Hurricane Dorian was a devastating Category 5 hurricane that hit the Bahamas exceptionally hard. Given the severity of the forecast, the European Space Agency directed both Sentinel-1 satellites to focus their collection over the Bahamas. This provided more re-visit times over this region in order to better monitor the impacts of the storm.
As soon as every Sentinel-1 scene was available during and after the hurricane, I ran a simple thresholding and median filter algorithm on the VH band of the Sentinel-1 image. Given how pronounced the flooding was, this simple algorithm worked quite well in delineating regions within the Bahamas that had been flooded. The flood map shown above is the output of this simple algorithm over the Grand Bahama, specifically looking at Freeport, the main city on the island.
Once the results were delivered, I tried a more complex approach, by training a UNet on the VV and VH bands, with training data collected from high resolution landcover data from coastal wetlands in the United States. Upon examination, the algorithm performed quite comparably to the simple algorithm in this case.