0% found this document useful (0 votes)
303 views9 pages

(2004) Wahl - Uncertainty of Predictions of Embankment Dam Breach Parameters PDF

Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
303 views9 pages

(2004) Wahl - Uncertainty of Predictions of Embankment Dam Breach Parameters PDF

Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 9

Uncertainty of Predictions of Embankment Dam

Breach Parameters
Tony L. Wahl1

Abstract: Risk assessment studies considering the failure of embankment dams often require the prediction of basic geometric and
temporal parameters of a breach, or the estimation of peak breach outflows. Many of the relations most commonly used to make these
predictions were developed from statistical analyses of data collected from historic dam failures. The prediction uncertainties of these
methods are widely recognized to be very large, but have never been specifically quantified. This paper presents an analysis of the
uncertainty of many of these breach parameter and peak flow prediction methods. Application of the methods and the uncertainty analysis
are illustrated through a case study of a risk assessment recently performed by the Bureau of Reclamation for a large embankment dam
in North Dakota.
DOI: 10.1061/共ASCE兲0733-9429共2004兲130:5共389兲
CE Database subject headings: Dam failure; Uncertainty analysis; Peak flow; Erosion; Dams, embankment; Risk management.

Introduction rather is idealized as a parametric process, defined by the shape of


the breach, its final size, and the time required for its development
Risk assessment studies considering the failure of embankment 共often called the failure time兲. Breaches in embankment dams are
dams often make use of breach parameter prediction methods that usually assumed to be trapezoidal, so the shape and size of the
have been developed from analysis of historic dam failures. Simi- breach are defined by a base width and side slope angle, or more
larly, predictions of peak breach outflow can also be made using simply by an average breach width.
relations developed from case study data. This paper presents an The failure time is a critical parameter affecting the outflow
analysis of the uncertainty of many of these breach parameter and hydrograph and the consequences of dam failure, especially when
peak flow prediction methods, making use of a previously com- populations at risk are located close to a dam so that available
piled database 共Wahl 1998兲 of 108 dam failures. Subsets of this warning and evacuation time dramatically affect loss of life. For
database were used by other investigators to develop many of the the purpose of routing a dam-break flood wave, breach develop-
relations examined. ment begins when a breach has reached the point at which the
The paper begins with a brief discussion of breach parameters volume of the reservoir is compromised and failure becomes im-
and prediction methods. The uncertainty analysis of the various minent. During the breach development phase, outflow from the
methods is presented next, and finally, a case study is offered to dam increases rapidly. The breach development time ends when
illustrate the application of several breach parameter prediction the breach reaches its final size; in some cases, this may also
methods and the uncertainty analysis to a risk assessment recently correspond to the time of peak outflow through the breach, but for
performed by the Bureau of Reclamation for a large embankment relatively small reservoirs the peak outflow may occur before the
dam in North Dakota. breach is fully developed. The breach development time as de-
scribed above is the parameter intended to be predicted by most
failure time prediction equations.
Breach Parameters The breach development time does not include the potentially
long preceding period described as the breach initiation phase
Dam-break flood routing models 关e.g., DAMBRK 共Fread 1984兲 共Wahl 1998兲, which can also be important when considering
and FLDWAV 共Fread 1993兲兴 simulate the outflow from a reservoir available warning and evacuation time. This is the first phase of
and through the downstream valley resulting from a developing an overtopping failure, during which flow overtops a dam and
breach in a dam. These models focus their computational effort may erode the downstream face, but does not create a breach
on the routing of the breach outflow hydrograph. The develop- through the dam that compromises the reservoir volume. If the
ment of the breach is not simulated in any physical sense, but overtopping flow were quickly stopped during the breach initia-
tion phase, the reservoir would not fail. In an overtopping failure,
1
Hydraulic Engineer, U.S. Dept. of the Interior, Bureau of the length of the breach initiation phase is important, because
Reclamation, Water Resources Research Laboratory D-8560, P.O. Box breach initiation can potentially be observed and may thus trigger
25007, Denver, CO 80225-0007. E-mail: [email protected] warning and evacuation. Unfortunately, there are few tools pres-
Note. Discussion open until October 1, 2004. Separate discussions
ently available for predicting the length of the breach initiation
must be submitted for individual papers. To extend the closing date by
one month, a written request must be filed with the ASCE Managing
phase.
Editor. The manuscript for this paper was submitted for review and pos- During a seepage-erosion 共piping兲 failure, the delineation be-
sible publication on June 25, 2002; approved on September 25, 2003. tween breach initiation and breach development phases is less
This paper is part of the Journal of Hydraulic Engineering, Vol. 130, apparent. In some cases, seepage-erosion failures can take a great
No. 5, May 1, 2004. ©ASCE, ISSN 0733-9429/2004/5-389–397/$18.00. deal of time to develop. In contrast to the overtopping case, the

