Professional Documents
Culture Documents
ABSTRACT
Both man-made earth dams and natural landslide dams pose enormous risks to the public because of the potentially
catastrophic floods generated by breaching of such dams. In order to manage the dam breaching risks, it is
necessary to evaluate the structural safety of dams, possible breaching modes and breach size, outflow hydrograph,
flood routing in the downstream river, and potential losses of lives and properties due to dam breach floods. It is
also necessary to prepare emergency plans and evaluate actions to betaken to mitigate the risks. In this paper,
results of recent research on three key issues in dam breaching risk assessment (i.e., failure modes and processes,
breaching parameters, and flood risk assessment) are presented, and a case study on Tangjishan Landslide Dam is
presented to illustrate the risk assessment procedure.
the breach initiation phase involves stages 1-4 while the often removed slowly by seeping water. Then, a pipe is
breach development phase involves stages 5-7. If a granular initially formed in the downstream slope after a significant
embankment contains a vertical cohesive core, the core will amount of material has been carried away. Once the pipe is
erode in a manner similar to that for a cohesive formed, the erosion process becomes rapid and the pipe
embankment. If the core is sloped with structural support advances upstream. For granular embankments, the
from the downstream shell, the core will fail structurally materials above the pipe collapse when the pipe is sufficiently
after the downstream shell is eroded away. large, which is followed by a similar process as in the case of
Seepage Erosion/Piping overtopping. Fig. 3 shows the breaching process of a granular
Another common cause of failure of embankment dams is embankment by seepage erosion/piping, in which the breach
seepage erosion/piping. McCook (2004) emphasized the initiation phase involves stages 1-4 while the breach
distinction between seepage erosion and piping to help development phase involves stages 5-6.
differentiate between the phenomena of flow through cracks For cohesive embankments, the portion of the materials
or structural contacts versus flow through granular media. above the pipe may keep stable, and hence the initial pipe in
Piping is defined as the progressive removal of soil particles the downstream slope continues to advance till the upstream
from a soil mass by percolating water, leading to the slope face. A penetrated pipe is finally formed through the
development of channels. Piping often occurs in granular embankment. Fig. 4 shows the whole breaching process of a
embankments. Cohesive embankments offer a greater cohesive embankment by seepage erosion/piping, in which
erosion resistance than granular fills do, and hence are less the breach initiation phase involves stages 1-3 while the
likely to suffer from piping. Seepage erosion occurs when the breach development phase involves stages 4-5.
water flowing through a crack or defect erodes the soil from
the walls of the crack or defect. Similar to overtopping, the 3
2
1
2
1
(a1)
(a1) 1-3
1-2
(b1)
(b1)
Collapse
坍塌
4
4 3 5
(a2)
(a2)
3-4
4-5
(b2)
(b2)
5 Fig. 4: Breaching Process of a Cohesive Embankment by
6
Seepage Erosion/piping: (a) Side View, (b) Downstream View
(a3) (Breach Initiation Phase: Stages 1-3 & Breach Development
Phase: Stages 4-5)
5-6 Failure Modes
(b3) Without channelized spillways or other protected outlets,
Fig. 3: Breaching Process of a Granular Embankment by landslide dams commonly fail by overtopping, while earth
Seepage Erosion/Piping: (a) Side View, (b) Downstream and rockfill dams fail by both overtopping and piping.
Materials above the Pipe Becomes Unstable and Collapses, Fig. 5 presents and compares the failure modes of landslide
View (Breach Initiation Phase: Stages 1-4 & Breach
Development Phase: Stages 5-6) dams and man-made earth and rockfill dams. Out of the
breaching characteristics by seepage erosion/piping are also 144 landslide dam failure cases with known failure modes,
different for granular and cohesive embankments. At the 91% (131 cases) failed by overtopping, 8% (12 cases) by
original stage of seepage erosion/piping, soil particles are piping, and 1% (1 case) by slope failure.
