Landslide land capability zonation using Dempster-shafer and frequency ratio models

Landslides are significant natural geologic hazard around the world. Expansion of urban and man-made structures into potentially hazardous areas leads to extensive damage to infrastructure and occasionally results in loss of life every year. Identification of factor affecting existing of landslide as well as its zonation in the given watershed is one of the basic tools for landslide control and selection of appropriate and effective solution as well. Thus, a research study with objective of recognizing factor affecting landslide and determination of lands with hypothential to its occurrence was conducted to prepare landslide zonation map for the Sorkhoon watershed using Dempster-shafer and frequency ratio models.
To reach this goal, after preparing of Landslide inventory map using field survey and aerial photo interpretation, data layers of distance from stream, distance from faults, elevation, slope, aspect, Topography wetness index (TWI), distance from roads, land use, lithology and Stream Power index (SPI) as Factors affecting landslides were selected and after applying Dempster-shafer and frequency ratio methods the final Landslide Hazard zoning was prepared. For calculating of weight of affective factors, was used the analytic hierarchy process in the software of expert choice. To validation of used methods the ROC curve was used.
The main factors that caused the landslides in this area based on field observations and Expertise opinions include lithology, distance from roads and slope, respectively, with scores (0.181, 0.163, 0.145) and vis-à-vis factors of elevation, slope and topography wetness index (TWI) respectively with scores (0.018, 0.036, 0.054) have the lowest impact on landslides. According to the results, frequency ratio models have obtained higher AUC (0.927) as compared to the Dempster-shafer (0.858) that shows the high correlation between hazard map and distribution map landslide inventory map and better evaluation of frequency ratio toward Dempster-shafer model.
The results of the validation showed that the frequency ratio model has higher efficiency and accuracy toward Dempster-shafer model for preparing of zonation map. Based on the results of the frequency ratio model 21128200 square meters (7.05 percentage) of the region located in the very low risk class, 67,144,500 square meters (20.45 percentage) of the area located in the low risk class, 90,113,400 square meters (27.45 percentage) located in the moderate Risk class, 91733400 square meters (27.94 percentage) of the area located in the high class and finally 56.160000 square meters (17.11 percentage) of the area located in the very high risk class.