Tài liệu Flood forecasting and awareness program for the thua thien hue province, vietnam

FLOOD FORECASTING AND AWARENESS PROGRAM FOR THE THUA THIEN HUE PROVINCE, VIETNAM by Hoang Thanh Tung A thesis submitted in partial fulfillment of the requirements for the Degree of Master of Engineering Examination Committee: Prof. Ashim Das Gupta (Chairman) Dr. Mukand Singh Babel (Co-Chairman) Dr. Dushmanta Dutta Nationality: Vietnam Previous Degree: Bachelor of Water Resources Engineering Hanoi Water Resources University, Vietnam Scholarship Donor: MARD/DANIDA Vietnam Scholarships Asian Institute of Technology School of Civil Engineering Bangkok, Thailand July 2004 Acknowledgements I wish to express my deepest gratitude and sincere appreciation to my advisor Professor Ashim Das Gupta and to my co-advisor Dr. Mukand Singh Babel for their patient guidance, valuable advices and continuous encouragements throughout the study. I also express my profound gratitude to committee member, Dr. Dushmanta Dutta for his valuable suggestions, recommendations and encouragements. I wish to acknowledge the various supports from the Standing Office of the Provincial Committee for Flood and Storm Control of Thua Thien Hue, Provincial Hydrometeorological Services of Thua Thien Hue, the Research Institute of Hydro-meteorology, the UNDP project office VIE/97/002 – Support to the Disaster Management System in Vietnam during my data collection. I am grateful to the Hanoi Water Resources University/DANIDA Vietnam for providing financial support throughout the study period. I wish to extend my deepest sense of gratitude to beloved parents, wife and daughters for their endless love and inspirations. Finally, appreciation is extended to everyone whose name is not mentioned but directly or indirectly helped me to complete this study. i Table of Contents Chapter Title I II III IV Page Introduction 1 1.1 General Introduction 1 1.2 Description of the Study Area 1 1.3 Description of the Huong River 1 1.4 Problems and Need of Study 2 1.5 Objectives of Study 3 1.6 Scope and limitation of Study 3 Literature Review 6 2.1 Flood Forecasting 6 2.2 Geographical Information Systems in Hydrology and Water Resources 9 2.3 Flood inundation maps 10 2.4 Integration of GIS, Remote Sensing and Hydrological/Hydraulic Models for Flood Warning 11 Theoritical Consideration 13 3.1 Flood Forecasting 3.1.1 Back Propagation Neural Network (BPNN) 3.1.1 Multi-variable Regression Analysis 13 13 16 3.2 Flood inundation mapping 3.2.1. Rainfall - Runoff Model HEC-HMS 3.2.2. VRSAP Models 17 17 18 3.3 GIS model 22 Data Collection and Methodology 23 4.1 Data collection 4.1.1 Hydrometeorological data 4.1.2 Bathymetric data 4.1.3 GIS and Remote Sensing data 23 23 26 26 4.2 Methodology 4.2.1 Development of a suitable flood forecasting method for downstream areas 4.2.2 Development of Flood inundation mapping 4.2.3 Recommendation of a program for improvement of public awareness for flood preparedness and mitigation 27 27 29 ii 33 Chapter Title V Page Results and Discussions 34 5.1 Data analysis 5.1.1 Rainfall data 5.1.2 Runoff data 34 34 35 5.2 Flood forecasting model for downstream area 5.2.1 Approach for flood forecasting: 5.2.2 Selection of suitable forecasting time 5.2.3 Selection of the best model for flood forecasting at Kim Long 37 37 37 38 5.3 Flood inundation mapping 5.3.1 Selection of flood warning scenarios 5.3.2 VRSAP model setup, calibration and verification 5.3.3 Development of flood inundation maps 44 44 46 50 5.4 Recommended program for improvement of public awareness for flood preparedness and mitigation 5.4.1 Institutional framework for disaster mitigation in Vietnam and knowledge of local people in the Thua Thien Hue Province for disaster preparedness 5.4.2 Recommended program for improvement of public awareness for flood preparedness and mitigation. VI Conclusions and Recommendations 60 60 63 71 6.1 Summary 71 6.2 Conclusions 71 6.3 Recommendations 6.3.1 Recommendation for implementation 6.3.2 Recommendation for future research 72 72 72 Apendixes 73 Apendix 1 74 I. Report on the current status of gauging stations network in the Huong River System Basin of Thua Thien Hue Province 75 II. Cross section data and schematization of the Huong River System for VRSAP model 76 Apendix 2 I. Statistical Tools II. Results from the calculation of 6 hours ahead - flood forecasting for downstream areas of the Huong River system at Kim Long by Multi-variable Regression method III. Results from the calculation of 6 hours ahead-flood forecasting for downstream areas of the Huong River system at Kim Long by BPNN method 80 81 Apendix 3 I. Calibration and verification results of the HEC-HMS model for Duong Hoa, 97 iii 86 90 Chapter Title Page Binh Dien, and Co Bi basins to calculate un-observed inflow at upstream boundary and lateral inflows II. Recommended specifications for upgrading of the gauging station network in the Huong River system basin References 94 98 104 iv List of Figures Figure Title Page 1.1 Administration map of the Thua Thien