Landslide susceptibility mapping at central Zab basin, Iran: A comparison between analytical hierarchy process, frequency ratio and logistic regression models

The purpose of this study is to compare the landslide susceptibility mapping models of logistic regression (LR), analytical hierarchy process (AHP) and frequency ratio (FR) applied in the central Zab basin (West Azerbaijan-Iran). Eight factors were used for landslide susceptibility mapping including slope, aspect, land cover, precipitation, lithology and the distance to roads, drainage, and faults that affect the occurrence of landslides. To get more precision, speed and facility in our analysis, all descriptive and spatial information was entered into GIS system. Satellite images (Landsat ETM + and SPOT 5) are also used to prepare for land use and landslide-inventory mapping respectively. Landslide events as used as dependant variable and data layers as independent variable, making use of the correlation between these two variables in landslide susceptibility. The three models are validated using the relative landslide density index (R-index) and the receiver operating characteristic (ROC) curves. The predictive capability of each model was determined from the area under the relative operating characteristic curve and the areas under the curves obtained using the LR, AHP, and FR methods are 0.8941, 0.8115, and 0.8634, respectively. These results indicate that the LR and FR models are relatively good estimators of landslide susceptibility in the study area. The interpretations of the susceptibility map reveal that precipitation, lithology and slope played major roles in landslide occurrence and distribution in the central Zab basin. In general, all three models were reasonably accurate. The resultant maps would be useful for regional spatial planning as well as for land cover planning.