Simulation of Dam-Break Flood Wave and Inundation Mapping: A Case Study of Attabad Lake
Simulation of Dam-Break Flood Wave and Inundation Mapping: A Case Study of Attabad Lake
Simulation of Dam-Break Flood Wave and Inundation Mapping: A Case Study of Attabad Lake
Volume
Wasim Karam et al., International Journal of Emerging 9. No.
Trends 6, June 2021
in Engineering Research, 9(6), June 2021, 703 – 714
International Journal of Emerging Trends in Engineering Research
Available Online at http://www.warse.org/IJETER/static/pdf/file/ijeter15962021.pdf
https://doi.org/10.30534/ijeter/2021/15962021
ABSTRACT 1. INTRODUCTION
The break study of Landslide or embankment dams is more To establish sustainable management of the water resources
essential due to uncertainty in their composition and lack of dams are much necessary. Dams are constructed to obstruct
knowledge of their response to other natural events because the flow path of a river stream and make good use of the water
they are natural and not properly designed for any disaster. that is ponded upstream of the structure. Dams can be
The paper aims to improve different methods of hydraulic classified in many ways, Based on the material of the dam,
modeling of dam break. In this present study dam break of dams are classified into embankment dams, concrete dams,
Attabad lake is simulated using the computational fluid and masonry dams [1]. More of the same a lake is a natural
dynamics technique. The numerical model (FLOW-3D) is pool of water caused by blockage of the river or underground
developed to solve the Reynolds averaged Navier-Stoke stream.
equation fully in 3D to predict peak flow depth at different
cross-sections, peak velocity, peak discharge, time to peak Dam serves a variety of purposes such as water supply for
depth, and time to peak discharge. The standard RNG irrigation, for drinking purposes, flood protection and
turbulence model is employed to simulate turbulence and then generation of electric power, recreation and navigation, etc.
flood inundation maps and velocity vectors for flow at villages [2]. However, despite all the safety that is incurred in its
are drawn. The results show that most of the flood wave design, there always lies a risk of breaching in dams. Dams
modeled through the Hunza river channel, is contained by the may fail due to large earthquakes, insufficient spillway
flood plain of Hunza River but for some of the villages i.e. capacity, piping, seepage, and extreme storms [3]. A dam
Miaun and chalat, which are situated inside the flood plain of break is a complete or incomplete catastrophic failure of a dam
Hunza River are at higher risk of inundation but the flood that causes an uncontrolled release of water. The flood wave
arrival time estimated for these villages is 31 and 44 min travels at a very high speed in the downstream reach of the
respectively which is enough time for the evacuation of the channel. Dam Breaks has some devastating consequences
population to safer areas while for some villages like Hassan including damage to property and infrastructure, loss of life.
Abad situated adjacent to Ali Abad is at higher risk of On the other hand, it also possess some serious ecological
inundation while the estimated flood arrival time for the impacts. [4]
village is 12 min which are not enough for the evacuation of
the population hence will need some extra flood protection The most common concern about dam break failures is that
structures for flood containment. The estimate of the peak they are abrupt which causes immediate flooding downstream
velocities implies higher shear stress in the river plain, risk of of the dam which leaves very little time for evacuation of
heavy erosion, damage to agricultural lands, residencies, and residents or protection of the property. In contrast flooding
morphological changes are projected. The analysis of the dam from the rainfall or snowmelt can be predicted days or weeks
break i.e. Peak depths, Peak Velocities, flood arrival time, and in advance. That’s why Dam breaks kill thousands of people
flood inundation maps should only be used as a guide in future across the globe. Some notable dam breaks are Malpasset
risk analysis and flood management. dam-break failure in France in 1959 (killed 500 people),
Vaoint dam failure in Italy in 1963 (killed 2600 people), Teton
Key words: Attabad Lake, Computational Fluid Dynamics, dam failure in Idaho United states in 1976 (killed 11 people
Dam Break Simulation, FLOW-3D, Flood inundation and 13000 cattle). Emergency action plans have to be prepared
Mapping, Navier stokes equation, Three dimensional Model. for every dam higher than 20 m with a reservoir volume over
15Mm3 [5]. Therefore accurate estimates of the flood arrival
time, flow depths, velocities are important for emergency
planning and flood hazard assessment because of the very least
703
Wasim Karam et al., International Journal of Emerging Trends in Engineering Research, 9(6), June 2021, 703 – 714
margin that we have in contrast to the flooding from rainfall or gradually varied [8]. This means that changes in channel
snowmelt. Since the most important part of flood hazard properties, such as cross-section shape, invert level, flow
assessment is the delineation of the flood plain and the depth, and pressure distribution, are negligible at shorter
identification of high-risk areas [6]. Hence predicting the distances. Which makes it hard for these models to simulate
high-risk areas is possible by drawing the flood inundation areas of rapid acceleration or deceleration. The numerical
maps. However, to predict these quantities we need to solve solution used by two-dimensional hydraulic models is usually
the governing equations they are unsteady and contain terms of based on a shallow water equation, from which the
pressure gradient, viscosity which need a lot of boundary Saint-Venant equation is derived for unidirectional flow.
