Investigation of effectiveness of different analytical hierarchy process methods in landslide susceptibility mapping

As with all natural disasters, studies on landslides are important steps to be prepared for landslides and increasing resilience to landslides. Different methods, such as Analytical Hierarchy Process (AHP), frequency ratio and logistic regression etc., are used in landslide susceptibility analysis, which is an important stage of landslide studies. In this study, a new, more objective and data-driven approach has been conducted by integrating fuzzy relations to AHP method in preparing landslide susceptibility mapping studies. At the first stage, the landslide inventory map of the study area, covering approximately 300 km2 within the borders of Seydikemer district of Muğla province, was updated and the current data were transferred into Geographical Information System (GIS) environment for the analyses. Totally 10 parameters maps, such as topographic elevation, slope, curvature, aspect, sediment transport capacity index, stream power index, topographical wetness index , distance to fault, distance to drainage and distance to ridges, have been prepared for the construction of the database. F-AHP (Fuzzy AHP) methods such as Fuzzy Extended Analysis Method (FEA) and Fuzzy Geometric Mean Method (FGM), are the combination of AHP and fuzzy logic methods and have been successfully used in preparing landslide susceptibility maps. Apart from these methods, in this study, the fuzzy relations were combined with AHP to prepare a data-driven model and its applicability has been investigated. Instead of using expert opinion, data-driven binary comparison values determined by fuzzy relationships were also used in preparing landslide susceptibility maps with FGM and FEA methods in order to make reliable comparisons. Additionally, a landslide susceptibility map was also prepared according to the Modified Analytical Hierarchy Process (M-AHP) method, in which the expert determined the parameter priorities at the beginning of the modeling. According to the performance evaluation of the four landslide susceptibility maps prepared, AUC values were calculated as 0.747 for FEA, 0.739 for FR-AHP, 0.738 for FGM and 0.727 for M-AHP, respectively. There are some limitations in evaluating the effects of some parameters used in landslide susceptibility analyses according to expert opinion. In the light of these data obtained, the FR-AHP method has been evaluated as a data-driven and more objective method that can be used in the preparation of landslide susceptibility maps with its applicability and high performance value.