HYDRAULIC SIMULATION OF FLASH FLOOD AS TRIGGERED BY NATURAL DAM BREAK

Civil Engineering Forum Volume XXII/1 - January 2013 HYDRAULIC SIMULATION OF FLASH FLOOD AS TRIGGERED BY NATURAL DAM BREAK Yanuar Tri Kurniawan Directorate General of Water Resources, Ministry of Public Works, INDONESIA yanuar_civil_gama@yahoo.com ABSTRACT On January 1 st, 2006 flash flood disaster (in Indonesia is known as banjir bandang) occurred in Kaliputih River, Jember District of East Java Province. This disaster resulted in more than 80 people were killed and hundreds were injured. The disaster was caused by natural dam break. The natural dam was formed by landslide due to heavy rainfall. After the January 2006 disaster, new cracks and crevices were found in the upstream area of Kaliputih River. Based on this condition, it cannot be disregarded to avoid repetition of similar disaster in the future. Therefore, it is required to conduct mitigation efforts in order to anticipate similar disaster in the future. One of the mitigation efforts is modeling simulation of the past event. The understanding which is obtained from the simulation can be used as reference to arrange other mitigation efforts plan and action. Modeling simulation of the January 2006 flood was conducted by involving 1-D model of HEC-RAS version 4.1.0 software. Flood hydrograph was obtained by analyzing related hydrologic aspects using Nakayasu method. The natural dam model was interpreted from field observation and related references. Some assumptions related to study constraints were taken. Model calibration was conducted by repeating simulation using fixed discharge and parameter values in a certain range. The observations were carried out to the maximum water surface and it was traced to the downstream river. Calibration model result showed that the height of natural dam significantly influence changes of water surface at control point. Tracing of flood result in reconstruction of January 2006 flood showed the conformity with the real event. It was observed from the arrival time of flood at certain location. From obtained results, it can be concluded that simulation modeling gave the acceptable results. Keywords: flash flood, simulation, natural dam. 1 INTRODUCTION On January 1 st, 2006 flash flood disaster (in Indonesia is known as banjir bandang) occurred in Kaliputih River, Jember District of East Java Province. This disaster resulted in more than 80 people were killed and hundreds were injured. According to an investigation, this disaster was caused by natural dam break. This natural dam was formed by landslide due to heavy rainfall for several days in a row. A few days after the event, a field investigation was conducted by ISDM Project Urgent Survey Team of Japanese Experts in cooperation with Directorate General of Water Resource and Agency of Research and Development, Ministry of Public Works. The investigation results stated that the trace of landslide heaps was found in upstream area of Kaliputih River. The landslide blocked mainstream of Kaliputih River and formed the natural dam. In 2007, a study was conducted by Naryanto, H.S., Wisyanto, and Marwata, B. from BPPT. The study results stated that after the January 2006 disaster, new cracks and crevices were found in the upstream area of Kaliputih River. Based on this condition, cannot be disregarded repetition of similar disaster in the future. Therefore, it is required to conduct mitigation efforts in order to anticipate of similar disaster in the future. The mitigation effort is aimed to reduce as much as possible the risk of losses which potentially occur. Flash flood (in Indonesia is known as banjir bandang) is flood disaster that occurs quickly or abruptly with large flood volume. Besides water, flash flood also mixed with fine and coarse material in the form of sand, pebbles, gravel, rock, and fallen trees. This disaster is very dangerous and destructive because of the high velocity and its mixture. Landslides, which are often found on steep slope in mountainous and hilly area, often show early indication of flash flood event. It s based on the estimation that landslide material fall and accumulated in river valley. The fallen landslide material block mainstream and form the natural dam (Utami and Hermawan, 1998). The high rainfall intensity will increases the flow and forms the reservoir in the upstream side of natural dam. If water surface in the reservoir exceeds natural dam (overtopping), it will erodes the top part of natural dam and causes the natural dam break. Similar study about the flood due to natural dam break was conducted by Chjeng-Lun Shieh, et al (2007) in Study on Warning Criteria of Rainfall and Hazard 1319

