Bivariate statistical index for landslide susceptibility mapping in the Rorachu river basin of eastern Sikkim Himalaya, India

Abstract

The main goal of the study is to prepare a landslide susceptibility map under Geographical Information System (GIS) environment using statistical index model to identify and demarcate the areas of future landslide occurrence. Firstly, landslide locations were identified with the help of previous reports, satellite images and intensive field study. For the preparation of landslide inventory, 80 landslide locations were identified and randomly separated to create training and validation datasets. Fifty landslides (62.5%) were used as training dataset and remaining 30 landslides (37.5%) were used for validating the model. Twelve landslide conditioning factors, including morphometric factors (slope angle, slope aspect, curvature, relative relief and drainage density) and non-morphometric factors (bedrock geology, soil, distance from drainage, distance from lineament, distance from road, Normalized Difference Vegetation Index (NDVI) and land use/land cover), were used to generate landslide susceptibility map of Rorachu river basin. Finally, the accuracy of the model was assessed by area under curve of Receiver Operating Characteristics (ROC) curve and landslide density method. The statistics of ROC curve showed that, the landslide susceptibility map using statistical index model has an accuracy of 91% which indicates a very good predictive capacity of the model. The result reveals that, landslide density of the Rorachu river basin is increasing with landslide susceptibility classes.