84 L.M. Zhang, M. Peng and Y. Xu
8% lasted ≤ 1 hour
90 0
37% 0 50 100 150 200 250 300 350 400
60 Age of landslide dams at time of failure (day)
two groups: geometric and hydrographic parameters. A dam breaching parameters is also compared. The dam erodibility
breach often has a trapezoidal shape, with the geometric is found to be the most important factor, influencing all
parameters of breach depth Hb, breach top width Wt, average the five breaching parameters. The reservoir shape
breach width Wave, and breach bottom width Wb as shown coefficient and the failure mode also play important roles
in Fig. 7. Any combination of three of these geometric in the prediction models.
parameters determines the breach shape and size. Table 2: Summary of Empirical Relations for Estimating the
Hydrographic parameters include peak outflow rate Qp Breaching Parameters of Earth Dams
and failure time Tf. After the onset of breaching, the outflow (a) Full-variable equations
through the breach increases till it reaches a peak and then
decreases till there is no longer any water in the reservoir Parameter Equation and Coefficients1
0.199 −1.274
or the breaching process ceases to develop. Failure time Tf Qp H Vw1/ 3
is defined as the period from the inception to the completion = 0.175 d e DT + FM + ER
gVw5/ 3 Hr Hw
of the breaching process (Singh and Snorrason 1984). Peak Failure mode
According to Fell et al. (2003), in most cases it has not outflow rate Dam type DT Erodibility ER
FM
been possible to identify the time of initiation of erosion, Qp 2
CW –0.503 Ove.3 –0.705 High4 –0.007
and the first signs of erosion tend to be in the progression CF2 –0.591 3
Pip. –1.039 Med.4 –0.375
phase. Therefore, from a practical standpoint, failure time HF2 –0.649 – – Low4 –1.362
is often recorded at the start of the breach development Hb H
phase. Failure time Tf in this study is also regarded as the = 0.453 − 0.025 d + DT + FM + ER
Hd Hr
breach development time. The time for initiation has been Failure mode
Breach Dam type DT Erodibility ER
recorded where it is possible, e.g., from increased seepage depth Hb FM
2
flows. CW 0.145 Ove.3 0.218 High4 0.254
CF2 0.176 Pip.3 –0.239 Med.4 0.168
Hd = dam height HF2 0.132 – – Low4 0.031
0.092 0.508
Hb = breach depth
Wt H Vw1/3
W t = breach top width = 1.062 d e DT + FM + ER
Hb Hr Hw
Wb = breach bottom width
Breach top Failure mode
Dam type DT Erodibility ER
width Wt FM
CW2 0.061 Ove.3 0.299 High4 0.411
CF2 0.088 Pip.3 –0.239 Med.4 –0.062
Wt
HF2 –0.089 – – Low4 –0.289
0.133 0.652
Hb
Wave H Vw1/ 3
= 0.787 d eDT + FM + ER
Hd Wb Hb Hr Hw
Average Failure mode
breach Dam type DT Erodibility ER
FM
Fig. 7: Geometric Parameters of a Dam Breach width Wave CW 2
–0.041 Ove.3 0.149 High4 0.291
CF2 0.026 3
Pip. –0.389 Med.4 –0.140
Breaching Parameters of Earth and Rockfill Dams HF2 –0.226 – – Low4 –0.391
Two sorts of prediction models have been developed: 0.707 1.228
Tb H Vw1/ 3
empirical formulas based on case study data and physically- = 0.304 d eDT + FM + ER
Tr Hr Hw
based prediction models. Xu and Zhang (2009) summarized
Breaching Failure mode
the development of these two sorts of models. Based on the Dam type DT
FM
Erodibility ER
time Tb
information from a dataset of 182 earth and rockfill dam CW 2
–0.327 Ove.3 –0.579 High4 –1.205
failure cases, Xu and Zhang (2009) proposed a CF2 –0.674 Pip.3 –0.611 Med.4 –0.564
multiplicative regression model to develop empirical HF2 –0.189 – – Low4 0.579
relationships between five breaching parameters (breach Note:
1
depth, breach top width, average breach width, peak outflow Hd = dam height, Hr = 15 m, Vw = released water volume, Hw =
rate, and failure time) and five selected dam and reservoir drop of water level, Tr = 1 hour, g = 9.8 m/s2.