Volume XXII/1 - January 2013 Civil Engineering Forum Zone Mapping for Landslide Dam Hazards, which simulated the flood due to natural dam break in Lung- Chuian River, Taiwan using HEC-RAS software. The flood propagation to the downstream was simulated by fixed bed hydraulic simulation method to estimate the hazard area. The natural dam break was simulated for 1 hour due to overtopping with 4 scenarios of rainfall intensity as inflow hydrograph. River geometry data was interpreted from topographic map and aerial photo because of limited fund and time. The simulation results showed that from 3 villages as the observation area, only 1 village potentially affected by the flood. The simulation result is very useful in providing information to arrange the emergency action plans. One of the mitigation efforts is by modeling simulation of the past event. The understanding which is obtained from the simulation can be used as reference to arrange plan and action of other mitigation efforts. This paper presents the result of the study related to the flash flood phenomena, and comprises of two focuses as follows; a). To reconstruct the January 2006 disaster through the natural dam break and flood modeling simulation using field observation approaches, secondary data, desk study, realistic and acceptable assumptions. b). To study the sensitivity of peak discharge resulted from natural dam break by changing breach parameters. The upstream area of Kaliputih River is located at southern slope of Argopuro Mountainous (see Figure 1). At the downstream river, Kaliputih River meets Dinoyo River. In this study, Kaliputih Watershed was divided to Kaliputih Hulu and Kaliputih Hilir Watershed. Modeling simulation was conducted at Kaliputih Hilir with the length of river is about 13.8 Km. Related to the study constraints, especially lacking of sufficient data and information, some assumptions were applied in this study, among others were: a) Before and after the disaster, the topography condition was assumed not to change significantly, b) The modeling simulation ignored the sediment transport in the river, c) The natural dam break was assumed to be triggered by overtopping, d) The modeling simulation ignored obstruction structures in the river. 2 METHOD OF SIMULATION The flash flood here in this study was studied by applying flow simulation technique involving 1-D model of HEC-RAS version 4.1.0 software. Figure 1. Watershed of kaliputih river. The simulation was carried out based on the hydraulic scheme as sketched in Figure 2. Several assumptions were adopted as follows; a) The methods was limited to the hydrology and hydraulic aspects; b) The simulation was carried out by using data available from field observation as well as from measurement, c) Some assumptions were taken relating of study constraints, d) The scenarios of the simulation were as follows: 1) Flow simulation on existing condition (without natural dam), 2) Flow simulation on the conditions of before natural dam formed, during natural dam formed, and the natural dam break. e) Boundary conditions in the simulation were as follows: 1) The flood hydrograph of January 1, 2006 was used as upstream boundary condition; 1320

Civil Engineering Forum Volume XXII/1 - January 2013 2) The friction slope (slope of the energy grade line) was used as downstream boundary condition due to lacking of data. The friction slope was set at 0.031; f) Modeling simulation was set from January 1, 2006 at 10.30 am to January 2, 2006 at 10.30 am or for 24 hours. Figure 3. River geometry identification through map. Figure 4. River geometry identification through survey. Figure 2. Sketch of hydraulic simulation. 