| dc.description.abstract | Landslides are one of the most widespread natural hazards occurring every year all over the world. In Uganda, they are common in Mt. Elgon areas. To minimize fatalities, there is need to know where landslides are likely to occur. In this study therefore, attempts were made to; characterize landslides basing on magnitude, relate pedologic characteristics of the derived magnitude classes and, map the spatial distribution of the landslide magnitude classes in Sironko and Bulambuli districts. Towards such efforts, a cross sectional study design and a quantitative approach were employed. Using snowball sampling, 45 landslide scars were visited, geocoded and investigated for scar• dimensions, from which data used in creating landslide magnitude classes using cluster analysis in R software version 3.4.4 was derived. At each landslide scar, a description of the soil morphological and physical properties was done. In addition, soil samples were picked for laboratory analysis of chemical properties of interest to this study. This data was compared with the derived landslide magnitude classes. To predict the spatial distribution of landslide magnitude classes, a response variable shape file for landslide classes and a predictor variables' raster file containing 18 layers of terrain, soil and geology data for the study area were prepared in SAGA GIS 2.3.1 and ArcGIS 10.5. Using this data, random forest modeling was implemented in R software and the output, further processed in ArcGIS to map distribution of the predicted landslide magnitude classes. The study shows that the area experiences three classes of landslide magnitude - low, moderate and high magnitude. These categories express themselves differently due to spatial and depth wise variation in soil physical and chemical properties. When the categories were predicted across the area, and using random forest modeling and terrain, geologic and pedologic covariates, it was found out that 507.15 km2 land area is landslide-free whilst 2.5 km2, 205.43 km2 and 5 .14 km2 land area is susceptible to low, moderate and high magnitude landslides, respectively. In this modeling effort, it was found out that slope influences landslide magnitude to a large extent compared to soil and longitudinal curvature. From this study, it was concluded that: (i) it is possible to characterize landslides basing on volume, area and flow length, using an objective classifier; (ii) a large proportion of the land (29%) in Sironko and Bulambuli is susceptible to moderate magnitude landslides, but an equally large area is not at threat of landslides; and (iii) soil, topography and geology influence the category of landslide experienced in Mt Elgon areas. It is recommended that high and moderate magnitude landslide susceptible areas be used for activities that require minimal land interference, such as is the case with forestry and conservation. | en_US |
