You are on page 1of 15

Predicting Peak Discharge from Gradually

Breached Embankment Dam


David C. Froehlich, Ph.D., P.E., D.WRE, M.ASCE 1

Abstract: Two nonlinear mathematical models are presented to predict peak discharge from a breached embankment dam based on
evaluation of measured outflows from 41 dam failures. Reported outflow rates from these failures have been examined carefully for accuracy
and are believed to be reliable. The first model is purely empirical, while the second is based on a semitheoretical approach that reduces the
maximum possible peak discharge that would be produced by instantaneous formation of a breach having prescribed dimensions. Results
given by the models are compared to those from other breach peak discharge calculation methods and are shown to produce considerably
more accurate predictions of peak discharge from breached embankment dams. DOI: 10.1061/(ASCE)HE.1943-5584.0001424. © 2016
American Society of Civil Engineers.
Downloaded from ascelibrary.org by La Trobe University on 07/06/16. For personal use only.

Introduction number of dam failures that have occurred in the United States is
unknown, but failures have been documented in every state. From
The National Inventory of Dams (NID) is a database maintained January 2005 through June 2013, state dam safety programs re-
by the U.S. Army Corps of Engineers (USACE) that contains facts ported 173 dam failures (about 20 per year) and 587 dam incidents
about more than 87,000 dams located in the United States and its (about 70 per year), which are events that, without intervention,
territories (USACE 2013). The data repository includes informa- would probably have resulted in uncontrolled releases of im-
tion about a dam’s location, size, purpose, type, last inspection, pounded water (ASDSO 2015).
regulatory facts, downstream hazard potential, and other technical The level of detail of hydrologic and hydraulic analyses needed
data. About 75,000, or nearly 86%, of these dams are formed by to evaluate the consequences of dam-breach floods depends on the
embankments constructed from natural erodible materials (earth downstream hazards they present. If fatalities are unlikely and prop-
and rock) that rely on their weight to hold back the force of water. erty damage potential is not significant, a simple procedure may
Because embankment dams are so numerous, potential flood haz- provide an adequate description of the extent and timing of down-
ards that would be created by uncontrolled releases of impounded stream flooding resulting from a failure. Uncomplicated flood-
water through a breach need to be evaluated to select spillway routing methods requiring small amounts of data and effort to
design floods and to prepare emergency action plans. estimate peak flow rates at locations downstream from a breached
Downstream exposure to danger created by a dam in the event dam have been presented by the Soil Conservation Service (SCS
of failure or misoperation leading to uncontrolled release of im- 1981), Hagen (1982), Costa (1985), Barker and Schaefer (2007),
pounded water (that is, a dam-breach flood) is classified in the NID and Environment Agency (2014). Each of these methods requires
as either a high, significant, or low hazard. Dams whose failure will an estimate of the maximum outflow rate from the reservoir.
create catastrophic flooding that would probably result in fatalities, Empirical and semitheoretical numerical models for determining
if sufficient warnings were not issued and other emergency actions the peak discharge from a gradual embankment dam failure have
not taken, are assigned a high hazard potential. Those that pose been developed by Kirkpatrick (1977), MacDonald and Langridge-
little threat to human life in the event of a failure, but would cause Monopolis (1984), Costa (1985), SCS (1985), Froehlich (1995),
large economic and/or environmental losses, are considered to have Webby (1996), Walder and O’Connor (1997), Thornton et al.
a significant downstream hazard potential. Low hazard potential (2011), Hooshyaripor et al. (2014), De Lorenzo and Macchione
classification is assigned to dams where failure would not endanger (2014), and Azimi et al. (2015).
lives and would cause only small amounts of damage limited pri- All of the methods cited for estimating peak discharge rely on
marily to the dam owner’s property. reported flow rates from past embankment dam failures, or from
small-scale laboratory experiments, to evaluate model coefficients.
Failures of large embankment dams (measured either by their
Techniques applied to determine some of the peak discharges used
height or by the volume of water they store) are comparatively rare,
for this purpose are undocumented, which by itself does not render
but can cause immense destruction and numerous fatalities when
a measurement unfit, but does warrant disbelief when the reported
they do occur. Small dams, for which spillway capacities may be
value is far from most others in an assembled data set and appears
able to accommodate floods having annual return periods of no
to be an exception. Other flow rates have been measured substantial
more than 100 years, fail considerably more often. The exact distances downstream from the failed dams and probably do not
represent accurately the true breach peak outflow rates, or the re-
1 ported peak discharges are impossibly large based on measured fi-
Consultant, 303 Frenchmans Bluff Dr., Cary, NC 27513. E-mail:
nal breach dimensions and the assumption of instantaneous breach
dcfroehlich@aol.com
Note. This manuscript was submitted on December 1, 2015; approved formation.
on April 14, 2016; published online on July 5, 2016. Discussion period Two nonlinear mathematical models are presented here to pre-
open until December 5, 2016; separate discussions must be submitted dict the peak discharge from a breached embankment dam based
for individual papers. This paper is part of the Journal of Hydrologic En- on examination of measured outflows from 41 past dam failures.
gineering, © ASCE, ISSN 1084-0699. Reported outflow rates from these failures have been examined

© ASCE 04016041-1 J. Hydrol. Eng.

J. Hydrol. Eng.
carefully for accuracy and are believed to be reliable. The first Breach Height and Reservoir Volume
model is purely empirical, while the second is based on a semi-
The bottom of a fully formed breach usually is the dam foundation,
theoretical approach. Results given by the models are compared to which is more resistant to erosion than the embankment material.
those from the breach peak discharge calculation methods cited However, the maximum height of a breach, Hb , might be limited
previously. Both models are shown to produce considerably more by the volume of water in the reservoir at the time of failure, or by
accurate predictions of peak discharge from a breached embank- the presence of a layer of erosion-resistant material located in the
ment dam than other approaches. embankment. The maximum height of a breach, or the minimum
bottom elevation of the trapezoidal breach approximation, needs to
be presupposed with reasonable certainty. The volume of water
Embankment Dam Failure Data contained in the reservoir at the time of failure, V w , is the amount
above the final breach bottom that will eventually flow out of the
Peak discharges and other pertinent data for 41 embankment dam reservoir. Depending on the embankment design and the elevation
failures were assembled from a variety of sources and are summa- of the breach bottom, V w might be substantially less than the res-
rized in Table 1. These data include a brief description of each dam, ervoir capacity.
the mode of failure, average width of the embankment (between
toes of the downstream and upstream slopes) W avg , water volume
above the breach bottom at the time of failure V w , height of water Depth of Impounded Water and Approach Flow Width
above the breach bottom at the time of failure Hw , height of Because most embankment dam breaches erode vertically to the level
the embankment above the breach bottom Hb , the width of flow of the excavated foundation, which is highly resistant to further
Downloaded from ascelibrary.org by La Trobe University on 07/06/16. For personal use only.

approaching the dam La , the measured peak discharge Qp , and downward breach growth, estimating the final bottom elevation of
the method used to calculate the peak discharge. The ranges of a potential trapezoidal breach, and thus Hw and Hb , for a potential
variables contained in the data set are given in Table 2. failure is straightforward. However, determining approach flow
width might require subjective evaluation for dams that do not span
a straight river valley perpendicular to its longitudinal axis. This is
Dam Description particularly true when the embankments form a completely closed
The name and general location of each dam is provided along with reservoir, such as the FP&L Martin Plant Dam described by the
the year construction was completed and the year of the failure. South Florida Water Management District (SFWMD 1980), or the
Materials used for embankment dams include natural soil or rock, Taum Sauk Dam described by the Federal Energy Regulatory Com-
or waste materials obtained from mining or milling operations. The mission (FERC 2006a, b). Values of La were determined for each of
embankment type is classified as either an earth-fill or rock-fill, the dam failures in Table 1 by measurement from satellite images.
depending on whether it has been built with only compacted soil,
or mostly compacted or dumped rock. Erosion resistance is deter- Peak Discharge
mined primarily by the type of materials from which the embank-
ment is made, and by the quality of construction. All peak discharges have been determined by indirect means rather
Embankments can by constructed of only a single material than by direct measurement using current meters or other flow
(homogeneous) or of different materials (zoned), which usually measurement devices. The several indirect methods used to calcu-
contain a compacted clay core to restrict seepage. Nearly all rock- late the peak flow rates are described as follows:
fill dams depend on an impervious upstream blanket or on a clay 1. The technique used most often, and perhaps the most accurate if
core to limit leakage. Some dams have been built with slender solid carried out by knowledgeable analysts, is the slope-area method
core walls, made from concrete, masonry, or steel, to control seep- (Dalrymple and Benson 1967). Accuracy of the calculated peak
discharge depends on reliability of measured high-water marks,
age through the embankment.
correctness of estimated roughness coefficients, appropriateness
of the one-dimensional flow supposition, and distance of the
Failure Mode measurement downstream from the breached dam.
2. Estimation based on reservoir volume change during a short
Breaches form in embankment dams in several different ways. For time interval is the second most frequently used approach,
some of the cases in Table 1, deciding on the mode of failure was which is based on the assumption of a nearly level water surface
difficult, especially if eyewitness accounts of the failure were not upstream of the dam and, in most cases, zero inflow to the re-
available. The International Commission on Large Dams (1974) servoir. Discharge accuracy is reduced by the latter assumption
finds that about one-third of all embankment-dam failures have for overtopping failures caused by large floods. Length of the
been caused by inadequate spillway capacities that result in over- time interval used in the calculation, and location of the gauge
topping of the embankments. Roughly another third were attrib- used to record water levels, also have unfavorable effects on
uted to internal erosion caused by concentrated seepage through accuracy of the computed discharge.
the embankments that erodes soil particles along the path of leak- 3. A number of peak discharges have been calculated as twice the
age, gradually enlarging the flow passage until failures occur. average outflow rate needed to empty the reservoir within a spe-
Other failures were created by embankment slides that lower the cified time period. Measurement accuracy depends largely on
dam crests and allow overtopping or overflow, foundation settle- reliability of the reported time, and on inflows to the reservoir
ment, and insufficient protection against wave action. On the basis during the failure. This approach is probably the least accurate
of these findings, failure modes have been classified either as of all methods used to determine peak discharges.
overtopping or as internal erosion (IE). Overtopping can be caused 4. Unsteady flow numerical modeling of reservoir outflows uses
in several different ways, which are identified as follows: inflows measured values of final breach dimensions and observed
exceeding spillway capacities (OF), outlet works failure (OG), breach formation times. This approach was used to estimate
wave action (OW), or lowering of the crest because of embank- peak discharge from the Big Bay Lake Dam (Yochum et al.
ment sliding (OS). 2008) and the Lake Dehli Dam (Holman 2011).

