strict limits than shown in Fig. 5 and reducing risks in
accordance with the as low as reasonably practicable 共ALARP兲 Fig. 5. Tolerable levels of risk 共adapted from T. W. Lambe and
principle. Associates 1982, 1989, with permission兲
ties. As is often the case, funds were limited and we were asked to
N CN Design Max. Fluid
Surface (+6345 ft.)
determine the optimal remedial measures to attain the corporate
objectives. Table 4 presents the failure consequences determined
Fluid Surface
jointly with the mining company management and operations
Pond (+6324 ft.) staff.
4 We performed stability analyses and internal erosion analyses
Design Max. Fluid
to determine, separately, the factor of safety against shear failure
Dam #2
Surface (+6240 ft.) and against internal erosion. Fig. 1 and 4 yielded the probability
Mine
of failure estimates for each failure mode which we then used to
Pond Scale in feet
Entrance Main obtain the combined failure probabilities for the following three
1 0 4,000
Dam conditions:
Fluid Surface 1. Before repairs to Dam No. 2 and Main Dam;
(+6236 ft.) South Dam
2. After Dam No. 2 but before Main Dam repairs; and
3. After repairs to Dam No. 2 and Main Dam.
The repairs to Dam No. 2 consisted mainly of relief wells to
Fig. 6. Plan view of fluid retention system dissipate the high pore pressures responsible for the low effective
stresses near the toe. The Main Dam would have required more
involved repairs, including a cutoff to control leaks. We evaluated
walking near the toe of Dam No. 2 proved difficult, with indi- each of the three conditions listed above for several fluid levels in
viduals often sinking almost to their knees during ill-fated at- Dam No. 2 and the Main Dam and different levels of geotechnical
tempts. Following the guidance in Table 1, we used the Fig. 1 surveillance. For each scenario, we determined the safety factor
Category II curve for Dam No. 2 and the Category IV curve for and corresponding annual probability of failure and calculated
the Main Dam. For example, for the Main Dam reservoir at average expected losses as the product of the annual probability
+6,234 ft., we computed a factor of safety of 1.53 and from the of failure times the cost of failure.
Fig. 1 Category IV curve, we selected a corresponding annual Table 3 shows some of the results of the risk assessment. Al-
probability of failure of 0.072. This 7.2⫻ 10−2 annual probability though the failure consequences estimates shown in Table 4 seem
of failure appears in Table 3 as the last entry for Condition 2. very approximate, they proved sufficiently detailed to differenti-
The mining company corporate management wanted to in- ate between the alternatives and yield helpful information for de-
crease the level of safety of the fluid retention system to reduce cision making. Using only four summary cost numbers we
the risk of a release that could flood the mine entrance and con- obtained a wide spread of average expected loss estimates—
taminate the pristine river downstream of the mine surface facili- enough spread to make some decisions obvious and make com-
Table 3. Risk Assessment Summary