2
control variables (dam height, reservoir shape coefficient, CW = dam with corewalls, CF = concrete faced dam,
dam type, failure mode, and dam erodibility). These HF = homogeneous/zoned-fill dam.
3
empirical relations are summarized in Tables 2(a) and 2(b). Ove. = overtopping, Pip. = piping.
4
The relative importance of each control variable to the dam High, Med., and Low = high, medium, and low erodibility,
respectively.
86 L.M. Zhang, M. Peng and Y. Xu
(b) Simplified equations dams is less sensitive to dam height, since a larger volume
Parameter Equation and Coefficients the models for landslide dams (Table 3) are applied to
Qp V 1/3
−1.276
Table 3: Summary of Empirical Relations for Estimating the
= 0.133 w e FM + ER Breaching Parameters of Landslide Dams
gVw5 / 3 Hw
Peak outflow (a) Full-variable equations
Failure mode FM Erodibility ER
rate Qp Para- Erodibility
Overtopping –0.788 High –0.089 Equation1
meter Coefficient, α
Piping –1.232 Medium –0.498
Qp H d -1.417 H d -0.265 Vd1/3 -0.471 High 1.276
– – Low –1.433 1/ 2 5/ 2
=( ) ( ) ( )
g Hd Hc Wd Hd
Hb H Qp Med. –0.336
= ER − 0.025 d V 1/ 3
Hd Hr ( l )1.569 e a
Hd Low –1.532
Breach depth Failure mode FM Erodibility ER
Hb – Hb H H V 1/3 High –0.316
Overtopping High 1.072 = ( d ) 0.882 ( d ) -0.041 ( d ) -0.099
Piping – Medium 0.986 Hc Hc Wd Hd
Hb Med. –0.520
– – Low 0.858 Vl1/ 3 0.139 a
( ) e
0.558 Hd Low -2
Wt V 1/ 3 FM + ER
= 0.996 w e
Hb Hw Wt H H V 1/ 3 High 1.683
= ( d )0.752 ( d )0.315 ( d )-0.243
Breach top Failure mode FM Erodibility ER Hc Hc Wd Hd
width Wt Wt Med. 1.201
Overtopping 0.258 High 0.377 Vl1/3 0.682 a
( ) e
Piping –0.262 Medium –0.092 Hd Low -2
– – Low –0.288
0.739 Wb H H V 1/ 3 High 0.775
V 1/ 3 = 0.004( d ) + 0.050( d ) − 0.044 ( d )
Wave FM + ER Hd Hc Wd Hd
= 5.543 w e Wb Med. 0.532
Hb Hw Vl1/ 3
Average + 0.088( )+ a
breach width Failure mode FM Erodibility ER Hd Low -2
Wave Overtopping –1.207 High –0.613
Tb H H V 1/ 3 High –0.635
Piping –1.747 Medium –1.073 = ( d )0.262 ( d )-0.024 ( d )-0.103
Tr Hc Wd Hd
– – Low –1.268 Tb Med. –0.518
0.654 1.246 Vl1/ 3 0.705 a
Tb H Vw1/ 3 ( ) e
= ER d Hd Low -2
Tr Hr Hw
Breaching Failure mode FM Erodibility ER Note:
time Tb –
1
Hd = dam height, Wd = dam width, Hc = 1 m, Vd = dam volume,
Overtopping High 0.038
Piping – Medium 0.066 Tr = 1 hour, g = 9.8 m/s2.
2
– – Low 0.205 No records are available for the low erodibility coefficient
cases.
