Presenting the results of landslide susceptibility mapping using partial least squares regression model: A case study of Khotang District, Koshi Province

Abstract

Landslides pose a significant threat in mountainous regions globally, causing substantial damage to infrastructure, disrupting livelihoods, and leading to loss of life. This study focuses on the Khotang District, Koshi Province, Nepal, a region highly susceptible to landslides due to its steep terrain, active tectonics and heavy monsoon rainfall. The research aims to assess landslide susceptibility in the district using Partial Least Squares Regression model (PLSR), a robust statistical technique capable of handling complex datasets with correlated variables. The research emphasizes the significant influence of topographic factors on landslide occurrence. Specifically, the Topographic Wetness Index (TWI) and Elevation were found to be the most influential variables, demonstrating the highest importance scores in the PLSR model. The model demonstrated excellent performance in predicting landslide susceptibility, with a balance between fit and generalization. It achieved a testing AUC of 0.740, indicating strong generalization ability and potential for practical applications. The findings of this study indicate the potential use of the PLSR for future landslide susceptibility mapping, owing to its robust predictive power. The study also enhances our understanding of the factors that influence landslide occurrences in the Khotang District. Furthermore, it provides a scientific basis for the implementation of effective mitigation measures to reduce landslide risks.