BUILDING A PROBABILITY FLOOD RISK MODEL, USING GIS, LOGISTIC REGRESSION AND FUZZY WEIGHTS

As flash flood events are becoming more frequent, the development of flood risk estimation methods is becoming more important than ever before. The aim of this paper is to develop a methodology in order to study the flood risk in a river catchment by taking into account various parameters that are organized in a spatial database (geological, geomorphological, topographic and land use data). The collected data are analyzed through the use of G.I.S software (ArcMap) in order to generate the appropriate steps for the formation of the flood risk model. Finally, we proceed to develop a probabilistic flood risk model (based on the logistic regression) using as variables the lithology, the slope, land use, average basin altitude. All the above variables are multiplied by the proper weightsin the form of triangular fuzzy numbers. These weights are related to the importance of each involved variable. This model has been applied in the area of Samos Island, Greece with success. The proposed methodology and the preliminary results, as exported for the entire island prove the suitability of this method in the creation of flood risk maps.