Breaching Parameters of Landslide Dams (b) Simplified equations
Breaching of landslide dams is sometimes studied by using Erodibility
Parameter Equation
empirical equations for man-made earthen dams. This is Coefficient, α
not reasonable because of the aforementioned distinct Qp H d -1.371 Vl1/ 3 1.536 a
High 1.236
differences between these two types of dams. Based on Qp 1/ 2 5/ 2
=( ) ( ) e Med. –
g Hd Hc Hd 0.380
records of 52 landslide dam cases with detailed breaching Low –
High 1.615
–
information in the HKUST landslide dam database,
Hb H V 1/3 0.500
–
empirical models for estimating the breaching parameters Hb = ( d ) 0.923 ( l ) 0.118 e a Med.
Hc Hc Hd 0.673
of landslide dams are developed following the similar Low –
procedure in Xu and Zhang (2009). The empirical relations High 0.588
Wt H V 1/ 3
are summarized in Table 3. Wt = ( d )0.911 ( l )0.271 ea Med. 0.148
Hc Hc Hd
Comparison of Breaching Parameters of Earth Dams Low –
and Landslide Dams High 0.624
Wb H V 1/ 3
Wb = 0.003( d ) + 0.070( l ) + a Med. 0.344
Figs. 8 and 9 compare the values of peak flow rate and Hd Hc Hd
Low –
breach duration for man-made earth and rockfill dams and –
landslide dams, respectively. The peak outflow rate, Qp, High
Tb H V 1/ 3 0.805
increases with dam height or lake volume outflow rate for Tb = ( d )0.293 ( l )0.723 ea Med.
–
Tr Hc Hd 0.674
both types of dams as shown in Fig. 8. Qp from landslide
Low –
Assessing Risks of Breaching of Earth Dams and Natural Landslide Dams 87
predict the peak outflow rate from landslide dams, the mean breaching duration (by 64% using Xu and Zhang’s models).
bias factors (i.e. the mean of the ratios of measured flow Separate models should be used.
rate to predicted flow rate) of the models are about 1.01,
20 Man-made arth and rockfill dams
which is close to unity. However, if the models for earth Landslide dams
Landslide dams induced by the Wenchuan earthquake
and rockfill dams (Table 2) are used to predict the peak 16
will be 0.34, which means that the equations for earth and 8
Man-made earth and rockfill dams Landslide dams
100000
(a)
(b)
1000000
Man-made earth and rockfill dams
Man-made earth and
Fig. 9: Comparison of Breaching Durations of Landslide Dams
Landslide dams
Landslide dams induced by the Wenchuan earthquake rockfill dams and Earth and Rockfill Dams: (a) Influence of Dam Height; (b)
Peak discharge (m 3 /s)
100000
Influence of Lake Volume
10000
1000
Landslide dams 5. RISK ASSESMENT AND MANAGEMENT
100
The procedure of general risk assessment has been described
by many researchers. In the case of dam failure risk
10
0.1 1 10 100 1000 10000
assessment, the following steps may be followed.
Lake volume (million m3) Information Collection and Observation
(b) When a dam is reported to be unsafe or a landslide dam
Fig. 8: Comparison of Peak Outflow Rates from Failure of has just formed, it is necessary to set up an observation
Landslide Dams and Earth and Rockfill Dams: (a) Influence of plan to closely monitor the safety of the dam, including
Dam Height; (b) Influence of Lake Volume but not limited to reservoir water level, inflow into the
The breaching duration of a landslide dam is generally reservoir, leakage from the dam, movements of the dam,
longer than that of a man-made dam with the same dam and the weather conditions.
height or lake volume, as shown in Fig. 9. The main reason At the same time, all necessary information required
is that the volume and width of the landslide dam are for evaluating the structural safety of the dam and possible
typically larger and the downstream slope is typically consequences of failure should be collected. Such
gentler, which reduces the erosion velocity during information includes geologic profiles and the material
breaching. The breaching duration is statistically only zoning within the dam, geotechnical properties of the dam
weakly related to the dam height as shown in Fig. 9(a). A materials, the river network downstream the dam,
higher dam dose not definitely have a longer breaching population distribution and statistics, land use, business
process, because the breach often does not develop to the operations etc. If such information is not available then
bottom of the dam. On average, the actual breaching necessary exploration or surveys are required.
