Landslide Susceptibility Analysis in Semen Bench District, South West Ethiopia

Abstract

Natural disaster management like landslide requires spatial information as a pillar of awareness and mitigation process. In that perspective, analysis of landslide susceptibility becomes crucial. Semen Bench district in south western Ethiopia is one of such prone and frequently disrupted by landslide due to its geo environmental condition. Therefore this study aims in investigating landslide susceptibility using GIS and remote sensing technologies. The logistic regression was employed to find the best fitting function to establish the relation between the landslide event and a set of nine predictive variables (Slope, Aspect, Elevation, Curvature, Drainage network, Rainfall, Land use, distance from Fault and Lithology). For ArcGIS susceptibility mapping, 160 pixels of dependent variable were generated as dichotomous variable (0 for absence and 1 for presence of landslide). The thematic map preparation and exporting to ASCII file was completed in the GIS environment. Five causative factors (Fault, Lithology, Aspect, Land use and Slope) became the statistically recognized as major influencing predictive parameters of landslide occurrence in the district. The coefficients values of the major factors are used in the spatial probability calculation by raster calculator. The final prediction result ranges from 0 to 0.93. Accuracy assessment for the prediction of the logistic regression was established by ROC AUC curve analysis. The value shows 0.909 which is quite acceptable. The estimated spatial probability value was then classified into four susceptibility classes: very low, low, medium and high occupying 14.8%; 24.15 %; 28.2% and 32.85% of the study area respectively. This result implies one third of the target area is under critical condition which needs serious land use planning and further investigations.