Integrating Dempster Shafer theory, certainty factors and topographic indices for landslide susceptibility analysis in Ha Quang district, Cao Bang province

Abstract

Ha Quang District, located within the Non Nuoc Cao Bang Geopark, is an area of high geological, ecological, and cultural significance and has been recognized by UNESCO as a member of the Global Geoparks Network. However, the region is highly susceptible to geological hazards, particularly landslides. This study applies Dempster-Shafer (DS) theory and the Certainty Factor (CF) method to evaluate landslide susceptibility using a Geographic Information System (GIS). A total of 196 landslide events were identified from historical records, Google Earth imagery, and field surveys to establish a landslide inventory map. Seven conditioning factors, including slope, Topographic Roughness Index (TRI), Topographic Wetness Index (TWI), Stream Power Index (SPI), Mass Balance Index (MBI), Normalized Difference Vegetation Index (NDVI), and rainfall, were integrated into the analysis. The belief map generated from the DS model was evaluated using receiver operating characteristic (ROC) analysis and the area under the curve (AUC), achieving an overall prediction accuracy of 74.5%. For comparison, the CF model yielded a success rate of 67%. The results demonstrate that the DS approach provides superior predictive performance by effectively handling uncertainty, thereby improving landslide susceptibility assessment compared to traditional GIS-based methods. The findings provide a valuable scientific basis for landslide risk mitigation, land use planning, and sustainable infrastructure development in mountainous regions.