Annual risk
Cost Annual
Scenario No. Consequences 共millions兲 probability Financial Life loss
Condition 1: Before Dam No. 2 repairs
Before Main Dam repairs
Dam No. 2 fails, release held by Main Dam Moderate $5 1.2⫻ 10−2 $60,000a None
Dam No. 2 fails, release causes Main Dam failure Major $30 5.9⫻ 10−3 $180,000a
8.2⫻ 10−2 $2,500,000b Many
Condition 2: After Dam No. 2 repairs
Before Main Dam repairs
Dam No. 2 fails, release held by Main Dam Moderate $5 1.5⫻ 10−5 $100 None
Dam No. 2 fails, release causes Main Dam failure Major $30 8.5⫻ 10−5 $2,600c
4.8⫻ 10−3 $140,000d Many
Main Dam fails 共filled to capacity兲 Major $30 8.5⫻ 10−1 $25,500,000 Many
Main dam fails 共fluid level at +6234 ft.兲 Major $30 7.2⫻ 10−2 $2,200,000 Many
Condition 3: After Dam No. 2 repairs
After Main Dam repairs
Dam No. 2 fails, release held by Main Dam Moderate $5 1.0⫻ 10−4 $5,000 None
Dam No. 2 fails, release causes Main Dam failure Major $30 4.2⫻ 10−8 $1a Many
1.0⫻ 10−4 $3,000b
Main dam fails 共filled to capacity兲 Major $$30 4.2⫻ 10−4 $12,000c Many
2.0⫻ 10−2 $600,000d
a
Main Dam reservoir empty before event.
b
Main Dam reservoir full before event.
c
Complete performance program.
d
No performance program.
Table 4. Failure Consequences
Event
Cost Lives
Magnitude Description 共dollars millions兲 lost
Major Sudden release of large quantities of stored fluid. Flooding $30 Many
occurs in the plant, the mine, and along the river. Results in
loss of life, loss of facilities, and shutdown of operations.
Severe Release of stored fluid sufficiently slow to permit evacuation $10 Several
of most personnel. Flooding occurs at the mine and the plant
but not along the river.
Moderate Slow uncontrolled release of fluid. Creates no significant $5 Few or none
flooding but produces offsite pollution and requires cleanup.
Minor Slow release of fluid. All fluid contained on site. $1 None
pelling cases for some of the options under consideration. The for routine engineering assignments. We trust that after reviewing
engineer should realize that using average expected loss esti- the methodology and studying the tailings dam example we suc-
mates, although appropriate for events with a large number of ceeded in convincing the reader of the following points:
repetitions 共e.g., automobile accidents兲, represents an approxima- 1. Quantification of expert judgment provides a practical alter-
tion for distinct cases such as the tailing dams in this example. native for determining probability of slope failure;
Table 3 indicates that repairing Dam No. 2 共i.e., relief wells, 2. Fig. 1 is consistent with fundamentals of probability theory
geotechnical performance program兲 can reduce the annual risk and better represents reality for high safety factors;
from $2,500,000 to $2,600. The annual risk from the Main Dam 3. Risk assessment based on quantification of expert judgment
with fluid at +6,234 ft 共its usual surface level兲 equals $2,200,000. provides a framework to arrive at rational management and
If the Main Dam reservoir level rises to its maximum design engineering decisions related to slope instability and other
capacity 共fluid at +6,240 ft兲, the annual risk increases to geotechnical problems; and
$25,000,000. If the owner implements the contingency plan for 4. Using the semiempirical relationships presented here, prac-
Dam No. 2, including repairing the Main Dam to comply with the ticing engineers can use this helpful risk management tool
performance criteria 共Lambe 1981兲, the annual risk decreases to a for slope instability applying their current skills.
negligible amount 共$1兲. If the owner fills the Main Dam to its In our experience, using subjective probability for risk assess-
design capacity after completing repairs, the risk increases to ments in geoengineering allows us to obtain 80% of the benefit at
$12,000. If storage requirements at the mine permit keeping Res- 20% of the cost, as suggested by Pareto’s principle or the 80/ 20
ervoir Nos. 1 permanently empty, the $3,000 annual risk from a rule, when compared to a rigorous evaluation based on analytical
Dam No. 2 failure does not justify the repair of the Main Dam. If procedures. The 80/ 20 principle, attributed to Italian civil engi-
the owner repairs the Main Dam, the annual risk of $12,000 for neer, sociologist, and economist Vilfredo Pareto 共1848–1923兲 and
full reservoir suggests that Reservoirs Nos. 1 and 4 can operate popularized by quality management pioneer Dr. Joseph Juran in
simultaneously at full capacity. The reduced risks, indicated by the United States US, states that 20% of something is always
the superscript C in Table 3, rest on a complete safety program for responsible for 80% of the results.
each dam. Without the safety programs, the risk for the repaired Ralph Peck used to tell students at the University of Illinois
Dam No. 2 increases from $2,600 to $140,000, and the risk from that if you understand a problem, you must be able to summarize
1
using a repaired Main Dam to capacity increases from $12,000 to it on one side of one piece of 8 2 in.⫻ 11 in. paper. Fig. 1 sum-
$600,000. marizes lifetime experiences of several geotechnical engineers
1
The most important conclusion we derived from the risk as- working with slope stability on one 8 2 ⫻ 11 piece of paper. Be-
sessment was that installing relief wells to decrease the pore pres- sides the application to slope stability proposed in this paper, we
sures near the toe made a lot of sense. In addition, the risk would like to see similar efforts to capture the experience of geo-
assessment provided insight into the following situations: technical engineering masters with other failure mechanisms for
1. Clearly, the Main Dam reservoir should not be allowed to the benefit of our profession.
rise to its maximum design level prior to repairs;
2. If the owner truly needed all the capacity in both reservoirs, Acknowledgments
remedial measures under consideration would allow them to
do so with a tolerable level of risk; The many projects summarized in Fig. 1 benefitted from the en-
3. Keeping the Main Dam reservoir empty would save the gineering skills of many of the writers’ colleagues, including Gre-
costly repairs and provide a fluid retention system with tol- gory Baecher, Ralph Grismala, Lilly Lee, Thom Neff, Anastasia
erable levels of risk; and Papadopoulos, José Ramos, Mario Troconis, and Robert V. Whit-
4. Geotechnical surveillance pays for itself. man. Lisa Bendixen and John Christian helped with insightful
discussions related to probability and risk. Several of the figures
in this paper were drawn by Mark Bethoney.
Conclusions
References
The main objective of this paper consists of demystifying risk
assessment by presenting a methodology that places quantitative Alonzo, E. 共1976兲. “Risk analysis of slopes and its application to slopes
risk assessment within reach of every geotechnical engineer, even in Canadian sensitive clays.” Geotechnique, 26共3兲, 453–472.
Baecher, G. B., and Christian, J. T. 共2003兲. Reliability and statistics in Lambe, T. W., Silva, F., and Lambe, P. C. 共1987兲. “Expressing the level of
geotechnical engineering, Wiley, Chichester, U.K., 319–322. stability of a slope.” The art and science of geotechnical engineering
Baecher, G. B., Paté, M.-E., and de Neufville, R. 共1980兲. “Dam failure in at the dawn of the twenty-first century: A volume honoring Ralph B.
benefit cost analysis.” J. Geotech. Engrg. Div., 106共1兲, 101–105. Peck, E. J. Cording, et al., eds., Wiley, New York, 558–588.
Benjamin, J. R., and Cornell, C. A. 共1970兲. Probability, statistics, and Meyerhof, G. G. 共1970兲. “Safety factors in soil mechanics.” Canadian
decision for civil engineers, McGraw-Hill, New York. Geotechnical Journal, 7共4兲, 349–355.
Christian, J. T., Ladd, C. C., and Baecher, G. B. 共1992兲. “Reliability and Morgan, G. C. 共1997兲. “A regulatory perspective on slope hazards and
probability in stability analysis.” Proc., Stability and Perf. of Slopes associated risks to life.” Landslide risk assessment, Proc., Int. Work-
and Embankments—II, Vol. 2, ASCE, New York, 1071–1111. shop on Landslide Risk Assessment, Honolulu, D. M. Cruden and R.
Christian, J. T., Ladd, C. C., and Baecher, G. B. 共1994兲. “Reliability Fell, eds., Balkema, Rotterdam, 285—295.
applied to slope stability analysis.” J. Geotech. Engrg., 120共12兲, Morgenstern, N. R. 共1995兲. “Managing risk in geotechnical engineering.”
2180–2207.
Proc., 10th Panamerican Conf. on Soil Mechanics and Foundation
Christian, J. T., and Urzua, A. 共1998兲 “Probabilistic evaluation of
Engineering, Vol. 4, Guadalajara, Mexico, Sociedad Mexicana de
earthquake-induced slope failure.” J. Geotech. Geoenviron. Eng.,
Mecanica de Suelos, Mexico, 102–126.
124共11兲, 1140–1143.
Roberds, W. J. 共1990兲. “Methods for developing defensible subjective
Cornforth, D. 共2005兲. Landslides in practice: Investigation, analysis, and
probability assessments.” Proc., Transportation Research Board, No.
