A Comparative Study Of Statistical And Machine Learning Approaches For Landslide Susceptibility Zonation In A Hilly Region

Abstract

The North-Western Himalayas are prone to Landslides due to extreme climatic conditions and human interventions. These areas are particularly susceptible to disasters due to steep slopes, complex geological features, heavy rainfall, and urban expansion. It's crucial to identify the areas that are most at risk to manage disasters effectively and plan for the community's safety. This study uses a GIS approach to analyse seventeen major factors contributing to landslides, including landscape, environmental conditions, and human activities. By employing methods such as Information Value and Weight of Evidence, the study examined the relationships between these factors and past landslides and refined the predictions using the Frequency Ratio. The AUC-ROC curve analysis of the validation with historical data shows that the resulting susceptibility map is a valuable tool for guiding development and mitigating disaster risks in these vulnerable areas. Future research will quantify the effects of anthropogenic influences on slope instability to better understand human-induced contributions to landslide risk in rapidly urbanizing mountain environments.