GIS based landslide susceptibility mapping using bivariate approach in Gozamin area, Northwestern Ethiopia

Abstract

Gis based landslide susceptibility mapping using bivariate approaches in Gozamin area, northwestern Ethiopia. The present study landslide susceptibility mapping was conducted in Gozamin area located at northwestern Ethiopian highland, about 325 km from the capital city Addis Ababa. The objective of the present study is the delineation of landslide prone areas and the development of a susceptibility map for the Gozamin area using GIS based Frequency ratio and Weight of evidence approaches. A total of 112 landslide inventories were collected and mapped through extensive field investigation and Google earth interpretation, which 79 (70 % landslide locations) were randomly selected for training the models, and the remaining 34 (30 % landslide locations) were used for validating the models. Slope, aspect, elevation, curvature, lithology, distance to drainage and lineament, relief, rain fall, landuse landcover and NDVI were landslide causative factors that used. Slope, aspect, elevation, curvature and distance to drainage were derived from DEM and relief was extracted from topographic map (1:50,000). Distance from river is extracted by multiple buffer ring. Lithology was prepared from field investigation, landuse landcover, and NDVI was extracted from sentinel 2 satellite image and rain fall map was calculated from tropical rain fall measuring mission (Trmm) data. The probability (FR) and weight contrast (WoE) of factors on landside occurrence were calculated and landslide susceptibility maps were produced in both model. The degree of influencing factors is different on landslide occurrence. The most influencing factors are distance to lineament, rain fall, relief, lithology, distance to river and aspect. Because of landslide in the study area people were displaced, houses were swept and buried, crops and cropland were destructed. The landslide susceptibility mapping in this study was validated landslide inventory data. Out of 34 landslide in the study area 82.5% falls on high and very high susceptibility zones, the remaining 17.5 falls on moderate and low classes for Frequency Model and 67.7% on high and very high susceptibility zones the rest 32.34 falls on moderate and low classes in Weight of Evidence model. This indicating that landslide susceptibility mapping using FR model is more accurate than WoE model for the study area