EnviroInfo 2002 (Wien) Environmental Communication in the Information Society - Proceedings of the 16th Conference Copyright © IGU/ISEP, Wien 2002. ISBN: 3-9500036-7-3 ANFAS: A Decision Support System for Simulating River Floods Poulicos Prastacos1 Abstract This paper discusses the ANFAS system, developed under a joint research effort between the EU and the PR of China. ANFAS is a decision support system that can be used by decision makers and stakeholders to simulate river floods and perform “what-if” scenarios. It is a web based distributed system that integrates the GIS databases, the mathematical models and the impact assessment methodologies. Both 1-D and 2-D models are included in the system and the interface between models and databases is done in a transparent way. The system includes extensive facilities that permit users to visualize the results on the web and/on their own personal computers. The system is presently applied in three pilot sites; the Vah River in Slovakia, the Loire River in France and the Yangtze River in China. 1. Introduction Rivers floods have caused extensive damages in the world over the past years resulting in human losses and extensive economic damages. Through the years governments have established several measures for diminishing the impacts of river floods. Since most flood prevention measures are capital intensive and/or may have political and economic consequences there is a need for tools that would permit decision makers to test their policies before adopting them. Since floods are affected by several factors (rainfall, terrain morphology, established protection measures, economic activity etc.) these tools must consider all cause and effect relationships and evaluate in a systematic way the various alternatives. Although several attempts have been made to develop mathematical models for flood simulation (Samuels 1999) there is a need for an integrated approach that brings together models, data and methodologies in a way that permits decision makers to use the available tools (Prastacos 2001). Flood simulation models are numeri1 Foundation for Research and Technology – Hellas, Institute of Applied and Computational Mathematics, Regional Analysis Division, P.O. Box 1527, GR-71110, Heraklion, Crete, Greece. Tel. +30 (810) 391767, Fax. +30 (810) 391761; e-mail: poulicos@iacm.forth.gr, http://www.iacm.forth.gr/regional 227 cal models often developed years ago that require excessive data manipulation to prepare inputs and visualize outputs. To address these problems the European Union and the People’s Republic of China have undertaken a large research project that leads to the development of a decision support system for flood simulation and implement it in three pilot sites. The overall objective of the ANFAS (datA fusioN for Flood Analysis and decision Support) project is to develop a Decision Support System for flood prevention and protection, integrating the most advanced techniques in data processing and management. The project started in 2000 and will be completed at the end of 2002. The research team includes eight European and five Chinese teams. The Information Society Technologies (IST) programme financed the work of European partners, while Chinese researchers are funded from the Ministry of Science and Technology (MOST) of China. 2. ANFAS framework The intended users of the system were defined to be not highly skilled hydraulic experts or modelers but rather the technical staff, managers and decision-makers, who have enough background in floodplain management and could interactively use the system and interpret its output. Some important features of the system include: 1. Decision support system capabilities; ANFAS includes facilities that permit users to define scenarios that consist of modifications in the hydraulic structures and perform “what-if” simulations. 2. Generic; ANFAS was designed as a generic system that is a system that can be used for flood simulation in a variety of rivers and situations. It therefore integrates both 1-D and 2-D numerical models for flood simulation, it is not constrained by data availability and resolution, and can be therefore implemented in data poor or data rich environments, 3. Integrated but modular; The system consists of different modules that are transparently integrated. Users can replace some modules (numerical models, impact methodologies, GIS system in their PC) with their own and/or use other datasets. 4. Web based distributed architecture; The complete system has been designed as a distributed system with data and models potentially residing on a different system. This permits the use of fast computers or HPCN clusters for solving large 2D models. Users access the system through their familiar web browser. 