Landslide susceptibility mapping using geospatial, analytical hierarchy process (AHP), and binary logistic regression (BLR) techniques – A study of Wadi Habban Basin, Shabwah, Yemen

Wadi Habban Basin is continuously prone to numerous geological hazards, such as landslides. This study aims to generate landslide susceptibility maps by integrating of remote sensing, Geographic-Information-System (GIS) techniques, Analytical-Hierarchy-Process (AHP), and Binary -Logistic-Regression (BLR) models. In this study, multiple datasets were utilized for delineating landslide susceptibility maps, including slope, elevation, lithology, aspect, proximity to faults, proximity to drainage, proximity to roads, proximity to lineaments, geomorphology, soil texture, rainfall, land use/land cover, Normalized Difference Vegetation Index, curvature, and stream power index. Spatially distributed maps and thematic layers for all the aforementioned parameters were generated using a combination of remote sensing data and ground-based observations within a GIS environment. A comparative analysis of the AHP and BLR models was conducted to evaluate their predictive capability. The BLR model classified 49 % of the area as high to very high susceptibility, compared to 26 % by AHP, and showed a stronger delineation of low-risk zones. ROC curve analysis indicated high predictive accuracy for BLR than AHP models, with AUC values of 90.4 % for BLR and 81.7 % for AHP. Validation using confusion matrices demonstrated that the (BLR) model achieved an overall accuracy of 91.5 %, with a precision of 93 % and a recall of 91 %. In comparison, the (AHP) model yielded an overall accuracy of 75 %, a precision of 87.5 %, and a recall of 75 %.Results confirm the robustness of the BLR model for effective landslide susceptibility mapping and highlight its potential for risk informed planning. This study provides valuable insights for disaster risk reduction, sustainable land-use management, and the application of targeted mitigation strategies in landslide-prone regions.