Hue Province 4 1.2 Causes of disasters in the Thua Thien Hue Province 5 1.3 Flood warning for the Thua Thien Hue Province 5 3.1 A typical three layer feed forward artificial neural network 13 3.2 An artificial neural network building block with activation function 13 3.3 Logistic activation function 14 3.4 Flowchart of the back propagation algorithm 16 4.1 Gauging network in the Thua Thien Hue Province 24 4.2 Cross section locations on the Huong River system in Thua Thien Hue Province 25 4.3 Flowchart of overall methodology of this study 28 4.4 Flowchart of methodology for development of suitable flood forecasting method for downstream areas 28 4.5 Flowchart of methodology for development of flood inundation maps 29 4.6 Flowchart of methodology of flood phase maps 30 4.7 Flowchart of methodology of flood effected crop 30 4.8 Flowchart of methodology of flood affected roads 31 4.9 Flowchart of methodology of flood affected population and villages 31 4.10 Flowchart of methodology of flood effected land types 32 4.11 Flowchart of methodology for recommendation of a program for improvement of public awareness for flood preparedness 33 5.1 Observed and calculated water levels by MVR model 42 5.2 Observed and calculated water levels by BPNN model 43 5.3 Hydrograph of the 1999 Flood on the Huong River at Kim Long 45 5.4 Calculated and observed water levels at Kim Long on the Huong River of the historical flood in November 1999 48 5.5 Calculated and observed water levels at Phu Oc on the Bo River of the historical flood in November 1999 48 5.6 Calculated and observed water levels at Kim Long on the Huong River of the severe flood in October 1983 49 v Figure Title Page 5.7 Calculated and observed water levels at Phu Oc on the Bo River of the severe flood in October 1983 49 5.8 Flood inundation map at alarm level I on the Huong River system 53 5.9 Flood inundation map at alarm level II on the Huong River system 54 5.10 Flood inundation map at alarm level III on the Huong River system 55 5.11 Flood inundation map at alarm level III+ on the Huong River system 56 5.12 Flooded roads in Hue City at alarm level III on the Huong River system 58 5.13 Land used risk map at alarm level III at different scenarios of Hue City 59 5.14 Flow chat of Institutional Frame Work for Disaster Mitigation in Vietnam 61 5.15 Flow chat of Disaster Warning in Vietnam 62 5.16 Sample of training materials on disaster mitigation at CCFSC 64 5.17 Sample of flood hazard map 65 5.18 Sample of flood preparedness and mitigation poster which is available at CCFSC and VRC 66 5.19 Sample of training materials for school children which are available at CCFSC, UNDP, and VRC 69 5.20 Sample of training materials for school children which are available at CCFSC, UNDP, and VRC 69 vi List of Tables Table 1.1 4.1 4.2 4.3 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9 5.10 5.11 5.12 5.13 5.14 5.15 5.16 Title Page The form characteristic of the Huong River basin Hydro-meteorological data GIS data Remote sensing data Statistical data of some rains caused by Northeast winds in the Thua Thien Hue Province Statistical data of some rains caused by tropical low pressures and storms in the Thua Thien Hue Province Possible occurrence of yearly peak discharge in months (1977 – 1999) Flood transmission time from Thuong Nhat to Kim Long Correlation matrix of independencies Summary of the linear –models Verification results of 3 models Verification results of 4 models Official flood alarm levels used in Vietnam Selected flood inundation scenarios in the Thua Thien Hue Province Selected flood inundation scenarios in the Thua Thien Hue Province Calculation results of flood inundation area in flood scenario Summary of inundation status during the historical flood (Nov. 1999) Estimated costs for Training of officials in charge of disaster preparedness and mitigation at provincial, district, and commune levels Estimated costs for events or grass-root training on flood awareness and mitigation for local people (commune level) Estimated cost for training of school children in primary schools vii 2 23 26 26 34 34 35 36 39 39 41 43 45 46 51 57 63 67 68 70 CHAPTER 1 INTRODUCTION 1.1 General Introduction Natural disasters and floods are natural events occurring annually in central Vietnam. It is not economically viable to protect against all the effects of the maximum possible flood for any given river basin. Further, physical works to contain or reduce floods may result in adverse consequences elsewhere. A realistic balance must be obtained between protecting life, property, and infrastructure from floods of a selected probability; and making adequate provision for evacuation and other emergency measures to cope with the effects of floods of greater magnitude. The recommended approach should be a combination of non-structural and structural flood control and flood damage protection. In the Central Part of Vietnam, due to the fact that the information available to provincial and local authorities and the community on likely flood levels and their expected time of occurrence was totally inadequate. Flood forecasting and warning systems urgently need to be upgraded and an improved public awareness and information program for protection of the population against natural disasters, including storms and floods, needs to be developed and implemented immediately. This study will focus on the enhancement of flood forecasting and awareness program for people in the flood prone area of Thua Thien Hue – one of the most flood affected provinces in central part of Vietnam. 