3 BOUNDARY CONDITIONS 3.1 Natural Dam In the January 2006 disaster, the actual natural dam geometry was unknown. In this study, the natural dam model was interpreted from field observation result and related references. The methods for modeling the natural dam were as follows: a) The bottom width of natural dam was assumed equal to the average width of river valley. It was interpreted from field observation and Google Earth image as is shown in Figure 3. b) The height of natural dam was roughly estimated from the landslide volume. By estimating the dimension of the landslide, it can be estimated that the landslide volume was about 50,000 m3. In this study, the height estimation of the natural dam was in ranging 15 meters to 25 meters. c) The height of natural dam was roughly estimated from the landslide volume. By estimating the dimension of the landslide, it can be estimated that the landslide volume was about 50,000 m 3. In this study, the height estimation of the natural dam was in ranging 15 meters to 25 meters. d) In this study, some modifications were taken concerning various aspects in the modeling of natural dam. The modifications are shown in Figure 5 and explained as follows: 1) Spillway with reasonable dimensions When water surface in reservoir reaches the natural dam, it will erode the top part of natural dam and forms natural spillway. 2) Gate at the bottom of natural dam This modification was aimed to simulate the blockage of river flow by the natural dam. Due to the limitation facility in the software, at the beginning of simulation, the natural dam had been modeled. At 10:30 am, the gate was in the fully open condition. The rain was started at 1:30 pm and then the landslide was assumed instantaneously occurred at 2:30 pm. The landslide was modeled by closing gate gradually. The gate dimensions were obtained by trial and error method in order to allow the flow before 2:30 pm was not distracted by the natural dam. In this study, the shape of gate was assumed in the rectangular shape. The width of gate is 4 m and the height is 3.5 m. 1321

Volume XXII/1 - January 2013 Civil Engineering Forum 3) The pilot flow This modification was caused by the limitation of the software. In the HECRAS software, the simulation can't be conducted at the zero discharge. The pilot flow was obtained by trial and error method. The pilot flow should be as small as possible in order to not significantly affect to the simulation result. In this study, the pilot flow was 0.05 m 3 /seconds. Figure 5. Natural dam characteristic by assumptions. Another modification was conducted towards the downstream of natural dam. This modification was applied by assuming of the lateral inflow in the steady flow form at 1 m 3 /s. This modification was aimed to the numerical stability in the simulation. The lateral inflow also should be as small as possible in order to not significantly affect the simulation result. The mechanism of the dam break is known as break parameters, which consists of: a) Bottom width of breach (b). In this study, bottom width of breach was set at 30 meters, it is equal to the width of river valley b) Breach slope (z). In this study, breach slope was set at (h/v)=1,88. It is characteristic of the internal angle friction of landslide material c) Break duration (t). In this study, break duration (t) was estimated by using the Froelich formula for dam break (Fread, 1988). In estimating the break duration, the characteristic of reservoir in the upstream side of natural dam was interpreted from topographic map by using ArcGIS software. The estimation result of the break duration was about 9 minutes. 3.2 River Geometry The river geometry data was obtained from Balai Sabo-Yogyakarta in the form of river cross section data. In this study, the cross sections of river were made wider using the topographic map. The coefficient of riverbed roughness used the n Manning value. In this study, the n Manning value was interpreted from photos which describe the situation of the Kaliputih River after the disaster. In determining of the n Manning value, the used reference is the Open Channel Hydraulics book (Chow, V.T., 1997). According to the book, natural river with sandy loam soil, lot of roots, shrubs, fallen trees in the stream because of landslide, and other sediment in the river bed has the n Manning value = 0.15. In this study, the n Manning values were set as follows: a) In the upstream side of natural dam, the n Manning value was set at 0.15; b) In the downstream side of natural dam, the n Manning value was set in ranging0.15 to 0.17. 