Landslide susceptibility mapping using weights of evidence model on the slopes of mount Elgon, Eastern Uganda

Abstract

The slopes of Mount Elgon in Eastern Uganda have become a hotspot for landslide disasters. A GIS-based bivariate method, particularly the weight of evidence model which is a quantitative statistical method, was applied to map landslide susceptibility in the Mount Elgon region. This study was aimed at evaluating the susceptibility of Bukalasi milli-watershed on the slopes of Mount Elgon to landslides, as an early warning strategy. A landslide inventory for the study area was prepared, and the weights of influence of selected landslide conditioning factors were calculated to present their relative importance in landslide susceptibility. Eight conditioning factors were used in this study namely; land use, lithology, rainfall, elevation, slope aspect, slope angle, plan curvature and profile curvature. Following the results of the Agterberg-Cheng test for conditional independence (probability = 62.5%), the hypothesis of conditional independence among the eight factors was accepted. Validation using the area under the receiver operating characteristics curve indicate satisfactory accuracy of the prediction rate (AUC=0.882) and success rate (AUC= 0.912) of the model. The final landslide susceptibility map highlights high susceptibility in the southern and western parts (22.50%) of the study area. It further shows that whereas Bukibumbi, Bundesi and Suume parishes are the most prone parishes as indicated by their classification under the high and very high susceptibility classes, Shibanga Parish is relatively the least prone to landslides disasters. It is, therefore, recommended that highly susceptible areas identified in this study should be prioritised during intervention programmes such as government relocation efforts. Furthermore, deforestation should have severe penalties since this study indicates that presence of forests reduce landslide susceptibility.