Assessing the impact of RCP4.5 and RCP8.5 scenarios on landslide susceptibility mapping using support vector machine: A case study of Penang Island, Malaysia

Abstract

This paper investigates the climate change influence on landslide susceptibility mapping (LSM) using a case study conducted on Penang Island in Malaysia, a region prone to landslides. This study was carried out due to limited research assessing the climate change effect on LSM, considering rainfall and temperature. The study employs climate factors, including temperature and rainfall, alongside 12 causative factors in a Support Vector Machine (SVM) model to develop LSM. The Statistical Downscaling Model (SDSM) is used to derive climate change projections under two Representative Concentration Pathways (RCP) scenarios (RCP4.5 and RCP8.5). Data preparation and normalization are performed using ArcGIS 10.7. Based on the results, future annual rainfall and daily temperatures are expected to rise under both scenarios, with RCP8.5 exhibiting more significant climatic changes. The LSM zonation is impacted more significantly under RCP8.5 due to the severity of climate change. LSM under an observation period achieves the best results (area under the curve (AUC) = 85.75, average statistical index (SI) = 94.48%, kappa = 0.885), followed by LSM under RCP4.5 (AUC = 84.38, average SI = 93.54%, kappa = 0.865) and LSM under RCP8.5 (AUC = 84.13, average SI = 93.34%, kappa = 0.860), demonstrating their reliability and adequate performance. These LSMs can assist local authorities in designating critical areas for monitoring and implementing an early-warning system to respond more effectively to landslide risks caused by climate change. However, the study's limitation is considering only two climate scenarios (RCP4.5 and RCP8.5). Future research should encompass a broader range of climate scenarios to develop the most reliable LSM, given the high uncertainty associated with climate change.