Abstract. Landslides rank as one of the most damaging natural disasters occurring in the Philippines, particularly in geologically and climatically complex regions such as Cebu Province. Accurate landslide susceptibility mapping is crucial for reducing disaster risks and promoting sustainable land use planning. This research evaluates the effectiveness of two commonly applied statistical models—Frequency Ratio (FR) and Logistic Regression (LR)—in producing landslide susceptibility maps for Cebu Province. Seven environmental conditioning factors were analyzed: slope, elevation, aspect, topographic wetness index (TWI), topographic position index (TPI), soil texture, and curvature. Landslide inventory data from 2009 to 2023 were compiled from news reports and validated using Google Earth imagery. Both FR and LR models were applied using the same set of factors, and their predictive performances were evaluated using ROC-AUC curve. Results show that both models effectively delineate landslide-prone areas, with slope emerging as the most influential factor. The LR model demonstrated marginally higher predictive accuracy with AUC of 0.9151 compared to the FR model with AUC of 0.8955, due to its ability to account for multivariate interactions among factors. The map produced by LR was compared to the existing map from MGB, and 257,142.11 hectares were found to fall within the agreement zone, having the same classification. The resulting susceptibility maps provide a scientific basis for local government units to enhance disaster preparedness, guide land use decisions, and prioritize risk mitigation efforts in Cebu Province.
Sañez et al. (Tue,) studied this question.