© ASCE 04016041-2 J. Hydrol. Eng.

J. Hydrol. Eng.
Downloaded from ascelibrary.org by La Trobe University on 07/06/16. For personal use only.

Table 1. Embankment Dam Breach Peak Discharge and Failure Characteristics


Volume of

© ASCE
Average water above Height of Approach Measured
Embank- embank- breach water above Height flow peak
Dam name and ment Year Year Failure ment width bottom breach bottom of breach width discharge Peak discharge
Number location typea completed failed modeb W avg (m) V w (Mm3 ) H w (m) H b (m) La (m) Qp (m3 =s) calculation method References
1 Apishapa, Colorado E, H, C 1920 1923 IE 82.4 22.8 28.0 31.1 200 6,850 Rate of reservoir volume “Failure of Apishapa”
change, 15 min time (1923), and Field et al.
interval (1923)
2 Baldwin Hills, E, H 1951 1963 IE 59.6 0.950 12.2 21.3 200 420 Twice the average flow California Department of
California rate needed to empty the Water Resources (1964),
reservoir in 1.25 h Jessup (1964), and Warne
(1964)
3 Banqiao, Henan E, H 1953 1975 OF 97 701 31.9 29.5 2,100 56,300 Twice the average flow Fujia and Yumei (1994),
Province, China rate needed to empty the Qing (1998), Henan Water
reservoir in 6.0 h Resources Authority (2005),
Courivaud (2007), and Xu
et al. (2009)
4 Bass Haven Lake, E, H — 1984 OX 22.9 0.641 4.90 9.20 — 240 Twice the average flow USCOLD (1988, pp. 46);
Texas rate needed to empty the Warren Samuelson, Dam
reservoir in 1.5 h Safety Program, Texas
Commission on
Environmental Quality,
written communication
(2015)
5 Belci, Bacău County, E, Z 1963 1991 OF 37.8 12.7 15.5 15.0 400 4,700 Twice the average flow Diacon et al. (1992), Vogel
Romania rate needed to empty the and Klein (1993), Graham

04016041-3

J. Hydrol. Eng.
reservoir in 1.5 h and Hilldale (2002), and
Marcello et al. (2009)
6 Bila Desna, Czech E, H 1915 1916 IE 29.6 0.290 10.7 14.6 170 320 Twice the average flow Jansen (1983, pp. 125–126),
Republic rate needed to empty the and Raška and Emmer (2014)
reservoir in 30 min
7 Bilberry, England E, Z 1845 1852 OS 62.5 0.327 23.6 23.0 800 725 Twice the average flow Saxena and Sharma (2005,
rate needed to empty the pp. 19–20), and Binnie
reservoir in 15 min (1981, pp. 49–69)
8 Big Bay Lake, E, H 1992 2004 IE 20.4 17.5 13.6 14.0 200 4,160 Numerical modeling of Yochum et al. (2008)
Mississippi breach outflow using
measured final breach
geometry and observed
formation time
9 Bradfield (Dale E, Z 1863 1864 IE 76.0 3.20 28.0 29.0 300 2,370 Twice the average flow “The Bradfield inundation.”
Dyke), England rate needed to empty the
(1874), Binnie (1981,
reservoir in 0.75 h pp. 256–265), Jansen (1983,
pp. 129–132), and Singh
(1996, pp. 80–81)
10 Butler Valley, E, H — 1982 OF 9.63 2.38 7.16 7.16 850 810 Slope-area measurement Byron N. Aldridge, U.S.
Arizona 600 m downstream Geological Survey, written
communication (1987)

J. Hydrol. Eng.
Downloaded from ascelibrary.org by La Trobe University on 07/06/16. For personal use only.

Table 1. (Continued.)
Volume of

© ASCE
Average water above Height of Approach Measured
Embank- embank- breach water above Height flow peak
Dam name and ment Year Year Failure ment width bottom breach bottom of breach width discharge Peak discharge
Number location typea completed failed modeb W avg (m) V w (Mm3 ) H w (m) H b (m) La (m) Qp (m3 =s) calculation method References
11 Castlewood, E, M 1890 1933 OF 47.4 6.17 21.6 21.3 180 3,570 Rate of reservoir volume “Castlewood dam” (1933),
Colorado change, 15 min time Field (1933), Houk (1933),
interval and Follansbee and Sawyer
(1948)
12 Centralia (Seminary E, H 1914 1991 OS 10.1 0.01333 5.50 6.10 40 71 Slope-area measurement Costa (1994) and Costa and
Hill) Reservoir No. 275 m downstream O’Connor (1995)
3, Washington
13 Delhi, Iowa E, H, C 1929 2010 OF 31.5 12.2 11.2 11.0 170 1,950 Numerical modeling of Fiedler et al. (2010, 2013),
breach outflow using and Holman (2011)
measured final breach
geometry and observed
formation time
14 FP&L Martin Plant, E, H 1977 1979 IE 27.7 125 5.09 9.14 4,100 2,750 Rate of reservoir volume SFWMD (1980)
Florida change based on three
stage recorders. Time
interval is not specified
15 Fred Burr, Montana E, H 1948 1948 IE 30.8 0.750 10.2 10.2 100 654 Slope-area measurement Boner and Stermitz (1967)
“a short distance”
downstream
16 French Landing, E, H 1925 1925 IE 34.3 3.87 8.53 14.2 350 929 Rate of reservoir volume “Undermining causes
Michigan change, 1 h time interval failure” (1925)

04016041-4

J. Hydrol. Eng.
17 Frenchman Creek, E, H 1952 1952 IE 37.3 16.0 10.8 12.5 600 1,420 Unknown. Probably rate Oltman (1955), Babb and
Montana of reservoir volume Mermel (1963); Mike Ulrick,
change Mont. Water Resources
Division, Dam Safety
Section, written
communication (1992)
18 Hästberga, Sweden E, H, C 1950 2010 OG 12.7 30.0 7.35 7.00 200 600 Rate of volume change Ekström (2012), and Rosvall
(15 min time interval) of and Dahlberg (2011)
a reservoir located 4 km
downstream
19 Hatchtown, Utah E, Z 1908 1914 OS 44.8 16.0 16.8 18.3 200 4,440 Twice the average flow Sterling (1914), “Ninth
rate needed to empty the biennial report.” (1915,
reservoir in two hours pp. 13–14)
20 Hell Hole, California R, Z 1964 1964 IE 103 30.6 35.1 56.4 470 17,000 Twice the average flow “Hell Hole Dam” (1965),
rate needed to empty the “The failure of Hell Hole
reservoir in one hour Dam” (1965), and Scott and
Gravlee (1968)
21 Ireland Reservoir E, H — 1984 IE 18.0 0.160 3.81 5.18 370 110 Slope-area measurement L. H. DeGrave, Colorado
No. 5, Colorado “a short distance” Department of Natural
downstream Resources, written
communication (1986)
22 Kelly Barnes, E, H 1948 1977 IE 19.4 0.777 11.3 12.8 80 680 Slope-area measurement Crisp et al. (1977), “Dam
Georgia 250 m downstream collapses” (1977), and
Sanders and Sauer (1979)