duration of landslide dams is 2.76 times of that predicted Evaluation of Structural Safety of the Dam and
using the equations for earth and rockfill dams. Possible Failure Modes
In summary, the direct application of the prediction The structural safety of the dam giving expected loading
models for man-made dams to landslide dams would lead conditions should now be assessed. Reliability theory may
to overestimation of the peak outflow rate (by about 200% be a valuable tool in this step considering the high level of
using Xu and Zhang’s models) and underestimation of the uncertainties in the dam materials and loadings. For many
88 L.M. Zhang, M. Peng and Y. Xu
landslide dams, it is certain that these dams will be All these measures are costly. Optimal decisions on
overtopped at a later time. The questions are when will risk mitigation measures can be made through a cost-benefit
overtopping occur and whether the dam will fail by piping analysis.
or not before the dam is overtopped. Risk mitigation
6. EXAMPLE: BREACHIING OF TANGJIASHAN
measures, such as the construction of division channels,
LANDSLIDE DAM, SICHUAN, CHINA
are usually very effective at an early stage.
Tangjiashan Landslide Dam, with a dam height of 82 -
Determination of Dam Breaching Parameters 124 m, a dam width of 612 m, a dam volume of 20.4 million
In this stage, the possible breach size and the corresponding m3 and a reservoir capacity of 316 million m3, is the largest
dam breaching hydrograph should be estimated. It is landslide dam induced by the Ms 8.0 Wenchuan earthquake
understandable that any estimates are associated with a high in May 2008. The dam mainly consists of three layers of
level of uncertainty. soil and rock materials: gravelly soil 5 - 15 m in thickness,
Flood Routing and Inundation Analysis strongly weathered cataclasite 10 - 15 m in thickness, and
How the dam breaching flood travels along the down-stream weakly weathered cataclasite as shown in Fig. 10. The
river should now be studied so that inundation zones can gravelly soil and strongly weathered cataclasite are of
be identified and the population to be affected by the flood medium erodibility.
can be warned ahead of the arrival of the flood. Analysis
tools such as BREACH (Fread, 1988) and HEC-RAS
(Institute for Water Resources, 2008) are useful in
preliminary flood routing analysis. Spillway
750
am, 2190 m3/s at 10:00 am, and 6500 m3/s (the peak outflow part of the original crest of the natural dam. In the second
rate) at 11:30 am. The water level decreased to 714.1 m at case, predictions of the breaching parameters are made
14:00 pm on 11 June from the maximum level of 743.1 m using the models for landslide dams shown in Table 3. In
at 1:30 am on 10 June. The corresponding lake volume the third case, predictions are made using the equations
decreased from 247 million m3 to 86 million m3. The final for earth and rockfill dams in Table 2. The mean values of
breach has a depth of 42 m, a top width of 145- 235 m and the breaching parameters for the three cases are presented
a bottom width of 80-100 m. in Table 4. As expected, the equations for earth dams grossly
overestimate the breach size and peak flow rate.
For simulating flood routing, 5303 cross sections are
used to represent the river downstream the dam, 26 of which
having detailed elevation and distance data (Liu 2008).
Manning’s roughness factor values for both the river
channel and the floodplains are taken from Chow (1959).
Figs. 12 and 13 show the flow processes at three locations
in the actual case and the medium edodibility case (the
second case). The peak flow rate in the second case is much
larger than that in the actual case and the breaching time
is slightly shorter than that in the actual case. One of the
reasons is that the lake elevation in the second case is
752 m, while it is 740 m in the actual case.
The flood travel time becomes shorter as the peak flow
rate increases. Thus, earlier warning is needed for larger
peak flow rate cases. The flood in Mianyang City takes
longer to fade due to the lower flow velocity there.
