Landslide susceptibility modeling derived from remote sensing, multi-criteria decision analysis, and GIS techniques: A case study in the Southeast Bohol Province, Philippines

Abstract

Abstract Several landslide hazard susceptibility mapping has been done in the Philippines for the past decade to augment the need for vulnerability assessment. Being a karst region, the southeast Bohol province is accustomed to frequent landslide occurrences, causing hazard risks in adjacent communities. These landslides are generally controlled by both extrinsic (e.g., road network, fault) and intrinsic (e.g., geomorphologic and geologic) factors. Field observation denotes several occurrences along steep slopes and stream banks. However, the challenges in updating the landslide hazard susceptibility maps arose from the scarcity of updated field information. With data access limitations, this study aims to generate accurate and precise landslide susceptibility models using remote sensing and statistical-based analysis processed in Geographic Information System (GIS). This study uses open-sourced medium-resolution satellite data and digital elevation models are utilized to generate the 8 landslide factor maps. These maps were analyzed through Analytical Hierarchy Process (AHP) and Fuzzy Logic Overlay. The generated landslide susceptibility models were validated using Forest-based Classification and Regression (FBCR), analysis of variance (ANOVA), and ordinary linear regression. The AHP-based model shows significant accuracy and compatibility with the actual susceptibility of the site since the p-value of the map is 0.031 in ANOVA, while the Fuzzy-based model cannot be considered accurate since it generated a 0.266 p-value result. Moreover, both models were analyzed in FBCR and resulted in a p-value of 2.2x10-16 in the ordinary linear regression validation, making both significant landslide prediction models. This result denotes that medium-resolution satellite data can generate accurate and compatible landslide susceptibility and prediction models, and this process will give disaster risk reduction managers an avenue to generate landslide models that are not bounded by data access limitations.