Flood Susceptibility Mapping Using a Support Vector Machine Models (SVM) and Geographic Information System (GIS)

Abstract

Preparing a flood susceptibility map is necessary and the first step in reducing the damage caused by floods. Due to a lack of information in most of the basins, many researches uses data mining techniques for hydrological studies, especially floods. The aim study is to identify areas with flood susceptibility using a support vector machine (SVM) in the Nekaroud basin. For this purpose, 12 geomorphologic, hydrological and physiographic parameters including slope, aspect, elevation classes, temperature, land use, rainfall, density and distance from the fault, density and distance from the drainages, density and distance from the road, which are provided in the ArcGIS, SAGA GIS and ENVI software’s environments. The GPS device was also used to acquire flood points. Finally, all variables and flood points were entered into the R software in ASCII format with the same pixel size (12.5 m). To evaluate model accuracy, ROC was used in the R software environment. The results of the evaluation showed that the SVM model has good accuracy in identifying flood susceptibility areas in the study area. In addition, the results of this study showed that flood susceptibility areas are more in the northern and northwest regions of the basin and in portions where the concentration of human settlements is higher, while the central regions of the basin with dense vegetation have a low sensitivity to flooding. The results of this study can help planners and researchers to do appropriate actions to prevent and reduce future flood risks. It can also be used to identify suitable and safe areas for construction development.

Keywords

• flood susceptibility map
• support vector machine
• data mining
• roc
• nekaroud basin