A Multi-Classifier Approach to Landslide Susceptibility Mapping Using Geospatial Data and Machine Learning in the Ghats Regions of Karnataka

Landslides pose a significant threat to life, infrastructure and the environment in the Western Ghats region of Karnataka. This study develops a comprehensive Landslide Susceptibility Mapping (LSM) model by integrating machine learning algorithms Random Forest, CART, Naive Bayes and Support Vector Machine (SVM) with remote sensing and geospatial data. The model incorporates critical geographic and environmental factors such as slope, aspect, elevation and rainfall to predict landslide susceptibility across six districts: Kodagu, Shimoga, Chikkamagaluru, Uttara Kannada, Dakshina Kannada and Udupi. The Random Forest model demonstrated the highest accuracy (94.2%) in predicting landslide-prone areas. Results indicate that Shimoga, Chikkamagaluru and Uttara Kannada have higher susceptibility, particularly in the Moderate and High categories, underscoring the need for targeted mitigation strategies. This study highlights the effectiveness of machine learning in geohazard mapping and provides a scalable approach that can be adapted for other regions vulnerable to landslides.