J. Hydrol. Eng.
Downloaded from ascelibrary.org by La Trobe University on 07/06/16. For personal use only.

Table 1. (Continued.)
Volume of

© ASCE
Average water above Height of Approach Measured
Embank- embank- breach water above Height flow peak
Dam name and ment Year Year Failure ment width bottom breach bottom of breach width discharge Peak discharge
Number location typea completed failed modeb W avg (m) V w (Mm3 ) H w (m) H b (m) La (m) Qp (m3 =s) calculation method References
23 Lake Avalon, New E, R 1894 1904 IE 42.7 31.5 13.7 14.6 370 2,320 Unknown Murphy (1905, pp. 172–
Mexico 173), and Middlebrooks
(1953)
24 Laurel Run, E, H — 1977 OF 40.5 0.555 14.1 13.7 160 1,050 Slope-area measurement Brua (1978), Chen and
Pennsylvania 1.6 km downstream Armbruster (1980),
“Johnstown” (1977), Hoxit
et al. (1982), and on-site
inspection by author
(September 1978)
25 Lily Lake, Colorado E, H 1913 1951 OW 13.2i 0.0925 3.35 3.66 60 71 Slope-area measurement W. J. Graham, U.S. Bureau of
at unknown location Reclamation, Denver, CO
written communication
(1987)
26 Little Deer Creek, E, H 1962 1963 IE 63.1 1.36 22.9 27.1 400 1,330 Slope-area measurement Collins (1964), Rostvedt
Utah at unknown location (1968, pp. B84-B87), and
Lindon (2004)
27 Lower Latham, E, H 1926 1973 IE 25.7 7.08 5.79 7.01 120 340 Slope-area measurement L. H. DeGrave, Colorado
Colorado at unknown location Dept. of Natural Resources,
written communication
(1986)
28 Lower Otay, R, C 1897 1916 OF 53.3 56.9 39.6 39.6 200 15,800 Twice the average flow Jansen (1983, pp. 151–153),

04016041-5

J. Hydrol. Eng.
California rate needed to empty the Justin (1932, pp. 41–43),
reservoir in 2.0 h Justin et al. (1950),
McGlashan and Ebert (1918),
and Silent (1916)
29 Lower Two E, H 1913 1964 IE 33.3i 29.6 11.3 11.3 350 1,800 Slope-area measurement Boner and Stermitz (1967)
Medicine, Montana 4 km downstream and USBR (1965, 1986)
30 Orós, Brazil E, R, Z 1960 1960e OF 110 660 35.8 35.5 2,000 58,000 Rate of reservoir volume “The breaching of the Orós
change, unknown time earth dam” (1960),
interval International Commission on
Large Dams (1974, pp. 68–
70 and 510–512), Jansen
(1983, pp. 166–167), and
Wahl (2014, p. 81)
31 Porter Hill, Oregon E, H 1990 1993 OS 12.0 0.015 5.00 5.80 80 30 Slope-area measurement Costa and O’Connor (1995)
150 m downstream
32 Prospect, Colorado E, H 1914 1980 IE 13.1 3.54 1.68 4.42 900 116 Rate of reservoir volume L. H. DeGrave, Colorado
change, unknown time Dept. of Natural Resources,
interval written communication
(1986)
33 Puddingstone, E, H 1926 1926 OF 43.0i 0.432 13.7d 13.7 420 960 Twice the average flow “Flood damage” (1926),
California rate needed to empty the International Commission
reservoir in 15 min on Large Dams (1974,
pp. 867–869), and Singh
(1996, pp. 94–95)

J. Hydrol. Eng.
Downloaded from ascelibrary.org by La Trobe University on 07/06/16. For personal use only.

Table 1. (Continued.)
Volume of

© ASCE
Average water above Height of Approach Measured
Embank- embank- breach water above Height flow peak
Dam name and ment Year Year Failure ment width bottom breach bottom of breach width discharge Peak discharge
Number location typea completed failed modeb W avg (m) V w (Mm3 ) H w (m) H b (m) La (m) Qp (m3 =s) calculation method References
34 Quail Creek, Utah E, Z 1986 1989 IE 56.6 30.8 16.7 21.3 550 3,110 Rate of reservoir volume “Investigation” (1989),
change, 15 min time Richard Hall, Utah Dept. of
interval Natural Resources, written
communication (1989),
O’Neill and Gourley (1991),
Carlson and Meyer (1995),
and Interagency Committee
on Dam Safety (2015)
35 Rito Manzanares, E, H — 1975 IE 13.4 0.120 4.57 7.32 75 83 Slope area measurements Watson (1975), and on-site
New Mexico at several locations from inspection by author
90 m to 13.5 km (September 2015)
downstream
36 Schaeffer, Colorado E, H — 1921 OF 80.8 4.44 31.9 30.5 350 4,930 Twice the average International Commission on
outflow rate needed to Large Dams (1974,
empty the reservoir in pp. 769–770), Follansbee and
30 min; Peak flow rate Jones (1922), and on-site
13 km downstream was survey by author
4,380 m3 =s from (September 1985)
slope-area measurement
37 Shimantan, Henan E, H — 1975 OF 58.0 167 27.4 25.8 900 25,300 Rate of reservoir volume Fujia and Yumei (1994),
Province, China change, unknown time Qing (1998, p. 26), and Xu

04016041-6

J. Hydrol. Eng.
interval et al. (2008)
38 South Fork, E, H 1853 1889 OF 64.0 18.9 24.6 24.4 300 8,500 Rate of reservoir volume “Another Reservoir” (1889),
Pennsylvania change, 30 min time Francis et al. (1891), Water
interval Supply Commission of
Pennsylvania (1914), Jansen
(1983, pp. 184–191),
McCullough (1987), and
on-site survey by author
(May 1989)
39 Taum Sauk E, Z 1962 2005 OF 46.8 5.39 36.5 36.3 250 8,180 Rate of reservoir volume FERC (2006a, b) and
Reservoir, Missouri change, 1 min time Rydlund (2006)
interval
40 Teton, Idaho E, Z 1975 1976 IE 250 310 77.4 86.9 500 65,120 Slope-area measurement Brown and Rogers (1977)
4 km downstream and Jansen (1983,
pp. 191–213)
41 Zhuguo, Henan E, Z 1970 1975 OF 99.0 18.5 23.8 23.5 300 11,200 Unknown Shuibo et al. (1993), and
Province, China Xu et al. (2008)
a
C = solid corewall (either concrete, masonry, or steel); E = earthfill; H = homogeneous, M = masonry; R = rockfill; and Z = zoned.
b
IE = internal erosion; OF = overtopping by floodwaters; OG = overtopping caused by outlet works failure; OS = overtopping caused by embankment sliding; OW = overtopping caused by wave action; OX =
intentional breaching by excavation (analyzed as an internal erosion failure).

J. Hydrol. Eng.
Table 2. Range of Variables in Dam Breach Peak Discharge Data Set where
Minimum Maximum 
1; for nonovertopping failure modes
Variable Units value value kM ðθ2 Þ ¼
θ2 ; for overtopping failure modes
Average embankment width (W avg ) m 9.63 250
Volume of water above breach bottom (V w ) Mm3 0.0133 701 is a failure mode factor,
Height of water above breach bottom (H w ) m 1.68 77.4
8
Height of breach (H b ) m 3.66 86.9
<1; for H b ≤ θ3
Approach flow width (La ) m 40 4,100 
m3 =s 30
kH ðθ3 ; θ4 Þ ¼ H b θ4
Measured peak discharge (Qp ) 65,120 : ; H b > θ3
θ3

is an embankment height factor, g = gravitational acceleration, and


Methods used to determine peak discharge from three dams
θi is the ith model parameter. The model given in Eq. (1) has
(Frechnman Creek, Lake Avalon, and Zhuguo) are unknown. How-
evolved from a previous analysis by the writer (Froehlich 1995)
ever, because the persons or organizations reporting the discharges
based on dimensional similitude and multiple regression analysis
are considered reliable sources of information, and because the
of a smaller set of peak discharge than is evaluated here. The rela-
flow rates do not appear to be extreme exceptions, the values have
tion includes physically relevant variables that were able to be col-
been accepted for use in the analysis carried out here.
lected from a wide variety of sources. However, noticeably absent
Some of the peak discharges in Table 1 differ from values re-
are data describing soil characteristics of the embankments, which
ported previously because of discovery of new information or
might have significant influence on the speed and extent of breach
Downloaded from ascelibrary.org by La Trobe University on 07/06/16. For personal use only.

reevaluation of existing data, as described in the following:


development. Because detailed soil characteristics are available for
1. Outflow from Hell Hole Dam used by the author in an earlier
so few of the dam failures reported in Table 1, they have not been
analysis (Froehlich 1995) has been increased to 17,000 m3 =s,
included in the data summary or evaluated in this analysis.
which is twice the average flow rate needed to empty the reser-
The generalized reduced gradient algorithm (Lasdon et al. 1976)
voir in 1 h.
was used to find the following optimal values of model parameters
2. Courivaud (2007, pp. 6–10) and Wahl (2014, p. 81) report that a
that provided a best fit to the 41 measurements of peak discharge:
recent evaluation of the Oros Dam failure drawdown records by
Électricité de France (EDF) as part an investigation by the Cen- θ1 ¼ 0.0175; θ2 ¼ 1.85; θ3 ¼ 6.1 mð20 ftÞ;
tre for Energy Advancement through Technological Innovation
1 1 1 1
(CEATI) Dam Safety Interest Group determined a peak outflow θ4 ¼ ; θ5 ¼ ; θ6 ¼ − ; and θ7 ¼
rate of 58,000 m3 =s, which is about 6 times greater than the 8 2 2 2
previously reported value (9,630 m3 =s), and which conforms Parameters in Eq. (1) were found by minimizing the root-mean-
much more closely with measurements of peak discharge from square error (RMSE) of the logarithms of Qp to reduce dependence
other large dam failures. on extremely large values, such as those of the Teton, Banqiao, and
3. Peak discharge caused by breaching of Rito Manzanares Dam Oros dam failures. Some of the parameters (θ3 to θ7 ) were adjusted
calculated by Watson (1975) has been revised based on reeva- slightly to nearby simple fractions. Inserting parameters and sim-
luation by the writer of roughness coefficients that characterize plifying Eq. (1) leads to the following expression for the expected
flow resistance of the downstream channel made during an value of Qp :
on-site inspection in September 2015.
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
4. The commonly reported peak discharge from the Banqiao
gV w Hw H 2b
Dam failure has been reduced to 56,300 m3 =s, which is twice Q̂p ¼ 0.0175 × kM × kH × ð2Þ
the average flow rate needed to empty the reservoir in 6.0 h. W avg
Reasons for the decrease are explained in Appendix I.
where