JOURNAL OF HYDRAULIC ENGINEERING © ASCE / MAY 2004 / 389


loading that causes a seepage-erosion failure cannot normally be flow from a breached dam. These methods are used for
removed quickly, and the process does not take place in full view, reconnaissance-level work and for checking the reasonability of
except that the outflow from a developing pipe can be observed dam-break outflow hydrographs developed from estimated breach
and measured. One useful way to view seepage-erosion failures is parameters. This paper considers the relations by Kirkpatrick
to consider three possible conditions: 共1977兲, SCS 共1981兲, Hagen 共1982兲, Bureau of Reclamation
1. Normal seepage outflow, with clear water and low flow rates; 共1982兲, MacDonald and Langridge-Monopolis 共1984兲, Singh and
2. Initiation of a seepage-erosion failure with cloudy seepage Snorrason 共1984兲, Costa 共1985兲, Evans 共1986兲, Froehlich 共1995b兲,
water that indicates a developing pipe, but flow rates are still and Walder and O’Connor 共1997兲.
low and not rapidly increasing. Corrective actions might still All of these methods, except Walder and O’Connor, are
be possible that would heal the developing pipe and prevent straightforward regression relations that predict peak outflow as a
failure. function of various dam and/or reservoir parameters, with the
3. Active development phase of a seepage-erosion failure in relations developed from analyses of case study data from real
which erosion is dramatic and flow rates are rapidly increas- dam failures. In contrast, Walder and O’Connor’s method is based
ing. Failure cannot be prevented. upon an analysis of numerical simulations of idealized cases
Only the length of the last phase is important when determining spanning a range of dam and reservoir configurations and erosion
the breach hydrograph from a dam, but both the breach initiation scenarios. An important parameter in their method is an assumed
and breach development phases may be important when consid- vertical erosion rate of the breach; for reconnaissance-level esti-
ering warning and evacuation time. Again, as with the overtop- mating purposes, they suggest that a range of reasonable values is
ping failure, there are few tools available for estimating the length 10 to 100 m/h, based on an analysis of case study data. The
of the breach initiation phase. method makes a distinction between so-called large-reservoir/
fast-erosion and small-reservoir/slow-erosion cases. In large-
Predicting Breach Parameters reservoir cases, the peak outflow occurs when the breach reaches
its maximum depth, before there has been any significant draw-
To carry out a dam-break flood routing simulation, breach param- down of the reservoir. In this case, the peak outflow is insensitive
eters must be estimated and provided as inputs to the dam-break to the erosion rate. In the small-reservoir case, there is a signifi-
and flood routing simulation model. Several methods are avail- cant drawdown of the reservoir as the breach develops, and thus
able for estimating breach parameters; a summary of the available the peak outflow occurs before the breach erodes to its maximum
methods was provided by Wahl 共1998兲. The simplest methods depth. Peak outflows for small-reservoir cases are dependent on
共Johnson and Illes 1976; Singh and Snorrason 1984; Bureau of the vertical erosion rate and can be dramatically smaller than for
Reclamation 1988兲 predict the average breach width as a linear large-reservoir cases. The determination of whether a specific
function of either the height of the dam or the depth of water situation is a large- or small-reservoir case is based on a dimen-
stored behind the dam at the time of failure. Slightly more sophis- sionless parameter incorporating the embankment erosion rate,
ticated methods predict more specific breach parameters, such as reservoir size, and change in reservoir level during the failure.
breach base width, side slope angles, and failure time, as func- Thus, so-called large-reservoir/fast-erosion cases can occur even
tions of one or more dam and reservoir properties, such as storage with what might be considered ‘‘small’’ reservoirs and vice versa.
volume, depth of water at failure, depth of breach, etc. All of This refinement is not present in any of the other peak flow pre-
these methods are based on regression analyses of data collected diction methods.
from actual dam failures. The database of dam failures used to
develop these relations is relatively lacking in data from failures
of large dams, with about 75% of the cases having a height less Developing Uncertainty Estimates
than 15 m 共Wahl 1998兲.
Physically based simulation models are available to aid in the In a typical risk assessment study, a variety of loading and failure
prediction of breach parameters. None are widely used at this scenarios are analyzed. This allows the study to incorporate vari-
time, but the most notable is the National Weather Service ability in antecedent conditions and the probabilities associated
共NWS兲-BREACH model 共Fread 1988兲. These models simulate with different loading conditions and failure scenarios. The un-
the hydraulic and erosion processes associated with flow over an certainty of key parameters 共e.g., material properties兲 is some-
overtopping dam or through a developing piping channel. times considered by creating scenarios in which analyses are car-
Through such a simulation, an estimate of the breach parameters ried out with different parameter values and a probability of
may be developed for use in a dam-break flood routing model, or occurrence assigned to each value of the parameter. Although the
the outflow hydrograph at the dam can be predicted directly. The uncertainty of breach parameter predictions is often very large,
primary weakness of the NWS-BREACH model, and other simi- there have previously been no quantitative assessments of this
lar models, is the fact that they do not adequately model the uncertainty, and thus breach parameter uncertainty has not been
headcut-type erosion processes that dominate the breaching of incorporated into most risk assessment studies.
cohesive-soil embankments 共e.g., Hanson et al. 2002兲. Recent It is worthwhile to consider breach parameter prediction un-
work by the Agricultural Research Service 共e.g., Temple and certainty in the risk assessment process because the uncertainty of
Moore 1997兲 on headcut erosion in earth spillways has shown breach parameter predictions is likely to be significantly greater
that headcut erosion is best modeled with methods based on en- than all other factors, and could thus dramatically influence the
ergy dissipation. outcome. For example, Wahl 共1998兲 used many of the available
relations to predict breach parameters for 108 documented case
Predicting Peak Outflow studies and plot the predictions against the observed values. Pre-
diction errors of ⫾75% were not uncommon for breach width,
In addition to the prediction of breach parameters, many investi- and prediction errors for failure time often exceeded one order of
gators have proposed simplified methods for predicting peak out- magnitude. Most relations used to predict failure time are conser-

