Comparison of Analytical Hierarchy Process (AHP) and Fuzzy AHP methods upon landslide susceptibility mapping using geospatial techniques

The implementation of effective landslide risk planning and mitigation strategies, essential for the safety of property and people, relies on reliable and accurate vulnerability maps. The devastating landslides of October 8, 2023 at Mbankolo, Yaounde (Cameroon), posed major challenges, requiring strong measures to mitigate future impacts. This study aims to map, for the first time, the areas likely to be affected by landslides by overlaying the landslide inventory with the triggering factor maps, using Analytical Hierarchy Process (AHP) and Fuzzy AHP, and to compare the performance of the methods in methodically describing the landslide susceptibility of the complex topography of Yaounde. “Landslide conditioning factor analysis using area under the curve (AUC) method,” rarely used in previous studies, refined the results. Analyses of the predictive models obtained after weighting the criteria showed that both models similarly define the spatial relationship between landslides and causal factors. Validation of the results using the success rate curve indicates that FAHP model (AUC = 84.50%) describes landslides slightly better than AHP model (AUC = 83.60%). Similarly, for complex topography of Yaounde, rigorous statistical (Chi-square and Kappa coefficient) studies demonstrate a very similar zoning of the models; a pixel classified as "high" by fuzzy AHP has a high probability of also being classified as such by AHP. This demonstrates that FAHP method can be used to develop relevant regional units for future mitigation and development strategies. This study thus provides a point of reference for decision-makers and land-use planners for mitigating landslides in the region.