1; for non-overtopping failure modes
Peak Discharge Relations kM ¼ ;
1.85; for overtopping failure modes
Two mathematical models are presented for predicting peak dis- 8
charge from gradually breached embankment dams. The form of < 1;
 1=8
for Hb ≤ Hs
kH ¼ Hb ; and
the first model is based on previously developed statistical relations : ; for H b > Hs
using measures of embankment and reservoir properties as pre- Hs

dictors. The second model reduces the maximum possible dis- 6.1 m ðfor SI unitsÞ
charge from an instantaneous reservoir breach having prescribed Hs ¼
dimensions. 20 ft ðfor U:S: customary unitsÞ

The effect of embankment height given by the factor kH changes


Empirical Formula noticeably for Hb < 6.1 m (20 ft) as shown in Fig. 1. The coeffi-
A nonlinear mathematical model, in which variables are normalized cient of determination Eq. (2) is 0.922, and the RMSE of ln Qp is
to create dimensionless quantities, for peak discharge from gradu- 0.277. Inclusion of La in the nonlinear model offered no significant
ally breached embankment dams is proposed as follows: improvement. A comparison of measured and predicted values of
 θ   Qp given by Eq. (2) for the 41 dams studied is shown in Fig. 2.
Vw 5 W avg θ6 Failure mode has a significant effect on peak discharge as
Q̂p ¼ θ1 × kM ðθ2 Þ × kH ðθ3 ; θ4 Þ × × shown by the factor kM , which predicts outflows produced by over-
H 3b Hb
 θ qffiffiffiffiffiffiffiffiffi topping to be 85% larger than those caused by internal erosion, all
Hw 7
× × gH 5b ð1Þ other factors being the same. Erosion of embankment material by
Hb overflows will remove masses of material whose weight holds the

© ASCE 04016041-7 J. Hydrol. Eng.

J. Hydrol. Eng.
dam in place against the hydraulic forces acting to move the dam. or multiple locations, or along the entire structure. When a breach
Even small sustained overtopping flows having depth of only a does eventually occur, lateral erosion of the opening can proceed
few centimeters can erode large amounts of soil from downstream faster than if the dam had not been overtopped because of the large
embankment slopes protected by vegetation within hours in single amount of downstream embankment material that has already been
removed. Furthermore, when spillway capacities are exceeded dur-
ing extreme floods and embankment overflows occur, additional
water volume will be added to the reservoir during the failure,
thereby increasing the rate and ultimate extent of breach enlarge-
ment. Generally, internal erosion failures do not occur during large
floods; therefore, reservoir inflows will be small when dams that
are not overtopped do fail.
The effect of embankment height on peak breach discharge is
shown by the factor kH to change significantly for H b < 6.1 m
(20 ft). Peak discharge from breaches of smaller dams will be
greater than would otherwise be expected. Most probably soil prop-
erties including degree of compaction, cohesion, and particle size
have more pronounced influence on the rate of erosion of the em-
bankments, thus causing breach growth to speed up and peak dis-
charge to be increased.
Downloaded from ascelibrary.org by La Trobe University on 07/06/16. For personal use only.

Semitheoretical Formula
Instantaneous and complete failure of a dam in an ideal channel
Fig. 1. Embankment height factor, kH both with and without tailwater and initially motionless water
is analyzed in a number of texts (see, for example, Stoker 1957,
pp. 333–341; Henderson 1966, pp. 304–309; Liggett 1994,
pp. 285–289, 293–298). The general problem considers a horizon-
tal frictionless channel with a wide rectangular cross-section of
constant shape extending longitudinally to infinity. With the dam
located at longitudinal channel distance x ¼ 0, initially water has
constant depth h0 for x > 0, and h1 for x < 0, with h0 ≤ h1 , and
zero velocity everywhere. At time t ¼ 0, the dam is suddenly
destroyed.
For the special case with h0 ¼ 0, which was analyzed first by
Ritter (1892), the discontinuity at x ¼ 0, t ¼ 0 is instantly wiped
out when the dam is removed resulting in a p positive
ffiffiffiffiffiffiffiffiffiffi wave moving
downstream in the dry channel with velocity 2gh1, and a negative
pffiffiffiffiffiffiffi
wave traveling upstream in the reservoir with velocity gh1. Water
depth p h ffiffiffiffiffiffiffi
along p
the
ffiffiffiffiffi portion of water in motion is given by
x ¼ ð2 gh1 − 3 ghÞt, which is a concave p upwards
ffiffiffiffiffiffiffi parabolic pro-
file tangent to the channel bottom p x ¼ 2 pgh
atffiffiffiffiffiffiffi ffiffiffiffiffi1 t. Velocity at any
location along the profile is V ¼ 2 gh1 − 2 gh. At the dam sec-
tion (x ¼ 0)

qffiffiffiffiffiffiffi
4 2 pffiffiffiffiffiffiffi 8
h ¼ h1 ; V ¼ gh1 ; and q¼V×h¼ gh31 ð3Þ
9 3 27

Fig. 2. Comparison of measured embankment dam breach peak


discharges and those predicted by Eq. (2) where q = unit discharge, which along with h and V, is independent
of time.
Because of resistance from bed friction and turbulence in real
channels, the downstream end of the profile is found to have
a rounded shape terminating in a near vertical face. However,
Schoklitsch (1917) finds good agreement between the theoretical
solution for the upstream end of the wave profile and experimental
measurements. Other experiments (Arbuthnot and Strange 1960)
also show that resistance has little effect on waves propagating up-
stream into the reservoir. For this reason, the theoretical maximum
discharge in a unit width channel will be calculated using Eq. (3).
The maximum peak discharge that would be produced by an
instantaneous partial breach (that is, a breach that does not encom-
pass the entire length of a dam) that forms in the shape of a trap-
ezoid (Fig. 3) is found by integrating q across the breach section to
Fig. 3. Trapezoidal dam breach representation
obtain the following relations:

© ASCE 04016041-8 J. Hydrol. Eng.

J. Hydrol. Eng.
8     qffiffiffiffiffiffiffiffiffi
>
> 8 La 0.28 4
>
< 27 Bavg B − m H b − H gH 3w ; for H w ≤ Hb
avg
5 w
Qpmax ¼      qffiffiffiffiffiffiffiffiffi ð4Þ
>
> 8 La 0.28 4 H 5=2
>
: ðBavg − mH b Þ − mH w 1 − b −1 gH 3w ; for H w > Hb
27 Bavg 5 Hw

where Bavg = average width of the final breach and m = average Critical Overtopping Depth
side-slope ratio (horizontal to vertical) of the trapezoidal opening.
Prediction of peak discharge from potential overtopping failures
The factor ðLa =Bavg Þ0.28 has been added to account for the effects
using Eqs. (2) and (5) requires an estimate of the critical overtop-
of a partial breach (that is, a breach that does not extend across the
ping depth at which breaches are considered to begin forming
entire width of the approach channel) based on the experimental
so that Hw and V w can be determined. The Federal Emergency
findings of Arbuthnot and Strange (1960) and Schmidgall and
Management Agency (FEMA) provides critical depths of overtop-
Strange (1961).
ping Hc ranging from 0 to 0.6 m that depend on the physical con-
Breach formation is considered to begin after water, either over-
dition of the embankment as described in Table 3 (FEMA 1987).
topping the embankment or seeping through it, has eroded the
Presumably, embankment slopes for these dams are covered by
downstream slope and crest width to the point of the upstream face.
normal protective measures such as grass and loose rock. However,
At this stage of a failure, outflows through the breach will increase
Downloaded from ascelibrary.org by La Trobe University on 07/06/16. For personal use only.

special overtopping protection that can withstand significantly


quickly. Recorded water levels in the reservoir or in the down-
stream channel will show sudden changes soon after the breach
has begun to form. The time taken from initiation of the breach
until it has formed completely is known as the breach formation
time tf (Froehlich 2008).
Wetmore and Fread (1984) presented a formula for peak dis-
charge through a breach that reduces the maximum possible value
based on tf . A similar equation for expected peak discharge is pro-
posed here as follows:
0 1β
B 1 C
Q̂p ¼ Qpmax × @ qffiffiffiffiffiA ð5Þ
g
1 þ α × tf Hb

where Qpmax is given by Eq. (4); α and β are parameters that need to
be determined; and values of Bavg , m, and tf are calculated using
the following expressions developed previously by the author
(Froehlich 2008):
Bavg ¼ 0.27 × kM × V 1=3 w ; where

