mdelgadoblasco
commented
3 years ago Hi Ben. It could be done, but the current approach would need some changes as we could not use statistics on flood images. In fact, the approach would be to consider each image separately, and apply the flood detection formula with respect to the dry period each time. Later we would need to sum up all the flood binary masks to get the total number of times a pixel was identified as flooded. This could be done for both S1 and S2 imagery, with the inconvenient that for S2 we may have many clouds in single images, instead to work on the cloud masked temporal mosaic.
Who could we generate temporal information on floods? The current process is to collapse all the after images into a single band and calculate the minimum at each pixel. This essentially identifies each pixel that has experienced a flood in the "after" time period. However, could we calculate the flood at each time period to identify how often each pixel has flooded?