Modeling and mapping landslide susceptibility zones using GIS based multivariate binary logistic regression (LR) model in the Rorachu river basin of eastern Sikkim Himalaya, India

Multivariate binary logistic regression (LR) model was used for the assessment of landslide susceptibility in the Rorachu river basin of eastern Sikkim Himalaya. For this purpose, a spatial database of 13 factors such as rainfall, slope, aspect, curvature, relief, drainage density, distance from drainage, distance from lineament, distance from road, geology, soil, Normalized Difference Vegetation Index (NDVI), and land use/land cover was constructed under Geographical Information System (GIS) environment. A landslide inventory map was prepared and converted into binary raster coded by 0 for absence and 1 for the presence of landslide. Total 946 landslide pixels were found out of which 725 landslide pixels (76.63%) were used as training dataset for the model and the model was validated using all landslide pixels. The coefficient value of geology was maximum followed by NDVI, soil and land use/land. The calculated probability value was used as Landslide Susceptibility Index (LSI). On the basis of the LSI value, the landslide susceptibility map was divided into five distinct categories of very low, low, moderate, high and very high susceptibility. The susceptibility classes depict that, 9.18% of total area was under very high and high susceptibility contains 75.48% area of total landslide. Finally the accuracy of the model was assessed by area under curve (AUC) of receiver operating characteristics (ROC) curve and landslide density method. The AUC value of 0.947 indicates a very good quality of the model and landslide density shows that the density is increasing with LSI classes.