Integration of GIS and Hydrodynamic Modelling to predict and simulate flood inundation risk in the Lower Bicol Floodplain, the Philippines Muhibuddin Bin Usamah Department of Geology, Hasanuddin University Kampus Unhas Tamalanrea, Jl. Perintis Kemerdekaan KM. 10 Makassar 90245, South Sulawesi, Indonesia Phone/Fax : +62-(0)411-580202 Email: muhy@unhas.ac.id, realmuhy@yahoo.com Web: http://www.unhas.ac.id Dinand Alkema Department of Earth System Analysis International Institute for Geo-Information Science and Earth Observation (ITC) Hengelosestraat 99, PO Box 6, 7500 AA Enschede, The Netherlands Phone: +31-(0)53 4874 286, Fax: +31-(0)53 4874 336 Email: alkema@itc.nl Web: http://www.itc.nl Abstract This study attempts to predict flood inundation risk in the Lower Bicol floodplain through the integration of GIS and hydrodynamic modelling using SOBEK software. Bicol floodplain is located on the Luzon island, the Philippines, which experiences serious and frequent floods and causes substantial suffering, loss of life and economic damage. Digital elevation models (DEM) of the floodplain were created in ILWIS through the GIS interpolation algorithms of spot heights and contour lines, which are obtained from topographic maps of different sources and scales. Other structure data available in the floodplain (such as road, embankment) were topologically adjusted and attributed for further analysis. The DEM of the floodplain was integrated with 1D schematization of river network to produce the final DEM of the study area. The final DEM was combined with the surface roughness map and discharge/water level at the four defined boundaries to be used for calculation in modelling phase using SOBEK. SOBEK integrates a one-dimensional (1D) modelling package with a two-dimensional (2D) hydrodynamic prediction package thus it has the functionality of including the one-dimensional river/channel flow and two-dimensional overland flow, which is called SOBEK 1D2D. The modeling outputs from SOBEK were transferred back to GIS for visualization and analysis. ILWIS was used for GIS processing in the pre and after modelling including GIS data preparation, integration, visualization and analysis. Snapshots at particular time interval was generated in SOBEK. The work resulted in flood model showing spatial distribution of flood extent, water depths, velocity and time of flooding for different flood events. The results of this study show that integration of GIS and hydrodynamic modelling is a good way of predicting flood to be used for flood prevention, risk assessment and flood management that can strengthen the local authorities in risk management. 1 Introduction Floods are becoming increasingly important hazards in all around the world. Every year, river and coastal flooding in the world results in major loss of life and property. The damage caused by the floods to the lives and properties is significantly higher than the damage done by any other natural hazards. In other words, major floods are the largest cause of economic losses from natural disasters and disaster-related deaths. As floods are among the most severe risks on human lives and properties, there is always a need to prevent more loss from the possible effects of floods in the community. One of the ways to prevent more loss and risk in the future is by understanding the behaviour of the flood. Flood modelling is one of the ways to understand the flood behaviour. The modelling result can be used to obtain flood information for flood emergency planning. 1.1 Study Area The Bicol River Basin occupies 312,000 ha at the southern end of Luzon island, in the province of Camarines Sur and Albay. It consists of one major river basin, which is Bicol River and 10 minor river basins. The Bicol River runs primarily from south to the north with the additional inflow discharge from Tributaries, e.g. Naga River. Near the estuary at the San Miguel Bay, the Libmanan River, coming from the west joins the Bicol River. The basin is about 130 km long and 25 km wide and is bounded along its length by the Ragay Hills to the southwest and a series of volcanoes to the northeast. The study area is focused on the Lower part of the basin, which is a coastal floodplain that lies between 130 to 140N and 1230 to 1240 E.  Figure 1. Location of the study area 2 Problem Statement The Bicol River Basin is a complex river system, much of which is low lying with low gradients. The floodplain is wide and suffers from perennial flooding problems caused by typhoons. Even in low magnitude-high frequency events, large areas are submerged for periods of weeks with depths of water often in excess of two meters (BCEOM Consultants, 1991) The river banks are normally very low and in some parts, the river is even higher than the ground surface of the back swamps. Most buildings and residential areas are within 1 buffer zone of two kilometres from the river, thus highly susceptible to floods. This study attempts to predict and simulate flood inundation risk in the Lower Bicol Floodplain, the Philippines through the integration of GIS and hydrodynamic modelling. The specific objectives of this study are:
