A GIS-Based Comparative Study of the Analytic Hierarchy Process, Frequency Ratio, and Logistic Regression Methods for Landslide Susceptibility Mapping along the Gilgit-Skardu Road

The Gilgit-Skardu road, located in the northern region of Pakistan, traverses a seismically active area due to the ongoing collision between the Eurasian and Indian Plates. This mountainous terrain is particularly susceptible to landslides, necessitating the creation of a landslide susceptibility map for effective hazard management. This study aimed to develop a comprehensive landslide inventory and utilized three GIS-based models—Analytic Hierarchy Process (AHP), Frequency Ratio (FR), and Logistic Regression (LR)—to identify areas at risk of landslides. Twelve causative factors were analyzed, including slope degree, aspect, plane curvature, profile curvature, proximity to roads, distance to streams, fault lines, geology, land cover, rainfall, elevation, and Normalized Difference Vegetation Index (NDVI). The study focused on delineating landslide-prone zones along the Gilgit-Skardu road. The susceptibility assessment was conducted using the three aforementioned GIS-based models, integrating remote sensing and geographic information system (GIS) techniques. The final susceptibility maps were generated based on a landslide inventory comprising 99 active landslides in the region. The AHP, FR, and LR models were evaluated for their predictive accuracy and ability to correlate landslide occurrences with the causative factors. Model performance was assessed using the Area Under the Curve (AUC) of the Receiver Operating Characteristic (ROC) curve, yielding AUC values of 74.96%, 74.16%, and 83.11% for the AHP, FR, and LR models, respectively. The LR model demonstrated superior predictive capability. This optimized landslide susceptibility model offers valuable insights for disaster mitigation and supports authorities in managing development programs in this vulnerable region.