Landslide susceptibility mapping using a machine learning approach and different environmental factors in the Western Ghats region (India)

Abstract A major landslide‐related concern in the Western Ghats is the increasing terrain instability resulting from the interaction between intense rainfall—aggravated by climate change—and inadequate environmental and urban planning. This study develops a detailed landslide susceptibility map for the Western Ghats region using a machine‐learning approach and multiple environmental factors. Four machine learning models—Random Forest (RF), Support Vector Machine (SVM), Stochastic Gradient Boosting (SGB) and Bagging—were applied, and model performance was validated using the Area Under the Curve (AUC) metric. Landslide susceptibility was assessed based on 22 conditioning factors, including topographical factors (elevation, slope, aspect and curvature), geological factors (lithology, lineament density and proximity to faults), hydrological factors (drainage density and proximity to rivers), climatic factors (rainfall and vegetation vigour), anthropogenic factors (proximity to roads, population density and land use) and other variables (seismic activity and soil type). The results showed that RF achieved the highest AUC (0.968), followed by SGB (0.959) and Bagging (0.943). The spatial analysis identified areas highly prone to landslides. District‐level factors such as deforestation, slope modification and heavy rainfall were significant contributors, with 72% of landslides occurring near roads. The susceptibility maps generated can support regional planning, early warning systems and disaster mitigation strategies by emphasising reforestation, regulated construction and targeted monitoring of high‐risk zones.