5. Visualization capabilities; Model results are translated to maps and diagrams. Additionally, users have the capability to easily identify issues that are of particular concern to them such as how long an area will be flooded, how long it will take for the flood to reach a particular area etc. The ANFAS system consists of four different modules. These are: 29.08.02, 5821 EI P 235 I2 PrastacosP.doc Copyright © IGU/ISEP, Wien 2002. ISBN: 3-9500036-7-3 228 1. Remote sensing/Computer vision; This is the component that prepares from a variety of sensors the topographic databases needed. 2. Geographic Information System; The GIS stores all input datasets and also through various scripts prepares in a transparent way the input files for the models, stores the results of the models in the various databases and is also used for visualization. Visualization is carried out the server or on the user’s PC. 3. Numerical models; This includes the numerical models for flood simulation. Two existing models CARIMA (1/1.5-D) and FESWMS (2-D) have been integrated. 4. Impact methodologies; In this module the results of the models are used to evaluate the impact of the flood on the surrounding area. The ANFAS integrated system with which users interact includes the GIS databases, the models and the impact methodologies coupled with the appropriate user interfaces (Figure 1). Preparation of the various GIS coverages is performed outside of the integrated system through either remote sensing analysis, digitization or data base manipulations. Once the various datasets have been prepared they are stored as standard shapefiles in the system data repository. Fig. 1: ANFAS data flow overview 29.08.02, 5821 EI P 235 I2 PrastacosP.doc Copyright © IGU/ISEP, Wien 2002. ISBN: 3-9500036-7-3 229 3. ANFAS architecture ANFAS is a web based distributed system. Users with appropriate privileges can access the system through the web browser and perform simulations. The system consists of a database server that handles all data sets, a modeling server that contains all the models and where all computations are performed and the ANFAS core server for handling all interactions and user management. The ANFAS web client is a web application that permits users to run the system. The various components can be implemented on different computers or even different locations. A complete description of the architecture is provided in Houdry et. al. (2001). Although, the web-based interface to ANFAS offers several advantages, it can limit users that are interested in analyzing model results on their own. It would be impractical to develop appropriate interfaces on the system that will permit ANFAS to handle all potential data uses. To provide freedom of choice appropriate procedures have been embedded in the system that permit users to download the results in standard shapefile format. The downloadable file can be viewed without any further processing by all commercial GIS packages. Additionally, users can download a library of software components (FLOOD-VIEW) that permits them to visualize the results under some specific GIS software. Presently such components have been developed for ESRI’s ArcView v. 3.2 using the Avenue script language. They permit users to perform thematic mapping, produce graphs displaying water depth at any point in the floodplain, display flood propagation in 3-D visualization etc. 4. Mathematical models The prediction of the flood evolution is carried out through numerical models. The most significant difference among the various models is on how they treat the dimension of the river and the floodplain. The approximation of the area can be in one dimension (1-D/1.5-D models), or in two or three dimension (2-D and 3-D models). Although the latter result in more accurate description of the flow they have excessive data requirements and heavy computational loads. In the 1-D models water flow in the river is assumed to be uni-dimensional and is represented by Saint Venant Equations. The floodplain is represented by a set of interconnected basins (cells) and water flow is governed by simple flow equations. In 2-D models, a finite element mesh is used to represent the river and the floodplain. Through the use of hydrodynamic equations water surface elevation and flow velocities are computed at each of the nodes of the mesh. 2-D models require detailed representation of the area topography and require excessive computer power even for modeling floods in relatively small areas. In ANFAS since the objective was to develop a generic system both 1-D and 2-D models have been integrated. These include the CARIMA 1-D model commercially available from SOGREAH, Grenoble, France and the FESWMS (Finite Ele- 29.08.02, 5821 EI P 235 I2 PrastacosP.doc Copyright © IGU/ISEP, Wien 2002. ISBN: 3-9500036-7-3 230 ment Surface Water Modeling System: 2-Dimensional Flow in a Horizontal Plane) available from FHWA (U.S. Federal Highway Administration). On the Loire pilot site both models have been implemented, while for the Vah pilot site only the FESWMS model has been applied since the end user felt that accurate results could not be obtained from a 1-D model. In the Chinese pilot site because of the large size of the area the 1-D model was used to model the flow in most of the area, whereas for a small portion of the area the 2-D model was applied. It must be pointed that the models used in China (1D model Susbed-NEW and 2-D model CSMS) are variants of the models used in the European pilot sites and also account for sedimentation an important issue for a large river such as the Yangtze. 