remedial/preventative options in soils, Wiley, Hoboken, N.J.
1288, National Research Council, Washington, D.C., 183–190.
D’Andrea, R. A., and Sangrey, D. A. 共1982兲. “Safety factors for proba-
bilistic slope design.” J. Geotech. Engrg. Div., 108, 1101–1118. Santamarina, J. C. 共1987兲. “Fuzzy sets and knowledge systems in geo-
technical engineering,” Dr. Philosophy thesis, Purdue Univ., West
Duncan, J. M., and Wright, S. G. 共2005兲. Soil strength and slope stability,
Lafayette, Ind.
Wiley, Hoboken, N.J.
Gilbert, R. B., and McGrath, T. C. 共1996兲. “Seven guidelines for manag- Stewart, R. A. 共2000兲. “Dam risk management.” Proc., Int. Conf. on
ing uncertainty in geoenvironmental design.” Uncertainty in the geo- Geotechnical and Geological Engineering (GeoEng2000), Mel-
bourne, Australia, Technomic Pub., Lancaster, Pa., 721–748.
logic environment: From theory to practice, Proc., of Uncertainty ’96,
Tavares, L. V., and Serafim, J. L. 共1983兲. “Probabilistic study on failure of
ASCE, Madison, Wis., 774–796.
Kulhawy, F. H., and Phoon, K. K. 共1996兲. “Engineering judgment in the large dams.” J. Geotech. Engrg., 109共11兲, 1483–1486.
evolution from deterministic to reliability-based foundation design.” Terzaghi, K. 共1929兲. “Effect of minor geologic details on the safety of
dams.” Technical Publication 215, American Institute of Mining and
Uncertainty in the Geologic Environment: From Theory to Practice,
Proc., of Uncertainty’96, ASCE, Madison, Wis., 774–796. Metallurgical Engineers, Geology and Engineering for Dams and Res-
Lambe, T. W. 共1985兲. “Amuay landslides.” Proc., 11th Int. Conf. on Soil ervoirs, New York, 31–46
Mechanics and Foundation Engineering, Golden Jubilee Volume, San T. W. Lambe & Associates. 共1982兲. Earthquake risk to patio 4 and site
Francisco, Balkema, Boston, 137–158. 400, Longboat Key, Fla.
Lambe, T. W., Marr, W. A., and Silva, F. 共1981兲. “Safety of a constructed T. W. Lambe & Associates. 共1989兲. Earthquake risk analysis for KSS tank
facility: Geotechnical aspects.” J. Geotech. Engrg. Div., 107共3兲, 339– area, Longboat Key, Fla.
352. Vick, S. G. 共1994兲. “Geotechnical risk and reliability—From theory to
Lambe, T. W., and Silva, F. 共1992兲. “Stability analysis of an earth slope.” practice in dam safety.” Proc., Earth, Engineers, and Education—A
Proc., Stability and Performance of Slopes and Embankments-II, Symp. in Honor of Robert V. Whitman, Cambridge, Mass., MIT, Cam-
ASCE, New York, 27–69. bridge, 45–58.
Lambe, T. W., and Silva, F. 共2003兲. “Evaluating the stability of an earth Whitman, R. V. 共1984兲. “Evaluating calculated risk in geotechnical engi-
structure.” Proc., 12th Pan-American Conf. on Soil Mechanics and neering.” J. Geotech. Engrg., 1102, 145–188.
Foundation Engineering, Cambridge, Mass., Verlag Gluckauf GmbH, Wu, T. H., and Kraft, L. M. 共1967兲. “The probability of foundation
Essen, 2785–2790. safety.” J. Soil Mech. and Found. Div., 93共5兲, 213–237.
DISCUSSIONS AND CLOSURES
for the first traverse of the TR carrying the Ares V over only the
Discussion of “Probability and Risk of critical part of the CW identified by Peck.
Slope Failure” by F. Silva, T. W. Lambe, and The above methodology provided a quantitative estimate of
W. A. Marr the probability of success at 99%, based on the authors’ experi-
December 2008, Vol 134, No. 12, pp. 1691–1699.
ence. This seemed consistent with our own subjective thinking. It
Downloaded from ascelibrary.org by MASS INSTITUTE OF TECHNOLOGY on 10/16/14. Copyright ASCE. For personal use only; all rights reserved.
DOI: 10.1061/共ASCE兲1090-0241共2008兲134:12共1691兲 appears that engineers might also use the paper to evaluate the
reliability of bearing stability on a surface with a near-zero slope
angle. We look forward to the authors’ comments.
J. H. Schmertmann, F.ASCE1; and G. Filz, F.ASCE2
1
Prof. Emeritus, Dept. of Civil and Coastal Engineering, Univ. of Florida,
Gainesville, FL 32611. E-mail: schmert@ufl.edu
2
Charles E. Via Jr. Prof. of Civil and Environmental Engineering, Virginia References
Tech, Blacksburg, VA 24061.
Peck, R. B. 共1969兲. “Advantages and limitations of the observational
method in applied soil mechanics.” Ninth Rankine Lecture, Géotech-
The discussers had the opportunity to use the subject paper in an nique, 19共1兲, 171–187.
important consulting problem involving bearing stability. We
thought the paper provided very useful experience-based judg-
ments, and applying it to our consulting problem allowed us to
possibly expand the potential use of the paper.
Peck 共1969, p. 174–176兲 explained the initial problem with the Discussion of “Probability and Risk of
potential instability of NASA’s tracked transporter 共TR兲. The TR Slope Failure” by F. Silva, T. W. Lambe, and
carried the Apollo, and later the shuttle rockets, several kilometers W. A. Marr
from the vertical assembly building 共VAB兲 to the A and B launch
December 2008, Vol. 134, No. 12, pp. 1691–1699.
pads along the A and B crawlerways 共CW兲 at the Cape Kennedy
DOI: 10.1061/共ASCE兲1090-0241共2008兲134:12共1691兲
Space Center in Florida. The CWs have now experienced roughly
100 loaded TR transverses to periodically stress and improve the
CW foundation conditions as well as experiencing approximately Robert Alperstein, M.ASCE1
1
Sr. Partner and Managing Member, RA Consultants LLC, 136 Aristotle
45 years of aging since construction. However, NASA’s proposed
Way, East Windsor, NJ 08512.
new Constellation series of rockets include the Ares V, and it will
increase the TR bearing load on the CWs by approximately 35%.
Our problem: to estimate the probability of success, or reliability, The authors make a valuable presentation of the relationship be-
of the TR carrying an Ares V rocket and traversing the weakest tween quality of engineering, factor of safety, and annual prob-
foundation area under the existing CW without any distress that ability of failure. The probabilities are mainly subjective
could delay a launch. We found the subject paper useful to make probabilities based on the authors’ and their colleagues’ judgment.
the first phase of this estimate, as described below. The value of the presentation was demonstrated for certain risk
We first made a comprehensive field investigation of the most assessment situations.
critical part of the CWs as identified by Peck 共1969兲. Then we This discusser agrees that quantifying risk is useful and under-
used the information produced to model the problem and predict standing its relationship to other engineering considerations is
pore pressure generation and settlement during a first traverse even more useful. However, recognizing the judgmental factors
with the new loading. Settlement seemed acceptable but bearing that are built into Fig. 1 of the paper is extremely important in
stability remained an issue. A 2D limit equilibrium, slope stability applying it to real world projects. This discusser questions the
analysis for bearing stability gave a FS= 1.25. Then a draft copy wisdom of calculating probabilities beyond a whole number
of the subject paper became available and gave us a methodology within any order of magnitude and in using weighting factors and
for estimating reliability. interpolations. This level of calculation precision appears to be
To do this we assumed that the paper applied to all slopes, well beyond the precision and accuracy of the basic input used in
including our very flat slope of only 3°. We later assumed it also developing Fig. 1.
applied to a 2D finite-element analysis. We also estimated from Further, extreme care should be exercised in establishing de-
the Table 1 共found in the paper under Discussion兲 guide that our sign criteria for various aspects of a project. Fig. 1 may not be
overall level of engineering for the problem fell between catego- sufficient for that purpose. The authors suggest that effective
ries I and II. Using these at FS= 1.25 with the experience curves stress considerations should be used in establishing a “best esti-
in Fig. 1 共found in the paper under Discussion兲, the theoretical mate” of design shear strengths.
curves in Fig. 2 共found in the paper under Discussion兲, and the During the late 1970s and early 1980s this discusser had sig-
$107 consequences curves in Fig. 5 共found in the paper under nificant responsibility for the geotechnical aspects of the Westway
Discussion兲, we estimated that the annual probability of a bearing project in New York City. The anticipated project cost was about
stability failure fell between 10−2 to 10−4. Because previous ex- $4.3 billion then, or of the same order as the Central Artery
perience 共Peck 1969兲 showed the first traverse as the most critical, Project in Boston. The project included major landfills onto thick
we concluded that the probability of success equaled at least 99% soft organic clayey sediments lying beneath the Hudson River.
764 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2010
J. Geotech. Geoenviron. Eng. 2010.136:769-771.

