281
WHAT HAPPENS IF THE CANAL BREAKS? TOOLS FOR ESTIMATING
CANAL-BREACH FLOOD HYDROGRAPHS
Tony L. Wahl1
ABSTRACT
A program of physical model tests and numerical unsteady-flow simulations has led to
the development of appraisal-level tools for predicting the characteristics of floods
caused by the breaching of homogeneous canal embankments. The procedures yield
estimates of the time needed for initiation and development of a breach, the magnitude of
the peak outflow, and the duration of the recession limb of the flood hydrograph. These
tools can help water managers identify canal reaches that have the potential to produce
floods with serious consequences. This can aid emergency management planning and
help to prioritize the need for more detailed investigations. This paper demonstrates the
use of the procedures and illustrates the importance of key input parameters, especially
the erodibility of the soil in the embankment. The method has not yet been tested against
real-world canal failures.
INTRODUCTION
The Bureau of Reclamation (Reclamation) is responsible for more than 8,000 miles of
irrigation canals in the western U.S., and failures of canal embankments have occurred
periodically throughout its history. When these canals were constructed, the adjacent
lands were primarily agricultural or undeveloped. Development of these lands has led to
greater interest in understanding the potential impacts and consequences of canal
embankment failures on surrounding areas. Threats to canals include animal burrows,
tree roots, penetrations by turnout pipes and utilities, embankment and foundation issues,
seismic events, internal erosion under static loading, hydrologic events, and operational
incidents.
Numerical modeling of breach outflows and downstream flooding can be used to evaluate
potential consequences of a canal breach. To facilitate appraisal-level investigations of
Reclamation’s canal inventory, a research program was undertaken to study the canal
breach process and develop tools for predicting canal breach outflow hydrographs (Wahl
and Lentz 2011). This work included both physical hydraulic modeling of the erosion
and breaching processes and numerical modeling of transient canal behavior during a
hypothetical breach event. Results from these studies were used to develop procedures
for estimating breach initiation and breach enlargement rates and associated canal breach
outflows. This paper illustrates the use of these procedures and demonstrates the
sensitivity of results to key input parameters.
1 Hydraulic Engineer, Bureau of Reclamation, Hydraulic Investigations and Laboratory Services, Denver, CO, 303-
445-2155, twahl@usbr.gov.
USCID -- April 15-19, 2013 -- Scottsdale, AZ
282 Seventh International Conference on Irrigation and Drainage
BACKGROUND
Although canal breaches have occurred throughout history, there have been remarkably
few efforts to generalize experiences from these events. Prior to this study, there was no
guidance specific to canals for predicting breach parameters or breach outflow
hydrographs. Dun (2007) provided the most notable prior work on the hydraulics of a
canal breach in a study of a navigation canal that failed in the United Kingdom in 2004.
Dun concluded that the hydraulics of canal breaches were significantly different from
breaches of traditional dams and storage reservoirs. For a traditional dam breach,
outflow is typically limited by the breach geometry and the reservoir storage, but for a
canal breach, outflow is also limited by the conveyance capacity of the reaches of canal
that deliver water to the breach site.
Nearly all canal embankments contain soils that may be conducive to headcut
development during erosion. Even canal embankment soils that do not demonstrate
plasticity contain enough fine materials to resist seepage loss of water and thus exhibit
enough apparent cohesion to allow headcuts to develop. Recognizing this general
characteristic, the typical stages of a canal breach can be described as follows:
1. Initial overtopping of the embankment, or development of a defect in an embankment
that allows erosive flow through the embankment or foundation (typically described
as internal erosion or “piping”).
2. Development of a headcut that begins on the downstream (outer) slope of the
embankment and migrates upstream toward the canal. In this stage, erosion is
primarily taking place downstream from the section (the hydraulic control) that
controls the outflow rate. The breach outflow rate is small and normal canal flow can
continue past the developing breach site.
3. Migration of the headcut through the hydraulic control, which enlarges the control
section rapidly and allows a dramatic increase in outflow. As the breach enlarges
during this stage, the size of the breach and the water level maintained in the canal are
the primary factors determining the outflow rate. In this phase, the breach outflow
becomes so large that flow reverses in the canal reach that was initially downstream
from the breach site.