3.1 The approach This research applies 1D2D hydraulic modelling to simulate the flood events with different return period of floods in the Lower Bicol Basin. In conducting 1D2D modelling, there are four input parameters needed (figure 2). They are digital elevation model to represent the topography or floodplains, landcover map that is transformed into surface roughness coefficient map, structures and hydrological data. Hydrological data used in this research are discharge and water level obtained from the result of the model developed in 1991 by NIPPON KOEI in Bicol River Basin Flood Control and Irrigation Development project (BCEOM Consultants, 1991). Discharge and water level hydrographs for the return period of 1.25, 5, 10 and 25 years were available. Values for return period of 2 and 50 years are derived from statistical interpolation and extrapolation  Figure 2. Parameter input for 1D2D hydraulic modelling This modelling uses SOBEK-rural, a product of the main SOBEK line of hydrodynamic water flow models developed at WL|Delft Hydraulics. This software package integrates a one-dimensional (1D) modelling package with a two-dimensional (2D) hydrodynamic prediction package thus it has the functionality of including the one-dimensional river/channel flow and two-dimensional overland flow, which is called "SOBEK 1D2D" (WL|DELFT Hydraulic, 2003). This 1D2D system is designed for the simulation of overland flooding or inundation when in normal conditions (in case of no flooding) hydrograph can be modelled as a one-dimensional (1D) network. If larger areas are inundated then assumptions for 1D flow are normally no longer valid and in that case, the system becomes truly two-dimensional (Frank, 2001). 3.2 Model Schematization The data input and editing interface in SOBEK is called NETTER. NETTER offers the possibility to set up the schematization on top of a background GIS map. NETTER also offers advanced analysis tools to show model result attached to the schematization. Its window based user interface makes it easy to input and edit geometric data. NETTER has two edit models. The first mode is the mode to set up the schematization, e.g. by adding new nodes. The second edit mode is the mode for editing the attribute data. In this mode, one can give attributes to the schematization objects. It is also important to know that SOBEK works with nodes which have different functionalities. NETTER also provides two separate menu bars to input and edit 1D feature like river and other drainage channels and 2D grid layer representing the floodplain. To start schematizing, it is important to have the DEM imported to the NETTER. The DEM that has been developed and exported into ArcInfo ASCII format file can directly be inputted to the schematization using "2D grid" tool. After importing the DEM, the flow boundary condition should be defined and be put on the 2D schematization. The flow boundary conditions will be connected using flow connection node in order to be able to define the 1D component. The complete schematization of the 2D network can be seen in figure below. To determine the exact values of all parameters, e.g. water level variation in time on specific locations in the 2D grid, some nodes called history stations are used.  Figure 3. Schematization of 2D Network with history stations 3.3 Input boundary condition In the schematization, there are four boundary nodes: three 1D nodes (up stream boundaries of the Bicol, Naga and Libmanan River) and one 2D Node, the sea level at San Miguel Bay. 4 Results and Discussion The model simulations were carried out for 2, 5, 10, 25 and 50 years return period of floods with 6 days of simulation time. With one hour interval, there were 120 ASCII raster water depths, water level and velocity files were obtained from each simulation. Those files were converted into GIS-ILWIS system to be able to do post simulation analysis to obtain flood indicator maps. The results of each simulation for each return period were analyzed to see the different flood behaviours of different flood events. Comparison is made based on the spatial extent, water depth, flow velocity and inundation time as different return periods of floods have different behaviours of those flood characteristics. 4.1 Comparison of 2, 5, 10, 25, and 50 year floods 4.1.1 Spatial Extent The model showed that there is a big difference in the flood extent from the most probable flood happen in a year to the least probable flood, i.e. from return period of 2 years to return period of 50 years. The slightest change occurred in the flood with 5 and 10 year return periods, where there is only one percent difference in terms of flood extent. The slight difference in the flood extent may be explained as due to the low elevation of the floodplain in the study area. In addition, the DEM used for this modelling was limited to the maximum of 5-meter height.  Figure 4. Flood Extent Map for return period of 2, 5, 10, 25, and 50 years. 4.1.2 Inundation depth With the less probable floods to occur, the inundation depth was deeper in the floodplain. Consequently, maximum depth of inundation was also increased in the smaller probability of floods to occur. Comparison between maximum depth of inundation for different return period of floods can be seen in figure 5.  Figure 5. Spatial distribution of water depths at peak-flow conditions for 2-, 5-, 10-, 25-, and 50-year return period of floods. The most extreme depth of water occurs in the flood with return period of 50 years, where there were 175 km2 areas with water depth was between 3 to 4 meters. Maximum depth which is about 6 meter only occurs in this event Results at some history stations in the model also show that the water reaches its peak flow at the 60th hour and continuously decreased until the fifth day of simulation (see figure 3 for location of history stations). The wetting system in the model was not defined thus after five days of simulations; it was obtained in the results that the water in the floodplains still remained. It varies from zero meters in the return period of 2 year to 70 cm for return period of 50 years (figure 6)  Figure 6. Water depth at history stations 56 for return period of 2, 5, 10, 25, and 50 years 4.1.3 Velocities Graphs of the dynamic simulations indicate the temporal variability of the flow velocities in main channel and at the inundated areas when water stream flow inundates the floodplain. Maximum simulated velocities decreased from large flood to small flood to occur. Maximum flow velocity at some points in the area was 1.58 m/s in the return period of 2 years to reach 3.47 m/s and 4.84 m/s in the return period of 25 and 50 years. However, for the return period of 5 and 10 years, maximum velocities have slight differences, about 2.35 m/s. The difference can be observed on the floodplain, where velocity in the return period of 5 years was slightly lower than that of 10 years return period.  