The Holland and Lincoln Tunnels and the PATH tubes crossed the
project alignment. Establishing design criteria for safety factors
and developing appropriate field and laboratory testing programs
was crucial to the project.
We developed the necessary criteria and programs by review-
ing about 25 cases reported in the literature where “strength” data
were available before construction, a design safety factor was
reported, and the embankments actually failed. All the reported
cases were either type I or II. The available “strength” data types
varied from case to case and included field vane tests, UU tests,
CIU 共with pore pressure measurements兲, and direct simple shear
共DSS兲 tests. Typical design safety factors ranged from about 1.2
to about 2. Calculated safety factors after failure ranged from

about 0.6 to 1.25. We found that even with design safety factors
in the range of 1.2 to 1.6 共using conventional UU tests兲 that the
probability of failure was higher than most of us wanted to be-
lieve.
Our field and laboratory testing programs included all the
aforementioned test types 共and CAU tests兲 and included consid-
eration of the in situ effective stresses and the stress history of the
various soil types. We found significant variations in strength Fig. 1. Data from discussers’ slope stability projects plotted on Fig.
characteristics, even with high-quality testing and consideration 1 from the original paper
of effective stresses. Generally, our confidence limits with regard
to strength were about⫾ 20%. earth structures based on a subjective characterization of the
Finally, our strength data and the apparent history we learned scope of engineering in design, construction, and maintenance.
from the literature were put in front of the most senior members While most engineering undertakings cannot justify rigorous re-
of the two firms responsible for the project and subjective judg- liability analysis, the proposed method can be applied easily and
ment was applied. This involved long and intense discussions, provides a quick estimate of the reasonableness and reliability of
sometimes resulting in “We need more tests!” or “We need dif- a design. Such analysis will help engineers to reduce bias and
ferent tests!” or “I like X tests better than Y tests!” The criteria convey to clients the significance of the uncertainty inherent in
that were established include a factor of safety of at least 1.3 for the analysis and design of earth structures. It may also be used to
static loadings within the project areas having the lowest conse- demonstrate to clients the direct benefit obtained from expendi-
quences of failure; and that the analysis should be done by total tures for better engineering and engineering involvement even on
stress analysis prior to or during construction considering the ef- relatively small projects. The method is particularly helpful in
fective consolidation stresses. Effective stress analysis could be explaining the relationship between the computed factor of safety
done during construction using pore pressures measured in the and the probability of failure 共risk兲 to clients without technical
field. backgrounds. Whereas factors of safety are a somewhat abstract
Unfortunately, the project was cancelled because of environ- concept, probability of failure is easily understood by most
mental concerns. Nevertheless, this discusser hopes that these in- people, even those with a nontechnical background.
sights and experiences provide additional support to the authors’
concepts, but at the same time reveals the importance and limita-
tions of subjective judgment in establishing “hard” quantification
of risk.
Discussion of “Probability and Risk of

Slope Failure” by F. Silva, T. W. Lambe, and
W. A. Marr
December 2008, Vol. 134, No. 12, pp. 1691–1699.
DOI: 10.1061/共ASCE兲1090-0241共2008兲134:12共1691兲
Daniel R. Vanden Berge, P.E., M.ASCE1; and

Alan J. Esser, P.E., M.ASCE2
1
Geotechnical Engineer, EDP Consultants, Inc., 9375 Chillicothe Rd.,
Kirtland, OH 44094-8501. E-mail: dvberge@edpconsultants.com
2
Principal Engineer, EDP Consultants, Inc., 9375 Chillicothe Rd., Kirt-
land, OH 44094-8501. E-mail: ajesser@edpconsultants.com
The authors have provided a much-needed and elegantly simple Fig. 2. Discussers’ data plotted against curves for lognormal distri-
method of relating factors of safety and probability of failure for bution from Fig. 2 of the original paper
JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2010 / 765




The discussers have applied reliability analysis to slope stabil- determining the probability of failure would be valuable to regu-
ity projects using avoidance and stabilization techniques. Reli- lators and forensic experts.
ability analysis has been particularly useful when the project goal To address these differences in slope stability and slope stabi-
is avoidance 共i.e, locating structures an appropriate distance from lization projects, the discussers propose the inclusion of Table 1
the top of a natural slope to yield a desired factor of safety兲, and presented in this discussion. Several of the items listed in the
thus reduces the risk of future damage from slope failure. For design columns of Table 1 of the original paper have been ex-
stabilization projects, correlating FS to p共f兲 helps the stakeholders panded or revised to make the table more adaptable to natural
in a project to understand and quantify the relationship between slope and slope stabilization projects. For example, the investiga-
the incremental cost and level of risk associated with any stabili- tion and testing categories of Table 1 of the original paper em-
zation effort. Probability of failure is estimated using the method phasized sampling, identification of weak zones, and pore
proposed by Duncan 共2000兲. Coefficients of variation of 8–10% pressure measurements in the foundation soils. In Table 1 of this
are typically used for effective friction angles, and COV of discussion, these columns are expanded to include the soils com-
33–40% are used for effective cohesion. Because groundwater prising the upper parts of a slope or embankment, not just the
levels are generally monitored over a relatively short time, foundation soils. The revisions also introduce the importance of
groundwater variability requires much judgment. The discussers properly identifying topographic features. In the case of slope
have used Duncan’s Three-Sigma Rule to estimate the probable failures, surface topography may be a pivotal indicator of the
variation in groundwater levels. Unit weight is generally not var- type, extent, and progress of failure. Category III is labeled Av-
ied because its effect on the factor of safety tends to be insignifi- erage in Table 1 of the original paper, which seems to contradict
cant. Data for our projects are plotted in Fig. 1 of this discussion. the accompanying description and the level of engineering. The
discussers concur that Category III engineering is regrettably the
By correlating FS and p共f兲, the discussers have dissuaded owners
standard for much of the profession in slope engineering. The
from thinking that “any FS greater than 1 is good,” and helped to
average should lie between categories II and III, and the discuss-
involve owners more effectively in decision making.
ers suggest changing Category III to “below average,” as has been
Applying Table 1 of the original paper, the discussers find that
done in Table 1 of this discussion.
the methodology used in their stability evaluations generally falls
A new column has also been added in Table 1 to account for
in Category II, with characteristics ranging across Categories I to
owner influence on the project outcome. “Construction” and “Op-
III depending on project and client objectives. Data from these
eration and Monitoring” have been combined under a new head-
projects plot close to the Category II line proposed by the authors,
ing titled “Implementation” with a single weighting value
but fit better with the lognormal distribution curves presented in assigned to the combined columns. This downplays the influence
Fig. 2 of this discussion. This effect follows directly from assum- of construction, operation, and monitoring for projects where
ing a lognormal distribution for Pf as suggested by Duncan. these are minor factors compared to their greater influence in the
For projects involving natural slopes and slope stabilization, construction of major earth structures.
determination of the appropriate category using Table 1 of the
original paper is somewhat difficult because of its focus on major
constructed works. The construction and operation and monitor-
ing tasks as presented in the table are certainly appropriate, as the References
authors intend, for large earth structures. But these do not apply
directly where the objective is to site structures a sufficient dis- Duncan, J. M. 共2000兲. “Factors of safety and reliability in geotechnical
engineering.” J. Geotech. Geoenviron. Eng., 126共4兲, 307–316.
tance from a slope to avoid damage from slope failures while
leaving the slope untouched. In such cases it is useful to deter-
mine factors of safety and probability of failure at various dis-
tances from a slope to engage the client and others as informed
participants in the decision-making process. This information can Closure to “Probability and Risk of Slope
be used by developers, planners, zoning, and regulatory bodies
Failure” by F. Silva, T. W. Lambe, and W. A.
that desire to protect the natural environment and property values.
Some stabilization methods do not have components that require
Marr
ongoing maintenance or long-term monitoring by the engineer. December 2008, Vol. 134, No. 12, pp. 1691–1699.
With these projects, the owner may desire to take on routine DOI: 10.1061/共ASCE兲1090-0241共2008兲134:12共1691兲
monitoring responsibilities after stabilization is complete.
The owner often influences the level of design, the implemen- Francisco Silva, M.ASCE1; T. William Lambe,
tation of the design recommendations, and the decision-making Hon.M.ASCE2; and W. Allen Marr, F.ASCE3
1
process. For instance, one project may have a sophisticated owner Consulting Civil Engineer, GeoEngineering and Environment, 12 Baskin
with an engineering background or technical expertise, while at Rd., Lexington, MA 02421 共corresponding author兲. E-mail: silva@
the opposite extreme another might have an uninformed owner alum.mit.edu
2
Consulting Geotechnical Engineer, 32550 Clay Gully Rd., Myakka City,
with no interest or understanding of slope engineering. Sophisti-
FL 34251-9457.
cated owners are more likely to make decisions based on safety 3
President and CEO, Geocomp Corp., 1145 Massachusetts Ave., Boxbor-
and performance criteria, whereas with indifferent owners cost ough, MA 01719.
containment is likely to dominate decision making. Although this
factor may be beyond the control of the engineer, its influence on
the probability of failure will be significant and must be recog- We enjoyed the creative use of our work by Schmertmann and
nized. If the owner’s influence significantly increases the prob- Filz for such a critical application as NASA’s tracked transporter
ability of failure, the engineer may decide that the project is too at Cape Kennedy. We agree that our method is appropriate for this
risky to become involved. Recognition of the owner’s influence in application. In fact our Fig. 1 共of the original paper兲 should apply