conditions making it very difficult to solve them manually for Traditionally, the mesh has been a fixed-space rectilinear grid
with the governing equations solved. But two-dimensional
such complex situations. Either, we can use scale-down
hydraulic models assume that flow is one- or two-dimensional,
models but scale-down models can describe a very limited
vertical variations in the flow properties are neglected [9].
amount of problems with very limited operating conditions.
Then comes the 3D models which can easily simulate the
Moreover, scale-down models in cases of dam-break are also downstream, lateral, and vertical components of flow.
not very efficient because replicating the complex bed
topography in the laboratory can be a very expansive and Robb and Vasquez investigated the capability of two fully
laborious job. To do so even for simple dam-break scenarios three-dimensional computational (3D) fluid dynamics models
and non-variant bed topography a lot of technical assistance and one two-dimensional (2D) depth-averaged model. They
and financial resources are required. Alternatively, Flood found out that the results from CFD models (3D) show quite an
inundation maps are obtained by simulating the flood through improvement over those from the depth-averaged model (2D).
the terrain with the given initial and boundary conditions using [4]
Numerical Models. Numerical models discretize the partial
Ozmen-Cagatay and Kocaman presented an experimental
differential equations and give an approximate solution to the
and numerical investigation of the dam-break flow in the
problem. With the help of numerical models we can analyze
presence of obstacles over initially dry bed conditions. A
any fluid dynamics problem for the actual flow domain and
comparison was made between the shallow water equation,
virtually any realistic operating conditions. Reynolds’s Averaged Navier Stoke Equation, and physical
model. The experimental setup contained a rectangular
The discretization techniques that are normally in use are
channel 8.9m long, 0.3m wide, and 0.34m high. FLOW-3D
the finite difference method, finite volume method, and finite
was used for numerical investigation solving RANS Equation
with k-e Turbulence Model and it was concluded that the
shallow water equation is capable of generating free surface
profiles and the process requires less CPU time for large
computational domains. The measuring technique adopted in
the experiment yields accurate results and the RANS model
reproduces the flow under investigation with reasonable
accuracy while the simple SWE model indicates some
discrepancies particularly in predicting the negative wave
propagation. [10]
704
Wasim Karam et al., International Journal of Emerging Trends in Engineering Research, 9(6), June 2021, 703 – 714
simulated water depths and velocities. Velocities in the profile River and hence forming a lake on the upstream side. A huge
and plan at the upper layer were higher than in the lower layer landslide happened near Attabad village on the 4th of January
which represents the physical phenomena. [12] in 2010 at 100 km upstream of the confluence of Hunza and
Gilgit rivers as shown in Figure 2.
