Mapping Flood Risk of Nsanje District in Malawi Using Sentinel-1 Data

Abstract

The all-time imaging ability of SAR systems which can penetrate cloud cover and free availability of Sentinel C-Band data are very useful in deriving critical spatial information for flood disaster management in tropical areas such as Malawi. This study shows how Sentinel-1 data has been used to map flood extents by utilizing VH and VV polarizations and enhance risk mapping during the years 2015 to 2020 in Nsanje district. Multi-dated Sentinel-1 images were acquired, pre-processed, and analyzed through Change Detection and Thresholding technique to interpret backscattered radiation thereby distinguishing flooded and non-flooded areas. The VV Polarization result shows that percentage of land that remained under flood water in Nsanje district was 3.97% in 2015, 5.91% in 2019 and 2.05% in 2020. On the other hand, VH Polarization result shows about 7.71% and 2.02% of land remained under flood water in 2019 and 2020 respectively.Sentinel-2 pre-flood image was used to determine major land cover classes in the district. MNDWI derived from Landsat 8 imagery of the same date as 2019 Sentinel-1 crisis imagery was used for flood extent validation. Overall Accuracy of flood extent delineation was 96% and 90% for VH and VV polarization respectively. Flooding frequency in specific areas was observed to determine physical vulnerability and hence figure out the risk to a possible repetition of flood disaster. Further, social data were used to aid a better understanding of the capacity to cope with a possible flood hazard in the study area.