Geo-spatial approach-based landslide susceptibility mapping using analytical hierarchical process, frequency ratio, logistic regression and their ensemble methods

Abstract

Kurseong municipality and its surrounding hill slope are famous high altitudinal residential areas in the Darjeeling Himalayan region of India. In Darjeeling Himalayas, huge landslides occur in the rainy season every year. This paper was aimed to delineate landslide susceptibility zone (LSZ) in Kurseong and its surrounding hill slope in Darjeeling Himalayas. Nine landslide inducing parameters (i.e., slope, altitude, rainfall, geological structure, distance from river channels, distance from lineament, soil type or characteristics, land-use/land-cover and aspect) were used for mapping LSZ applying analytic hierarchy process (AHP), frequency ratio (FR), binary logistic regression (BLR) models and their ensemble (AHP–FR; AHP–BLR; and FR–BLR) with the aid of ArcGIS, SPSS and ERDAS software. The landslide susceptible maps produced in GIS environment by the AHP, FR, BLR and their ensemble models were classified into five susceptibility classes such as very low, low, moderate, high and very high. The area under curve (AUC) of receiver operating characteristics (ROC) and kappa statistics were used to analyze the accuracy of the LSZ maps. The AUC values of prepared maps were 78.86% (AHP), 80.22% (FR), 80.67% (BLR), 83.44% (AHP–FR), 84.39% (AHP–BLR) and 84.73% (FR–BLR). Overall kappa statistics were 0.789 (AHP), 0.812 (FR), 0.799 (BLR), 0.837 (AHP–FR), 0856 (AHP–BLR) and 0.868 (FR–BLR), indicating the good accuracy of the models. The mapping of LSZ will be useful for land-use planning and future landslide hazard mitigation strategies.