Landslide Susceptibility Modeling Using a Machine Learning Algorithm in Halila Catchment, Gamo Zone, Southern Ethiopia.

Abstract

Landslides are major disasters causing significant property damage and loss of life. This study focuses on identifying the areas most vulnerable to landslides and the key factors contributing to landslides in the Halila catchment of the Gamo zone using machine learning algorithms. The primary goal was to determine the causative factors and develop landslide susceptibility maps using support vector machine (SVM) and random forest (RF) algorithms. Based on Google Earth imagery, approximately 302 landslides were identified and randomly divided into training (70%) and validation (30%) datasets within the ArcGIS environment. Ten landslide-related factors were selected: slope, aspect, curvature, distance from streams, and distance from lineaments, lithology, land use/land cover, soil type, elevation, and rainfall. The resulting landslide susceptibility maps were classified into very low, low, moderate, high, and very high susceptibility classes using both support vector machine and random forest. These maps were validated using the receiver operating characteristic (ROC) curve; with the area under the curve (AUC), the results indicate that the SVM model achieved a success rate of 0.816 and a predictive rate of 0.809. In comparison, the Random Forest (RF) model demonstrated slightly better performance with a success rate of 0.821 and a predictive rate of 0.811. To mitigate the impact of landslides, recommendations include implementing strict land-use planning and zoning regulations, enforcing and updating building codes, stabilizing vegetation and slopes, and installing effective drainage systems.