Cloud-Independent Flood Mapping and Surface Water Dynamics Using Sentinel-1 SAR in the Lower Volta Basin, Ghana

Abstract

Flood mapping in tropical river basins is frequently constrained by cloud cover during major flood events. This study integrated Sentinel-1 Synthetic Aperture Radar (SAR) with Sentinel-2, CHIRPS, ERA5-Land, SRTM, JRC Global Surface Water, ESA WorldCover and GHSL data to characterise flood dynamics and exposure in the Lower Volta Basin, Ghana. Sentinel-1 VV and VH backscatter were used to map seasonal water, event-based temporary flooding, cumulative and recurrent inundation, and flood-period surface change. Threshold sensitivity, Otsu benchmarking and a 600-point Sentinel-1/Sentinel-2 comparison were used to assess classification performance. Event-composite temporary water covered 5.868 km², while the cumulative footprint reached 7.386 km², of which 4.487 km² (60.7%) was recurrently detected. A broader 2,032.146 km² backscatter-change zone indicated hydrological surface responses extending beyond directly detectable open water. Flooding was concentrated mainly in low-relief wetland and floodplain environments, with herbaceous wetland accounting for 62.08% of the cumulative land-cover overlay. The Sentinel-1/Sentinel-2 comparison achieved 95.833% overall agreement, a 95.697% F1-score, and κ = 0.917. However, this represents inter-sensor consistency rather than independent field validation. The results demonstrate the value of combining direct open-water detection, flood recurrence, surface-change analysis and exposure information for flood assessment in cloud-prone tropical basins.