5. The Geographic Information system – Visualization of the results All geographic files in the system are of shapefile format. This is a well-known format originally defined by ESRI, Inc. that all GIS packages can read. Internally the map server that displays the results on the web is using its own optimized format. The data flow between models and databases is handled through the “model encoding” and “model decoding” components. The “model encoding” reads the geographic and attribute databases and prepares the file needed for running the numerical model. All of the input data files, with the exception of the mesh, are prepared automatically without any interference from the end user. The mesh, needed to run the 2-D model, is prepared outside of the system through specialized software, the mesher. The “model decoding” component translates these model results files into a standard structure (dbf format) that can be handled by any GIS software. One of the most important requirements that end users specified was to be able to easily visualize model outputs and most importantly visualize the results in terms that could be of use to them. Hence, visualization of model results was a key concern in the system development. Two types of visualization capabilities are provided in ANFAS. The first one is embedded in the system and permits end users to display maps on the web that show the flood extent and the “alert time”, that is the time it will take for the flood wave to reach any particular area. The second type is handled outside of the system using the downloadable components discussed earlier. With the results of the 2-D models users can perform 3-D visualizations. The 3-D visualization is carried out using the 3-D Analyst extension of ArcView. Appropriate scripts have been developed that permit users to display the flood wave propagation in 3-D and also to produce fly-through. On the Chinese version of the system a Geobeans component is used to handle the 3-D visualization. 29.08.02, 5821 EI P 235 I2 PrastacosP.doc Copyright © IGU/ISEP, Wien 2002. ISBN: 3-9500036-7-3 231 6. Pilot sites, data preparation and scenarios specification To test the generic aspect and the functionality of the complete system ANFAS is implemented in three pilot sites. An extensive effort was undertaken to assemble datasets for each of the pilot sites through satellite interpretation and aerophotography and organize them under ArcView (Prastacos et el. 2001). For the European sites Digital Terrain Models with vertical resolution of about 20 cm were developed and used for the 2-D numerical model calibration. The set of datasets to implement a system such as ANFAS include: Terrain topography (DTM, soils, roughness, etc. Hydraulic information on the river (cross sections, basins, dykes etc.) Infrastructure (roads, railways, bridges) - Socio-economic information (needed for impact assessment) On each pilot site a scenario to be tested was specified. A scenario can include assumptions about the hydrologic conditions (maximum discharge), and/or assumptions about changes in the hydraulic structures (dykes, spillways etc.) or the terrain (new road, bridge etc.). A complete discussion on the various scenarios to be implemented can be found in Courtois et. al. (2002). In Loire the ANFAS system is tested and validated in the “Val d’Ouzouer” area located in the Middle Loire valley. Length of the river is about 30 km and total area is 132 km². Part of the data used for estimating CARIMA were available from an earlier attempt to implement an 1-D model in the area. However, these could not be used for implementing the 2-D model. The DTM was very coarse and therefore a precise DTM, with vertical resolution of 20 cm, was obtained through aerophotography carried out in July/August 2001. Land uses, soil classification (used to estimate roughness), and location of several hydraulic structures were obtained from the interpretation of optical (SPOT) and radar (Radarsat and ERS) satellite images. The scenarios tested in Loire are related to the functioning of the fuse dyke under different hydrologic conditions. A collapse of the dyke results in flooding the Ouzouer valley, but reduces the risks of flooding in the downstream urban valleys. Since the dyke was constructed in 1886 there is a concern that the soil is so compacted that there will be a delay before its destruction. Hence the ANFAS system will be used to estimate flooding under different assumptions about how long it will take for the fuse dyke to collapse. The hydrological conditions to be used correspond to discharge levels of 50, 70, 100, 170 and 200 years flood return periods. The part of the Vah river between the Hricov and Nosice dams is the pilot site in Slovakia. This represents a stretch of 37 km of river with a floodplain of 0.5 to 1 km width. Total area is 75 km2. No efforts have been made in the past for developing a flood model for the area. Water Research Institute and Vah River Authority carried out an extensive data compilation to assemble the needed datasets for implementing the FESWMS model. 