1.2 Description of the Study Area Thua Thien Hue Province is Located at the latitudes 16°14' - 16°15' North, longitudes 107°02' - 108°11' East, is 127 km long and 60 km wide on average, with mountains accounting for up to 70% of the natural land. Geographically, Thua Thien Hue borders Quang Tri Province to the North, Da Nang City to the South, with Laos P.D.R., separated by the Truong Son range, to the West, and over 120 km of seacoast to the East. Thua Thien Hue Province covers a natural area of 5,009 sq.km, and has a population of 1,050,000 in 1999 (The administrative map of the Thua Thien Hue Province is in Figure 1.1). Thua Thien Hue lies in the tropical monsoon zone influenced by the convergent climate of the subtropical North and the tropical South. There are two distinct seasons: the rainy season, with storms and hurricanes, lasts from September to December; and the dry season, with little rain, lasts from January to August. Floods and Storms are the main disasters in the Thua Thien Hue Province (causes of disaster in the Thua Thien Hue Province is briefly described in Figure 1.2) 1.3 Description of the Huong River The Huong River is the greatest river of the Thua Thien Hue Province and located from 16o00’ to 16o45’ of the north latitude and from 107o00’ to 108o15’ of the east longitude, its west is the Truong Son mountains, its north is Bach Ma mountains, its south is contiguous to the Da Nang city and its east is Eastern sea. 1 The Huong River has a basin area of 2,830 km2 that representing 56% of total area of Thua Thien Hue province and playing an important role on water resource as well as the inundation status of the province. Over than 80% of this basin area is hills and mountains with their heights from 200 to 1,708 m. A 5% of the basin area is the coastal dunes with altitude from 4 – 5 m to 20 – 30 m; the remainder area is 3,700 ha which can be cultivatable. Main flow of the Huong River originates from a high mountain area of the Bach Ma range where is from 900 to 1,200 m of altitude. From its origin to the Tuan cross river, the main flow is called as the Ta Trach and from the Tuan cross river, it is called as the Huong River (or the Perfume River), the river that expresses many bold romantic feature on life, culture and natural landscape of the Hue ancient capital. The form characteristic of the Huong River system and its great branches is shown in the table below: Table 1.1 The form characteristic of the Huong River basin Name of river or branches River class Huong Main Hai Nhut I ravine CaRum I BaRan Co Moc I ravine Huu Trach I Bo I Dai Giang Tributary Length Basin Ave. (km) area height of (km2) basin (m) River bed incline (%) Ave. width of basin (km) 44.6 104 15 2830 75.3 330 4.8 10.6 29 219.3 458 62.3 18 88.3 51 94 27 729 938 River Curve network coefficient density (km/km2) 0.6 1.65 11.3 0.58 1.45 1.30 326 384 3.7 9.5 14.6 12.7 0.64 0.64 1.51 1.85 The Huong River water converges in the Tam Giang - Cau Hai lagoon that its length is of 67 km, average width is of 2.2km and its depth changes from 1. 5m, and then goes to the sea through mainly the Thuan An and Tu Hien estuaries. 1.4 Problems and Need of Study The floods in November and in December 1999 in the central part of Vietnam clearly underlined the vulnerability of this part of the country to natural disasters. While floods occur every year, those in 1999 were of a particularly severe nature and the extent of injury and damage was extremely widespread. The government at central and local levels took immediate action during and after the floods to assist the local communities to cope with the crisis and to help restore food supplies, essential services, and essential infrastructure. Further more, An Integrated Natural Disaster Mitigation Policy for Central Vietnam has been prepared which involves both structural and non-structural measures. However, the non-structural measures are more concentrated and preferable because of the current condition of fund limitation in Vietnam. In these non-structure measures, improvement of flood forecasting and enhancement of public awareness on disaster preparedness and mitigation are those of highest priority. Therefore, this study is 2 not only useful for the author but also useful for the implementation of the Early Flood Warning Systems for Central Provinces of Vietnam. 