Landslide Susceptibility Mapping around Alkumru Dam (Siirt, Türkiye) Using Machine Learning and Ensemble Models

This study assessed the landslide susceptibility of the Alkumru Dam Reservoir and its surrounding areas in Siirt Province, Türkiye, using six machine learning algorithms—Support Vector Machines, Logistic Regression, Maximum Entropy, Naive Bayes, Random Forest, and Artificial Neural Networks—alongside an Ensemble model. The analyses were conducted using MaxEnt and R software. A comprehensive landslide inventory and eleven conditioning factors were employed during the modeling process. Model performance was evaluated using the Area Under the Receiver Operating Characteristic Curve (AUC) metric. All models demonstrated high predictive accuracy (AUC = 0.877–0.959), with the Naive Bayes model achieving the highest performance (AUC = 0.959). The spatial distribution of susceptibility maps was highly consistent across all algorithms. High and very-high-susceptibility zones were consistently identified on the southern, western, and southwestern slopes of the Alkumru Dam Reservoir and the right slope of the Botan River Valley—areas characterized by weak lithological units with low mechanical strength. Quantitative validation confirmed that 52.2%-76.8% of mapped landslides fall within high- and very-high-risk zones. The resulting susceptibility maps provide a valuable scientific basis for sustainable land-use planning, disaster risk management, and dam safety in the region.