7000
Dam Site
Fig. 11: Flood Routing in the River Downstream Tangjisshan 6000 Tongkou Town
Mianyang City
Landslide Dam /s) 5000
3
Flo
4000
HEC-RAS 4.0 is used to simulate the dam breaching w
rat
3000
progression and flood routing in the downstream river. Dam e
(m
2000
breaks and dike breaks can be simulated using this program
1000
given such breaching parameters as breach size, breaching
0
duration, and breaching progression. 6/10 6/10 6/10 6/10 6/10 6/10 6/10 6/10 6/10 6/10 6/11 6/11 6/11
To simulate breaching progression, three cases are 5:00 7:00 9:00 11:00 13:00 15:00 17:00 19:00 21:00 23:00
Time
1:00 3:00 5:00
9 June) after the water level reached elevation 740 m (the Fig. 13: Flow Rates in the Medium Erodibility Case
weir elevation of the discharge channel). While in the Dekay and McClelland (1993) suggested a model to
second and third cases, no discharge channel is considered, calculate the loss of life caused by flooding based on a
and it is assumed that the dam breaches 47 hours after the regression analysis of casualty data in 29 flooding cases.
water level reaches elevation 752 m, which is the lowest Three parameters, namely population at risk (PAR),
90 L.M. Zhang, M. Peng and Y. Xu
warning time (WT) and flood force, are used in the model: second largest city in Sichuan Province, with a peak flow
LOL = PAR /{1 + 13.277( PAR ) 0.440
EXP[0.759(WT ) − of 7110 m3/s at 17:18 of 10 June. There was not much
(1) damage to property in Mianyang City either, as the design
3.790( Force) + 2.223(WT )( Force)]}
flood standards are 10800 m3/s for 20-year returned floods,
where “Force” = the flood severity. Equation (1) may 13000 m3/s for 50-year returned floods, and 14600 m3/s
be simplified by substituting zero or 1.0 for Force. If the for 100-year returned floods (Liu 2008).
PAR is located in a canyon, where the floodwater is likely
One may ask what would have happened if no
to be very deep and swift, “Force” equals 1.0. On the other
engineering measures had been taken to lower the failure
hand, if the PAR is located on a plain, where the floodwater
is likely to be shallow and slow, Force equals zero. risk (e.g. excavation of the division channel) and the
population at risk had not been evacuated. To answer this
The value of PAR is not equal to the total population.
Only those who are exposed to a certain level of flood are question, the losses of life in three cases of assumed warning
considered as population at risk. In order to obtain PAR, a times are evaluated: warning at the start of breaching,
physically-based human-water interaction model (Lind et warning when the peak flow occurs at the dam site, and no
al. 2004) for human beings to stay stable in a flow is applied. warning at all. The results of loss of life in the three cases
In the model human beings may loss stability when the are shown in Tables 5 and 6. From the results in the two
product of water depth, h, and velocity, v, reaches a certain tables, the following observations can be made:
limit. Based on experimental data, Lind et al. (2004) (1) The PAR is very sensitive to the peak flow rate. It
obtained the critical conditions of hv to maintain human changes from 17312 people in the real case to
stability in water flow: (hv) cr ranges from 0.65 to 44809 people in the second case, and 232607
2.13 m2/s with a sample mean of 1.22 m2/s and a COV of people in the third case, with the peak flow rates
0.27. In the present study, the critical condition of human of 6526 m 3/s, 12316 m 3/s, and 36547 m 3/s,
stability is selected as 1.22 m2/s. Based on the results of respectively.
the inundation analysis, the PAR values in the real case (2) The LOL is very sensitive to the warning time. If
and the two assumed erodibility cases (see Table 4) are the warning time is sufficient, very limited LOL
shown in Tables 5 and 6. might occur, whereas the LOL becomes large if
Table 5: Loss of Lives in the Real Case with Different no warning time is available. In the real case, the
Warning Times Beichuan Town was severely flooded, but no
Warning at Start Warning When Peak No casualty was caused because the PAR was
Population PAR of Breach Flow Occurs at Dam Warning
Location
(Person) (Person) evacuated ahead of time.