1.5 Objectives of Study The overall objective of this study is to develop a flood forecasting and awareness program for the Thua Thien Hue Province, Vietnam. To obtain this objective, some sub-objectives that need to be achieved are as follows: • To develop a flood forecasting model for downstream areas. • To develop flood inundation mapping for the Thua Thien Hue Province that including a series of inundation maps corresponding to several flood water level scenarios at downstream. • To develop a program for improvement of public awareness for flood preparedness and mitigation. Figure 1.3 shows the relationship between components of the proposed flood forecasting and public awareness program for the Thua Thien Hue Province 1.6 Scope and limitation of Study The scope of this study is limited to the Huong River system basin of the Thua Thien Hue Province and includes: • To collect and analyze data and information on hydrology, topology, socioeconomic and current gauging network in the Huong River system of the Thua Thien Hue Province. • To select suitable forecasting model and then calibrate and verify the selected model for downstream areas of the Huong River system. • To calibrate and verify the developed VRSAP (a hydrodynamic model) and the HEC-HMS models for the Huong River system. • To develop flood inundation maps for the Thua Thien Hue Province that including a series of scenarios based on GIS analysis and simulation results of the above model. • To develop a program for improvement of public awareness on flood preparedness and mitigation for the Thua Thien Hue Province. 3 Figure 1.1 Administration map of the Thua Thien Hue Province 4 Figure 1.2 Causes of disasters in the Thua Thien Hue Province (Source: Strategy and Action Plan for Mitigating Water Disaster in Vietnam, UNDP Vietnam) Flood inundation maps Hydrological/Hydraulic models GIS and Satellite Images flooding area flood depths H, Q Base Station Provincial Flood Warning Office Flood forecasting H, Q Neural Network/or simple flood forecasting program Fast Estimation of effected population effected road, land use,.. Safe access place during flood time Flood warning SCADA control/monitoring software Public Awareness Training Remote gauging stations Flood forecasting Local people 1 Figure 1.3 Flood warning for the Thua Thien Hue Province 5 GIS analysis GIS analysis CHAPTER 2 LITERATURE REVIEW Flood forecasting and warning including the integration of GIS and hydrologic models has been described and studied by many researchers, so quite a large number of literatures available in AIT and also those accessed through Internet have been reviewed. This chapter summaries the review of related literatures in structured way shown below: Literature Review Flood forecasting GIS in hydrology and water resources Flood inundation mapping Integration of GIS, Remote Sensing and Hydrological/Hydraulic Models for Flood Warning 2.1 Flood Forecasting Flood forecasting methods range from the very simple to complex methods. However, there are currently two main approaches employed in flood forecasting. • The first is a mathematical modeling approach. It is based on modeling the physical dynamics between the principal interacting components of the hydrological system. In general, a rainfall-runoff model is used to transform point values of rainfall, evaporation and flow data into hydrograph predictions by considering the spatial variation of storage capacity. A channel flow routing model is then used to calculate water movement down river channels using kinematics wave theory. Example of this approach is Mike 11 – FF (this model includes 3 modules: NAM model for simulation of rainfall – runoff process in sub-basins, Hydrodynamic River Module (HD) for the simulation of discharge and water levels along rivers and flood plains, and Flood Forecasting Module (FF) enabling easy management of real time data as well as issuing of forecasts. • The second main approach to flood forecasting is modeling the statistical relationship between the hydrologic inputs and outputs without explicitly considering the physical process relationships that exits between them. Examples of stochastic models used in hydrology are the autoregressive moving average 6 model (ARMA) of Box & Jenkins (1976) and Markov model (Montgomery,1976). ARMA models work on the assumption that an observation at a given time is predictable from its immediate past, i e., a weighted sum of a series of previous observation. Markov methods also rely on past observation but the forecasts consist of the probabilities that the predicted flow will be within specified flow intervals, where the probabilities are conditioned on the present state of the river. Artificial neural networks offer a significant departure from the conventional approach to problem solving and have been applied successfully to a variety of application areas including pattern recognition, classification, optimization problems and dynamic modeling. Although neural networks were historically inspired by the biological functioning of the human brain, in practice the connection is more loosely based on a broad set of characteristics which they both share, such as the ability