Flood susceptibility zonation using new ensemble Bayesian-AHP methods (Case study: Neka Watershed, Mazandaran Province)

Flood susceptibility mapping is the first step in flood management programs. Flood prediction can help reducing its following damages. The main objective of this study is identification of prone areas to flooding using new ensemble Bayesian-AHP methods in the Neka-Sari watershed, Iran. Flood inventory map was prepared due to statistical analyses. A total of 240 (70 %) and 102 (30 %) out of 342 observed events were used as training and validation data set, respectively. Based on literature review, as well as extensive field studies, a total of 11 parameters in relation to flood occurrences were selected for flood mapping, including slope percent, elevation, distance to river, drainage density, NDVI, lithology, land use, topography wetness index (TWI), stream power index (SPI), rainfall, and curvature. The weights of each factor were determined by AHP method. Also, the relation between factor classes and flood events and the weight of each lasses were estimated using Bayesian theory. Finally, by integration of factors and their classes in ArcGIS, flood susceptibility map was obtained with five classes. In order to evaluate the obtained model, ROC curve was employed. Results showed that the ensemble model had a high accuracy (76.10 %) in flood susceptibility mapping. Also, slope percent, elevation, and land use have the highest effect on flood events with values of 0.260, 0.195, and 0.146, respectively. According to the results, 24.17 and 37.15 % of the study area are categorized in high and very high susceptibility classes, respectively. The presented combined model can be used for further studies on natural hazard mapping and disaster management.