390 / JOURNAL OF HYDRAULIC ENGINEERING © ASCE / MAY 2004


Fig. 1. Predicted and observed breach widths 共Wahl 1998兲, plotted arithmetically 共top兲 and on logarithmic scales 共bottom兲

vatively designed to underpredict the reported time more often The uncertainty analysis was performed using the database
than they overpredict, but overprediction errors of more than one- presented in Wahl 共1998兲, with data on 108 case studies of actual
half of an order of magnitude did occur several times. embankment dam failures, collected from numerous sources in
The first question that must be addressed in an uncertainty the literature. The majority of the available breach parameter and
analysis of breach parameter predictions is how to express the peak flow prediction equations were applied to this database of
results. The case study datasets used to develop most breach pa- dam failures, and the predicted values were compared to the ob-
rameter prediction equations include data from a wide range of served values. Computation of breach parameters or peak flows
dam sizes, and thus, regressions in log–log space have been com- was straightforward in most cases. A notable exception was the
monly used. Fig. 1 shows the observed and predicted breach peak flow prediction method of Walder and O’Connor 共1997兲,
widths as computed by Wahl 共1998兲 in both arithmetically scaled which requires that the reservoir be classified as a large- or small-
and log–log plots. In the arithmetic plots, it would be difficult to reservoir case. In addition, in the case of the small-reservoir situ-
draw in upper and lower bound lines to define an uncertainty ation, an average vertical erosion rate of the breach must be esti-
band. In the log–log plots, data are scattered approximately mated. The Walder and O’Connor method was applied only to
evenly above and below the lines of perfect prediction, suggesting those dams that could be clearly identified as large-reservoir
that uncertainties would best be expressed as a number of log
共where peak outflow is insensitive to the vertical erosion rate兲 or
cycles on either side of the predicted value. This is the approach
small-reservoir with an associated estimate of the vertical erosion
taken in the analysis that follows.
rate obtained from observed breach heights and failure times. Two
The other notable feature of the plots in Fig. 1 is the presence
other facts should be noted:
of some significant outliers. Possible sources of these outliers
include the variable quality of the case study parameter observa- 1. No prediction equation could be applied to all 108 dam fail-
tions being used to test the predictions and the potential for mis- ure cases, due to the lack of required input data for the spe-
application of some of the prediction equations in the analysis cific equation or the lack of an observed value of the param-
described here due to lack of detailed firsthand knowledge of each eter of interest. Most of the breach width equations could be
case study situation. Such problems should not affect a careful tested against about 70 to 80 cases, the failure time equations
future application of these prediction equations to a specific case, against 30 to 40 cases, and the peak flow prediction equa-
and we do not wish for them to affect the present analysis of the tions against about 30 to 40 cases.
uncertainties of the methods themselves. Admittedly, much of the 2. The testing made use of the same data used to originally
scatter and the appearance of outliers are probably due to the develop many of the equations 共since the 108-dam database
inherent variability of the data caused by the variety of factors was compiled from these and other sources兲, but each equa-
that influence dam breach mechanics, and this variability should tion was also tested against additional cases, the number
be preserved as we analyze the uncertainties of the prediction varying depending on the method. This should provide a fair
equations. To exclude the truly anomalous data 共the statistical indication of the ability of each equation to predict breach
outliers兲 and retain the characteristic variability, an objective out- parameters for future dam failures. 共It is difficult to say ex-
lier exclusion algorithm was applied 共Rousseeuw 1998兲. The se- actly how many additional cases were analyzed for each
lected algorithm has the advantage that its performance is itself method, since the exact number of failures used to develop
insensitive to the presence of the outliers, which overcomes a each method is not indicated clearly in literature for all
common problem encountered when attempting to exclude outli- methods, and some are based on a combination of statistical
ers. analysis of case studies and physically based theory.兲

