Landslide Mapping and Susceptibility Assessment of Chittagong Hilly Areas, Bangladesh

Landslides are natural phenomena in mountainous areas that cause damage to properties and death to people around the world. In Bangladesh, landslides have caused enormous economic loss and casualty in Chittagong Hilly Areas (CHA). In this dissertation, a landslide inventory of CHA was prepared using Google Earth and field mapping. Google Earth-based mapping helped in recording landslides in inaccessible areas like forests. In contrast, field mapping helped in mapping landslides in accessible areas like areas near road networks. For absence data sampling of landslide susceptibility mapping, this research proposed the Mahalanobis distance (MD) based absence data sampling and compared it with the slope-based absence data sampling. Three Upazilas (subdistricts) of Rangamati district, Bangladesh was used as the study area. Fifteen landslide causal factors, including slope aspect, plan curvature, and geology, were used in the random forest model for landslide susceptibility mapping. The area under the success and prediction rate curves, statistical indices including the Kappa index, showed that both the absence data sampling method provided similar accuracy. But based on the Seed Cell Area Index (SCAI) MD based landslide susceptibility map was more consistent and did not overestimate the landslide susceptibility like the slope-based model. Finally, this study assessed the impact of three land use/land cover (LULC) scenarios: a. existing (2018); b. Proposed LULC (Planned); and c. Simulated (2028) LULC on landslide susceptibility of Rangamati municipality of Rangamati district. The random forest model was used, and it showed that high susceptibility zones would increase in both proposed and simulated LULC scenarios. It indicated that LULC change would increase the landslide susceptibility in the study area. The increase of landslide susceptibility is comparatively low in the proposed LULC, indicating the importance of implementing planned LULC in the study.