1.3; for overtopping failures
kM ¼ ;
1.0; other failure modes

1.0; for overtopping failures Fig. 4. Comparison of measured embankment dam breach peak dis-
m¼ ; and charges and those predicted by Eq. (5)
0.7; other failure modes
sffiffiffiffiffiffiffiffiffi
Vw
tf ¼ 63.2 × ð6Þ
gH 2b Table 3. Critical Overtopping Depth for Embankment Dams of Varying
Condition
Just as in evaluation of Eq. (1), nonlinear optimization was used Critical
to minimize the root-mean-square error (RMSE) of logarithms Condition overtopping
of values of Qp . using the 41 measurements of peak discharge of dam Description of dama depth H c (m)
presented in Table 1. The following expressions provide optimal Good Practically no seepage, no noticeable 0.61
values of αand β∶ settlement, and embankment slopes in
good condition
α ¼ 0.000045 and β ¼ 500 × ½ðW avg × H 2b Þ=V w 2=3 ð7Þ Fair Moderate seepage, some settlement of 0.3
crest, and some erosion on embankment
With α and β obtained from Eq. (7), the coefficient of determina- slopes
tion Eq. (5) is 0.917, and the RMSE of ln Qp is 0.313. Poor Excessive seepage, significant slump of 0.0
A comparison of measured and predicted values of Qp given by crest, cracks in embankment, and
Eq. (5) for the 41 dams analyzed is shown in Fig. 4. The effect of erosion of slopes
embankment width on peak discharge is shown by its influence on Note: Data from FEMA (1987).
the exponent β, which increases along with W avg . Accordingly, a
Presumably, no special overtopping protection is used to resist erosion of
with all other factors held constant in Eq. (5), increasing W avg re- embankment downstream slopes. Normal embankment coverage includes
duces Qp , as would be expected. grass and loose rock.

© ASCE 04016041-9 J. Hydrol. Eng.

J. Hydrol. Eng.
larger overflow depths than those given in Table 3 is installed on from actual full-scale dam failures and from small-scale experimen-
some embankment dams. These defensive measures include roller- tal studies. Even the semitheoretical formulations by Walder and
compacted concrete, precast concrete block systems, continuously O’Connor (1997) and by De Lorenzo and Macchione (2014), as
reinforced concrete slabs, gabions, reinforced rockfill, and geotex- well as the neural network model of Hooshyaripor et al. (2014),
tiles (Task Committee on Overtopping Protection 1994, p. 116; rely on observed failure data to establish model parameters.
FEMA 2014; Gosden et al. 2014). Ratios of predicted to measured Qp given by all of the methods,
along with those from Eqs. (2) and (5), are compared visually in the
box-and-whisker diagrams shown in Fig. 5. A box-and-whisker
Comparison of Peak Discharge Predictions
diagram provides a convenient way of depicting graphically the
Several other methods that have been developed for predicting prediction error (that is, the predicted value minus the measured
peak discharge from gradually breached embankment dams are value) through their five-number summaries: the smallest observa-
compared using the assembled data. Mathematical expressions tion (sample minimum, not including outliers), lower quartile (Q1),
for determining the peak discharge from a gradual embankment median (Q2), upper quartile (Q3), and largest observation (sample
dam failure developed by Kirkpatrick (1977), MacDonald maximum, not including outliers). Outliers are those measure-
and Langridge-Monopolis (1984), Costa (1985), SCS (1985), ments not within 1.5 × IQR of the lower and upper quartiles, where
Froehlich (1995), Webby (1996), Walder and O’Connor (1997), IQR ¼ Q3 − Q1 = interquartile range (IQR). The minimum and
Thornton et al. (2011), Hooshyaripor et al. (2014), De Lorenzo maximum outliers are also shown in the diagram, along with the
and Macchione (2014), and Azimi et al. (2015) are presented average. Spacing between different parts of the boxes helps indicate
in Table 4. The RMSE of values of ln Qp predicted for the failures the degree of dispersion (spread) and skew (asymmetry) of the pre-
Downloaded from ascelibrary.org by La Trobe University on 07/06/16. For personal use only.

in Table 1 are also provided in Table 4. diction errors.


Most of the formulas have been developed by multiple regres- From the box-and-whisker diagrams, predictions given by
sion analysis or nonlinear model fitting based on assembled data Eq. (2) are shown to be consistently accurate (the IQR extends from

Table 4. Summary of Formulas for Predicting Peak Discharge from Gradually-Breached Embankment Dams
Source Formulaa RMSEb

pffiffiffi 0.3 m; for SI units 1.166
Kirkpatrick (1977) Qp ¼ 0.4 × gðH w þ ηÞ2.5 , where η ¼
1.0 ft; for U.S. customary units
MacDonald and Langridge-Monopolis (1984) Qp ¼ 1.175 × ðV w H w Þ0.41 , where V w ¼ m3 , and H w ¼ m 0.730

Costa (1985) Qp ¼ 0.763 × ðV w H w Þ0.42 , where V w ¼ m3 , and H w ¼ m 0.696



maxf0.000421 × ½ðV w H w Þ=ðW avg H b Þ1.35 ; 1.77 × H 2.5
w g; if H w < 31.4 m 1.612
SCS (1985) Qp ¼
16.6 × H 1.85
w ; if H w ≥ 31.4 m
where V w ¼ m3 , H w ¼ m, H b ¼ m, and W avg ¼ m
Froehlich (1995) Qp ¼ 0.607 × V 0.295
w H 2.24 3
w , where V w ¼ m , and H w ¼ m 0.488
pffiffiffi
Webby (1996) Qp ¼ 0.0443 × gV 0.365
w H 1.405
w 0.386
pffiffiffi 0.47 0.15 0.94
Walder and O’Connor (1997) Qp ¼ 0.031 × gV w H w H b 0.608

Thornton et al. (2011) Eq. (1): Qp ¼ 0.12 × L1.785 a 1.688


Eq. (2): Qp ¼ 0.863 × V 0.335 w H 1.833
b W −0.663
avg 0.591
0.493 2.205 0.226
Eq. (3): Qp ¼ 0.012 × V w H b La 0.628
pffiffiffi
Hooshyaripor et al. (2014)c (1) Qp ¼ 0.0068 × gV 0.543 w H 0.871
w 0.756
The following formulas represent the neural network. Unity-based normalization is used 1.628
to bring all input and output values into the range ½−1,1
(2) Qp ¼ ðQp0 þ 1Þ½ð78100 − 2.12Þ=2 þ 2.12, where Qp0 ¼ −2.5417 ×
tanhð−9.0885 × V w0 − 5.7519 × H w0 þ 3.357Þ − 2.8054 × tanhð2.4520 × V w0 þ 1.5365 ×
H w0 − 1.0821Þ þ 0.1740 × tanhð2.6086 × V w0 þ 1.967 × H w0 þ 3.0178Þ − 0.0062 ×
tanhð1.4053 × V w0 − 7.6159 × H w0 − 3.386Þ − 1.1008
V w0 ¼ 2½ðV w − 0.0037Þ=ð660 − 0.0037Þ − 1, H w0 ¼ 2½ðH w − 1.37Þ=ð77.4 − 1.37Þ − 1,
V w ¼ Mm3 , H w ¼ m, and Qp ¼ m3 =s

De Lorenzo and Macchione (2014) 0.321 × g0.258 ðve V w Þ0.485 H 0.802
b ; for over topping failures 0.714
Qp ¼
0.347 × g0.263 ðve V w Þ0.474 H 2.992
w H b−2.151 ; for piping failures

0.07 m=s for SI units
where ve ¼
1.46 ft=s for U.S. customary units
pffiffiffiffiffiffiffiffiffi
Azimi et al. (2015) Qp ¼ 0.0166 × gV w × H w 0.365
a
Unless noted otherwise, any consistent set of units can be used. For example, for SI units, Qp ¼ m3 =s, g ¼ 9.807 m=s=s, V w ¼ m3 , H w ¼ m, and H b ¼ m;
and for U.S. customary units, Qp ¼ ft3 =s, g ¼ 32.174 ft=s=s, V w ¼ ft3 , H w ¼ ft, and H b ¼ ft.
b
Root-mean-square error between predicted values of ln Qp and the values actually measured.
c
Coefficients of the neural network model given in Hooshyaripor et al. (2014) are incorrect. The proper coefficients obtained from Farhad Hooshyaripor
(written communication, September 2015) are used here.

© ASCE 04016041-10 J. Hydrol. Eng.