4. The breach eventually enlarges to the point that the hydraulic control shifts from the
breach opening to the two canal sections. Critical-depth flow occurs in the leg of the
canal upstream from the breach and also in the leg of the canal downstream from the
breach. The breach may continue to widen, but the outflow rate cannot increase. As
the canal drains, the flow rate through the two critical sections drops and the breach
outflow rate is reduced.
One potential modification of this staged breach process is a situation in which the
embankment is weak enough to allow the overtopping channel or initial pipe to enlarge
so rapidly that steps 2 and 3 are not distinct from one another but are effectively
combined into one step in which erosion and enlargement of the hydraulic control section
USCID -- April 15-19, 2013 -- Scottsdale, AZ
Tools for Estimating Canal-Breach Flood Hydrographs 283
occurs simultaneously with headcut development and advance. This would not change
the hydraulic control shift that still occurs during the last step of the process.
The research studies carried out at Reclamation have focused on the last three steps of the
process outlined above. These studies assumed that the occurrence of the first step in the
process is given; there has been no attempt to model the initiation of piping, which is a
complex process that can occur through a large variety of specific mechanisms (e.g., Von
Thun 1996; Engemoen 2012). These studies have also been based on the conservative
assumption that there is no intervention, such as early shutdown of the canal or closing of
check gates at the upstream and downstream ends of a reach experiencing a breach event.
This provides results that are appropriate for the worst-case scenario of a breach that
develops so rapidly that intervention is not possible.
PHYSICAL MODELING
Physical modeling to support this research was described in detail by Wahl and Lentz
(2011). The facility used in the hydraulics laboratory (Figure 1) recreated a typical canal
flow situation prior to development of a breach. Water could be provided into both ends
of a non-erodible canal with an erodible test section in the middle. Each test started with
normal canal flow past the test embankment, and as the breach developed, the flow into
both ends of the model canal was increased to maintain boundary conditions at the breach
site that were representative of a fast-developing breach in a long canal reach (i.e., a
relatively steady canal water surface). The upper limit of inflow provided to each end of
the canal was the theoretical critical-flow discharge capacity of the canal sections.
Figure 1. Overview of canal breach model test
facility, looking in the upstream direction.
The three tested embankments were constructed in the model as simulated fill sections in
a canal reach that is elevated above the surrounding landscape. Soil used to construct the
embankments was a silty sand (SM) obtained from a local landscape materials supplier.
USCID -- April 15-19, 2013 -- Scottsdale, AZ
284 Seventh International Conference on Irrigation and Drainage
To simulate the wide range of erodibility properties that can occur in real canal
embankments, we varied both the water content at compaction and the level of
compaction effort. The test soil contained about 10% clay fines and exhibited some
plasticity (PI=5), so its erodibility was sensitive to the placement conditions. The
erodibility of each test embankment was measured during embankment construction and
after the completion of each breach test using submerged jet testing (Hanson and Cook
2004; ASTM D5852). Across three breach tests, the erodibility of the embankments
varied by about three orders of magnitude as indicated by detachment rate coefficients
obtained from the jet tests.
The breach tests exhibited three canal breach development scenarios, all initiated by
erosion through a pre-formed pipe in the embankment (a #4 rebar embedded in the
embankment and removed to start the test). The first test with a well-compacted and
erosion-resistant embankment produced a very slow headcut migration and breach
widening process, without a sudden and catastrophic breach outflow. This test was
representative of a scenario in which there would likely be adequate time to shut down
the canal and reduce the severity of the breach outflow. The second test demonstrated the
breach behavior of a poorly-compacted and very erodible embankment, with rapid
headcut development, headcut migration, and breach widening. The third test illustrated
an intermediate situation in which the embankment was very erodible, but the initial pipe
was located so high in the embankment that flow through it was small and initial headcut
development and migration were slow. However, when the headcut finally migrated into
the canal prism, failure and breach widening were nearly as rapid as that seen in the
second test.