to any situation where factor of safety is defined as available fies the establishment of design criteria, or performance criteria in
shear strength divided by mobilized shear stress. Fig. 1 共of the the terminology used by Lambe et al. 共1981兲, as a key early step
original paper兲 should not be used with a factor of safety defined for any design. We continue to encounter numerous examples of
as maximum allowable force divided by the applied force as this projects that run into serious difficulties due to the absence of
definition is not consistent with our method. clearly stated performance criteria early in the design process. For
We much appreciate the detailed discussion and additional risk-informed design, the engineer could select a level of safety
data provided by Vanden Berge and Esser. They used a lognormal based on tolerable risk and use Table 1 共of the original paper兲 and
relationship between factor of safety and probability of failure. Fig. 1 共of the original paper兲 to help select the corresponding
Our Fig. 2 共of the original paper兲 shows how lognormal distribu- target factor of safety for the design of the earth structure. For
tions compare to our recommended curves and the text explains more traditional designs, the engineer can better communicate to
our reasons for departing from a lognormal distribution at the low the owner the level of risk achieved using Table 1 共of the original
and high values of factor of safety. However, we would like to paper兲 and Fig. 1 共of the original paper兲 to estimate the probabil-
emphasize that, although we compared our Fig. 1 共of the original ity of failure for the design factor of safety.
paper兲 to various probability distributions to check for general Regarding accuracy and precision of probabilities from Fig. 1
compliance with probability fundamentals, Fig. 1 共of the original 共of the original paper兲, we implied with words like “semiempir-
paper兲 shows semiempirical data without concern for the type of ical” that the specific values of probability of failure in Fig. 1 共of
distribution it may represent. “It is what it is,” and we purposely the original paper兲 are estimates and therefore not presented as
avoided any fitting to predefined distributions. accurate or precise. Determining the accuracy of failure probabil-
The writers recognize that engineers work with owners with ity estimates for geotechnical events quickly becomes a daunting
widely different perspectives on the appropriate the level of en- task, and probability estimates from rigorous application of theo-
gineering for their projects. While intrigued by the concept that retical principles do not necessarily produce more accurate num-
the type of owner should have a direct impact on the probability bers than the subjective probabilities presented in our paper. As a
of failure, we believe the characteristics summarized in Table 1 well-known geotechnical engineer and probability expert jest-
共of the original paper兲 would adequately reflect the owner’s per- ingly mentioned to the senior author about three decades ago, “it
spective. We caution that publicly classifying project owners as will be much harder to prove us wrong.” But it’s not the point of
uninformed, indifferent, or unsophisticated during the normal ex- our paper to provide an accurate or precise measure of probability
ecution of an engineering assignment can also negatively influ- of failure for a given situation. Our intent is to provide an ap-
ence the final outcome in a different, but still undesirable way. We proximate means to easily estimate the likely probability of fail-
should also clarify that the characteristics summarized in Table 1 ure to help manage geotechnical risks. In this regard, the writers
共of the original paper兲 refer to both the foundation soils and the offer the following comments:
compacted soils used to construct the earth structure. 1. Much of the benefit of applying risk assessment methodolo-
Vanden Berge and Esser suggest that our recommendations do gies to engineering projects derives from the execution of the
not apply to slopes where one is trying to locate a facility far process itself and not solely from the numerical end result.
enough away to avoid the risk. We believe that our method does The systematic examination of failure mechanisms and their
handle this case; indeed many of the data points were for cases likelihood of occurrence results in a learning process for the
involving slopes with constructed facilities located above the engineering team that is difficult to duplicate without the
slope. Instead of focusing on the minimum factor of safety of the tools summarized in our paper. We emphasize that the risk
slope, which may correspond to a shallow sliding surface, one assessment methodologies complement, but do not replace,
would focus on failure surfaces that pass through the foundation the traditional deterministic engineering assessment involv-
of the facility and use the probability of failure associated with ing field exploration, determination of soil properties, engi-
that factor of safety to estimate the level of safety of the facility. neering analyses, and a measure of judgment to distill
The farther one places the facility back from the slope the higher rational recommendations from all of these important com-
the factor of safety and the lower the probability of failure. ponents. During the last three decades we observed that the
Sadly, we agree with Messrs. Vanden Berge and Esser that the estimation of failure probabilities often deterred engineers
average level of engineering should lay above our Category III, from performing a systematic evaluation of risk. We devel-
but would not relabel Category II as “average” at this time. We oped the paper’s Fig. 1 共of the original paper兲 to eliminate, or
found the characteristics defined in Category III to more closely at least soften, this obstacle.
align with what we see as typical geotechnical practice on the 2. After applying Fig. 1 共of the original paper兲 to many field
“project down the street.” Based on our collective experience we situations, we learned not to ascribe undue importance to the
felt there are two levels of practice above “average.” “Above absolute value of the failure probabilities used for a risk as-
average” projects are those where engineers go above the normal sessment. As mentioned in the paper, we should use defen-
and the ordinary and expend extra effort and money because the sible values. However, while we agree with Mr. Alperstein’s
consequences of poor performance are significant to the project sense that the absolute numbers might not have a high level
owner. A typical firm might do a few of these projects a year. of accuracy or precision, we find that approximate numbers
Then there are those requiring “best” practices where the conse- provide useful comparisons of various alternatives. One of
quences of poor performance are very large. Typical firms might the significant benefits of our method is to use it as a rational
work on one of these projects every decade. These are projects means to assess the relative value of alternatives. Do I spend
like nuclear power plants, major dams, and foundations for tall more for engineering to get to a higher level of safety and
buildings, that call upon the best we can bring to the table to lower probability of failure? Do I use a higher factor of
minimize risk for the least cost. Increasing the average level of safety to reduce the probability of failure? For these types of
practice above our Category III should constitute a goal of our questions, we think the relative changes in probability of
profession and particularly of our professional organizations. failure from Fig. 1 共of the original paper兲 are sufficiently
Robert Alperstein contributed keen insight based on major precise. When using Fig. 1 共of the original paper兲 for this
projects in the New York metropolitan area. He correctly identi- purpose, having more precision to account for the differences

between the alternatives provides tangible benefits. It’s
analogous to the engineer who carries all the intermediate
calculations to three significant digits, then rounds the final
result to two significant digits.
Alperstein’s discussion of 25 cases in the New York area is
quite interesting. We urge him to publish these cases as the results
of their work could be very beneficial to future undertakings in
the area.
In closing, we again thank the discussers for their comments
and ideas. Their contributions enrich the value of our paper. We
welcome feedback from others, which may be provided via
e-mail to the corresponding author at silva@alum.mit.edu.
References
Lambe, T. W., Marr, W. A., and Silva-Tulla, F. 共1981兲. “Safety of a