WT LOL WT LOL WT LOL
(h) (person) (h) (person) (h) (person) (3) If there is no warning time, the fatality rate (LOL/
Beichuan 30000 14926 7.83 0 0.17 428 0 691
Tongkou 7300 0 8.67 0 1.00 0 0 0 PAR) could be over 5.3% even though a large
Hanzeng 10000 0 8.83 0 1.17 0 0 0
Qianlian 16300 903 9.33 0 1.67 1 0 129
number of people are located far away from the
Longfeng
Shima
15000
20500
1484
0
9.83
10.33
0
0
2.17
2.67
0
0
0
0
175
0
dam site and in low force (plain) areas.
Mianyang 1127000 0 10.67 0 3.00 0 0 0
Total 1226100 17312 0 430 996 7. CONCLUSIONS
Both man-made earth dams and natural landslide dams
Table 6: Loss of Lives in the Medium Erodibility Case with
pose enormous risks to the public because of the potentially
Different Warning Times
catastrophic floods generated by dam failures. This paper
Warning at Start Warning When Peak
No Warning
Population PAR of Breach Flow Occurs at Dam looks into three key issues in dam breaching risk
Location
(Person) (Person)
WT LOL WT LOL WT LOL assessment, namely failure modes and processes, breaching
(h) (Person) (h) (Person) (h) (Person)
Beichuan 30000 28619 4.83 0 0.08 792 0 1007
parameters, and flood risk assessment. The following
Tongkou
Hanzeng
7300
10000
1267
2294
5.50
5.83
0
0
0.67
1.00
24
13
0
0
159
229
conclusions can be drawn:
Qianlian
Longfeng
16300
15000
4498
4991
6.33
6.67
0
0
1.50
2.00
4
2
0
0
342
364
(1) Man-made earth and rockfill dams most
Shima 20500 3140 7.17 0 2.33 0 0 276 commonly failed by overtopping and piping,
Mianyang 1127000 0 7.50 0 2.50 0 0 0
Total 1226100 44809 0 835 2377 whereas over 90% landslide dams failed by
In the real case, no fatality was caused in the dam- overtopping. Landslide dams seldom failed by
breaching event because of early warning and mass piping (8% only).
evacuation, although the water level increased more than (2) Breaching parameters are key inputs for breaching
7 m at Beichuan Town. The flood arrived at Mianyang, the flood risk assessment. Separate equations have
Assessing Risks of Breaching of Earth Dams and Natural Landslide Dams 91
been developed for estimating the breaching DeKay, M.L. and McClelland, G.H. (1993). Predicting loss
parameters of earth dam and landslide dams. Soil of life in cases of dam failure and flash flood. Risk
erodibility is an important parameter that governs Analysis, 13(2), 193-205.
the breaching process. The two types of dams differ Fread, D.L. (1988). BREACH: An erosion model for
significantly. If the equations for man-made dams earthen dam failures. National Weather Service, Office
are used for landslide dams, the peak flow rate of Hydrology, Silver Spring, Md.
could be overestimated on average by nearly 200% ICOLD (2010). Dams and the world’s water. Int.
while the breaching duration could be Commission on Large Dams, Paris. Online at http://
underestimated by about 60%. www.i col d-ci gb. net / i m a ges/ PDF_E N/ a ms%20
(3) The population at risk and the flood travel time &%20the%20World%20Water.pdf.
are very sensitive to the peak flow rate. Early Institute for Water Resources (2008). User’s manual of
warning is shown to be an effective way of risk HEC-RAS river analysis system, Version 4.0. U.S. Army
mitigation. If the warning time is sufficient, very Corps of Engineers, Blacksburg, Mississippi.
few casualties might be resulted, whereas the loss
Lind, N., Hartford, D. and Assaf, H. (2004). Hydrodynamic
of life may become large if no warning time is
models of human instability in a flood. Journal of the
available.
