Evaluating the Different Statistical Models for Flood Susceptibility Mapping in Guilan Province

Due to the lack of information in most of the watersheds, many researchers attempt to use spatial analysis within Geographic Information System (GIS) in hydrological and Flood Prone (FP) area studies. The present study was designed to compare the efficiency of three models i.e. Support Vector Machine (SVM), Generalized Linear Model (GLM) and Generalized Additive Model (GAM) for preparing the flood susceptibility mapping in Guilan province, Iran. For this purpose, slope, aspect, plan curvature, elevation, distance from the river, drainage density, geology, land use, Topographic Wetness Index (TWI) and Stream Power Index (SPI) layers were derived in GIS (ArcGIS and SAGA-GIS). Using 220 flood locations, 70% and 30% out of total flood locations were then used to calibrate and to validate the performance of the models, respectively. The evaluation results of the models accuracy using the area under the curve (AUC) and Kappa indices showed that in terms of AUC, the SVM with 0.835 and the GAM with 0.827, and the GLM with of 0.79 performed very good and good classes, respectively. In terms of Kappa index, the SVM with 0.58, GAM with 0.53 and GLM with 0.48 are performed good and acceptable classes, respectively. Therefore, based on the mentioned indices, the SVM superior to other two models for identifying the flood susceptibility areas.