Data collected from the three tests were used to relate the soil erodibility parameters
(detachment rate coefficient and critical shear stress) and hydraulic attack (estimated
shear stresses and energy dissipation rates) to observed headcut migration and breach
widening rates. The relations between these variables were found to be consistent with
observations from breach testing of traditional embankment dams (Hunt et al. 2005;
Temple et al. 2005; Hanson et al. 2011). This led to the development of simplified
mathematical models for predicting headcut advance, piping hole enlargement, and
breach widening rates. The first two models are relevant to estimating the time required
for breach initiation (the time preceding headcut advance through the hydraulic control),
which affects the amount of time available for detection of a breach in progress and
warning of the downstream population at risk. The last model can be used to estimate the
rate of breach enlargement after breach initiation has been completed.
NUMERICAL MODELING
The physical model tests provide a means to predict how a breach will develop. The
other significant question is what breach outflow hydrograph will be produced through
this opening. This is dependent on both the characteristics of the breach and the transient
behavior of the water within the canal reach in which the breach occurs, since drawdown
of the canal and development of a varying water surface profile in the canal will change
the head acting on the breach opening and the amount of flow that can be delivered to the
breach site. To quantify these effects, one-dimensional unsteady flow modeling was
USCID -- April 15-19, 2013 -- Scottsdale, AZ
Tools for Estimating Canal-Breach Flood Hydrographs 285
undertaken using HEC-RAS (Wahl and Lentz 2011). Numerous canal breach scenarios
were simulated with varying canal sizes, breach times, canal reach lengths, and breach
locations within the canal reach. This led to the development of dimensionless
relationships that yield estimates of breach hydrograph parameters (peak outflow and
recession time) as a function of breach development time, breach location within the
canal reach, and canal hydraulic properties.
CANAL BREACH OUTFLOW PREDICTION PROCEDURE
The essential characteristics of a canal breach hydrograph are the time required for
breach initiation, the time required for breach development, and the resulting breach
outflow hydrograph. The hydrograph may be defined by the peak outflow magnitude, the
time at which peak outflow occurs, and the time required for the hydrograph to recede.
The physical embankment breach tests and HEC-RAS modeling conducted in this
research project provide a basis for estimating all of these characteristics of a canal
breach event.
Breach Initiation
Breach initiation may take place through one or a combination of three different
processes: headcut advance caused by overtopping flow; headcut advance due to flow
through an existing piping channel that is not enlarging significantly; or continuous
enlargement of an existing piping defect. Models for all three processes were developed
(Wahl and Lentz 2011), but only the first two based on headcutting are presented here, as
they are believed to be more reliable at this time.
Breach Initiation by Headcut Advance due to Overtopping Flow. Consider the canal
embankment shown in Figure 2, which is depicted as a fill section deeper than the canal
prism. Flow overtops the embankment with head Hov. The unit discharge over the
embankment can be estimated from a broad-crested weir equation as q=2.6Hov
1.5 with Hov
in ft and q in ft3/s/ft. Assuming that headcutting initiates at the toe of the embankment,
the time for breach initiation is the time required for the headcut to advance the distance
L back to the upstream edge of the embankment crest. The headcut advance rate can be
estimated from (Temple et al. 2005)
( )1/ 3
h C qH
dt
dX = (1)
where:
dX/dt = headcut advance rate (ft/hr);
C = headcut advance rate coefficient (s1/3/hr);
q = unit discharge (ft3/s/ft); and
Hh = headcut height (ft).
Hanson et al. (2011) showed (and the physical hydraulic model testing of canal breaches
confirmed) that C can be estimated as C=0.44kd, with kd being the detachment rate
coefficient obtained from a submerged jet erosion test with units of ft/hr/psf. In the event
USCID -- April 15-19, 2013 -- Scottsdale, AZ
286 Seventh International Conference on Irrigation and Drainage
that a jet test is unavailable, values of kd may be estimated using Table 1 (Hanson et al.
2011) which relates kd to the clay content, compaction effort, and water content of the
soil during compaction (relative to the optimum water content that yields maximum dry
density during a compaction test). Note that this table gives values of kd in metric units,
but they may be converted using the factor shown with the table.