to learn and generalize, distributed processing and robustness (Tawatchai, 2003). The basic function of a neural network, which consists of a number of simple processing nodes or neurons, is to map information from an input vector space onto an output vectors space. These neurons are distributed in layers which can be interconnected in a variety of architectural configurations. Information is delivered to each neuron via the weighted connections between them. Information processing within each neuron normally comprises two stages. In the first stage, all incoming information is converted into activation, where the most common activation function is simply the sum or the weighted inputs. In the second stage, a transfer functions, such as the sigmoid, converts the activation into an output value. A neural network learns to solve a problem by modifying the values of the weighted connections through either supervised or unsupervised training. In supervised training, neural networks are provided with a training set consisting a number of input patterns together with the expected output, and adjustments to the weights are based on the differences between observed and expected output values. These adjustments are calculated using a gradient descent algorithm (optionally modified for higher performance with refinements such as a momentum or second derivative term). Currently, the most widely applied network is the multilayer perception using a supervised training algorithm, known as back-propagation. Once trained, the network is validated with a testing dataset to assess how well it can generalize to unseen data. In unsupervised training, the artificial neural network attempts to identify relationships inherent in the data without knowledge of the outputs and is often used in classification problems. To date, there have only been a handful of neural network applications that address the hydrological forecasting problem. Research has shown that neural networks have great potential as substitutes for rainfallrunoff models (Abrahart and Kneale, 1997). Zhizun (1997) has demonstrated the success of neural networks, trained with a genetic algorithm water level prediction. Tawatchai (1998) used Back propagation feed forward neural network model for forecasting daily river discharge of the Upper Nan River at Tha Wang Pha Gauging Station in Thailand. The Upper Nan River Basin has the drainage area of 2,200 km2 at Tha Wang Pha gauging station, four rainfall gauging stations, one pan evaporation station and one stream gauging station. In this basin rainfall varies significantly over the basin area and with time; discharge at Tha Wang Pha station also varies significantly. This model was calibrated with the flood in 1993 and verified with flood in 1994. The results were very satisfactory both in calibration and verification. This showed that the back propagation feed forward neural network model is well suitable with this study. However, the NN in flood forecast is limited to floods of magnitude equal or smaller than that 7 considered in training; large error may occur if NN model is applied to forecast of much larger magnitude. Tawatchai (1999) used Back propagation feed forward neural network model (BPNN) and Auto Regressive and Moving Average model (ARMA) for forecasting hourly discharges and water levels of the Chao Phraya River at Bangkok Memorial Bridge. The area in this study is the Lower Chao Phraya River Basin, from Amphoe Bang Sai (km 112) to Fort Chula (Km1) at the mouth of the river including Bangkok Metropolis area. This part of the Chao Phraya River is considerably affected by the tide. There are many canals associated with the river along its length, some take water from and return it to the main stream creating a complicated network. The BPNN model was calibrated and verified by using the hourly measured water levels and discharges at Bangkok Memorial Bridge. Stochastic modeling of error time series known as ARMA model was used to improve the performance of BPNN model by adaptive error correction technique. The results were very satisfactory. Bartual (2002) had developed a short term river flash flood forecasting by using the artificial neural networks (ANN) model for the upper Serpis River basin (460 km2), with the outlet in Beniarres reservoir (29hm3). The system is monitored by 6 rain gauges, providing 5 min rainfall intensities, while reservoir inflows are derived from depth measurements in the reservoir every half hour and real time data from controlled discharges in the spillway. In order to produce 1, 2, 3 hours forecasts, the model make use of the distributed rainfall information, together with observed discharges in the preceding hours. Several ANN topologies have been tested and compared, including linear and non linear schemes, being in all cases three layer feed-forward networks. Best results are obtained with different architectures for each forecasting horizon, basically due to the decreasing dependence of future inflows with respect to preceding values