3.3 Flood Hydrograph To conduct the modeling simulation, flood data was required from the catchment area as input data at upstream boundary. The flood data was obtained by analyzing related hydrologic aspects. In this study, the hydrologic analysis was carried out through the following descriptions: a) Calculation of hourly rainfall distribution from daily rainfall data available using the Alternating Block Method (ABM) Rainfall Distribution, b) Daily rainfall data available was obtained from the Gentong Hydrology Station, daily rainfall on January 1, 2006 was amounted to 160 mm, c) The rainfall intensity was calculated using the Mononobe formula, d) The duration was set to 10 hours, it was based on information from local people, e) The effective rainfall was calculated using the runoff coefficient, which was adopted from the Mononobe Table. For steep mountainous and river in mountain area, the runoff coefficient was set to 0.85, f) Elaborating of hourly rainfall distribution to be the flow using The Nakayasu Synthetic Unit Hydrograph (HSS Nakayasu) Method, g) The base flow was calculated using HSS GAMA I. From calculation result, it was obtained equal to 1.77 m 3 /s, h) In elaborating the rainfall to be the flow, the Nakayasu parameter was obtained from the characteristics of Kaliputih Hulu Watershed as catchment area. The effective rainfall distribution and calculation result of the flood hydrograph on January 1, 2006 from Kaliputih Hulu Watershed is shown in Figure 6. 1322

Civil Engineering Forum Volume XXII/1 - January 2013 Upstream Boundary (RS 13+784) Natural Dam (RS 12+980) Figure 6. Sketch of hydraulic simulation. 4 RESULTS AND DISCUSSIONS 4.1 Model Calibration The model calibration is adjustment of uncertainty model parameter to the obtained solution. The uncertainty is caused by applying immeasurable data or containing errors in measurement. In this study, the model calibration was conducted by repeatedly simulation using fixed discharge and the parameter values in a certain range. The obtained solution was compared of its conformity to the measured data. In this study, the calibration of model using the parameters as follows: a) The height of natural dam (hd) in ranges of 15 meters to 25 meters; b) The break duration (t) in ranges of 6 minutes to 18 minutes. The observation was carried out to the maximum water surface. The flood mark from field observation was used as the measured data. The observation location is located at a distance of 6 Km from the natural dam (see Figure 7). Figure 8 shows the results of the model calibration. As in Figure 8, the maximum water surface close to the measured data in ranges of the natural dam height (hd) of 22 meters to 24 meters. In the model calibration process, the break duration (t) was not significantly influenced. 4.2 Reconstruction of January 2006 Flood In this study, reconstruction of the January 2006 flood the height of natural dam (hd) was set at 23 meters and the break duration (t ) was set at 15 minutes. The used data is shown in Table 1. Table 1. Reconstruction of the January 2006 flood Parameters Magnitudes Base of natural dam, (+m) 697.02 Base of Spillway, (+m) 717.02 Center top of natural dam, (+m) 720.02 Water surface as triggering of natural dam break, (+m) 720.15 Base of breach, (+m) 698.02 Base of breach width, (m) 30 Break duration, (minutes) 15 Downstream Boundary (RS 0+000) Figure 7. Control point location. Control Point (RS 6+984) Figure 8. Calibration result at control point. The simulation results show that the natural dam break was started at 8:27 pm. Peak discharge resulting from the natural dam break was 891.94 m 3 /s and the maximum velocity was 4.74 m/s at a distance of 60 meters from natural dam. 4.3 Flood Routing of Reconstruction of January 2006 Flood The routing of flood results show that the flood reached the Km 4.06 (RS 8+920) at 8:52 pm or 0.42 hours (25 minutes) from the start of the natural dam break. The peak discharge arrived at 8:56 pm or in 1323

Volume XXII/1 - January 2013 Civil Engineering Forum 0.48 hours (29 minutes). This results close to the real event. It was based on information from the local people. In Table 2 is shown the routing of flood at 5 locations and in Table 3 is shown the quantitative comparison between the simulation results and the data. Stationing (RS) Table 2. Routing of flood at 5 locations Distance from natural dam Max water surface (MWSE) Peak discharge Time to achieve MWSE Max Velocity Max water depth Time arrival of flood (km) (+m) (m 3 /s) (hour) (m/s) (+m) (m) (hour) 12+918 0.06 695.35 891.53 0.22 4.74 689.77 5.58 0.07 8+920 4.06 436.13 699.72 0.48 2.61 432.51 3.62 0.42 6+984 6.00 328.56 562.04 0.72 1.70 324.03 4.53 0.65 3+513 9.47 177.40 378.81 1.30 0.94 174.09 3.31 1.15 0+000 12.98 89.64 283.60 1.97 1.61 86.57 3.06 1.70 The natural dam blocked the mainstream river and formed a reservoir in the upstream side. When the natural dam break occurred, the water was released