Presentation of new ensemble method of Bayesian and logistic regression models in landslide susceptibility assessment in the Khalkhal Township

The aim of current research is to assess of landslide susceptibility in the Khalkhal Township, southern Ardabil using an ensemble and new method namely Bayesian and logistic regression (BT-LR) models. At first, landslide inventory map was prepared and then effective factors on landslide occurrence were identified. These factors are slope degree, plan curvature, slope aspect, elevation, landuse, lithology, distance from fault, distance from river, distance from road, and drainage density. In the next step, weight of factors and their class were calculated by logistic regression and Bayesian theory, respectively. Finally, landslide susceptibility map produced by hybrid of BT-LR model were divided to four susceptibility classes such as low (24.64%), moderate (25.95%), high (24.44%), and very high (24.94%). Due to accuracy of the mentioned map was used of 30% landslide locations while don’t applied in modeling by ROC curve. Results of accuracy showed that the hybrid model presents a good accuracy with area under the curve (AUC) value of 80.70%. So, it proposed to apply of this landslide susceptibility map for landuse and regional planning in the Khalkhal Township.