J. Hydrol. Eng.
Downloaded from ascelibrary.org by La Trobe University on 07/06/16. For personal use only.

Fig. 5. Box-and-whisker diagrams of ratios of predicted to measured values of embankment dam breach peak discharge given by several methods

about 0.8 to 1.2), with median and average values both near one, A previous evaluation of dam breach peak discharge by the
and with little skew. The IQR of predictions from Eq. (5) is slightly author (Froehlich 1995) was based on measured flows from only
smaller, but the results are somewhat skewed. A single outlier is 22 embankment dam failures. The effect of embankment dam
generated by Eq. (2), and Eq. (5) produces none. The two formulas width on peak discharge could not be determined with certainty
also yield the smallest RMSE values, which is not unexpected from this limited set of data, although with all other factors held
considering that the expressions were fit specifically to the 41 mea- constant, increased embankment width would certainly lead to a
sured peak discharges. smaller peak outflow because of the additional volume of material
Predictions given by Froehlich (1995), Webby (1996), and that needs to be eroded. The expanded data set evaluated here has
Azimi et al. (2015) provide generally good results, although all allowed the effect of embankment width to be determined with
of these formulas tend to underpredict peak discharge, are slightly greater certainty, and the effect is taken into account in both of the
skewed, and, except for Webby (1996), generate high outliers. Even developed expressions.
though the formula given by Hooshyaripor et al. (2014) has the Although care should always be taken when applying a math-
smallest IQR of all the methods, it underestimates peak discharge ematical model to dams whose characteristics are outside the range
by about 50% and produces a comparatively large RMSE. of those used to develop the expression, the predicted peak dis-
All of the other methods produce results that are lower in ac- charge given by the semitheoretical relation is bounded by the po-
curacy than those just mentioned. Notably poor predictions are tential maximum flow that would be produced by instant formation
given by the SCS (1985) and the Thornton et al. (2011) formulas. of a breach. For this reason, one should not hesitate to apply the
Semitheoretical models developed by Walder and O’Connor (1997) formula to dams for which stored water volumes or embankment
and by De Lorenzo and Macchione (2014) both overestimate Qp , heights exceed those of the 41 failures included in the evaluation
have a comparatively large interquartile range, and generate high data set. It is also for this reason that this peak discharge formula is
outliers. The neural network model of Hooshyaripor et al. (2014) preferred by the author even though the purely empirical expression
yields several high outliers and produces a comparatively large provides a slightly better fit to the assembled data.
RMSE. Several uncomplicated flood-routing methods requiring small
amounts of data and effort to estimate peak flow rates at locations
downstream from a breached dam are currently used to develop
approximate inundation maps. Each of these methods requires
Summary and Conclusions an estimate of the maximum outflow rate from the reservoir.
Two nonlinear mathematical models were developed to predict the The findings of this paper will help improve peak discharge calcu-
lation and delineation of areas likely to be affected by dam-breach
peak discharge from a breached embankment dam based on analy-
floods.
sis of measured peak outflow rates from 41 dam failures ranging in
size from small to large. The first model is purely empirical, pro-
viding optimal values of coefficients to fit measured discharges Appendix I. Banqiao Dam Peak Breach Outflow Rate
best. The second model is based on a semitheoretical approach that
reduces the maximum possible peak discharge through a breach of During 4–8 August 1975, an extreme storm occurred in Henan
prescribed dimensions that forms instantly. Both formulas are Province, China, producing a record-setting 5-day total rainfall of
shown to provide more accurate estimates of peak outflow rates 1,631 mm. Two large dams, Banqiao and Shimandan, failed from
from failed dam used in the analysis than several other approaches. overtopping as the result of flooding produced by the storm.

© ASCE 04016041-11 J. Hydrol. Eng.

J. Hydrol. Eng.
The commonly reported peak outflow from the Banqiao Dam reservoir level gives a maximum possible total outflow rate of
failure is 78,100 m3 =s based on a single recording gauge measure- 80,750 m3 =s (with a small additional quantity of water overflow-
ment of the reservoir level and level-pool flood routing calculation ing the 1.3 m high parapet wall running along the crest of the
(Yi 1998; Henan Water Resources Authority 2005). dam), which assumes an instantaneous breach at the maximum
From information provided by Xu et al. (2008), the final reservoir level.
breach in the Banqiao Dam is approximated closely as a trapezoid Banqiao Dam breach formation was actually rather gradual,
with a top width of 372 m at elevation 116.3 m, and a bottom taking about 5.5 h. The reservoir is reported to have “drained com-
width of 210 m at elevation 86.8 m, giving Bavg ¼ 291 m, Hb ¼ pletely” in about 6 h. For these reasons, peak breach discharge was
29.5 m, and m ¼ 2.75. The maximum water-surface elevation certainly considerably less than the maximum possible value, which
recorded was 117.9 m, giving H w ¼ 31.1 m. From Eq. (4), using is only slightly greater than the commonly reported peak outflow rate
measured values of Bavg and m, the maximum possible outflow obtained from level-pool rate-of-drawdown calculations.
though the breach is Qpmax ¼ 76,900 m3 =s. Adding to this amount Estimated peak discharge through the breach is given by Eq. (5)
the combined spillway discharge (about 3,850 m3 =s) at the peak using measured values of Bavg , m, and tf as follows:

0 1500×½ðW 2 2=3
avg ×H b Þ=V w 

B 1 C
Q̂p ¼ Qpmax × @ qffiffiffiffiffiA
Downloaded from ascelibrary.org by La Trobe University on 07/06/16. For personal use only.

g
1 þ 0.000045 × tf Hb
2 3500 × f½59.6 m×ð30.3 mÞ2 =701 Mm3 g2=3
1
¼ 76,900 m3 =s × 4 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi5
1 þ 0.000045 × ð5.5 h × 3,600 s=hÞ 9.807 m=s=s
29.5 m

¼ 76,900 m3 =s × 0.6610.913 ¼ 52,700 m3 =s ð8Þ

After subtracting the combined spillway discharge (which was H c = critical overtopping depth;
about 600 m3 =s at the time of the peak discharge based on level- Hs = Height of small embankment dam;
pool drawdown rate) from the calculated total peak outflow Hw = height of water above breach bottom at time of breach
(78,100 m3 =s), the estimated peak discharge through the breach initiation;
remains significantly less than the commonly reported value, and h = water depth;
close to the value used here (56,300 m3 =s). kH = factor that accounts for embankment height;
An explanation for overestimation of the peak outflow rate at kM = factor that accounts for failure mode;
Banqiao Dam based on level-pool calculations may lie with the La = approach flow width;
position of the gauge used to record reservoir water-surface levels, m = average breach side-slope ratio (horizontal:vertical);
which is unknown. If it was located within the rapid drawdown Q
⌢ p = measured dam-breach peak discharge;
zone near the breach, the measured water level would have re- Qp = predicted dam-breach peak discharge;
flected a much faster outflow rate than if the gauge had been sit- Qpmax = maximum possible peak discharge from a breach of
uated far away from the breach. Need for a more comprehensive specified dimensions that forms instantly;
evaluation of reservoir drawdown is another reason. Goodell and q = unit width flow rate;
Wahlin (2009) studied differences in calculated peak discharge tf = breach formation time;
given by a level-pool flood routing approach and by dynamic flood V = water velocity;
routing (that is, numerical solution of the one-dimensional cross- V w = volume of impounded water above breach bottom at time
section averaged flow equations.) Their findings presented graphi- of failure;
cally suggest that dynamic flood routing might produce a peak W avg = average width of embankment above breach bottom;
discharge about 20% less than level-pool routing. α = coefficient in semitheoretical formula;
For the aforementioned reasons, the commonly reported peak β = coefficient in semitheoretical formula; and
discharge produced by the Banqiao Dam failure seems excessive. θ = coefficient in empirical formula.
Consequently, the significantly smaller value calculated as twice
the average outflow rate over a 6-h period is used in the analysis References
presented here.
“Another reservoir has burst.” (1889). Eng. News, 22(35), 193.
“Castlewood Dam failure floods Denver.” (1933). Eng. News-Rec.,
101(32), 174–176.
Notation “Dam collapses in southeast spur federal inspection plan.” (1977).
Eng. News-Rec., 199(19), 13.
The following symbols are used in this paper: “Failure of Apishapa earth dam in southern Colorado.” (1923). Eng.
Bavg = catchment area; News-Rec., 91(9), 357–358.
g = gravitational acceleration; “Flood damage to Puddingstone earth-fill dam.” (1926). Eng. News-Rec.,
Hb = height of trapezoidal dam breach; 96(22), 913.

© ASCE 04016041-12 J. Hydrol. Eng.

J. Hydrol. Eng.
“Hell Hole Dam isn’t a complete washout.” (1965). Eng. News-Rec., Dalrymple, T., and Benson, M. A. (1967). “Measurement of peak discharge
174(10), 28–29. by the slope-area method.” Chapter A2, Techniques of water-resources
“Investigation of the cause of Quail Creek dike failure.” (1989). Rep. of investigations, U.S. Geological Survey, Washington, DC.
Independent Review Team. De Lorenzo, G., and Macchione, F. (2014). “Formulas for the peak
“Johnstown is inundated again by a record 500-year flood.” (1977). discharge from breached earthfill dams.” J. Hydraul. Eng., 10.1061/
Eng. News-Rec., 199(4), 9. (ASCE)HY.1943-7900.0000796, 56–67.
“Ninth biennial report of the state engineer to the Governor of Utah– Diacon, A., Stematiu, D., and Mircea, N. (1992). “An analysis of the Belci
1913-1914.” (1915). Arrow Press, Salt Lake City. dam failure.” Int. Water Power Dam Constr., 44(9), 67–72.
“The Bradfield inundation.” (1874). Chambers’s J. Popular Lit. Sci. Arts, Ekström, I. (2012). “Recent dam incidents and failures in Sweden.” Proc.,
538, 241–244. 6th Int. Conf. on Scour and Erosion, Société Hydrotechnique de France,
“The breaching of the Orós earth dam in the State of Ceará, Paris, 281–288.
North-East Brazil.” (1960). Water Water Eng., 64(774), 351–355. Environment Agency. (2014). “Reservoir flood maps (RFM) guide:
“The failure of Hell Hole Dam.” (1965). West. Constr., 40(4), 65–70. Explanatory note on reservoir flood maps for local resilience forums–
“Undermining causes failure of French Landing Dam.” (1925). Eng. version 3.” Rep. No. LIT 6882, Dept. for Environment, Food and Rural
News-Rec., 94(18), 735–736. Affairs, Bristol, U.K.
Arbuthnot, G. L., Jr., and Strange, J. N. (1960). “Floods resulting from FEMA (Federal Emergency Management Agency). (1987). “Risk-based
suddenly breached dams—Conditions of minimum resistance— approach to dam safety assessment.” Denver.
Hydraulic model investigation.” Rep. No. 1, U. S. Army Engineer FEMA (Federal Emergency Management Agency). (2014). “Technical
Waterways Experiment Station, Corps of Engineers, Vicksburg, MS. manual: Overtopping protection for dams.” Washington, DC.
ASDSO (Association of State Dam Safety Officials). (2015). “Dam failures FERC (Federal Energy Regulatory Commission). (2006a). “Report of
and incidents.” 〈http://www.damsafety.org/news/?p=412f29c8-3fd8-4529 findings on the overtopping and embankment breach of the Upper
Downloaded from ascelibrary.org by La Trobe University on 07/06/16. For personal use only.

