Evaluating Methods and Polarizations of S-1 Sar for Time-Series Flood Hazard Mapping Akaki Catchment

No Thumbnail Available



Journal Title

Journal ISSN

Volume Title


Addis Ababa University


Lack or absence of data is the main limiting factor for studying flood hazard and risk in many basins across the globe. The Google Earth Engine (GEE) has a potential to narrow data gaps by providing ease of access to remote sensing data and enabling automatic and rapid generation of flood inundation map for Near Real-Time monitoring. In this study, GEE was used to assess and analyze the Sentinel-1 (S-1) Synthetic Aperture Radar (SAR) dataset for flood inundation mapping in the Akaki catchment which hosts Addis Ababa city in the central part of Ethiopia. Ground Control Points were collected at the time of satellite overpass for evaluating the accuracy of the generated flood inundation maps. Change detection and Histogram thresholding methods were compared using co-polarized (VV) and cross-polarized (VH) images. A new method which is Root of Normalized Image Difference (RNID) was developed for change detection. Major flood affected roads in Addis Ababa city and Land Use Land Cover (LULC) classes were detected from April to November of 2017 to 2020. The result shows the RNID method performed better than the histogram thresholding for flood inundation mapping in the study area. The VH polarization performed better than the VV polarization to detect the lower signal backscatter intensity generated from the flooded surface. An overall accuracy of 95% and kappa coefficient 0.86 was obtained when applying the RNID method and VH polarization. In the Akaki catchment, the remote sensing mapping showed that the flood commonly starts in May and recedes in November, but flood was frequent and widespread from June to September. At the downstream of the new expressway, the riverine and pluvial flood frequently occurred in the past four years. The flood inundation map also showed that several major roads of Addis Ababa are affected by flooding. The irrigated and built-up area were the most affected land use classes with an inundation extent of 1057.05 ha (21.28% of the total irrigated land) and 544 ha (1.44% of the total urban area) respectively. The S-1 SAR was found useful for time series flood inundation mapping and the new change detection method (RNID) performed better in urban and peri-urban flood mapping, but the accuracy of the flood map varies with the flood detection method and the image polarization.



Gee, S-1 Sar, Flood Mapping, Addis Ababa, Akaki Catchment, Hazard