constructed facility: Geotechnical aspects.” J. Geotech. Engrg. Div.,
107共3兲, 339–352.
Fig. 1. Comparison of single-point and double oedometer tests on

samples of silty sand from 1.5 m depth
Discussion of “Oedometer Behavior of an
Artificial Cemented Highly Collapsible Soil” samples. One specimen was used to measure the collapse poten-
by G. M. Medero, F. Schnaid, and W. Y. Y. tial at a vertical stress of 200 kPa in a single-point oedometer test
Gehling as described by the authors. The other two specimens were used
June 2009, Vol. 135, No. 6, pp. 840–843.
in a double oedometer test in which one specimen was maintained
DOI: 10.1061/共ASCE兲1090-0241共2009兲135:6共840兲 at its natural water content throughout while the final specimen
was soaked at a vertical stress of 11 kPa prior to beginning the
test. The three specimens from the block sample recovered from a
William J. Neely, P.E., M.ASCE1 depth of 1.5 m were found to be practically identical. Specimen
1
Senior Vice President and Chief Engineer, Retaining Wall Services, Inc.,
dry densities averaged 16.83⫾ 0.04 kN/ m3, natural water con-
1660 Hotel Circle North, Ste. 304, San Diego, CA 92108. E-mail:
wneely@retainingwallservices.com
Using an artificially cemented geomaterial, the authors found col-

lapse potential measured in single-point oedometer tests to be
higher than for initially identical specimens subjected to double
oedometer tests. The discusser’s experience with both single and
double oedometer testing of several collapsible natural soils indi-
cates there is usually very little difference between the results
from the two types of test, provided relatively undisturbed soil
samples are recovered from the field.
As the authors correctly point out, the major difficulties in
studying collapsible soils using field samples are associated with
disturbance of the structure of the soil and the inherent variability
of natural deposits. For many collapsible soils these difficulties
may be largely avoided by using hand-dug or machine-excavated
test pits to recover large block samples from which one or more
oedometer specimens may be trimmed. The size of the block
sample 共typically 300 mm on side兲 aids in minimizing distur-
bance to the soil structure, except near the edges of the block;
variability is reduced by trimming individual test specimens as
close to one another as possible.
As part of an investigation of a site on collapsible soils in
Johannesburg, South Africa, the discusser carried out a series of
single-point and double oedometer tests on specimens of silty fine
sand trimmed from undisturbed block samples cut from the side-
walls of inspection pits. Block samples were recovered from
depths of 1.5 m and 2.5 m below site grade. Three specimens for Fig. 2. Comparison of single-point and double oedometer tests on
consolidometer testing were cut from each of the two block samples of silty sand from 2.5 m depth

cemented artificial materials, with collapse in the uncemented
material being complete at relatively low stress levels. For the
cemented material collapse potential reaches a maximum at less
than 100 kPa and approaches zero at about 200 kPa, while the
natural soil exhibits peak collapse potential at 250 kPa, with con-
tinuing measurable collapse potential to vertical stresses of
1,000 kPa or more.
Closure to “Oedometer Behavior of an

Artificial Cemented Highly Collapsible Soil”

by G. M. Medero, F. Schnaid and W. Y. Y.
Gehling
June 2009, Vol. 135, No. 6, pp. 840–843.
DOI: 10.1061/共ASCE兲1090-0241共2009兲135:6共840兲
Fig. 3. Variation of collapse potential with applied vertical stress
G. M. Medero1; F. Schnaid2; and W. Y. Y. Gehling3
1
Lecturer, School of the Built Environment, Heriot-Watt Univ., William
Arrol Bldg, Edinburgh, EH14 4AS, UK. E-mail: g.medero@hw.ac.uk
tents were in the range 16.8–16.9%, and initial void ratios aver- 2
Assoc. Prof., Dept. of Civil Engineering, Federal Univ. of Rio Grande do
aged 0.853⫾ 0.005. The specimens trimmed from the block Sul, Av. Osvaldo Aranha 99, 90035-190, Porto Alegre, RS, Brazil.
sample at 2.5 m depth exhibited similar uniformity, with dry den- E-mail: fernando@ufrgs.br
sities of 15.44⫾ 0.01 kN/ m3, natural water contents between 20.1 3
Assoc. Prof., Dept. of Civil Engineering, Federal Univ. of Rio Grande do
and 20.3%, and initial void ratios of 1.115⫾ 0.002. Sul, Av. Osvaldo Aranha 99, 90035-190, Porto Alegre, RS, Brazil.
The results of the single-point and double oedometer tests are E-mail: wai@ufrgs.br
presented in Fig. 1 for the sample at 1.5 m and in Fig. 2 for the
sample from a depth of 2.5 m. For both samples, the change in
void ratio upon inundation at a vertical stress of 200 kPa in the The writers appreciate the discusser’s comments and interest in
single oedometer test is practically the same as that measured in our paper. Regarding the points raised, clarifications are given
the corresponding double oedometer test as the difference in void herein.
ratio between the void ratio versus log 共vertical stress兲 curves for The writers remarked in the paper that it is difficult to inves-
one specimen maintained at its natural water content throughout tigate the parameters that control the collapse phenomenon by
and an otherwise identical specimen that was soaked at 11 kPa at using natural samples because of the heterogeneity of natural de-
the beginning of the test, at the same value of vertical stress. The posits and the difficulty of taking samples from the field without
double oedometer tests also indicate that very little collapse oc- destroying the structure. From the writers’ experience with highly
curs upon inundation at stress levels below those corresponding to porous residual collapsible soils, it is problematic to obtain undis-
the existing vertical stress in the field. At high stress levels turbed samples from the field. Moreover, one of the biggest prob-
共 ⬎ 1,000 kPa兲 both void ratio versus log 共vertical stress兲 curves in lems inherent in testing natural collapsible soils is the fact that it
the double oedometer test tend to merge with that of the single- is impossible to control parameters like initial void ratio and level
point test. of cementation. By controlling the preparation conditions it is
The data in Figs. 1 and 2 have been used to illustrate the way possible to evaluate the influence of the variation of both initial
in which collapse potential varies with vertical stress. The double void ratio and cement content. The authors highlighted in the
oedometer curves are used for this purpose since collapse poten- conclusions that although the samples produced by the technique
tial is measured at only one value of vertical stress in the single- may not reproduce the overall structure complexity of natural
point test. The results are presented in Fig. 3 where the values of soils, it is useful to investigate the variables that control soil col-
collapse potential from the single-point tests are seen to lie close lapsibility 共void ratio, cementation degree, and suction level兲.
to the curve for the corresponding double oedometer test. The This understanding is relevant to the development of effective
collapse potential of the silty sand at Johannesburg increases rap- constitutive models design to describe the collapse phenomenon
idly with increasing vertical stress, reaching a maximum of about taking into account these key parameters.
4% around 250 kPa, before declining at higher stress levels as the The discusser presented two sets of results of double and
metastable structure is gradually destroyed. Although the authors’ single oedometer tests in which the obtained collapse potential
manmade geomaterial is not intended to mimic all aspects of the was similar using both the double and single oedometer tests. The
behavior of collapsible natural soils, it is important to note some writers pointed out that the use of the double oedometer test to
of the differences between their artificial material and at least one predict collapse potential may underestimate the value of collapse
natural soil so that improvements may be made to the experimen- potential, mainly at low vertical stresses. The discusser’s results
tal material. The amount of collapse in the natural silty sand is compare just one single-point oedometer test wetted at 200 kPa
significantly less than for the artificial material, whether cemented with a double oedometer results. The writers would like to under-
or not, although this may not be the case for all natural materials. line that in a double oedometer test the samples are not exposed
The greater collapse may be the result of weaker bonding be- to the stress and wetting paths that a collapsible soil would be
tween individual particles in the artificial material or penetration exposed to in the field. This should be considered and this method
of sand particles into the expanded polystyrene particles. The au- of testing is not necessarily representative of an in situ stress and
thors’ Fig. 2 shows very different behavior for cemented and un- wetting paths.