of the series as the time horizon is increased, while rainfall information increases its importance as a predictor. The ANN architectures finally proposed are achieved through pruning algorithms. Training is performed using the quasi-Newton approach. Abrahart and See (1999) in their research outlined some neural network data fusion strategies for continuous river level forecasting where data fusion is the amalgamation of information from multiple sensors and/or different data sources. The objective of data fusion is to provide a better solution than could otherwise be achieved from the use of single source data. The simplest data-in-data-out fusion architecture involves the combination of input data from multiple sources. Continuous river level forecasts derived from a set of conventional, fuzzy logic, and neural network models, are amalgamated via a feed-forward neural network. The data fusion methods are demonstrated using data from sites located in contrasting basins: the Upper River Wye and the River Ouse. The potential improvements that can be achieved through the implementation of these types of amalgamation methodologies could have significant implications for the design and construction of automated flood forecasting and flood warning systems. Bárdossy and Kontur (2001) in their research conclude that soft computing methods (fuzzy rules, neuronal nets and probabilistic reasoning) offer the possibility of a model free forecasting system. The application of these techniques is investigated using different input variables and changing data quality assumptions. Past observations are used to set up fuzzy rules for forecasting. Besides the traditional parameters of discharge and precipitation other factors such as the influence large scale meteorological features is investigated. The methods are applied for different forecasting time horizons and the 8 change of the influence of the input is documented. A cross validation and a split sampling approach are used to validate the results. The methodology is applied to discharges at the Neckar River in South-West Germany. Results are compared with traditional linear/ nonlinear forecasting methods. The DHI Water & Environment in cooperation with six government agencies in Thailand during the past eight years the MIKE 11 modeling system has been set up successfully in seven river basins in Thailand for the purpose of planning, design and operation of flood control measures as well as real time flood forecasting. The forecast results are encouraging especially in the two latter river basins where automatic telemetry systems have been established to provide the required real time data on rainfall and river water levels. In the other four basins, where low-cost data collection and transmission systems are operated manually, the models produce reliable forecasts when the required data are made available and applied in real time. This is, however, not always the case due to occasional operational problems with the manual data collection, transmission and model operation. Therefore the trend in Thailand regarding flood forecasting is currently towards automatic telemetry systems combined with advanced automatic model execution and forecasting (Nielsen and Klinkhachorn, 2002). 2.2 Geographical Information Systems in Hydrology and Water Resources The GIS are computational tools to store, recover, process and visualize spatial data. A very interesting concept presented by Gurnell and Montgomery (1998) defines the GIS as information integration technologies which may include aspects of cartography, remote sensing, demography, economy, landscape attributes, computational science, etc. The application of the GIS in Hydrology and Water Resources can be classified in two categories: management and analysis. • The uses referred to with management include data storage, recovery and visualization. For example, certain attributes may be stored within the GIS, such as the sitting of wells, the hydrographical network, reservoirs, etc. The manager and planners of this resource can use this information for taking decisions to do with water resources and land planning. • The analytical applications of the GIS refer to modeling. The real potential of the GIS lies in its analytical capacity, which allows, among other possibilities, the generation of new layers of information. Ya et al. (1998) developed a map-based flow simulation model under a geographic information system (GIS) based on the concepts of object – oriented programming. The hydrology simulation model is composed of three elements, which are (1) equations that govern the hydrologic processes, (2) maps that define the study area and (3) database tables that numerically describe the study area and model parameters. GIS was used to integrate these three separate model-components. The map-based surface/subsurface water flow simulation model was developed and successfully applied to simulate surface and subsurface flow on the Niger River Basin in West Africa. In the process of constructing