-b5c9-8d47364c1f3e〉 (Oct. 22, 2015). Dam—Taum Sauk pumped storage project.” FERC No. 2277, Taum
Azimi, R., Vatankhah, A. R., and Kouchakzadeh, S. (2015). “Predicting Sauk Investigation Team, Washington, DC.
peak discharge from breached embankment dams.” E-Proc., 36th IAHR FERC (Federal Energy Regulatory Commission). (2006b). “Taum Sauk
World Congress, Hague, Netherlands. upper dam breach, FERC No. 2277—Technical reasons for the breach
Babb, A. O., and Mermel, T. W. (1963). “Catalog of dam disasters, failures, of December 14, 2005.” FERC Independent Panel of Consultants
and accidents.” U.S. Dept. of the Interior, Bureaus of Reclamation, (IPOC) Rep., Washington, DC.
Denver. Fiedler, W., King, W., and Schwanz, N. (2010). “Independent panel of
Barker, B., and Schaefer, M. (2007). “Dam safety guidelines—Technical engineers—Report on breach of Delhi Dam.” Miscellaneous Rep.,
Note 1: Dam break inundation analysis and downstream hazard classi- National Dam Safety Review Board, Federal Emergency Management
fication.” Washington State Dept. of Ecology, Water Resources Pro- Agency, Washington, DC.
gram, Dam Safety Office, Olympia, WA. Fiedler, W., King, W., and Schwanz, N., and Holman, W. (2013). “Lake
Binnie, G. M. (1981). Early Victorian water engineers, Thomas Telford, Delhi Dam failure.” Changing times—The challenges and risks of
London. managing aging infrastructure under a new financial reality, U.S.
Boner, F. C., and Stermitz, F. (1967). “Floods of June 1964 in northwestern Society on Dams, Denver, 675–692.
Montana.” U.S. Geological Survey, Washington, DC. Field, J. E. (1933). “Data on Castlewood Dam failure and flood.” Eng.
Brown, R. J., and Rogers, D. C. (1977). “A simulation of the hydraulic News-Rec., 101(36), 279–280.
events during and following the Teton Dam failure.” Proc., Dam-Break Field, J. E., Jones, B. M. and Floyd, O. N. (1923). “Failure of Apishapa
Flood Routing Model Workshop, U.S. Water Resources Council, earth dam in Colorado–II.” Eng. News-Rec., 91(11), 418–424.
131–163. Follansbee, R., and Jones, E. E. (1922). “The Arkansas River flood of
Brua, S. A. (1978). “Floods of July 19–20, 1977 in the Johnstown area, June 3–5, 1921.” U.S. Geological Survey, Washington, DC.
western Pennsylvania.” Rep. No. 78-963, U.S. Geological Survey, Follansbee, R., and Sawyer, L. R. (1948). “Floods in Colorado.” U.S.
Washington, DC. Geological Survey, Washington, DC.
California Department of Water Resources. (1964). “Investigation of the Francis, J. B., Worthen, W. E., Becker, M. J., and Fteley, A. (1891). “Report
failure of Baldwin Hills Reservoir.” Sacramento, CA. of the committee on the cause of the failure of the South Fork
Carlson, D. D., and Meyer, D. F. (1995). “Flood on the Virgin River, Dam—With discussion.” Transactions, 24, 432–469.
January 1989, in Utah, Arizona, and Nevada.” Water-Resources Inves- Froehlich, D. C. (1995). “Peak outflow from breached embankment dam.”
tigations Rep. No. 94-4159, U.S. Geological Survey, Salt Lake City. J. Water Resour. Plann. Manage., 10.1061/(ASCE)0733-9496(1995)
Chen, C.-L., and Armbruster, J. T. (1980). “Dam-break wave model— 121:1(90), 90–97.
Formulation and verification.” J. Hydraul. Div., 106(5), 747–767. Froehlich, D. C. (2008). “Embankment dam breach parameters and their
Collins, W. E. (1964). “Report on the cause and failure of the Little Deer uncertainties.” J. Hydraul. Eng., 10.1061/(ASCE)0733-9429(2008)
Creek Dam, Provo River Project, Utah.” Miscellaneous Rep., U.S. Dept. 134:12(1708), 1708–1721.
of the Interior, Bureau of Reclamation, Denver. Fujia, T., and Yumei, L. (1994). “Reconstruction of Banqiao and Shimantan
Costa, J. E. (1985). “Floods from dam failures.” Rep. No. 85-560, U.S. dams.” Int. J. Hydropower Dams, 1(4), 49–53.
Geological Survey, Denver. Goodell, C. R., and Wahlin, B. (2009). “Dynamic and level pool reservoir
Costa, J. E. (1994). “Multiple flow processes accompanying a dam-break drawdown: A practical comparison for dam breach modeling.” Proc.,
flood in a small upland watershed, Centralia, Washington.” Water- 33rd IAHR Congress, Water Engineering for a Sustainable Environ-
Resources Investigations Rep. No. 94-4026, U.S. Geological Survey, ment, International Association for Hydraulic Research, Madrid, Spain,
Denver. 6843–6850.
Costa, J. E., and O’Connor, J. E. (1995). “Geomorphically effective Gosden, J. D., Ambler, T., and Courtnadge, A. P. (2014). “Improving the
floods.” Natural and anthropogenic influences in fluvial geomorphol- overtopping resistance of existing flood detention reservoirs.” Main-
ogy, J. E. Costa, A. J. Miller, K. W. Potter, and P. R. Wilcock, eds., taining the safety of our dams and reservoirs: Proc., 18th Biennial
American Geophysical Union, Washington, DC. Conf. of the British Dam Society, Belfast, Northern Ireland, 426–437.
Courivaud, J. R. (2007). “Analysis of the dam breaching database.” CEATI Graham, W. J., and Hilldale, R. C. (2002). “Spillway gate failure or mis-
Rep. No. T032700-0207B, CEATI International, Dam Safety Interest operation: Representative case histories.” U.S. Bureau of Reclamation,
Group, Montréal. Denver, CO, 1–12.
Crisp, R. L., Fox, W. E., Robinson, R. C., and Sauer, V. B. (1977). “Report Hagen, V. K. (1982). “Re-evaluation of design floods and dam safety.”
of failure of Kelly Barnes Dam, Toccoa, Georgia.” Federal Investigative Proc., 14th Int. Commission on Large Dams (ICOLD) Congress, Vol. 1,
Board Rep., U.S. Army Corps of Engineers, Washington, DC. International Commission on Large Dams, Paris, 475–491.

© ASCE 04016041-13 J. Hydrol. Eng.

J. Hydrol. Eng.
Henan Water Resources Authority. (2005). The August 1975 catastrophic Rosvall, G., and Dahlberg, P. (2011). “Dammbrott, Hästberga, Hässleholms
flood disaster in Henan, Yellow River Water Conservancy Press, kommun, Skåne län, den 7 November 2010.” Final Rep., Statens
Zhenzhou, China. Haverikommission (Swedish Accident Investigation Board),
Henderson, F. M. (1966). Open channel flow, MacMillan, New York. Stockholm, Sweden.
Holman, W. E. (2011). “Lake Delhi Dam—Design alternatives report– Rydlund, P. H. (2006). “Peak discharge, flood profile, flood inundation,
Appendix B: Hydrologic and hydraulic studies report.” Stanley and debris movement accompanying the failure of the upper reservoir
Consultants Rep., Lake Delhi Combined Recreational Facility and at the Taum Sauk pump storage facility near Lesterville, Missouri.”
Water Quality District, Delhi, India. Scientific Investigations Rep. 2006-5284, U.S. Geological Survey,
Hooshyaripor, F., Tahershamsi, A., and Golian, S. (2014). “Application of Reston, VA.
copula method and neural networks for predicting peak outflow from Sanders, C. L., and Sauer, V. B. (1979). “Kelly Barnes Dam failure flood
breached embankments.” J. Hydro-Environ. Res., 8(3), 292–303. of November 6, 1977.” Hydrologic Investigations Atlas HA-613, U.S.
Houk, I. E. (1933). “Failure of Castlewood rock-fill dam.” West. Constr. Geological Survey, Washington, DC.
News Highway Build., 8(16), 373–375. Saxena, K. R., and Sharma, V. M. (2005). Dams: Incidents and accidents,
Hoxit, L. R., Maddox, R. A., Chappel, C. F., and Brua, S. A. (1982). A. A. Balkema Publishers, Leiden, Netherlands.
“Johnstown-western Pennsylvania storm and floods of July 19-20, Schmidgall, T., and Strange, J. N. (1961). “Floods resulting from suddenly
1977.” U.S. Geological Survey, Washington, DC. breached dams—Conditions of high resistance—Hydraulic model in-
Interagency Committee on Dam Safety. (2015). “Evaluation and moni- vestigation.” Rep. No. 2, U.S. Army Engineer Waterways Experiment
toring of seepage and internal erosion.” FEMA P-1032, Federal Station, Corps of Engineers, Vicksburg, MS.
Emergency Management Agency, Washington, DC. Schoklitsch, A. (1917). “Über Dammbuchwellen.” Sitzungberichten
International Commission on Large Dams. (1974). Lessons from dam der Königliche Akademie der Wissenschaften, 126(Part IIa),
incidents, Paris. 1489–1514.
Downloaded from ascelibrary.org by La Trobe University on 07/06/16. For personal use only.