boundaries was created using the LS-DYNA finite-element pro-
Discussion of “Near-Field Effects on gram. Material damping was set as zero in the model. Four hy-
Array-Based Surface Wave Methods with pothetical soil profiles, cases 1 to 4, defined in the paper in Fig. 1
Active Sources” by S. Yoon and G. J. Rix were used for numerical simulations. Total numbers of receivers
used in the array were 10, 15, and 20. The receiver-to-receiver
March 2009, Vol. 135, No. 3, pp. 399–406.
DOI: 10.1061/共ASCE兲1090-0241共2009兲135:3共399兲
共RR兲 distance was kept at 1 m with various source-to-first re-
ceiver 共SR兲 distances of 1, 3, 5, 7, and 10 m. Dispersion curves
were generated for each soil profile with respect to the total num-
A. M. W. Aung1 and E. C. Leong2 ber of receivers and SR distances. Comparison of the dispersion
1
Project Officer, School of Civil & Environmental Engineering, Nanyang
Technological Univ., Blk N1, 50 Nanyang Ave., Singapore 639798 curves with the authors’ Fig. 2 is shown in Fig. 1 of this discus-
2
Assoc. Prof., School of Civil & Environmental Engineering, Nanyang sion for 20 receivers, using the notation and format used by Yoon
Technological Univ., Blk N1, 50 Nanyang Ave., Singapore 639798 共2005兲. For instance, 1SR-1RR20 means source-to-first receiver
distance of 1 m with 20 receivers in a uniformed array of 1 m

receiver-to-receiver distance. Fig. 1 shows that the overall trends
The authors presented a comprehensive account on the near-field of dispersion curves and the apparent curves are in close agree-
effects in surface wave methods with active sources that employ ment for each soil profile.
multiple receivers in a linear array. The authors concluded that the The Case 1 soil profile in the paper consisting of a homog-
near-field effects are primarily associated with the underestima- enous half-space will be used for discussion purposes as the dis-
tion of measured Rayleigh phase velocities compared to the plane persion curve defined by plane Rayleigh phase velocity is not
Rayleigh phase velocities. This discussion will examine the nor- affected by the soil profile and is least ambiguous to discuss the
malized parameters suggested by the authors to approximate deviation of Rayleigh phase velocity from the plane Rayleigh
maximum probable errors in measured Rayleigh phase velocities phase velocity. For the Case 1 profile, the trend shown in Figs.
using the finite-element method and its application to surface 2共a–c兲 generally agrees with the results of Yoon 共2005兲 in which
wave tests. Rayleigh phase velocities are underestimated over the low-
The discussers performed numerical simulations of array- frequency range. However, Rayleigh phase velocities largely con-
based continuous surface waves tests 共CSWS兲 involving an active verge to plane Rayleigh phase velocities from 40 Hz onward,
source to investigate near-field effects. A two-dimensional axi- irrespective of the number of receivers or SR distance. Therefore,
symmetric finite-element mesh incorporating nonreflecting the results indicate that Rayleigh phase velocities may be mea-
Fig. 1. Dispersion curves from numerical simulations: 共a兲 Case 1—20 receivers; 共b兲 Case 2—20 receivers; 共c兲 Case 3—20 receivers; and 共d兲 Case
4—20 receivers

Fig. 2. Dispersion curves and normalized parameters for Case 1 profile
sured with a high degree of accuracy at frequency range beyond tance from the excitation source due to the effect of geometric
40 Hz. The results also suggest that the accuracy of Rayleigh spreading and material damping.
phase velocity measurement is more strongly influenced by fre- To illustrate the practical limitations of a surface wave test,
quency rather than the number of receivers or SR distance. numerical simulations were performed for the Case 1 profile with
The dispersion results from the Case 1 profile shown in Figs. zero material damping using a maximum harmonic loading of
2共a–c兲 are replotted using the normalized Rayleigh wave velocity 70 kg, typical of mass vibrator used in surface wave test. The
and normalized array center distance defined by the authors in displacement time histories at receiver locations of 1, 3, and 15 m
Fig. 2共d兲. The trend in Fig. 2共d兲 is in general agreement with the for a frequency run of 10 Hz are shown in Fig. 4共a兲. The maxi-
trend obtained by the authors in that the deviation of Rayleigh mum displacement amplitude A observed at various receiver lo-
phase velocity from the plane Rayleigh phase velocity reduces cations are plotted against the square root of the distance r from
with increasing normalized array center distance. These results the source in Fig. 4共b兲. Fig. 4共a兲 shows that the signal amplitude
also suggest that near-field effects may be manifested both in decreases with increasing distance. No signals can be detected
terms of over and underestimation of Rayleigh phase velocities beyond 15 m. Fig. 4共b兲 depicts the well-known attenuation of
even in an idealized soil profile such as a homogenous half-space. amplitude with the square root of distance for surface waves.
Fig. 2共d兲 also shows that the error in Rayleigh phase velocity is These observations suggest that it may not be practical to locate
less than 5% for normalized array center distance greater than 2.
This observation agrees well with the conclusion made by the
authors.
In making reference to the normalized array center distance, it
is noteworthy that receivers must be located at a great distance
from the source for one to measure Rayleigh phase velocities with
minimal errors in the low-frequency range. Fig. 3 shows SR dis-
tances for various numbers of receivers corresponding to the nor-
malized array center distance of 2 at different frequencies. It can
be seen that source-to-first receiver distance must be about 70 m
to minimize the near-field effects at 5 Hz.
However, it may be difficult to implement a field surface wave
test with array configurations corresponding to normalized array
center distance of 2 and above. Apart from limitations imposed by
site conditions and the test equipment, it may also be impractical
to receive signals at receivers if they are located at a great dis- Fig. 3. SR distances for normalized array center distance of 2

