Flood Spatiotemporal Patterns in the Lower Nyando River Basin in Kisumu County, Kenya

Keywords: Floods, Google Earth Engine, Remote sensing, SAR
Abstract Type: Poster Abstract

Abstract

The lower Nyando River basin in Kisumu County, Kenya is plagued by recurrent bi-annual devastating floods threatening lives, livelihoods, infrastructure and the riparian ecosystem. There is inadequate baseline information on recent flood spatial extent needed to develop flood mitigation plans in the study area. This study combined synthetic aperture radar (SAR) and optical remote sensing to construct recent flood history in Google Earth Engine and ArcGIS Pro. Supervised land use classification using random forest was performed on the before and after optical images. Closest images to peak floods were selected. The classes were Water, Builtup_Bareland, and vegetation. Before and after radar data were acquired in the in Interferometric Wide Swath (IW) MODE, ascending, and a single polarization used, VH. Filters such as slope and permanent water bodies performed. The difference images were used as flooded areas and the raster calculator tool used to merged and categorized into 5 classes; Very high, high, medium, low and very low. They were 7, and 8 maps from SAR and optical images respectively. Sentinel-1 SAR GRD, C-band has the capability to penetrate clouds and operate day and night providing a reprieve in flood extent mapping in tropical/cloudy regions. However, its low temporal resolution of 12 days is a limiting factor especially in flash floods cases. The flood frequency map provides baseline data needed by flood mitigation planners identifying what regions need more attention to ensure equity in resource use and allocation and aid emergency response.