JOURNAL OF HYDRAULIC ENGINEERING © ASCE / MAY 2004 / 391


A step-by-step description of the uncertainty analysis method to 60% for two of the breach width equations 共Bureau of Recla-
follows: mation 1988; Von Thun and Gillette 1990兲 and four of the peak
1. Plot predicted versus observed values on log–log scales. flow equations 关Kirkpatrick 1977; SCS 1981; Bureau of Reclama-
2. Compute individual prediction errors in terms of the number tion 1982; Singh and Snorrason 1984 共the first of the two equa-
of log cycles separating the predicted and observed value, tions shown in Table 1兲兴. All of these prediction equations are
e i ⫽log10(x̂)⫺log10(x)⫽log10(x̂/x), where e i is the predic- based solely on the dam height or water depth above the breach
tion error, x̂ is the predicted value, and x is the observed invert, suggesting that dam height by itself is a poor predictor for
value. breach width or peak outflow.
3. Apply the outlier-exclusion algorithm to the series of predic-
tion errors computed in Step 2. The algorithm is described
by Rousseeuw 共1998兲.
Summary of Uncertainty Analysis Results
• Determine T, the median of the e i values. T is the estima-
tor of location.
• Compute the absolute values of the deviations from the The four methods for predicting breach width 共or volume of ma-
median, and determine the median of these absolute devia- terial eroded, from which breach width can be estimated兲 all had
tions 共MAD兲. absolute mean prediction errors less than one-tenth of an order of
• Compute an estimator of scale, S MAD⫽1.483* (MAD). magnitude, indicating that on average their predictions are on
The 1.483 factor makes S MAD comparable to the standard target. The uncertainty bands were similar 共⫾0.3 to ⫾0.4 log
deviation, which is the usual scale parameter of a normal cycles兲 for all of the equations except the MacDonald and
distribution. Langridge-Monopolis equation, which had an uncertainty of
• Use S MAD and T to compute a Z score for each observation, ⫾0.82 log cycles.
Z i ⫽(e i ⫺T)/S MAD , where the e i ’s are the observed predic- The five methods for predicting failure time all underpredict
tion errors, expressed as a number of log cycles. the failure time on average, by amounts ranging from about one-
• Reject any observations for which 兩 Z i 兩 ⬎2.5. fifth to two-thirds of an order of magnitude. This is consistent
• If the samples are from a perfect normal distribution, this with the previous observation that these equations are designed to
method rejects at the 98.7% probability level. Testing conservatively predict fast breaches, which will cause large peak
showed that application to normally distributed data would outflows. The uncertainty bands on all of the failure time equa-
lead to an average 3.9% reduction of the standard devia- tions are very large, ranging from about ⫾0.6 to ⫾1 order of
tion. magnitude, with the Froehlich 共1995a兲 equation having the small-
4. Compute the mean, ē, and the standard deviation, S e , of the est uncertainty.
remaining prediction errors. If the mean value is negative, it Most of the peak flow prediction equations tend to overpredict
indicates that the prediction equation underestimated the ob- observed peak flows, with most of the ‘‘envelope’’ equations
served values, and if positive the equation overestimated the overpredicting by about two-thirds to three-quarters of an order of
observed values. Significant over or underestimation should magnitude. The uncertainty bands on the peak flow prediction
be expected, since many of the breach parameter prediction equations are about ⫾0.5 to ⫾1 order of magnitude, except the
equations are intended to be conservative or provide enve- Froehlich 共1995b兲 relation which has an uncertainty of ⫾0.32
lope estimates, e.g., maximum reasonable breach width, fast- order of magnitude. In fact, the Froehlich equation has both the
est possible failure time, etc. lowest prediction error and smallest uncertainty of all the peak
5. Using the values of ē and S e , one can express a confidence
flow prediction equations.
band around the predicted value of a parameter as
兵 x̂•10⫺ē⫺2S e ,x̂•10⫺ē⫹2S e 其 , where x̂ is the predicted value.
The use of ⫾2S e approximately yields a 95% confidence
band. Application
Table 1 summarizes the results. The first two columns identify
the method being analyzed, the next two columns show the num- To illustrate the application of the uncertainty analysis results, a
ber of case studies used to test the method, and the next two case study is presented. In January 2001 the Bureau of Reclama-
columns give the prediction error and the width of the uncertainty tion conducted a risk assessment study for a large embankment
band. The last column shows the range of the prediction interval dam in North Dakota 共Fig. 2兲. Two potential failure modes were
around a hypothetical predicted value of 1.0. The values in this considered: 共1兲 Seepage erosion and piping through foundation
column can be used as multipliers to obtain the prediction interval materials, and 共2兲 seepage erosion and piping through embank-
for a specific case. ment materials. No distinction between the two failure modes was
Although the detailed data are not shown in Table 1, prediction made in the breach parameter analysis, since most methods used
errors and uncertainties also were determined prior to applying to predict breach parameters lack the refinement needed to con-
the outlier exclusion algorithm to determine its effect. Outlier sider differences in breach morphology for such similar failure
exclusion reduced the values of S e by at least 5% up to about 20% modes. Breach parameters were predicted using most of the meth-
in most cases. Since this exceeds the 3.9% reduction one would ods discussed earlier in this paper, and also by modeling with the
expect when applying the algorithm to a normally distributed NWS-BREACH model.
dataset, it suggests that true outliers were excluded rather than The potential for failure and the downstream consequences
just occasional extreme values that one would expect in normally from failure increase significantly at higher reservoir levels, al-
distributed data. The use of outlier exclusion did not materially though the likelihood of occurrence of high reservoir levels is
change the results of the study 共i.e., the same methods had the low. The reservoir rarely exceeds its top-of-joint-use elevation
lowest uncertainty before and after outlier exclusion兲. One no- 共the water surface elevation corresponding to the maximum
table fact is that the outlier exclusion algorithm reduced S e by 30 amount of storage allocated to joint use, i.e., flood control and

392 / JOURNAL OF HYDRAULIC ENGINEERING © ASCE / MAY 2004


Table 1. Uncertainty Estimates for Breach Parameter and Peak Flow Prediction Equations
Number of case studies
Mean Width of
Before After prediction uncertainty Prediction interval
outlier outlier error band, ⫾2S e around hypothetical
Reference Equation exclusion exclusion 共log cycles兲 共log cycles兲 predicted value of 1.0
Breach width equations
Bureau of Reclamation 共1988兲 B avg⫽3h w 80 70 ⫺0.09 ⫾0.43 0.45–3.3
MacDonald and V er⫽0.0261(V w h w ) 0.769 earthfill 60 58 ⫺0.01 ⫾0.82 0.15– 6.8
Langridge-Monopolis 共1984兲 V er⫽0.00348(V w h w ) 0.852 nonearthfills
共e.g., rockfills兲
Von Thun and Gillette 共1990兲 B avg⫽2.5h w ⫹C b 78 70 ⫹0.09 ⫾0.35 0.37–1.8
Froehlich 共1995a兲 B avg⫽0.1803K o V w0.32h 0.19
b 77 75 ⫹0.01 ⫾0.39 0.40–2.4

Failure time equations


MacDonald and t f ⫽0.0179V er
0.364
37 35 ⫺0.21 ⫾0.83 0.24 –11
Langridge-Monopolis 共1984兲
Von Thun and Gillette 共1990兲 t f ⫽0.015h w highly erodible 36 34 ⫺0.64 ⫾0.95 0.49– 40
t f ⫽0.020h w ⫹0.25 erosion resistant
Von Thun and Gillette 共1990兲 t f ⫽B avg /(4h w ) erosion resistant 36 35 ⫺0.38 ⫾0.84 0.35–17
t f ⫽B avg/(4h w ⫹61) highly erodible
Froehlich 共1995a兲 t f ⫽0.00254(V w ) 0.53h ⫺0.9
b 34 33 ⫺0.22 ⫾0.64 0.38 –7.3
Bureau of Reclamation 共1988兲 t f ⫽0.011(B avg) 40 39 ⫺0.40 ⫾1.02 0.24 –27

