| dc.description.abstract | Landslides have wreaked severe devastation in Uganda's highlands and hilly districts especially Bududa. The Bududa avalanche on March 1, 2010, destroyed socio-economic infrastructure and killed over 365 people while displacing hundreds more. Landslide susceptibility mapping refers to the ascertainment of the possibility of slope failures in a given geological setting as a result of landslides. The purpose of this is to determine the areas that are prone to landslide occurrence in a particular location. These studies have been carried out on the slopes of Mount Elgon some of which have been on a wide scale, while others have not incorporated various vital factors. To add on, most of these studies are based on approaches such as Analytical Hierarchical Process (AHP), Fuzzy Logic, among others. Fuzzy AHP, an approach that combines the AHP and Fuzzy Logic has proven to provide more reliable and accurate results based on previous literature.
This study analyzes landslide susceptibility mapping using this Fuzzy AHP approach on a more localized study area i.e Bududa district. Factors considered in this analysis are rainfall, slope, soil type, curvature, distance to streams, Stream Power Index (SPI), Topographic Wetness Index (TWI) and Land cover. To this effect, the various datasets were obtained and for land cover, a 10m high resolution sentinel-2 image was downloaded from USGS website for the year 2021 and preprocessing carried out. A Digital Elevation Model (DEM) i.e. SRTM was obtained and the same year, preprocessing of the sinks and peaks carried out. Bududa district as the study area was clipped out for all the datasets. Land cover classification was carried out on the Sentinel-2 imagery using the maximum likelihood classifier due to its accuracy. The slope, TWI, SPI, Curvature maps were obtained from the DEM. Streams were also delineated from the DEM and the distance from streams map obtained using Euclidean distance tool. These maps were then resampled to match the 10m resolution of land cover and then reclassified into classes of five. Expert opinion was used to determine the relative importance of the factors using Fuzzy AHP. The factor maps were then overlaid basing on their criteria weights and the resultant susceptibility classified from least to highest.
Results showed that weights of these factors on a scale of 0-1 were rainfall (0.26), soil (0.21), slope (0.14), land cover (0.12), distance from streams (0.05) SPI (0.08), TWI (0.06) and curvature (0.04). The confusion matrix after classification had an accuracy of 80.3%. After the final susceptibility map was obtained, it was found out that 2.1% was extremely susceptible, 27.6% of the study area was susceptible to landslides, 49.1% of the study area was moderately susceptible to landslides, 19.6% had low susceptibility and 1.5% had least susceptibility. Using past landslide inventories, a Receiver Operating Characteristics curve was obtained giving an area under the curve of 84.1%. The results demonstrate that land slide susceptibility mapping using combined Fuzzy AHP can be used in early planning and disaster preparedness through resettlement. The results reveal areas of high susceptibility for which early warning programs can be initiated. The method adopted can also be unscaled to determine other similar areas. | en_US |