Roga and Pandey used two-dimensional turbulent
non-premixed combustion model. A standard k-ω turbulence
model was used for modelling turbulence. It was observed
from the modelling results that very little amount of OH were
found after combustion. A complete combustion process was
modelled using 2D CFD model. [14]
After all efforts from the local agencies the immediate risk
of dam break due to overtopping was reduced but still there are
many other mechanisms with which the dam can possibly
break. Owing to the geophysics and the location of the Attabad
lake, it can break because of two reasons (1) Attabad lake is
located in Zone 3 as per zonation of seismic activity and is
always susceptible to larger earthquakes so the waterlogged
material of the landslide causing the impounding of water may
lose its strength due to strong ground shaking (liquefaction)
Figure 2: Case Study Location on Map of Pakistan and hence the dam may break (2) another landslide into the
body of the lake may occur, causing the overflow of water over
1.1 Study Area the impounding material and hence the dam material may
The paper focuses on performing a dam-break flow breach. Previously the estimate of the break of Attabad Lake
simulation of Attabad Lake in Pakistan. The lake is formed has been attempted using the One-dimensional HEC-RAS
due to a natural landslide disaster that blocked the Hunza model. In this study, we analyzed the dam break using a 3D
705
Wasim Karam et al., International Journal of Emerging Trends in Engineering Research, 9(6), June 2021, 703 – 714
2.1 FLOW-3D
3. SOLUTION PROCEDURE
Figure 7: Free Surface Elevation relative to local origin
3.1 Topography
the underlying terrain was enclosed in 4 ASTER global DEMs
Normally the topography can be taken from the DEMs of that are downloaded from earth explorer, USGS. The DEMs
the particular location in 30m resolution. The topography of are then merged in ArcGIS to make a single tagged image
706
Wasim Karam et al., International Journal of Emerging Trends in Engineering Research, 9(6), June 2021, 703 – 714
format (.tif) file. FLOW-3D takes the topography as stereo the data in FLOW-3D. Probes can either be a line or a point.
lithography (.stl) file so the resultant .tif file is to be converted Point probes are assigned at 15 different locations to derive the
into .stl format. There is also a utility available in FLOW-3D fluid properties at those points after simulation. Probes are
for the conversion from DEM format to topography (.stl) points at which the model gives us the data output. The data
format but the file that is extracted from the ArcGIS after output that we can get from these probes includes Velocity,
mosaicking the 4 DEMs has .tif format. The .tif file is Hydraulic head, Pressure, Scour, Free surface elevation, Flow
converted to .stl file through a computer program. The merged depth and other hydraulic parameters of fluid and the
DEM was then converted to stereo lithography (.stl) format. topography.
As shown in Figure 5 the file is in 3D and is in .stl format. But
as we can see the file represents downstream of the lake. 3.4 Assumptions/Boundary Conditions
Meshing and applying the equation to the DEM as a whole can To simulate the Attabad dam-break using FLOW-3D, the
lengthen simulation time. Handling the 3D model is not easy following conditions are assumed,
due to its larger size. Therefore we have to reduce the size by
cutting the terrain and restraining the terrain to area The soil mass is replaced by an X-plane baffle with the
representing only the pathway of Hunza River. So for this X-Coordinate at -2303. This baffle will be completely
purpose, The DEM was edited in Google SketchUp with the removed as a situation of a dam break.
help of its handy FREDO tools. The resultant image of the
DEM can be seen in Figure 6 and 7. The earthen dam is assumed to be completely removed at
the instant the dam breaks.
3.2 Reservoir Modelling
The base flow of the Hunza River is not considered in the
As the .stl file is imported to FLOW-3D, Water depth in the simulation.
DEM was modeled with the help of defining a fluid region.
The fluid region was defined in a manner to give a maximum No input discharge to the reservoir is taken into account.
depth of 109 in the lake. The coordinates of the depths and the
Manning’s n value for the area is assumed to be 0.03.
lake are taken with respect to the local origin. The location of
the fluid region along with the spatial distribution of flow The dam break is instantaneous.
depth referenced over Google Earth image can be seen in
Figure 8. 3.5 Specific conditions (Related to FLOW-3D)
The following conditions were assumed related specifically
to FLOW-3D,
SI units
3.5 Meshing
The quality of the results depends upon the mesh size. Finer
the mesh size greater the accuracy. With finer meshes, the
simulation time may increase but the results are accurate.