rapidly through the crack. This condition increases the water volume and the peak discharge drastically in short period. In propagating to the downstream, the peak discharge was damped due to the influence of river floodplain changes and deceleration of flood velocity. Table 3. Quantitative comparison between the simulation results and the data Observation The roar of flood was started at around 8.30pm (assumed as natural dam break) Flood reached RS 8+920 in 0.5 hours (assumed as time arrival of flood) Mark of max. flood 1. Pondok pesantren building at Kemiri Village, Panti subdistrict - River bed - Mark of max. flood River bed Simulation Data result at 8.30 pm at 8.27 pm Deviation -3 minutes 0.5 hour 0.42 hour -0.08 hour 324.03 m 328.61 m 328.56 m Relative deviation (%) - Water depth 4.58 m 4.53 m 0.05 m -1.09 2. Trees at downstream boundary - River bed 86.57 m - Mark of max. 89.57 m 89.63 m flood - Water depth 3 m 3.06 m 0.06 m 2.00 4.4 Impacts of Dam Break to Discharge A comparison between the flood hydrograph of reconstruction of the January 2006 flood and the flood hydrograph on existing condition (without natural dam) is shown in Figure 9. The comparison was conducted at 2 locations. A. First location (RS 6+984), 6 Km from natural dam; B. Second location (RS 0+000), 12.98 Km from natural dam. Figure 9. Flow hydrographs at two locations It is seen from Figure 9 at Point A, due to the presence of natural dam break, the peak discharge increased approximately 5 times as much of original peak discharge. Similar phenomena occurred at Point B where the increase of the peak discharge was about 2.4 times. 4.5 Sensitivity of the Discharge due to the Change of Breach Parameters In this analysis, the height of natural dam was set to 22 meters and the nonlinearity rate of the breach growth was set to 2.5. The sensitivity analysis results show that the peak discharge resulting by the natural dam break is sensitive to changes of the break duration (see Table 4 and Figure 10). Table 4. Simulation results of varies breach parameter Scenarios Bottom breach width (m) Breach parameter Duration (minutes) Breach slope Peak discharge (m 3 /s) 1 30 9 1.88 1,078 2 30 9 0.47 1,136 3 30 9 0.94 1,129 4 30 9 1.41 1,114 5 30 12 1.88 814 6 30 15 1.88 677 7 30 18 1.88 575 8 22.5 9 1.88 1,061 9 15 9 1.88 1,022 10 75 9 1.88 1,014 1324

Civil Engineering Forum Volume XXII/1 - January 2013 a. Development of modeling simulation using other scenarios. It is aimed to obtain the worst possibility in extreme condition, b. The further study of modeling simulation involving the transport sediment simulation to obtain the closer result to the real event, c. Development of modeling simulation using the 2- D simulation method to obtain the area affected by the flooding. Figure 10. Sensitivity of peak discharge to changes of the breach parameters 5 CONCLUSIONS The followings are the conclusions of the study; a. The modeling of natural dam break and flow simulation with some assumptions were applied with acceptable results, b. The peak discharge resulted from the natural dam break is sensitive to changes of the break duration (t), c. Results of the tracing of flood in reconstruction of January 2006 flood showed the conformity with the real event, d. Assumptions and justifications which were taken in this study, allowing errors and deviations in the result compared to the real event. By considering the results of these studies then some related suggestions are forwarded as follows: ACKNOWLEDGMENTS The author would like to express gratitude to the Directorate General of Water Resources, Ministry of Public Works upon data and field survey facilities those have been provided to the author. REFERENCES Chjeng-Lun Shieh, dkk, 2007, Study on Warning Criteria of Rainfall and Hazard Zone Mapping for Landslide Hazards, 2nd International Conference on Urban Disaster Reduction, November 27-29, 2007. Naryanto, H.S., Wisyanto, Marwanta, B., 2007, Landslide and Flash Flood Potential and their Analysis of Disaster Event at 1 January 2006, Argopuro Mountaineous Area, Jember District, Alami, Vol. 2, 2007. Salukh, F.I., 2004, Analysis of Flood Routing of Flash Flood due to Dam Failure (Case Study on Tilong Dam, Kupang District, East Nusa Tenggara Province), Master Thesis (in Bahasa), Magister in Natural Disaster Management, Post Graduate Programme, Universitas Gadjah Mada, Yogyakarta. 1325

Volume XXII/1 - January 2013 Civil Engineering Forum [this page intentionally left blank] 1326