Peak flow equations


Kirkpatrick 共1977兲 Q p ⫽1.268(h w ⫹0.3) 2.5 38 34 ⫺0.14 ⫾0.69 0.28 – 6.8
SCS 共1981兲 Q p ⫽16.6(h w ) 1.85 38 32 ⫹0.13 ⫾0.50 0.23–2.4
Hagen 共1982兲 Q p ⫽0.54(S•h d ) 0.5 31 30 ⫹0.43 ⫾0.75 0.07–2.1
Bureau of Reclamation 共1982兲 Q p ⫽19.1(h w ) 1.85 envelope eq. 38 32 ⫹0.19 ⫾0.50 0.20–2.1
Singh and Snorrason 共1984兲 Q p ⫽13.4(h d ) 1.89 38 28 ⫹0.19 ⫾0.46 0.23–1.9
Singh and Snorrason 共1984兲 Q p ⫽1.776(S) 0.47 35 34 ⫹0.17 ⫾0.90 0.08 –5.4
MacDonald and Q p ⫽1.154(V w h w ) 0.412 37 36 ⫹0.13 ⫾0.70 0.15–3.7
Langridge-Monopolis 共1984兲
MacDonald and Q p ⫽3.85(V w h w ) 0.411 envelope eq. 37 36 ⫹0.64 ⫾0.70 0.05–1.1
Langridge-Monopolis 共1984兲
Costa 共1985兲 Q p ⫽1.122(S) 0.57 35 35 ⫹0.69 ⫾1.02 0.02–2.1
Costa 共1985兲 Q p ⫽0.981(S•h d ) 0.42 31 30 ⫹0.05 ⫾0.72 0.17– 4.7
Costa 共1985兲 Q p ⫽2.634(S•h d ) 0.44 31 30 ⫹0.64 ⫾0.72 0.04 –1.22
Evans 共1986兲 Q p ⫽0.72(V w ) 0.53 39 39 ⫹0.29 ⫾0.93 0.06 – 4.4
Froehlich 共1995b兲 Q p ⫽0.607(V w0.295h w1.24) 32 31 ⫺0.04 ⫾0.32 0.53–2.3
Walder and O’Connor 共1997兲 Q p estimated by computational and 22 21 ⫹0.13 ⫾0.68 0.16 –3.6
graphical method using relative
erodibility of dam and volume of
reservoir
Note: All equations use metric units 共m, m3, m3/s兲. Failure times are computed in hours. Where multiple equations are shown for application to different
types of dams 共e.g., earthfill versus rockfill兲, a single prediction uncertainty was determined, with the set of equations considered as a single algorithm.

conservation purposes兲, and has never exceeded an elevation of Dam Description


440.7 m. Four potential reservoir water surface elevations at fail-
ure were considered in the study: The case study dam is located a few kilometers upstream from a
• Top-of-joint-use, elevation: 436.67 m, reservoir capacity of city with a population of about 15,000. It was constructed by the
about 45.6⫻106 m3 , Bureau of Reclamation in the early 1950’s. The dam is operated
• Elevation 438.91 m, reservoir capacity of about 105 by Reclamation to provide flood control, municipal water supply,
⫻106 m3 , and recreational and wildlife benefits.
• Top-of-flood-space 共the design maximum reservoir level The dam is a zoned-earth fill with a height of 24.7 m above the
reached during the temporary storage of flood runoff兲, eleva- original streambed. The crest length is 432 m at an elevation of
tion 443.18 m, reservoir capacity of about 273⫻106 m3 , and 448.36 m and the crest width is 9.14 m. The design includes a
• Maximum design water surface, elevation: 446.32 m, storage central compacted zone 1 of impervious material, and upstream
of about 469⫻106 m3 . and downstream zone 2 of sand and gravel, shown in Fig. 3. The
For illustration purposes, only the results from the top-of-joint- abutments are composed of Pierre Shale capped with glacial till.
use and top-of-flood-space cases are presented here. The main portion of the dam is founded on a thick section of

JOURNAL OF HYDRAULIC ENGINEERING © ASCE / MAY 2004 / 393


Breach Width
Predictions of average breach width are summarized in Table 2.
Table 2 also lists the predictions of the volume of eroded embank-
ment material made using the MacDonald and Langridge-
Monopolis equation, and the corresponding estimate of average
breach width.
The uncertainty analysis described earlier showed that the
Reclamation equation tends to underestimate the observed breach
width, so it is not surprising that it yielded the smallest values.
The Von Thun and Gillette equation and the Froehlich equation
produced comparable results for the top-of-joint-use scenario, in
which reservoir storage is relatively small. For the top-of-flood-
space scenario, the Froehlich equation predicts significantly larger
Fig. 2. Aerial photo of the dam and reservoir considered in the case breach widths. This is not surprising, since the Froehlich equation
study application relates breach width to an exponential function of both the reser-
voir storage and reservoir depth. The Von Thun and Gillette equa-
tion accounts for reservoir storage only through the C b offset
alluvial deposits. Beneath the dam, a cutoff trench was excavated parameter, but C b is a constant for all reservoirs larger than
to the shale on both abutments, but between the abutments, foun- 12.3⫻106 m3 , as was the case for both scenarios.
dation excavation extended to a maximum depth of 7.6 m, and did Using the MacDonald and Langridge-Monopolis equation, the
not provide a positive cutoff of the thick alluvium. The alluvium estimate of eroded embankment volume and associated breach
beneath the dam is more than 37 m thick in the channel area.
width for the top-of-joint-use scenario is also comparable to the
There is a toe drain within the downstream embankment near
other equations. However, for the top-of-flood-space scenario, the
the foundation level, and a wide embankment section to help
prediction is much larger than any of the other equations, and in
control seepage beneath the dam, since a positive cutoff was not
fact is unreasonable because it exceeds the dimensions of the
constructed. Based on observations of increasing pressures in the
dam.
foundation during high reservoir elevations and significant boil
The prediction intervals developed through the uncertainty
activity downstream from the dam, eight relief wells were in-
analysis are sobering for the analyst wishing to obtain a definitive
stalled along the downstream toe in 1995 and 1996. To increase
result, as the ranges vary from small notches through the dam to
the seepage protection, a filter blanket was constructed in low
a complete washout of the embankment. Even for the top-of-
areas downstream from the dam in 1998.
joint-use case, the upper bounds for the Froehlich equation and
the Von Thun and Gillette equation are equivalent to about one-
Results—Breach Parameter Estimates half of the length of the embankment.
Predictions were made for average breach width, volume of
eroded material, and failure time. Side slope angles were not pre- Failure Time
dicted because equations for predicting breach side slope angles Failure time predictions are summarized in Table 3. All of the
are rare in literature; Froehlich 共1987兲 offered an equation, but in equations indicate increasing failure times as the reservoir storage
his later paper 共1995a兲, he suggested simply assuming side slopes increases, except the second Von Thun and Gillette relation,
of 0.9:1 共horizontal:vertical兲 for piping failures. Von Thun and which predicts a slight decrease in failure time for the top-of-
Gillette 共1990兲 suggested using side slopes of 1:1, except for flood-space scenario. For both Von Thun and Gillette relations,
cases of dams with very thick zones of cohesive materials where the dam was assumed to be in the erosion resistant category.
side slopes of 0.5:1 or 0.33:1 might be appropriate. The predicted failure times exhibit wide variation, and the rec-
After computing breach parameters using the many available ommended values shown at the bottom of Table 3 are based on
equations, the results were reviewed and judgment applied to de- much judgment. The uncertainty analysis showed that all of the
velop a single predicted value and an uncertainty band to be pro- failure time equations tend to conservatively underestimate actual
vided to the risk assessment study team. These recommended failure times, especially the Von Thun and Gillette and Reclama-
values are shown at the bottom of each column in the tables that tion equations. Thus, the recommended values are generally a
follow. compromise between the results obtained from the MacDonald