FLOW-3D uses a Rectangular/Cylindrical Cartesian Mesh to
apply the governing equations to the domain in which the
Figure 8: Model of lake referenced over Google Earth Image scenario is enclosed. Initial and boundary conditions are
applied to the mesh. In FLOW-3D rectangular mesh is used.
3.3 Probes Installation So the domain is enclosed in about a total of 15 mesh blocks.
Where two meshes coincide, there a continuative boundary
To obtain the data from the analysis we must spot locations condition is used. A continuative boundary condition means
in the terrain. These locations are assigned with probes to get
707
Wasim Karam et al., International Journal of Emerging Trends in Engineering Research, 9(6), June 2021, 703 – 714
whatever the scenario is there in the connected mesh block the inundation maps, and flow depth hydrographs) is extracted
other mesh block just keeps it continuous. The boundary from the simulation files.
condition at the junction of the two mesh blocks is marked as
“continuative”, which means that the numerical scheme is 4. RESULTS AND DISCUSSION
continuative on this boundary and in a case, fluid is passing
A total of 4714 sec of the simulation was done for the dam
through this boundary, the flow properties through this
break and the flood wave was modeled in the downstream
boundary will be continuous within the fluid. The result of the
terrain from the point of the dam break. The results obtained
meshing of lake is shown in Figure 9.
from the simulations are
1. Flow Depth Hydrographs
2. Flood inundation maps
3. Velocity Vectors at different time frames after the Dam
breaks
708
Wasim Karam et al., International Journal of Emerging Trends in Engineering Research, 9(6), June 2021, 703 – 714
(D) (E)
Figure 10: Flow Depth Hydrographs of the downstream villages Bridge having a higher base time because of the slow arrival of
(A) Karim Abad (B) Ghulmet (C) Thol (D) Chalat (E) Nomal the attenuated water mass throughout the valley.
4.3 Flood Inundation Maps
The Flood inundation maps show the extent of flooding that
could happen if the dam breaks. Flood inundation maps are
created with the help of reference images of Google Earth that
are textured on the DEMs topography in FLOW-3D For flood
inundation maps, first we model the Peak Flow Depth at a
particular section that happens at a correspondent simulation
time. For example, the flood inundation map for the Karim
Abad village is derived for the simulation time of 1800 sec
because max flow depth at this section occurs at 1800th sec of
the simulation time. Similarly then by drawing the Flow Depth
Contours we know the exact depth at the particular location of
the particular section. Flood inundation maps are derived for
max depth at the particular probe installed at the village
cross-section in the topography. The flood inundation map of
Karim Abad shows in Figure 11 that the smaller villages of
Ganish, Altit, and Garlet are at higher risk of inundation than
most of the Karim Abad, Ali Abad, and Sumayar areas which
are at more elevated areas than these villages. The maximum
section of the village of Hassan abad situated under the flood
Figure 11: Flow Hydrograph at Karim Abad and Nomal Bridge plain between the villages of Aliabad and Miachar are at
greater risk of inundation.
709
Wasim Karam et al., International Journal of Emerging Trends in Engineering Research, 9(6), June 2021, 703 – 714
In Figure 12 we can see that the part of Sikandar Abad chalat village are at greater risk of inundation. Max part of the
agricultural fields and most of the Miaun Village as modeled chalat village situated at the right bank of the Hunza River is
are at risk of inundation. The Miaun village is situated on the completely inundated as modeled. While engulfing most of the
right bank of the Hunza River opposite to thol. The flood chalat valley road and also the Karakoram highway at the
arrival time for Miaun village is 31 min which can be regarded meander bend at the end of the chalat valley. The arrival time
as enough time for the population of Miaun to evacuate to near of flood wave at the chalat valley is 44 min which is enough
situated safer areas subject to availability. While for most of time for the evacuation of the population to safer sites.
