Chapter 9 - Landslide susceptibility modeling using a generalized linear model in a tropical river basin of the Southern Western Ghats, India

Abstract

The Western Ghats, extending for about 1,500km parallel to the southwest coast, is the most prominent orographic feature of the Indian Peninsula. The unique physiographic-climatic-geological conditions make the western slopes of Western Ghats highly susceptible for the occurrence of frequent and widespread landslides. This research aims to produce landslide susceptibility map of a tropical river basin (Pamba River Basin) of the Southern Western Ghats. For this, a spatial database with 14 geo-environmental variables such as lithology, landuse/land cover, soil texture, slope angle, slope aspect, distance from the roads, distance from the streams, distance from the lineaments, NDVI, terrain ruggedness index, topographic wetness index, stream power index, profile curvature, and plan curvature are developed, and a landslide inventory map of 241 landslide locations is created from high resolution satellite images. Initially, a multicollinearity test is conducted to eliminate highly correlated variables, and accordingly, terrain ruggedness index is omitted from final analysis. Thereafter, generalized linear model, a modified flexible version of ordinary regression is used for final landslide susceptibility modeling. The landslide susceptibility map is classified into five susceptible zones: stable area (47.16%), the least susceptible area (24.77%), moderate susceptible area (13.68%), high susceptible area (9.3%), and extremely susceptible area (5.09%). The predicted result is validated with receiver operating characteristics curve with area undercover values and other statistical measures, including sensitivity, specificity, and accuracy. The predicted model shows an accuracy of 84.7% (AUC 0.847) in training phase and 89.2% (AUC 0.892) in validation phase. The study also reveals that slope angle, stream power index, distance from road, planform curvature, and NDVI are the significant factors, which contribute to the occurrence of landslides in the study area. The predicted map is trustworthy and can be used for future land use planning and hazard mitigation in the study area.