Figure 2. Canal embankment parameters for estimating headcut
advance rate due to overtopping flow.
Combining these equations, the time for breach initiation in hours is:
( )( 1.5 )1/ 3
ov
initiation 0.44 2.6 d h k H H
t = L (2)
where:
L = required headcut advance distance, from toe of exterior slope (ft);
Hh = potential height of headcut (ft);
Hov = overtopping head (ft); and
kd = detachment rate coefficient (ft/hr/psf).
One could argue that the headcut should be assumed to initiate at the top of the slope to
conservatively shorten the migration distance required, but in that case the head acting on
the headcut would be initially small. The headcut would eventually deepen to approach
Hh, and it is believed that the time required for this to occur is comparable to the time
needed for headcut migration from the toe back to the head of the slope.
Table 1. — Approximate values of kd in cm3/(N-s) as a function of compaction
conditions and % clay (Hanson et al. 2011). [1 cm3/(N-s) = 0.5655 ft/hr/psf]
% Clay
(<0.002 mm)
Modified
Compaction
(56,250 ft-lb/ft3)
Standard
Compaction
(12,375 ft-lb/ft3)
Low
Compaction
(2,475 ft-lb/ft3)
≥Opt WC% <Opt WC% ≥Opt WC% <Opt WC% ≥Opt WC% <Opt WC%
Erodibility, kd, cm3/(N·s)
>25 0.05 0.5 0.1 1 0.2 2
14-25 0.5 5 1 10 2 20
8-13 5 50 10 100 20 200
0-7 50 200 100 400 200 800
USCID -- April 15-19, 2013 -- Scottsdale, AZ
Tools for Estimating Canal-Breach Flood Hydrographs 287
Breach Initiation by Headcut Advance due to Piping Flow. Analysis of this case is
similar to the previous situation, except that the overtopping flow is replaced by orifice
flow through a piping defect in the embankment. The elevation of this defect and its
diameter and length must be specified to allow estimation of the flow rate through the
pipe. The starting diameter should be a practical value relating to the size of piping
defect that might prompt notice of the piping condition by project personnel and begin
the cycle of potential operational responses to a canal emergency. The key variables are
illustrated in Figure 3.
Figure 3. Canal embankment parameters for estimating headcut advance
rate due to piping flow.
The flow rate through the pipe can be estimated by applying the energy equation
( )
pipe
pipe
pipe
2
pipe
4 1
2
d
L
f
d gH
Q
+
=
π
(3)
where:
Q = discharge (ft3/s);
dpipe = pipe diameter (ft);
g = acceleration due to gravity (ft/s2);
Hpipe = head across pipe (ft);
f = friction factor, assumed to be 0.05 for a relatively rough pipe interior; and
Lpipe = length of pipe (ft).
The unit discharge effective in advancing the headcut can then be estimated by
converting the flow through the round pipe into the unit discharge of an equivalent
square jet, q=(π/4)1/2(Q/dpipe)=0.886Q/dpipe. The time required for headcut advance is
then computed as
( )( )1/ 3
pipe
initiation 0.44k 0.886QH / d
t L
d h
=
(4)
USCID -- April 15-19, 2013 -- Scottsdale, AZ
288 Seventh International Conference on Irrigation and Drainage
Note that the distance L is shown in Figure 3 as the distance to the upstream crest, not the
full distance to the upstream end of the pipe. This leads to a shorter, more conservative
estimate of the breach initiation time and is consistent with the observed behavior of the
test embankments, which seemed to experience collapse of the bridge over the pipe at
about the time that headcutting reached the upstream side of the crest.
Breach Initiation by Pipe Enlargement. A model for pipe enlargement was developed by
Wahl and Lentz (2011), but was extremely sensitive to the values of kd and the critical
shear stress of the soil, τc, as well as the choice of a starting condition for the piping
erosion analysis. The model may be of interest for future research.
Breach Development. The breach development phase is characterized by headcut
advancement through the upstream (canal side) slope of the embankment down to its toe,
followed by widening of the breach in both directions until the breach becomes wide
enough that it no longer serves as the hydraulic control. At this point, control of the flow
shifts to the critical-flow sections that will exist in the upstream and downstream canals.