the villages after Ghulmet the flood wave is contained by the
flood plain and the villages in this section other than Miaun are
comparatively at lower risk. While Figure 13 shows that a
large part of the Sonikot road and Karakoram highway near the
710
Wasim Karam et al., International Journal of Emerging Trends in Engineering Research, 9(6), June 2021, 703 – 714
711
Wasim Karam et al., International Journal of Emerging Trends in Engineering Research, 9(6), June 2021, 703 – 714
4.4 Velocity Vectors valley, the velocity again increases going through the Hunza
From the velocity data obtained from the probes, we know River. Which shows that the flood wave entered a choked
that the highest velocity in the main channel of the Hunza valley after the chalat and thus the velocity starts increasing
River due to flood wave occurred at Ahmedabad just below again.
the dam break site while the minimum velocity occurs at The flow while going through the meander bend just before
Jafarabad, Sikandar abad, and Chalat which is 21, 21, and 22 the chalat valley undergoes centrifugal action and the wave is
m/s respectively. That is due to the huge meander bend that split after hitting the mountain as the velocity vectors at the
intercepts the flow while entering into chalat valley and the start of the chalat village can be seen in Figure 17.
water entering into a wider area in chalat. After the chalat
712
Wasim Karam et al., International Journal of Emerging Trends in Engineering Research, 9(6), June 2021, 703 – 714
Time Time to
Flood Depth at
Peak Peak to Max Peak
Arrival Peak
S. No Village Name Discharge Depth Peak Velocity Discharg
Time Velocity
(m3/sec) (m) Depth (m/sec) e
(sec) (m)
(sec) (sec)
5. LIMITATIONS:
Every model has its limitation based on the assumptions that prediction of the parameters like inundation maps and
it takes to solve the basic principle equation. The uncertainty estimates arrival times can never be completely accurate
in the solution of dam break problem using CFD arises initially because no single model can fully explain any particular
due to the resolution of the Bed topography derived from the complex scenario like a dam break. Every model has its
dam break, it is quite hard to replicate it because of the variant limitations. Also, the uncertainty of the failure mechanism of
bed topography and also the change in the density of the water the breaching process of the embankment material, though the
with time as it flows downstream. These limitations have a process in this study was not taken into account, makes the
direct impact on the results of the solution such as flood travel results less reliable. However, 3D CFD Models do solve most
time and flood inundation maps. Since to make the results of the problems associated with the assumptions to the models
more reliable high resolution, newly developed satellite data but it does that by the cost of lengthened simulation time and
should be used. Though the results are key in the identification higher range of data input. However, the results of the CFD
of areas that are at risk if the Attabad Lake is breached but the models also depend upon the resolution of the topographic
calculated discharges, travel time of wave, water depths, and data i.e. Resolution of the DEM, incorporation of the land
Velocities should only be used as a guide in future studies and cover, and properties of water i.e. density of the flood wave as
emergency flood management practices. Manning’s n value it flows downstream, etc. Using the FLOW-3D CFD model,
cannot be replicated with ultimate accuracy in any flow flood inundation maps, velocity vectors, and flow depth
problem and for such large-scale problems like Dam-Break hydrographs for the dam break of Attabad lake were plotted.
DEM and its resolution, initially assuming the dam to break Results of the simulations show that the peak depth at various
instantaneously while we are not certain about the dam failure sections of the villages downstream of the lake caries between
mechanism and accurately replicating the Manning’s n value 92 to 15 m. These values indicate higher inundation extents as
directly alter the simulation results. shown in inundation maps of the villages. The peak velocities
6. CONCLUSIONS vary between 44 to 21 m/sec it indicates higher shear stress at
the downstream valley and through the sections where the
Dam break causes heavy damages to infrastructure, flood passes can experience an extreme washout of
landscapes, agriculture, and the environment but the
713
Wasim Karam et al., International Journal of Emerging Trends in Engineering Research, 9(6), June 2021, 703 – 714
agricultural land, residencies, and markets and can danger the simulation. Engineering Applications of Computational
lives of the people situated in the valley downstream. The Fluid Mechanics, 5(4), 541–552.