this map-based model, techniques were developed to address and solve some GIS related problems such as treatment of spatially-referenced time-series data, feature-oriented map operations, dynamic segmentation of an arc, and integration of flows along a line. Maidment et al. (1996) developed a spatial hydrology model which simulates the water flow and transport on a specified region of the earth using GIS data structures. Constructing a GIS hydrology model at first is to determine what variables will be 9 calculated on how many spatial units for a defined number of time periods. Construction of complex models must proceed by partitioning the total problem into a series of submodels that interact with a common database. It is the capacity of GIS for rigorously defining this database which makes possible complex models connecting various parts of the hydrologic cycle within a particular region. Leipnik et al. (1993) have explained about the implementation of GIS for water resources planning and management. GIS is designed to store information about the location, topology and attributes of spatially referenced objects and many database queries are performed through it. They have described in detail about the stages in the implementation of GIS and understanding of these stages help in using GIS in water resources planning and management effectively. Dubey and Nema (2000) in their research concluded that remote sensing data along with field data when combined on GIS platform results into useful informative map for planning land and water resource actions, keeping in mind the resource availability and utilizability. The spatial distribution of resources availability, utilization and its quantification in very short time with least affords makes this information widely acceptable to variety of fields specially where time factor is a constraint. 2.3 Flood inundation maps Three approaches are conceivable for the estimation of the flood hazard areas in the basin under consideration (Infrastructure Development Institute, 2003): • Estimation based on geomorphological survey • Estimation based on experienced floods, and • Estimation based on hydrologic/hydraulic models To reduce flood damage, a flood risk map must contain accurate information on the magnitude of flood and distinct flood hazard area. The size of flood is usually expressed in terms of discharge rate, flood height or rainfall intensity. The flood hazard area is usually distinguished on a map by a colored area display or contour line of equal water depth. In some cases, point depth of water is more useful for local residents. From the technical aspects, these three estimation methods are characterized as follows: a. Estimation based on geomorphological survey In this method, attempts are made to grasp qualitative flood prone areas by the classification of the topography, using the fact that topographical classification such as alluvial fans, deltas, nature levees and back marshes are closely related to the frequency of floods and the inundated duration. The characteristics of this method are as follows: • Results are obtained by reflecting the floods which had occurred in the past. Therefore, the inundation regimes when a flood of relatively large scale occurs could be grasped to some extent. • Though it may be necessary to obtain the help of experts in the fields of topography and geology in the geomorphological classification, this method is less costly than other methods. • Although it is possible to make a preliminary estimate with regards to the inundated areas and the inundation regimes (duration, velocity of flow, etc), judgment cannot be made on inundation level and the differences of the regimes among the floods of different scales. 10 b. Estimation based on experienced flood This is a method for finding out flood prone areas and the flood level based on the past flood conditions. The results of this investigation form the most important part in the public announcement of flood prone areas. The characteristic of this method are as follows: • Inundated areas and the flood level could easily be known. • Since it is experienced flood, reliability of estimation is high. Therefore, this method is advantageous in the public announcement with sufficient persuasive power. Moreover, very valuable lessons could be drawn with regards to the response of residents against floods. • Since the scale of floods is various, corrections should be made by using hydrologic/hydraulic models. c. Estimation based on hydrologic/hydraulic models In this method, the scale of flood is determined by calculations on probability of the rainfall or discharge, and the water levels corresponding to the flood scale is estimated by using hydrologic/hydraulic models. The characteristics of this method are as follows: • The flood water level can be estimated for given flood scale. Therefore, it is possible to use it in designing the height of foundation embankment required for building etc, or in estimating the inundation depth of