Akademik Veri Yönetim Sistemi
Natural Hazards, cilt.114, sa.3, ss.2571-2604, 2022 (SCI-Expanded)
This study investigates the effects of hybrid modeling approaches, factor standardization, and categorical mapping on the performance of landslide susceptibility maps. For this purpose, three models of frequency ratio (FR), weights of evidence (WoE), and certainty factor (CF) and six hybrid models based on the combination of FR, WoE, and CF with logistic regression (LR) and analytical hierarchy process (AHP) were used to analyze landslide susceptibility in Van, Turkey. Eleven factors including elevation, slope, aspect, profile curvature, plan curvature, lithology, land use/cover, normalized difference vegetation index, topographic wetness index, terrain ruggedness index, distance to faults, and distance to roads were used. The analysis results were categorized into five groups as Very High, High, Medium, Low, and Very Low landslide susceptibility using four classification methods of equal interval, natural breaks, geometric interval, and quantile; thus, 36 landslide susceptibility maps were obtained. We compared the accuracy of the maps using area under the curve for success and prediction rates, percentage distribution of landslides, and landslide density index. The results indicated that 26 out of 36 maps had good prediction accuracy but differed in predictive abilities. Furthermore, the comparisons revealed hybrid modeling improved prediction performance, and the AHP-based hybrid models gave better results than the LR-based hybrid models. The CF for factor standardization and the geometric interval and quantile for categorical mapping gave better results than the other methods. In the study area, however, the CF-AHP-EI, based on the categorization of the CF-AHP hybrid model using equal interval, provided the best result. The CF-AHP-EI landslide susceptibility map shows that 5.8% of Van has Very High and 34.4% High landslide susceptibility.