Fig. 3. Cross section through the case study dam

394 / JOURNAL OF HYDRAULIC ENGINEERING © ASCE / MAY 2004


Table 2. Predictions of Average Breach Width
Top of joint use, elevation of 436.68 m Top of flood space, elevation of 443.18 m
Predicted breach 95% prediction Predicted breach 95% prediction
Equation width 共m兲 interval width 共m兲 interval
Bureau of Reclamation 共1988兲 39.0 17.7–129 58.5 26.2–193
Von Thun and Gillette 共1990兲 87.5 32.3–157 104 38.4 –187
Froehlich 共1995a兲 93.6 37.5–225 166 66.4 –398
MacDonald and Langridge-Monopolis 共1984兲 146,000 22,200–991,000 787,000 118,000–5,350,000
Volume of erosion 共m3兲
Equivalent breach width 共m兲 85.6 12.8 –582a 462a 69.2–3140a

Recommended values 共m兲 90 35–180 165 60– 400


a
Exceeds actual embankment length.

and Langridge-Monopolis and Froehlich relations. Despite this when the breach reaches its maximum size, before significant
fact, some very fast failures are documented in literature, and this drawdown of the reservoir has occurred. Despite the refinement
possibility is reflected in the prediction intervals determined from of considering large- versus small-reservoir behavior, the Walder
the uncertainty analysis. and O’Connor method was found to have uncertainty similar to
most of the other peak flow prediction methods 共about ⫾0.75 log
cycles兲. However, among the 22 case studies to which the method
Results—Peak Outflow Estimates
could be applied, only four proved to be large-reservoir/fast-
Peak outflow estimates are shown in Table 4, sorted in order of erosion cases. Of these, the method overpredicted the peak out-
increasing peak outflow for the top-of-joint-use scenario. The flow in three cases, and dramatically underpredicted in one case
lowest peak flow predictions come from those equations that are 共Goose Creek Dam, South Carolina, failed 1916 by overtopping兲.
based solely on dam height or depth of water in the reservoir. The Closer examination showed some contradictions in the data re-
highest peak flows are predicted by those equations that incorpo- ported in literature for this case. On balance, it appears that the
rate a significant dependence on reservoir storage. Some of the Walder and O’Connor method may provide reasonable estimates
predicted peak flows and the upper bounds of the prediction limits of the upper limit on peak outflow for large-reservoir/fast-erosion
would be the largest dam-break outflows ever recorded, exceed- cases.
ing the 65,000 m3/s peak outflow from the Teton Dam failure. For this application, results from the Froehlich method were
共Storage in Teton Dam at failure was 356⫻106 m3 ). The length of considered to be the best estimate of peak breach outflow, and the
the reservoir 共about 48 km兲 may help to attenuate some of the results from the Walder and O’Connor method provided an upper
large peak outflows predicted by the storage-sensitive equations, bound estimate.
since there will be an appreciable routing effect in the reservoir
itself that is probably not accounted for in the peak flow predic-
NWS-BREACH Simulations
tion equations.
The equation offered by Froehlich 共1995b兲 clearly had the best Several simulations runs were made using the NWS-BREACH
prediction performance in the uncertainty analysis, and is thus model 共Fread 1988兲. The model requires input data related to
highlighted in Table 4. This equation had the smallest mean pre- reservoir bathymetry, dam geometry, the tailwater channel, em-
diction error and narrowest prediction interval by a significant bankment materials, and initial conditions for the simulated pip-
margin. ing failure.
The results for the Walder and O’Connor method are also The results of the simulations are very sensitive to the eleva-
highlighted. As discussed earlier, this is the only method that tion at which the piping failure is assumed to develop. In all cases
considers the differences between the so-called large-reservoir/ analyzed, the maximum outflow occurred just prior to the crest of
fast-erosion and small-reservoir/slow-erosion cases. This dam the dam collapsing into the pipe; after the collapse of the crest, a
proves to be a large-reservoir/fast-erosion case when analyzed by large volume of material partially blocks the breach and the out-
this method 共regardless of the assumed vertical erosion rate of the flow becomes weir controlled until the material can be removed.
breach—within reasonable limits兲, so the peak outflow will occur Thus, the largest peak outflows and largest breach sizes are ob-

Table 3. Failure Time Predictions


Top of joint use, elevation of 436.68 m Top of flood space, elevation of 443.18 m
Equation Predicted failure time 共h兲 95% prediction interval Predicted failure time 共h兲 95% prediction interval
MacDonald and Langridge-Monopolis 共1984兲 1.36 0.33–14.9 2.45a 0.59–26.9
Von Thun and Gillette 共1990兲, t f ⫽ f (h w ) 0.51 0.25–20.4 0.64 0.31–25.6
Von Thun and Gillette 共1990兲, t f ⫽ f (B,h w ) 1.68 0.59–28.6 1.33 0.47–22.6
Froehlich 共1995a兲 1.63 0.62–11.9 4.19 1.59–30.6
Bureau of Reclamation 共1988兲 0.43 0.10–11.6 0.64 0.15–17.4

Recommended values 1.5 0.25–12 3.0 0.3–17


a
Predicted erosion volume exceeded total embankment volume; total embankment volume was used in the failure time equation.