For purposes of this appraisal-level model, the period of headcut advance into the canal is
assumed to be short compared to the time for breach widening and is incorporated into
the estimate of the widening time by assuming that widening begins from a breach width
of zero. The breach is assumed to have vertical sidewalls during the widening phase and
a rectangular cross-section, as observed in physical model tests and real embankment
failures.
To estimate the breach development time, it is necessary to first define the ending
condition for this phase. We need to determine the maximum theoretical flow that can be
provided to the breach site by the upstream and downstream canals. This is
accomplished by iteratively solving a system of three equations applying to critical flow
(Clemmens et al. 2001):
c
c
T
Q gA
3
= (5)
c
c
c T
y H A
1 2 = − (6)
2
1
2
1 1 2gA
H = h + Q (7)
where:
yc = critical depth,
Ac = area of the critical section,
Tc = top width of the critical section,
h1 = normal flow depth in the canal,
H1 = total energy head in the canal at normal flow, and
A1 = area of the canal at normal depth.
USCID -- April 15-19, 2013 -- Scottsdale, AZ
Tools for Estimating Canal-Breach Flood Hydrographs 289
For the design normal-depth flow condition of the canal, the flow depth h1 is known and
a value of H1 can be computed using Eq. 7. Next, assume a starting value for critical
depth, yc, such as yc=0.7H1. For this critical depth, the cross-sectional area, Ac, and top
width, Tc, of the canal may be computed. The critical discharge can then be computed
from Eq. 5 and a refined estimate of yc computed with Eq. 6. H1 should be kept constant,
so the iteration between Eqs. 7 and 8 is continued until convergence is obtained. The
maximum theoretical breach outflow, Qc,max, will be two times the critical discharge
computed with Eq. 5, assuming that both canals have the same cross section. This flow
must pass through the breach opening in the canal embankment, and we will assume
again that it does so in a critical-flow condition. The critical flow depth through the
rectangular breach opening will be estimated as (2/3)yn, where yn is the normal depth of
flow in the canal. (This is a crude estimation that ignores any head loss that occurs in the
canal as flow approaches the breach). For a rectangular channel, the critical flow depth is
yc=(q2/g)1/3, so the unit discharge at the end of breach widening is q=([2yn/3]3g)1/2 and the
final width of the breach is
y g
Q
b
n
c
3
,max
max
3
2
=
(8)
The breach widening rate is estimated using a relation developed by Hunt et al. (2005)
and confirmed in the physical model tests discussed previously.
2 [0.7 ( 1/ 3 /1.49)2 ]
d w c c k g y n
dt
db = γ −τ (9)
where db/dt is the change in breach width per unit time, the constant 1.49 comes from the
Manning equation in English units, and Manning’s n is taken to be 0.020 in the breach
opening. With the final breach width and widening rates known, the time required for
breach widening is
2 [0.7 ( 1/ 3 /1.49)2 ]
max
f
d w c c k g y n
t b
γ −τ
= (10)
The critical shear stress, τc, may be assumed to be zero to obtain a conservatively short
estimate of the breach widening time. Once the breach widening time is estimated, it is
converted to a dimensionless quantity, t*f=tf/tref, with tref being a reference time based on
the hydraulic depth of the canal, D, and the wave celerity, c
D g
Dg
D
c
t D / ref = = = (11)
In this equation the hydraulic depth, D, is defined to be the canal flow area divided by the
wetted top width, and g is the acceleration due to gravity.
USCID -- April 15-19, 2013 -- Scottsdale, AZ
290 Seventh International Conference on Irrigation and Drainage
Numerical modeling of hypothetical canal failures (Wahl and Lentz 2011) was used to
develop relations for predicting the dimensionless peak outflow, Q*peak=Qpeak/Qc,max.
Figure 4 shows the dimensionless peak outflow versus the dimensionless breach
development time. Data points at or just below the upper envelope curve come from
simulations in which the hypothetical breach site is a long distance upstream from the
next downstream check structure along the canal, so there is a significant volume of
water in the downstream canal that can drain back upstream to add to the breach outflow.
