An Ensemble Modeling of Frequency Ratio (FR) with Evidence Belief Function (EBF) for GIS-Based Landslide Susceptibility Mapping: A Case Study of the Coastal Cliff of Safi, Morocco

To map the susceptibility to landslides in the coastal cliffs of Safi, Morocco, the current study intends to construct and compare a spatial integration of two bivariate statistical models of frequency ratio and evidential belief function with their unique ensembles (EBF-FR). For this, a thorough map of the 206 currently active landslides was created using satellite images and verified in a field study. Then, a random percentage of the landslide inventory was split between the training and testing models, 70% and 30%, respectively. For landslide susceptibility modeling, sixteen conditioning factors including elevation, slope, slope aspect, profile curvature, terrain roughness index, sediment transport index, topographic wetness index, stream power index, topographic position index, distance to rivers, rainfall, lithology, distance to faults, land use/land cover, distance to roads, and a normalized difference vegetation index were taken into account. The generated susceptibility maps were validated using the receiver operating characteristic (ROC) curve and the area under the curve (AUC). The integrated EBF-FR model outperformed both traditional bivariate techniques with a ROC-AUC value of 0.96, and all ensemble EBF-FR and individual models demonstrated higher prediction performance. The slope and the brown clay class in the lithology both considerably contribute landslides occurrence in the examined area, according to the examination of the conditioning factors and their prediction rate (PR). The proposed models' accuracy is deemed to be extremely satisfactory when taking the ROC-AUC into account, and they can be employed for landslide risk management and mitigation in the investigated area.