Landslide Susceptibility Zonation Using Geospatial Techniques: the case of Tello woreda, Ethiopia.

Abstract Landslide is a common disaster in highlands of Ethiopia including the present study area, Tello woreda, SNNPR, during rainy seasons. To reduce such hazard landslide susceptibility zonation is a fore head step. It describes the likelihood or probability of new landslide occurrence in an area based on the assumption past and present is a key for future. With the aim to classify the study area in to landslide susceptibility zones, Logistic regression model was integrated to GIS techniques. Data was acquired from field survey, remote sensing and from different organizations. 178 landslide polygons were created from field and Google earth data. Eight parameters Slope, Aspect, curvature, Elevation, Land use/cover, Lithology, road network and River stream network were included in this study. By combining the rasterized landslide map and causative factors, landslide distribution and densities of parameter classes were calculated through BSA. Equal number of nonlandslide points (from landslide free area) to landslide centroids was randomly selected as a sample for LR model. Using this data Multicollinearity between parameters was analyzed. The result from Multicollinearity analysis shows that, no interdependency between parameters; all are significant. Next to that on R Stastical software the data set was subset into two as training and testing datasets. Using the training data set intercept and coefficients for parameters were generated. The roc curve drawn by using the validation dataset indicates 89.34% area under curve. This shows as performance of the model is very good. Elevation class 1500-2000m has highest influence for the occurrence of landslide in the area. Lastly, intercept and coefficients from R software were taken to GIS environment , susceptibility map was generated and then reclassified to 05 susceptibility classes as 37.08 % very low, 23.98 % low, 14.78 % moderate, 11.56 % high and 12.61 % falls in very High susceptibility classes. Finally, the LSM produced can be used by decision-makers for land use planning and landslide mitigation purpose.