Figure 7. Spatial distribution of velocities at peak-flow conditions for 2-, 5-, 10-, 25-, and 50-year return period of floods. 4.1.4 Warning Time Warning time maps shows how much time is needed by the people before the flood reaches their home. The results of the simulation show that the bigger return period of floods, the less warning time for the people until the water reaches their house. Figure 8 shows spatial distribution of warning time before the water reaches the area.  Figure 8. Warning Time Map for return period of 2, 5, 10, 25, and 50 years. 5 Model Validation and Calibration 5.1 Validation of the inundation extent Since no measured flood data available in the whole Bicol Basin, the simulation results for return periods of 5, 10, and 25 years were compared with existing flood extent map for the same return periods. Those existing maps were available from the model developed by NIPPON KOEI in 2003 from Bicol River Basin and Watershed Management Program. The model had been used to test a range of model scenarios for the purpose of flood alleviation and also for other management works for the Bicol basin (Nippon KOEI CO, 2003). Comparison between model result and the model reference developed by Nippon Koei in percentage and in surface area which are flooded and not flooded can be seen in table 1.
5.2 Calibration of the Water Level Model Since there is no real measurement data, the maximum water level of the model result was validated with maximum water level at some sub basins from the 2003 model from Bicol River Basin and Watershed Management Program. Scatter plot of the model results and the reference model shows agreement or more or less similar results. Comparison of the results of 2-, 5- and 25-year return periods with the reference model clearly showed positive relation, does not deviate from the expected line x=y.  Figure 9. Comparison between simulated water level and reference model for return period of 5 years  Figure 10. Comparison between simulated water level and reference model for return period of 10 years  Figure 11. Comparison between simulated water level and reference model for return period of 25 years Overall, good agreement between model result and hydraulic model from Bicol River Basin and Water Management Program was obtained for the calibration events. However, it should be noted that not all areas are calibrated due to the unavailability of calibrating data. 6 Conclusion and Recommendation 6.1 Conclusion The integration of GIS and combined 1D2D hydrological modelling opens up possibilities for studying flood control measures, flood forecasting and development of flood evacuation plans. Information from the model can be used as a source for flood risk assessment and can be useful for decision making. For basin planning and management, this model can be used as reference for identifying free lands for future development in the Bicol River Basin, especially in the lower part on which this research is focused. In addition, detailed visualization from 1D2D flood model can be an important information for flood management programme in giving recommendations for future planning. This study shows the application of combined 1D-2D modelling. The objectives of this research were achieved and several maps had been derived to show the difference behaviour of floods at different return periods. 6.2 Limitation of the research and further direction Distinguishing different return period of floods is mainly done by using one variable, which is a different discharge and water level value for different return period of floods. However, it is a well known matter of fact that the geomorphology is also changed from time to time as the changing of man-made artefacts like embankments, etc. Not to mention 50 years, within 5 or ten years, there could be changes in the geomorphologic features within one area, especially in the low-land area where erosion and sedimentation can take place simultaneously as well as man-made modifications of the terrain. In addition, sedimentation from Mayon Volcano has accelerated after several eruptions that resulted in changes of the morphology of the river, particularly in the channel itself. These kinds of changes were not included in the modelling parameters. Therefore, integrating hydrological modelling with geomorphic modelling that shows prediction of geomorphic change will add extra information for such work. This geomorphic modelling includes, for instance, sedimentation from upper part of the basin. By this, real situation will be more represented. 7 Acknowledgement This research was conducted under ITC's SLARIM research project: Strengthening Local Authorities in Risk Management. Authors thank to Cees van Westen as principal investigator of SLARIM project. The main data of this research would not have been available without the generous help of Gilberto Abion from NEDA Region V, Legazpi, the Philippines and Ms. Estella J. Gumabon from NAMRIA office in Fort Bonifacio, Manila, the Philippines. 8 References
Page 1 of 1
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||