(a) Displacement-Time History (b) Attenuation of Amplitude with Distance
Fig. 4. Attenuation of displacement amplitude: 共a兲 displacement-time history; 共b兲 attenuation of amplitude with distance
receivers far away from the excitation source in field surface curves based on an assumed soil profile and iterative matching of
wave tests. On the other hand, Rayleigh phase velocities may not synthetic curve with the field dispersion curve by varying stiff-
be accurately measured especially in the low-frequency range if ness distribution of soil profile. Although this approach could in-
the receivers are located close to the excitation source. Therefore, volve high computational cost at times, Matthews et al. 共1996兲
it is likely that over- or underestimation of Rayleigh phase veloci- highlighted the principal advantage of the finite-element approach
ties are inherent in field surface wave tests. as having the ability to model near-field effects and complex sub-
On this note, the authors suggested that lower-bound values in surface geometries.
the plots of the normalized parameters for different types of soil In conclusion, near-field effects are inevitable in surface wave
profiles may be used to estimate the maximum probable error in tests with active sources and are largely unknown in that they
the measured Rayleigh phase velocities depending on the normal- may be manifested in terms of over- or underestimation of Ray-
ized array center distance. However, it is important to highlight leigh phase velocities depending on soil profiles. It may be im-
that scatter of normalized parameters can vary greatly depending practical to conduct a surface wave test with large normalized
on soil profiles. Near-field effects may also be present both in array center distance due to geometric and material damping. It is
terms of over and underestimation of measured Rayleigh phase an equally difficult task to estimate the probable errors in terms of
velocity. In addition, soil profiles are largely unknown in most upper- or lower-bound values using plots of the normalized pa-
field tests. Therefore, one cannot be certain of the magnitude of rameters. The finite-element method, which accounts for near-
lower-bound values to be taken as reference. It could be the case field effects, is one possible method to invert dispersion curves
of having to take the upper-bound values in situations where the from surface wave tests with active sources.
measured Rayleigh phase velocities are overestimated.
The accuracy of measured Rayleigh phase velocities is espe-
cially important if the inversion algorithm to deduce the soil pro- References
file is based on the assumption of plane Rayleigh phase velocities.
It is also unfortunate that inaccuracy of determination of Rayleigh
Matthews, M. C., Hope, V. S., and Clayton, C. R. I. 共1996兲. “The use of
phase velocities due to near-field effects occurs in low frequency surface waves in the determination of ground stiffness profiles.”
waves because it may invariably affect the depth of investigation Proc., Instn Civ. Engrs Geotech. Engng, 119共2兲, 84–95.
in surface wave tests. In view of this, finite-element approach as Yoon, S. 共2005兲. “Array-based measurement of surface wave dispersion
used in this discussion may be used for inversion. The finite- and attenuation using frequency-wave number analysis.” Ph.D. disser-
element method involves the generation of synthetic dispersion tation, Georgia Institute of Technology, Atlanta, Ga.

Table 1: Earth Structure Categories and Characteristics
DESIGN
LEVEL OF OPERATION AND
Analyses and CONSTRUCTION
ENGINEERING Investigation Testing MONITORING
Documentation
I (Best) • Evaluate design and • Run lab tests on undisturbed • Determine FS using • Full time supervision by • Complete performance
performance of nearby specimens at field effective stress parameters qualified engineer program including comparison
structures conditions based on measured data • Construction control tests by between predicted and
Facilities with • Analyze historic aerial • Run strength test along field (geometry, strength, pore qualified engineers and measured performance (e.g.,
high failure photographs effective and total stress pressure) for site technicians pore pressure, strength,
consequences • Locate all non-uniformities paths • Consider field stress path in • No errors or omissions deformations)
(soft, wet, loose, high or • Run index field tests (e.g., stability determination • Construction report clearly • No malfunctions (slides,
low permeability zones) field vane, cone • Prepare flow net for documents construction cracks, artesian heads)
• Determine site geologic penetrometer) to detect all instrumented sections activities • Continuous maintenance by
history soft, wet, loose, high or low • Predict pore pressures and trained crews
• Determine subsoil profile permeability zones other relevant performance
using continuous sampling • Calibrate equipment and parameters (e.g., stress,
• Obtain undisturbed samples sensors prior to testing deformation, flow rates) for
for lab testing of foundation program instrumented section
soils • Have design report clearly
• Determine field pore document parameters and
pressures analyses used for design
• No errors or omissions
0.20 0.20 • Peer review 0.20 0.2 0.20
II (Above • Evaluate design and • Run standard lab tests on • Determine FS using • Part-time supervision by • Periodic inspection by
performance of nearby undisturbed specimens effective stress parameters qualified engineer qualified engineer
Average) structures • Measure pore pressure in and pore pressures • No errors or omissions • No uncorrected malfunctions
• Exploration program strength tests • Adjust for significant • Selected field measurements
Ordinary facilities tailored to project • Evaluate differences differences between field • Routine maintenance
conditions by qualified between laboratory test stress paths and stress path
engineer conditions and field implied in analysis that
0.40 conditions 0.40 could affect design 0.40 0.40 0.40
III (Average) • Evaluate performance of • Index tests on samples from • Rational analyses using • Informal construction • Annual inspection by qualified
nearby structures site parameters inferred from supervision engineer
• Estimate subsoil profile index tests • No field measurements
Unimportant or from existing data and • Maintenance limited to
temporary borings emergency repairs
facilities with low
failure
consequences 0.60 0.60 0.60 0.60 0.60
IV (Poor) • No field investigation • No laboratory tests on • Approximate analyses using • No construction supervision • Occasional inspection by non-
samples obtained at the site assumed parameters by qualified engineer qualified person
• No construction control • No field measurements
Little or no tests.
engineering 0.80 0.80 0.80 0.80 0.80

Probability Risk Paper As Published Discussions Updated Plots

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Probability Risk Paper As Published Discussions Updated Plots

Încărcat de

Drepturi de autor:

Formate disponibile

Probability and Risk of Slope Failure

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2008 / 1691

1692 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2008

Level of Analyses and

1694 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2008

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2008 / 1695

Fig. 3. Risk assessment methodology 共Bendixen 2003, personal

• The magnitude or severity of the consequences if the event

Overview Risk Assessment—Example from Engineering

Tolerable Level of Risk

In any risk assessment, the engineer eventually has to face the

1696 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2008

Table 3. Risk Assessment Summary

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2008 / 1697

1698 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2008

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2008 / 1699

764 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2010

J. Geotech. Geoenviron. Eng. 2010.136:769-771.

to about 2. Calculated safety factors after failure ranged from

Discussion of “Probability and Risk of

Daniel R. Vanden Berge, P.E., M.ASCE1; and

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2010 / 765

J. Geotech. Geoenviron. Eng. 2010.136:769-771.

J. Geotech. Geoenviron. Eng. 2010.136:769-771.

J. Geotech. Geoenviron. Eng. 2010.136:769-771.

J. Geotech. Geoenviron. Eng. 2010.136:769-771.

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2010 / 769

J. Geotech. Geoenviron. Eng. 2010.136:769-771.

770 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2010

J. Geotech. Geoenviron. Eng. 2010.136:769-771.

e-mail to the corresponding author at silva@alum.mit.edu.

Lambe, T. W., Marr, W. A., and Silva-Tulla, F. 共1981兲. “Safety of a

Fig. 1. Comparison of single-point and double oedometer tests on

Using an artificially cemented geomaterial, the authors found col-

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2010 / 771

J. Geotech. Geoenviron. Eng. 2010.136:769-771.

Closure to “Oedometer Behavior of an

Artificial Cemented Highly Collapsible Soil”

772 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2010

J. Geotech. Geoenviron. Eng. 2010.136:769-771.

distance of 1 m with 20 receivers in a uniformed array of 1 m

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2010 / 773

J. Geotech. Geoenviron. Eng. 2010.136:769-771.

Fig. 2. Dispersion curves and normalized parameters for Case 1 profile

774 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2010

J. Geotech. Geoenviron. Eng. 2010.136:769-771.

(a) Displacement-Time History (b) Attenuation of Amplitude with Distance

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2010 / 775

J. Geotech. Geoenviron. Eng. 2010.136:769-771.

S-ar putea să vă placă și