Points lying well below the envelope curve are for simulations in which the breach site
was closer to the downstream end of the canal reach. Figure 5 shows the percentage of
the envelope value that was actually developed as a function of the dimensionless
distance from the breach site to the downstream end of the reach. Note that the curve
shown in Figure 5 is modified from that shown in Wahl and Lentz (2011) so that the
curve passes through 50% at a dimensionless distance of 1. Thus, if the breach is located
very near the downstream end of the reach, then the downstream channel is short and
contributes almost nothing to the peak outflow, so the maximum possible outflow is 50%
of the value obtained from the envelope curve. Note also that in the numerical
simulations the distance from the breach to the upstream end of the canal reach had much
less effect on the peak outflow than did the downstream distance. Combining the two
relations shown on these figures produces one equation for estimating the peak outflow:
( ) ( )
= = − 1/ 4
ds
1/ 6
f
peak ,max peak ,max *
1 0.5
*
( * ) 1.9
t L
Q Q Q Qc c (12)
where t*f is the dimensionless breach development time defined earlier and L*ds is the
downstream canal reach length nondimensionalized by the hydraulic radius, Lds/Rh. The
value of L*ds is never allowed to be less than 1.
Figure 4. Dimensionless peak outflow from hypothetical canal breaches as a function of
dimensionless breach development time.
0.10
1.00
10 100 1000 10000 100000 Q*peak
t*f
Q*peak
Envelope curve
1.9(t*f)-1/6
USCID -- April 15-19, 2013 -- Scottsdale, AZ
Tools for Estimating Canal-Breach Flood Hydrographs 291
Figure 5. Effect of downstream canal reach length on peak breach outflow. Lds is the
length of the downstream canal and Rh is the hydraulic radius.
The peak discharge is assumed to occur at the end of the breach widening phase. The
other parameter of significant interest is the time required for the breach outflow to
recede back toward the normal canal flow rate. (Since we assume that the canal is not
shut down during a hypothetical “fast” breach, the canal continues to supply water from
upstream at the normal rate.) To describe the recession curve, the duration for the flow to
drop back to a flow rate of Qnormal+0.5(Qpeak-Qnormal) can be estimated with Eq. 13 (Wahl
and Lentz 2011), which defines the curve shown in Figure 6.
0.66 f
f
recession *
123 t
t
t =
(13)
Figure 6. Hydrograph recession time as a function of breach development time.
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1 10 100 1000 10000 100000
Q*peak/Q*envelope
Dimensionless downstream canal length, L*ds=Lds/Rh
1-0.5(L*ds)-0.25
t*recession = 123(t*f)-0.66
0.01
0.1
1
10
100
10 100 1,000 10,000 100,000 1,000,000 t*recession = trecession/tf
t*f = tf/tref
USCID -- April 15-19, 2013 -- Scottsdale, AZ
292 Seventh International Conference on Irrigation and Drainage
SPREADSHEET MODEL
The set of equations described above has been programmed into a spreadsheet model that
allows a user to describe the canal properties, the embankment dimensions, and the
embankment materials, and then estimate breach initiation time, breach development
time, and the breach outflow hydrograph. The next step in the development of this tool is
to validate it against actual canal failures. This will require case studies with the
necessary input data and good estimates of the actual breach outflow hydrograph.
EXAMPLE APPLICATION
To illustrate the use of the procedures described in this paper, consider a hypothetical
example as follows:
• Earthen canal with design discharge of 800 ft3/s;
• Canal cross section in the reach of interest is trapezoidal with 15 ft base width, 2:1
(H:V) side slopes, bed slope = 0.000379 (2 ft/mile), and Manning’s n = 0.028.
Normal depth of flow for the design discharge is 8.32 ft.
• The canal reach being considered is a 3-mi-long fill section, with gated check
structures at each end of the reach. The check structures are assumed to remain at
their normal operating positions during a breach event (worst-case, very rapid
breach scenario).