JOURNAL OF HYDRAULIC ENGINEERING © ASCE / MAY 2004 / 395


Table 4. Predictions of Peak Breach Outflow
Top of joint use, elevation of 436.68 m Top of flood space, elevation of 443.18 m
Predicted peak outflow 95% prediction Predicted peak outflow 95% prediction
Equation 共m3/s兲 interval 共m3/s兲 interval
Kirkpatrick 共1977兲 818 229–5,570 2,210 620–15,100
SCS 共1981兲 1,910 439– 4,590 4,050 932–9,710
Bureau of Reclamation 共1982兲 共envelope兲 2,200 439– 4,620 4,660 932–9,780
Froehlich „1995b… 2,660 1,410–6,110 7,440 3,940–17,100
MacDonald/Langridge-Monopolis 共1984兲 4,750 714 –17,600 11,700 1,760– 43,400
Singh/Snorrason 共1984兲, Q p ⫽ f (h d ) 5,740 1,320–10,900 5,740 1,320–10,900
Walder and O’Connor „1997… 6,000 960–21,400 12,200 1,950–43,500
Costa 共1985兲, Q p ⫽ f (S * h d ) 6,220 1,060–29,200 13,200 2,240– 61,900
Singh/Snorrason 共1984兲, Q p ⫽ f (S) 7,070 570–38,200 16,400 1,310– 88,400
Evans 共1986兲 8,260 496 –36,300 21,300 1,280–93,700
MacDonald/Langridge-Monopolis 共1984兲 15,500 776 –17,100 38,300 1,910– 42,100
共envelope兲
Hagen 共1982兲 18,100 1,270–38,100 44,300 3,100–93,000
Costa 共1985兲, Q p ⫽ f (S * h d ) 共envelope兲 25,300 1,010–30,900 55,600 2,220– 67,800
Costa 共1985兲, Q p ⫽ f (S) 26,100 521–54,700 72,200 1,440–152,000

tained if the failure is initiated at the base of the dam, assumed to are large for all methods, and thus it may be worthwhile to incor-
be at an elevation of 423.67 m. This produces the maximum porate uncertainty analysis results into future risk assessment
amount of head on the developing pipe, and allows it to grow to studies when predicting breach parameters using these methods.
the largest possible size before the collapse occurs. Table 5 shows Predictions of breach width generally have an uncertainty of
summary results of the simulations. For each initial reservoir el- about ⫾1/3 order of magnitude, predictions of failure time have
evation, a simulation was run with the pipe initiating at an eleva- uncertainties approaching ⫾1 order of magnitude, and predictions
tion of 423.7 m, and a second simulation was run with the pipe of peak flow have uncertainties of about ⫾0.5 to ⫾1 order of
initiating about midway up the height of the dam. magnitude, except the Froehlich peak flow equation, which has an
There is a wide variation in the results depending on the as- uncertainty of about ⫾1/3 order of magnitude.
sumed initial conditions for the elevation of the seepage failure. The uncertainty analysis made use of a database of informa-
The peak outflows and breach widths tend toward the low end of tion on the failure of 108 dams compiled from numerous sources
the range of predictions made using the regression equations in literature 共Wahl 1998兲. Those wishing to make use of this da-
based on case study data. The predicted failure times are within tabase may obtain it in electronic form 共Lotus 1-2-3, Microsoft
the range of the previous predictions, and significantly longer Excel, and Microsoft Access兲 on the Internet at http://
than the very short 共0.5 to 0.75 h兲 failure times predicted by the www.usbr.gov/pmts/hydraulicsគlab/twahl/
Bureau of Reclamation 共1988兲 equation and the first Von Thun The case study presented here showed that significant engi-
and Gillette equation. neering judgment must be exercised in the interpretation of pre-
dictions of breach parameters. The results from use of the physi-
cally based NWS-BREACH model were reassuring because they
Conclusions fell within the range of values obtained from the regression-based
methods. However, at the same time, they also helped to show
This paper has presented a quantitative analysis of the uncertainty that even physically based methods can be highly sensitive to the
of various regression-based methods for predicting embankment assumptions of the analyst regarding breach morphology and the
dam breach parameters and peak breach outflows. The uncertain- location of initial breach development. The NWS-BREACH
ties of predictions of breach width, failure time, and peak outflow simulations demonstrated the possibility for limiting failure me-
chanics that were not revealed by the regression-based methods.

