Flood Susceptibility Assessment Using Geospatial Techniques and Predictive Modeling in Bunyala Sub-County, Busia County, Kenya

Abstract

Flooding remains one of the most destructive natural hazards globally and is a major contributor to disaster-related losses in Kenya. Bunyala Sub-County in Busia County is particularly vulnerable due to recurrent flooding associated with River Nzoia, River Yala, and Lake Victoria backflow. Despite previous geospatial studies in the area, limited attention has been given to temporal changes in flood zones, the influence of flood drivers, and the application of predictive models in flood susceptibility mapping. This study assessed flood susceptibility in Bunyala Sub-County using Geographic Information Systems (GIS), Remote Sensing, and predictive modeling. Multi-temporal Landsat imagery (2000, 2010, and 2020), Digital Elevation Model (DEM), rainfall, GPS, and OpenStreetMap datasets were processed using ArcGIS Pro, Google Earth Engine, and SPSS to map flood zones and evaluate the influence of slope, elevation, distance to water bodies, rainfall, and land use/land cover (LULC). Logistic regression was used to determine the relationship between flood occurrence and the selected flood drivers, while a weighted overlay approach was applied to generate a Flood Susceptibility Index (FSI) raster map, which was further reclassified into a final flood susceptibility map. Results revealed a net increase of +3.70 km² in flood zones between 2000 and 2020, with a decline of -8.72 km² between 2000 and 2010 followed by an increase of +12.42 km² between 2010 and 2020. LULC emerged as the strongest-associated flood driver in the model, followed by slope and distance to water bodies. The LR model achieved an overall classification accuracy of 78.7% and an AUC value of 0.841, indicating good predictive performance. The final flood susceptibility map classified 62% of the study area as high susceptibility, 34% as moderate susceptibility, and 4% as low susceptibility. The study demonstrates the effectiveness of integrating GIS, remote sensing, and predictive modelling in flood susceptibility assessment and provides valuable information for land use planning, flood risk management and climate adaptation in flood-prone areas.