Evaluating the Performance of Multi-criteria Decision-making Techniques in Flood Susceptibility Mapping

Abstract

Abstract Performances of multi-criteria decision-making techniques in prediction of flood susceptibility are varied. We evaluated performances of ARAS, CODAS, COPRAS, EDAS, MOORA, TOPSIS, VIKOR, and WASPAS in predicting flood susceptibility of Barpeta district of Assam, India. Elevation, slope, proximity to river, geomorphology, drainage density, rainfall, land use/land cover, lithology, soil, stream power index, topographic wetness index and plan curvature were used as flood conditioning factors. The results show higher flood susceptibility over areas characterized by gentle slopes, low elevation and high proximity to drainage. Performances of the models were evaluated using 216 locations (flood and non-flood conditions) randomly classified into training (70%) and validation (30%) through area under receiver operating characteristic (ROC) curve (AUC). TOPSIS model showed better success (AUC = 0.965) and prediction rate (AUC = 0.962) than other models. Among the best performing models, highest percentage of area under high flood susceptibility was predicted by TOPSIS. Therefore, TOPSIS can be effectively used for flood risk management in areas having similar geographical conditions.