• The canal embankments on both sides of the canal are 15 ft tall from land-side toe
to crest, and the freeboard between the crest and the normal operating water
surface is 2 ft. The crest width is 16 ft, and the external embankment slope is 2:1.
• The embankment is constructed from a silty sand (SM) with 4% clay. The
embankment was constructed in about 1910 and is believed to have been
compacted by animal traffic (low compaction effort) at a water content that was
equal to or wetter than optimum.
• Locations of greatest concern are near the downstream end of the reach and about
1 mile upstream from the downstream check structure. Several homes are located
near the toe of the embankment at each of these locations.
The canal was operated for two years at a reduced discharge of 500 ft3/s, and the flow
depth during this time was only 6.61 ft. When the canal is returned to service this year at
the original design flow rate, a potential failure mode is piping through muskrat burrows
located at the water line corresponding to the previous years’ operations (3.71 ft below
the embankment crest). We will assume that a muskrat burrow has a starting diameter of
2 inches and passes straight through the embankment.
Before considering specific breach locations and material parameters, we can use Eqs. 5-
7 to compute the maximum theoretical peak outflow, which will be the reference
discharge for any breach scenario. The maximum critical-flow discharge in one canal
USCID -- April 15-19, 2013 -- Scottsdale, AZ
Tools for Estimating Canal-Breach Flood Hydrographs 293
reach is 2054 ft3/s, before the canal drains significantly, and thus, the maximum
theoretical breach outflow is Qc,max = 4108 ft3/s.
Soil erodibility parameters are estimated by referring to Table 1. For the compaction
conditions described, the table suggests kd = 200 cm3/(N-s) = 113.1 ft/hr/psf. We will
assume that the critical shear stress for this material is 0 psf. Applying the equations
describing the model for headcut advance caused by flow through the pipe (muskrat
hole), the initial flow through the pipe is 32 GPM, and the time needed for breach
initiation is 34 minutes. For comparison, a scenario in which the canal is misoperated so
that the banks are overtopped by 3 inches yields a breach initiation time of 33 min. If we
return to the piping scenario and the initial pipe is raised 1 ft higher in the embankment,
the breach initiation time increases to 38 min; if lowered by 1 ft the time reduces to
33 min, so the result is relatively insensitive to the initial pipe elevation.
The breach widening phase of the process is analyzed next. The breach widening rate is
estimated to be 181 ft/hr, using Eq. 9. The maximum breach width needed to release the
theoretical peak outflow previously calculated is only 55 ft, so the time needed for breach
widening is only 18.4 min.
To predict the peak outflow from the breach, we must select a location for the breach.
We consider two possible locations, one at the end of the reach (Lds=5 ft; L*ds=1), and the
second located 1 mile upstream from the end of the reach (L*ds≈1000). Applying Eqs.
11-12 we obtain a peak outflow of 1050 ft3/s at the downstream site and 1910 ft3/s at the
upstream site.
To test the sensitivity of the results to the soil erodibility parameters, let us revisit Table 1
and assume that the embankments were compacted dry of optimum. This changes the
estimated value of kd to 800 cm3/(N-s), or 452 ft/hr/psf. Assuming again that the piping
failure initiates at 3.71 ft below the embankment crest, the breach initiation time is now
reduced to 8.5 min and the breach widening time is only 4.6 min. The peak outflow for a
breach at the downstream end of the reach is increased to 1320 ft3/s and the peak outflow
for a breach 1 mile upstream is 2410 ft3/s. Figure 7 shows a predicted breach hydrograph
for this latter case. The figure includes a plot of the estimated product of flow depth and
velocity (DV) at the breach opening. This parameter can be useful for assessing the
lethality of the flood and its potential to cause property damage, although if the flood is
able to spread rapidly downstream from the breach, the DV values will drop accordingly
and the potential for damage will diminish. A one or two-dimensional flood routing
simulation may be needed to predict inundation depths and flooding severity at a distance
from the breach site.
Table 2 summarizes results for the scenarios discussed above, and one other involving an
assumption that the embankment was constructed with standard compaction effort near
optimum water content.