arrival time varies between 5 to 68 min at villages Ahmed https://doi.org/10.1080/19942060.2011.11015393
Abad and Nomal respectively. This indicates that the villages [11]. Toombes, L., & Chanson, H. (2011). Numerical
nearer to the dam break don’t have enough time to evacuate Limitations of Hydraulic Models. 10th Hydraulics
although thanks to attenuation that the other downstream Conference, (July), 2322–2329.
villages have got enough time to evacuate the vulnerable areas. https://doi.org/10.1016/j.jalz.2016.06.1613
The model results can play a key role in risk analysis and flood [12]. Zarein, M. (2015). Modeling Dam-Break Flows Using
management practice in the identification of the vulnerable a 3d Mike 3 Flow Model, (January).
areas to the flood wave resulting from the failure of the [13]. George, A. C., & Nair, B. T. (2015). Dam Break
embankment dam of Attabad Lake. Analysis Using BOSS DAMBRK. Aquatic Procedia,
4(Icwrcoe), 853–860.
ACKNOWLEDGEMENTS:
https://doi.org/10.1016/j.aqpro.2015.02.10
The authors would like to acknowledge the facilitation from [14]. S. Roga and K. M. Pandey, “Computational Analysis of
Department of Civil Engineering, Abasyn University Supersonic Flow Regime Using Ramp Injector with
Peshawar, Pakistan. The assistance provided by Engr. Faisal Standard K- ω Turbulence Model” .World Academy of
Rehman, Assistant Professor Department of Civil research in Science and Engineering, vol. 2, no. 1, pp.
Engineering, UET Peshawar, Pakistan in terrain modelling is 31–40, 2013.http:// doi.org/10.1.1.348.5862.
highly appreciated.
REFERENCES
[1]. Zhang, L. & Peng, M. & Chang, D.S. & Xu, Y. (2015).
Dam Failure Mechanisms and Risk Assessment, First
Ed. John Wiley and Sons, Singapore 473 pp.
10.1002/9781118558522.
[2]. T. L. Wahl, “Dam Breach Modeling – an Overview of
Analysis Methods,” 2nd Jt. Fed. Interagency Conf. Las
Vegas, NV, pp. 1–12, 2010.
[3]. Khosravi K. “Dam Break Analysis and Flood
Inundation Mapping : The Case Study of Sefid-Rud
Dam,” no. August 2019. DOI:
10.1016/B978-0-12-815998-9.00031-2
[4]. Robb, D. M., & Vasquez, J. A. (2015). Numerical
simulation of dam-break flows using depth-averaged
hydrodynamic and three-dimensional CFD models.
22nd Canadian Hydrotechnical Conference, (June).
[5]. Mohammad Rostami, M. S. (2015). Human Life Saving
by Simulation of Dam Break using Flow-3D. Trend in
Life Sciences, 4(3), 308–316
[6]. Gharbi, M., Soualmia, A., Dartus, D., & Masbernat, L.
(2016). Comparison of 1D and 2D hydraulic models
for floods simulation on the Medjerda River in
Tunisia. Journal of Materials and Environmental
Science, 7(8), 3017–3026. https://doi.org/10.1080/153
[7]. Andrei, A., Robert, B., & Erika, B. (2017). Numerical
Limitations of 1D Hydraulic Models Using MIKE11
or HEC-RAS software - A case study of Baraolt
River, Romania. IOP Conference Series: Materials
Science and Engineering, 245(7).
https://doi.org/10.1088/1757-899X/245/7/072010
[8]. Henderson, F.M. (1966). Open Channel Flow. MacMillan
Company, New York, USA, P. No 304-313
[9]. Betsholtz, A., & Nordlöf, B. (2017). Potentials and
limitations of 1D, 2D and coupled 1D-2D flood
modeling in HEC-RAS. Lund University, 128.
https://doi.org/10.1016/S0300-9440(03)00139-5
[10]. Ozmen-Cagatay, H., & Kocaman, S. (2011). Dam-break
flow in the presence of obstacle: Experiment and CFD
714