Jansen, R. B. (1983). “Dams and public safety.” U.S. Dept. of the Interior, Scott, K. M., and Gravlee, G. C., Jr. (1968). “Flood surge on the Rubicon
Bureau of Reclamation, Denver. River, California–Hydrology, hydraulics, and boulder transport.”
Jessup, W. E. (1964). “Baldwin Hills dam failure.” Civ. Eng., 34(2), U.S. Geological Survey, Washington, DC.
62–64. SCS (Soil Conservation Service). (1981). “Simplified dam-breach routing
procedure.” U.S. Dept. of Agriculture, Washington, DC.
Justin, J. D. (1932). Earth dam projects, 2nd Ed., Wiley, New York.
SCS (Soil Conservation Service). (1985). “Earth dams and reservoirs.”
Justin, J. D., Hinds, J., and Creager, W. P. (1950). Engineering for
U.S. Dept. of Agriculture, Washington, DC.
dams—Volume III— Earth, rock-fill, steel, and timber dams, Wiley,
SFWMD (South Florida Water Management District). (1980). “Embank-
New York.
ment failure, Florida Power and Light Company, Martin Plant cooling
Kirkpatrick, G. W. (1977). “Guidelines for evaluating spillway capacity.”
reservoir.” Interim Final Draft Rep., West Palm Beach, FL.
Water Power Dam Constr., 29(8), 29–33.
Shuibo, P., et al. (1993). “Chinese-Finnish cooperative research work on
Lasdon, L. S., Waren, A. D., Jain, A., and Ratner, M. (1976). “Design and
dam break hydrodynamics—Part I. Investigation report on dam safety
testing of a generalized reduced gradient code for nonlinear program-
research in China.” National Board of Waters and the Environment,
ming.” Technical Rep. SOL 76-3, Systems Operations Laboratory,
Helsinki, Finland.
Dept. of Operations Research, Stanford Univ., Stanford, CA.
Silent, R. A. (1916). “Failure of the lower Otay Dam.” Eng. News, 75(3),
Liggett, J. A. (1994). Fluid mechanics, McGraw-Hill, New York.
334–336.
Lindon, M. (2004). “The Little Deer Creek Dam failure.” J. Dam Saf., 3(4), Singh, V. P. (1996). Dam breach modeling technology, Kluwer Academic,
20–24. Dordrecht, Netherlands.
MacDonald, T. C., and Langridge-Monopolis, J. (1984). “Breaching char- Sterling, G. (1914). “Failure of the dam of the Hatchtown Reservoir, Utah.”
acteristics of dam failures.” J. Hydraul. Eng., 10.1061/(ASCE)0733 Eng. News, 71(23), 1274–1275.
-9429(1984)110:5(567), 567–586. Stoker, J. J. (1957). Water waves: The mathematical theory with applica-
Marcello, A., Frigerio, A., and Mazzà, G. (2009). “Case histories: A world- tions, Interscience, New York.
wide view.” Dam-break problems, solutions and case studies, D. de Task Committee on Overtopping Protection. (1994). Alternatives for
Wrachien and S. Mambretti, eds., WIT Press, Southampton, U.K., overtopping protection of dams, ASCE, Reston, VA.
273–334. Thornton, C. I., Pierce, M. W., and Abt, S. R. (2011). “Enhanced predic-
McCullough, D. G. (1987). The Johnstown flood, 2nd Ed., Simon and tions for peak outflow from breached embankment dams.” J. Hydrol.
Shuster, New York. Eng., 10.1061/(ASCE)HE.1943-5584.0000288, 81–88.
McGlashan, H. D., and Ebert, F. C. (1918). “Southern California floods of USACE (U.S. Army Corps of Engineers). (2013). “National inventory of
January, 1916.” U.S. Geological Survey, Washington, DC. Dams Trifold Brochure.” EP 360-1-23, Geospatial Systems Directorate,
Middlebrooks, T. A. (1953). “Earth-dam practice in the United States.” Alexandria, VA.
Trans. Am. Soc. Civ. Eng., 118, 697–722. USBR (U.S. Bureau of Reclamation). (1965). “Schedule, general
Murphy, E. C. (1905). “Destructive floods in the United States in 1904.” provisions, specifications, and drawings—Lower Two Medicine
U.S. Geological Survey, Washington, DC. Dam–Blackfeet Indian Irrigation Project, Montana.” Denver.
Oltman, R. E. (1955). “Floods of April 1952 in the Missouri River basin.” USBR (U.S. Bureau of Reclamation). (1986). “Guidelines to decision
U.S. Geological Survey, Washington, DC. analysis.” Denver.
O’Neill, A. L., and Gourley, C. (1991). “Geologic perspective and cause USCOLD (U.S. Committee on Large Dams). (1988). Lessons from dam
of the Quail Creek Dike failure.” Bull. Assoc. Eng. Geol., 28(2), incidents, ASCE, Reston, VA.
127–145. Vogel, A., and Klein, R. (1993). “Der Bruch des Staudammes Belci in
Qing, D. (1998). The river dragon has come!: The Three Gorges Dam and Rumänien.” Geotechnik, 16(1), 12–17.
the fate of China’s Yangtze River and its people, M. E. Sharpe, Armonk, Wahl, T. L. (2014). “Evaluation of erodibility-based Embankment Dam
NY. breach equations.” Hydraulic Laboratory Rep. HL-2014-02, U.S. Dept.
Raška, P., and Emmer, A. (2014). “The 1916 catastrophic flood follow- of the Interior, Bureau of Reclamation, Denver.
ing the Bílá Desná dam failure: The role of historical data sources Walder, J. S., and O’Connor, J. E. (1997). “Methods for predicting peak
in the reconstruction of its geomorphologic and landscape effects.” discharge of floods caused by failure of natural and constructed earthen
Geomorphology, 226, 135–147. dams.” Water Resour. Res., 33(10), 2337–2348.
Ritter, A. (1892). “Die Fortpflanzung der Wasserwellen.” Vereine Warne, W. E. (1964). “The Baldwin Hills Dam failure.” West. Constr.,
Deutscher Ingenieure Zeitschrift, 36(2), 947–954. 39(2), 78–80.
Rostevdt, J. O. (1968). “Summary of floods in the United States during Water Supply Commission of Pennsylvania. (1914). Rep. of the Water Sup-
1963.” U.S. Geological Survey, Washington, DC. ply Commission of Pennsylvania–1912, State Printer, Harrisburg, PA.

© ASCE 04016041-14 J. Hydrol. Eng.

J. Hydrol. Eng.
Watson, A. T. (1975). “Report on dam failure on Rito Manzanares.” M. V. Khire, and A. N. Alshawabkeh, eds., ASCE, Reston, VA,
New Mexico State Engineer Office Rep., New Mexico Office of the 162–169.
State Engineer, Dam Safety Bureau, Santa Fe, NM. Xu, Y., Zhang, L., and Jia, J. (2009). “Catastrophic failure of Banqiao
Webby, M. G. (1996). “Discussion of ‘Peak outflow from breached em- Dam.” Proc., Case Studies: Learning from Dam Incidents and Failures,
bankment dam.’ by D. C. Froehlich.” J. Water Resour. Plann. Manage., Centre for Energy Advancement through Technological Innovation
10.1061/(ASCE)0733-9496(1996)122:4(316), 316–317. (CEATI) International, Montreal, 24–25.
Wetmore, J. N., and Fread, D. L. (1984). “The NWS simplified dam break Yi, S. (1998). “The world’s most catastrophic dam failures: The August
flood forecasting model for desk-top and hand-held microcomputers.” 1975 collapse of the Banqiao and Shimantan Dams.” The river dragon
Federal Emergency Management Agency, Washington, DC. has come!, D. Qing, ed., M.E. Sharpe, New York.
Yochum, S. E., Goertz, L. A., and Jones, P. H. (2008). “Case study of
Xu, Y., Zhang, L., and Jia, J. (2008). “Lessons from Catastrophic
the Big Bay Dam failure: Accuracy and comparison of breach predic-
Dam Failures in August 1975 in Zhumadian, China.” Geocongress
tions.” J. Hydraul. Eng., 10.1061/(ASCE)0733-9429(2008)134:9(1285),
2008: Geosustainability and geohazard mitigation, K. R. Reddy,
1285–1293.
Downloaded from ascelibrary.org by La Trobe University on 07/06/16. For personal use only.

© ASCE 04016041-15 J. Hydrol. Eng.

J. Hydrol. Eng.

You might also like