house. • Arbitrary probability can be selected as the scale of flood for study. • Change of flood level due to river improvement and land subsidence can be taken into account. • The accuracy of this method is largely affected by the accuracy of topographical data. Therefore, compared with other methods, relatively large expenditure and time are necessary for accurate estimation. • Various constants are included in the runoff model, and as a result, considerable errors would be included in the runoff calculations if the constants are inappropriate. However, when the models are calibrated using the records of past floods, the reliability becomes high. Therefore, investigation of past floods is required without fail in adopting this method. As mentioned above, each method has its advantages and disadvantages in estimating the inundation areas. Among them, the most reliable and persuasive method is the estimation based on past flood records. However, the data and information obtainable through the past flood survey are, in general, limited. Studies by hydrologic/hydraulic simulation models and geomorphologic surveys need to be conducted to complement the information required for the flood inundation mapping. 2.4 Integration of GIS, Remote Sensing and Hydrological/Hydraulic Models for Flood Warning The graphical information system (GIS) and remote sensing has proven to be a very effective tool for describing, analyzing, modeling and integrating forecasted flood levels with other related information such as topographic, thematic and attribute information. It offers new opportunities to develop and implement a user-friendly, interactive decision 11 support system for flood forecasting and warning as well as identifying the affected areas using dynamic spatial modeling. Aziz et al. (2002) had successfully developed a flood warning system for Sunderganj area of Bangladesh by integrating Danish Hydrodynamic Model MIKE 11 and GIS. MIKE 11 and GIS are successfully integrated in Arc View GIS environment for retrieving near-real time flood level information. The study illustrates that an effective warning system can release warning in advance 72, 48 and 24 Hrs. It can change the existing perspective of flood preparedness and mitigation substantially and render information for better decision making for saving lives of people in Suderganj area of Bangladesh. Mekong River Commission (MRC) has developed flood risk hazard maps throughout the Basin to improve flood preparedness in the four riparian countries in the Lower Mekong Basin (Mahaxay, 2002). This work is one part of a larger initiative being undertaken by the MRC to strengthen regional cooperation for flood management and mitigation. The inundation maps (database) have been compiled from various sources including remote sensing data, field surveys, topography, hydrology, hydraulic model, etc. It is useful in many fields of application. The development of flood risk hazard maps can be briefly describes as the followings: • Initial information on flood extents was derived from RADARSAT imagery supported in some areas by field surveys. The available information focused largely on the high and peak flood levels, with very little information about the propagation of the flood. They provide no information on duration. Flood depth information is also limited to the maximum depth. • A hydraulic model analysis was used to provide more comprehensive information on the flood extents for lower flood levels, as well as information on water levels, depths and durations. The VRSAP (a Mathematical Hydraulic Model named Vietnam Rivers and Plains) hydraulic model was able to simulate maximum daily water levels at 1977 points (nodes) in the lower Mekong region from July 1 2000 until December 31 2000. Using these point data, a water surface could be calculated for each day by interpolating data using a nearest neighbour analysis. In addition to this point data, water levels in over 903 polygons (cells) were also provided that can be transformed to a water surface. • Flood depth maps were derived by subtracting the Digital Elevation Model (DEM) from the daily simulated water surface. An ArcView Avenue program was written to automatically generate these DEMs (grids). The result of this exercise was produced for 184 (number of days) inundation depth grids. • Flood duration can be derived from the inundation grids by counting the number of days that each cell had an inundation depth (normally greater then zero). • The compiled information was combined into integrated data sets. The two components of the Integrated Databases are the ‘Specific Floods’ information on the floods in 1995, 1996, 2000 and 2001 and the ‘Flood Hazard’ . information on flood scenarios which are defined in to 3 flood classes: minor flood – a flood with a probability of occurrence of about once every 1.2 years; medium flood – a flood with a probability of occurrence of about once every 5 years; and, major flood – a flood with a probability of occurrence of about once every 20 years. 12