USCID -- April 15-19, 2013 -- Scottsdale, AZ
294 Seventh International Conference on Irrigation and Drainage
Figure 7. Predicted breach outflow hydrograph for the hypothetical example.
Table 2. Hypothetical canal breach hydrograph predictions.
Embankment
compaction Failure initiation
kd
(ft/hr/psf)
Breach
initiation
time
(min)
Widening time to
reach peak
outflow
(min)
Peak outflow if
breach at
downstream end
(ft3/s)
Peak outflow if
breach is 1 mile
upstream
(ft3/s)
Low effort,
optimum
water
content
Headcut advance due to flow
through 2” animal burrow
pipe 3.71 ft below
embankment crest
113.1 34 18.4 1050 1910
Overtopping by 3” 113.1 33
Low effort,
dry Headcut advance due to piping 452 8.5 4.6 1320 2410
Standard
effort,
optimum
water
content
Headcut advance due to piping 56.6 68 37 930 1700
CONCLUSIONS
Prediction of canal breach outflow hydrographs requires modeling of both breach
development processes and the transient response of the canal. Physical model testing
and analytical work has produced methods for estimating breach initiation time, breach
development time and breach width. Numerical modeling of canal and breach dynamics
has produced relations for predicting breach outflow hydrograph characteristics as a
function of breach development time and breach location within a canal pool relative to
nearby check structures that regulate the canal flow. These components have been
assembled to create an integrated mathematical model that can be used to make appraisal-
0
10
20
30
40
50
60
70
80
0
500
1000
1500
2000
2500
3000
0 0.2 0.4 0.6 0.8 1
DV at breach opening, ft2/s
Breach Outflow, cfs
Time, hours
Estimated breach hydrograph, following completion of breach initiation
Discharge, cfs
DV, ft^2/s
USCID -- April 15-19, 2013 -- Scottsdale, AZ
Tools for Estimating Canal-Breach Flood Hydrographs 295
level estimates of canal breach outflow hydrographs as a function of canal hydraulic
properties and embankment material properties. The method has not yet been tested
against real-world canal failures.
REFERENCES
ASTM, 2003. Annual Book of ASTM Standards, Section 4: Construction, Vol. 04.08.
Philadelphia, Pa.: ASTM.
Dun, R.W.A., 2007. An improved understanding of canal hydraulics and flood risk from
breach failures. Water and Environment Journal. 21(1): 9-18.
Engemoen, W. 2012. Reclamation’s processes for evaluating potential internal erosion
concerns in embankment dams. 6th International Conference on Scour and Erosion,
August 27-31, 2012, Paris, France.
Hanson, G.J., and K.R. Cook, 2004. Apparatus, test procedures, and analytical methods
to measure soil erodibility in situ. Applied Engineering in Agriculture, 20(4):455-462.
Hanson, G.J., D.M. Temple, S.L. Hunt, and R.D. Tejral, 2011. Development and
characterization of soil material parameters for embankment breach. Applied
Engineering in Agriculture, 27(4):587-595.
Temple, D.M., G.J. Hanson, M.L. Neilsen, and K.R. Cook. 2005. Simplified breach
analysis for homogeneous embankments: Part 1. Background and model components. In
Technologies to Enhance Dam Safety and the Environment, Proceedings of the Annual
Meeting of the United States Society on Dams, Denver, Colorado, 151‐161.
Von Thun, J.L., 1996. Understanding piping and seepage failures - the no. 1 dam safety
problem in the west. ASDSO West Regional Conference, April 14-16, 1996, Lake
Tahoe, NV.
Wahl, T.L. and D.J. Lentz, 2011. Physical hydraulic modeling of canal breaches,
Hydraulic Laboratory Report HL-2011-09, U.S. Dept. of the Interior, Bureau of
Reclamation, Denver, CO.
Wahl, T.L., 2012. Numerical modeling to predict canal breach outflow hydrographs.
2012 World Environmental and Water Resources Congress, Environmental and Water
Resources Institute of the American Society of Civil Engineers, Albuquerque, NM, May
20-24, 2012.
USCID -- April 15-19, 2013 -- Scottsdale, AZ