29.08.02, 5821 EI P 235 I2 PrastacosP.doc Copyright © IGU/ISEP, Wien 2002. ISBN: 3-9500036-7-3 232 A very accurate DTM was prepared through a LIDAR (Light Detection And Ranging) survey. The grid size was 1 x 1 m and average vertical error less than 15 cm. In the scenarios to be tested discharge levels corresponding to 2, 5, 10, 20, 100, 200 and 500 years flood return periods will be used. The interest is to map flood extent and perform impact assessment analysis. Additionally, the system will be used to test the vulnerability to floods of a new highway to be constructed in the left bank of the river. The pilot site in China is the Jing Jiang Reach in the Yangtze River (Figure 2). In the past 150 years, all of the huge floods in the Yangtze River affected the Jing Jiang reach, which is 340 km long. The Jing Jiang retention area is a particular feature for flood management. In 1952, an area of about 930 km2 was completely enclosed by dykes in order to be used as potential water retention area (total volume is estimated to be 5.4 billions m3) in case of huge flood events to cut off part of the flow discharge. The entry of water in the retention area is controlled by the 1.1 km long North gate (maximum discharge of 7,700 m3/s), as the outflow to the Dongting lake network is controlled by the South gate (maximum discharge 3,600 m3/s). The decision to open or not the North gate relies on various factors, based on the water level recorded at strategic points on Jing Jiang reach. When the decision of flooding the Jing Jiang retention area is taken, the population is evacuated, and a breach is voluntary open in the Yangtze dyke with dynamite, to permit the water to permit the water to reach the North gate. The North gate has functioned three times during the 1954 flood event, and the 40,000 persons living at this time inside the retention area had to be evacuated. During the 1998 flood event, the 500,000 persons living inside the retention area were preventively evacuated, but finally the decision was to not open the North gate. Fig. 2: The Yangtze River pilot site 29.08.02, 5821 EI P 235 I2 PrastacosP.doc Copyright © IGU/ISEP, Wien 2002. ISBN: 3-9500036-7-3 233 The application of ANFAS in the area concentrates in the Jing Jiang retention area. Because of the complex relationships between the Yangtze, its tributaries and the Dongting lake using flood return periods from upstream is not possible. Since in China past floods are considered critical hydrological events the system will simulate the big floods that occurred in 1954, 1981, 1996 and 1998. To demonstrate the effects after opening of the flood retention area to the water level in the main channel of the Yangtze River, different opening (inflow discharge to the retention area from 2,000 m3/s to full open) will be simulated. 7. Conclusions The ANFAS integrated system represents an effort to develop a decision support system for river flood simulations. Of course, flood simulation is not a new subject; the major innovation that the system brings forward is that it permits end users to perform flood simulations and test “what-if” scenarios without being mathematics or informatics experts. The intricacies of the mathematical models and the necessary data processing are hidden from the users and extensive visualization facilities are provided so that end users can immediately see the results in maps and 3-D animations. ANFAS represents a system developed jointly by a large research team that included a group of scientists from different fields and different countries and continents. It represents a large effort to address the problem of river floods in a multidisciplinary way using experiences from both the European countries and China, the nation where the problem of river floods is the most intense. The results achieved from this collaborative effort prove that environmental problems should be addressed in a multidisciplinary way. Bibliography Courtois, C., Petrovic P., Lucac M. and X. Zhang (2002): Specification of the scenarios to be used for flood simulations, Deliverable D4.3 ANFAS project. Houdry, P., Rhin C. and N. Deverge (2001): System Design Document; Deliverable D5.1, ANFAS project Prastacos, P (2001):. Architectures for distributed information systems for water resources management, in: Systems Analysis Modelling and Simulation, 41 (1), pp. 123-148. Prastacos, P., Courtois, C., Diamandakis, M and P. Petrovic (2001): Data Acquisition and Preparation; GIS data structure Deliverable D1.2/D1.3 ANFAS project. Samuels, P. (1999), RIBAMOD, River Basin Modelling, management and flood mitigation, Final Report, RIBAMOD Project. 29.08.02, 5821 EI P 235 I2 PrastacosP.doc Copyright © IGU/ISEP, Wien 2002. ISBN: 3-9500036-7-3