Table 5. Results of National Weather Service-BREACH Simulations


of Seepage-Erosion Failures Notation
Initial water Initial Breach The following symbols are used in this paper:
surface elevation Peak Time-to-peak width at B avg ⫽ average breach width 共m兲;
elevation of piping outflow, outflow, t p time t p
C b ⫽ offset factor in the Von Thun and Gillette breach
共m兲 failure 共m兲 共m3/s兲 共h兲 共m兲
width equation, varies as a function of reservoir
Top of joint use volume;
436.68 423.7 2,280 3.9 15.7 ē ⫽ average prediction error;
436.68 430.1 464 2.1 6.5 e i ⫽ individual prediction errors, log cycles;
h b ⫽ height of breach 共m兲;
Top of flood space
h d ⫽ height of dam 共m兲;
443.18 423.7 6,860 4.0 24.7
h w ⫽ depth of water above breach invert at time of
443.18 430.1 1,484 1.4 10.3
failure 共m兲;

396 / JOURNAL OF HYDRAULIC ENGINEERING © ASCE / MAY 2004


K o ⫽ overtopping multiplier: 1.4 for overtopping; 1.0 for Froehlich, D. C. 共1987兲. ‘‘Embankment-dam breach parameters.’’ Hy-
piping; draulic Engineering, Proc. 1987 ASCE National Conf. on Hydraulic
Engineering, New York, 570–575.
MAD ⫽ median of absolute deviations from T;
Froehlich, D. C. 共1995a兲. ‘‘Embankment dam breach parameters revis-
Q p ⫽ peak breach outflow 共m3/s兲;
ited.’’ Water Resources Engineering, Proc. 1995 ASCE Conf. on Water
S ⫽ reservoir storage 共m3兲; Resources Engineering, New York, 887– 891.
S e ⫽ standard deviation of the errors; Froehlich, D. C. 共1995b兲. ‘‘Peak outflow from breached embankment
S MAD ⫽ estimator of scale derived from the median of the dam.’’ J. Water Resour. Plan. Manage. Div., Am. Soc. Civ. Eng.,
absolute deviations, analogous to standard deviation; 121共1兲, 90–97.
T ⫽ median of the errors, an estimator of location; Hagen, V. K. 共1982兲. ‘‘Re-evaluation of design floods and dam safety.’’
t f ⫽ failure time 共h兲; Proc., 14th Congress of Int. Commission on Large Dams, Interna-
V er ⫽ volume of embankment material eroded 共m3兲; tional Commission on Large Dams, Paris.
V w ⫽ volume of water stored above breach invert at time Hanson, G. J., Cook, K. R., and Temple, D. M. 共2002兲. ‘‘Research results
of failure 共m3兲; of large-scale embankment overtopping breach tests.’’ 2002 ASDSO
x̂ ⫽ predicted value of parameter; Annual Conf., Association of State Dam Safety Officials, Lexington,
Ky.
x ⫽ observed value of parameter; and
Johnson, F. A., and Illes, P. 共1976兲. ‘‘A classification of dam failures.’’ Int.
Z i ⫽ standardized error. Water Power Dam Constr., 28共12兲, 43– 45.
Kirkpatrick, G. W. 共1977兲. ‘‘Evaluation guidelines for spillway ad-
equacy.’’ The evaluation of dam safety, Engineering Foundation
References Conf., ASCE, New York, 395– 414.
MacDonald, T. C., and Langridge-Monopolis, J. 共1984兲. ‘‘Breaching
Bureau of Reclamation. 共1982兲. Guidelines for defining inundated areas characteristics of dam failures.’’ J. Hydraul. Eng., 110共5兲, 567–586.
downstream from Bureau of Reclamation dams, Reclamation Planning Rousseeuw, P. J. 共1998兲. ‘‘Chapter 17: Robust estimation and identifying
Instruction No. 82-11, U.S. Department of the Interior, Bureau of outliers.’’ Handbook of statistical methods for engineers and scien-
Reclamation, Denver, 25. tists, 2nd Ed., H. M. Wadsworth Jr., ed., McGraw–Hill, New York,
Bureau of Reclamation. 共1988兲. ‘‘Downstream hazard classification 17.1–17.15.
guidelines.’’ ACER Tech. Memorandum No. 11, U.S. Department of Singh, K. P., and Snorrason, A. 共1984兲. ‘‘Sensitivity of outflow peaks and
the Interior, Bureau of Reclamation, Denver, 57. flood stages to the selection of dam breach parameters and simulation
Costa, J. E. 共1985兲. ‘‘Floods from dam failures.’’ U.S. Geological Survey, models.’’ J. Hydrol., 68, 295–310.
Open-File Rep. No. 85-560, Denver, 54. Soil Conservation Service 共SCS兲. 共1981兲. ‘‘Simplified dam-breach rout-
Evans, S. G. 共1986兲. ‘‘The maximum discharge of outburst floods caused ing procedure.’’ Tech. Release No. 66 共Rev. 1兲, 39.
by the breaching of man-made and natural dams.’’ Can. Geotech. J., Temple, D. M., and Moore, J. S. 共1997兲. ‘‘Headcut advance prediction for
23共4兲, 385–387. earth spillways.’’ Trans. ASAE, 40共3兲, 557–562.
Fread, D. L. 共1984兲. DAMBRK: The NWS dam-break flood forecasting Von Thun, J. L., and Gillette, D. R. 共1990兲. ‘‘Guidance on breach param-
model, National Weather Service, Office of Hydrology, Silver Spring, eters.’’ Internal Memorandum, U.S. Dept. of the Interior, Bureau of
Md. Reclamation, Denver, 17.
Fread, D. L. 共1988兲 共revised 1991兲. BREACH: An erosion model for Wahl, T. L. 共1998兲. ‘‘Prediction of embankment dam breach
earthen dam failures, National Weather Service, Office of Hydrology, parameters—A literature review and needs assessment.’’ Dam Safety
Silver Spring, Md. Rep. No. DSO-98-004, U.S. Dept. of the Interior, Bureau of Reclama-
Fread, D. L. 共1993兲. ‘‘NWS FLDWAV model: The replacement of DAM- tion, Denver.
BRK for dam-break flood prediction.’’ Dam Safety ’93, Proc., 10th Walder, J. S., and O’Connor, J. E. 共1997兲. ‘‘Methods for predicting peak
Annual ASDSO Conf., Association of State Dam Safety Officials, Lex- discharge of floods caused by failure of natural and constructed earth
ington, Ky., 177–184. dams.’’ Water Resour. Res., 33共10兲, 12.

JOURNAL OF HYDRAULIC ENGINEERING © ASCE / MAY 2004 / 397

You might also like