American Water Resources Association, 40(1), 89-96.
(4) Each step of dam risk assessment is highly
uncertain. The reliability of any estimate should Liu, F. (2008). The simulation of one-dimensional dam
be aware of in making risk mitigation decisions. breach and emergency calculation of the Sichuan quake
lake. Bachelor thesis, Tsinghua University, Beijing. (in
ACKNOWLEDGMENTS Chinese)
The research reported in this paper was supported by the McCook, D.K. (2004). A comprehensive discussion of
Research Grants Council of the Hong Kong SAR (622207), piping and internal erosion failure mechanisms.
the National Science Foundation of China (50828901) and Association of State Dam Safety Officials (ASDSO)
the Ministry of Science and Technology (2009BAK56B05). Annual Meeting. Phoenix.
REFERENCES Peng, M., Zhang, L.M. and Huang, R.Q. (2010). Risk
analysis of Tangjiashan landslide dam. GeoFlorida
Bonnard, C. (2009). LARAM School - Introduction and
2010: Advances in Analysis, Modeling & Design, Dante
landslide vulnerability. Lecture notes in LARAM School,
O. Fratta, Anand J. Puppala, and Balasingam
Italy.
Muhunthan (editors). ASCE GSP No. 199, doi 10.1061/
Chang, D.S. and Zhang, L.M. (2010). Simulation of the 41095(365)222.
erosion process of landslide dams due to overtopping
Ru, N.H. and Niu, Y.G. (2001). Embankment dam -
considering variations in soil erodibility along depth.
incidents and safety of large dams. Water Power Press,
Natural Hazards and Earth System Sciences, 10, 933–
Beijing, (in Chinese).
946.
Singh, K.P. and Snorrason, A. (1984). Sensitivity of outflow
Chen, L. (2008). Press release - states of flood control and
peaks and flood stages to the selection of dam breach
drought relief in China. Beijing, Aug. 2008. Online:
parameters and simulation models. J. Hydrol. 68(1-4),
http://www.gov.cn/wszb/zhibo133/wzsl.htm. Accessed
295–310.
on Aug. 10, 2009).
Wahl, T.L. (1998). Prediction of embankment dam breach
Chow, V.T. (1959). Open channel hydraulics. McGraw-
parameters. Report DSO-98-004. U.S. Department of
Hill Book Company, NY.
the Interior Bureau of Reclamation, Dam Safety Office,
Costa, J.E. and Schuster, R.L. (1988). The formation and Denver.
failure of natural dams. Geological Society of America
Xu, Y. and Zhang, L.M. (2009). Breaching parameters of
Bulletin, 100, 1054–68.
earth and rockfill dams. J. Geotech. Geoenviron. Eng.,
Cui, P, Zhu, Y.Y., Han, Y.S., Chen, X.Q. and Zhuang, J.Q. ASCE, 135(12), 1957-1970.
(2009). The 12 May Wenchuan earthquake-induced
Xu, Y., Zhang, L.M. and Jia, J.S. (2009). Catastrophic
landslide lakes: distribution and preliminary risk
failure of Banqiao Dam. Proc. Case Studies: Learning
evaluation. Landslides, 6(3), 209-223.
92 L.M. Zhang, M. Peng and Y. Xu
from Dam Incidents and Failures, 24-25 March 2009 Proc. 2nd International Symposium on Geotechnical
in Los Angeles, International Dam Safety Interest Safety ands Risk (IS-Gifu 2009), Honjo et al. (eds), Gifu,
Group, Montreal, QC Canada. Japan, 11-12 June 2009, Taylor & Francis Group,
Zhang, L.M. (2009). Challenges in multi-hazard risk London, 237-244.
assessment and management: Geohazard chain in Zhang, L.M., Xu, Y. and Jia, J.S. (2009). Analysis of earth
Beichuan Town caused by Great Wenchuan Earthquake. dam failures-A database approach. Georisk, 3(3), 184-
189.