Comparison of bivariate and multivariate statistical and heuristic-based landslide susceptibility models: An example from Ayvali{dotless}k (Bali{dotless}kesir,Northwestern Turkey) [İki ve çok degişkenli i̇statistik ve sezgisel tabanli{dotless} heyelan duyarli{dotless}li{dotless}k modellerinin karşi{dotless}laşti{dotless}ri{dotless}lmasi{dotless}: Ayvali{dotless}k (Bali{dotless}kesir,Kuzeybati{dotless} Türkiye) örnegi]

Landslides are one of the most destructive natural hazards which frequently occur after earthquakes in our country and in the world. from engineering point of view,prediction of landsliding before its occurence has a great importance to mitigate the landslide related damages,and determination of landslide prone areas by the methods,based on probability,has spread out both in our country and in the world in the last two decades. in this study,a comparison of the most common landslide susceptibility mapping methods,namely bivariate,multivariate statistical and heuristic methods,were carried out. for this purpose,ayvali{dotless}k (bali{dotless}kesir) and its near vicinity were selected as study area,and in total 45 landslides were mapped. morphologic,geologic and land-use data were produced in geographical information systems (gis) by using available topographical and relevant thematic maps. in the area,slope gradient and aspect,lithology,weathering conditions of the rocks,stream power index (spi),topographical wetness index (twi),distance from drainage,density of structural features,land-cover and vegetation cover density were considered as the parameters causing the landslides. all of the parameters were standardized in a common scale by using fuzzy membership functions. then,the contribution of each of these parameters for the landslide occurrence were investigated by likelihood ratio,logistic regression and analytical hierarchy methods,and the weight values of the parameters were calculated. considering the weight values determined by each method,landslide susceptibility maps were produced,and the performances of the produced maps were tested by comparing landslide locations using area under curvature (auc) approach. based on this,the auc values were determined to be 0.76,0.77 and 0.89 for likelihood ratio,logistic regression and analytical hierarchy models,respectively. accorrding to these results,analytical hierarcy model was considered to be the best landslide susceptibility method for the study area.