A time series of Envisat/ASAR images was acquired for monitoring the inundation evolution in Doñana National Park wetlands. Flood mapping from the ASAR data alone was unfeasible due to the complex casuistic of Doñana’s covers backscattering. A digital terrain model and a vegetation map were then utilized to complement the ASAR data. The use of irregular filtering neighborhoods adapted to the terrain elevation contours drastically improved the ASAR images filtering. Pixels highly likely to belong to the different cover classes were selected by combining the vegetation map, the DTM-based sub-basin segmentation and previous knowledge on the covers backscattering. The regions were grown based on the Mahalanobis distance, yielding accurate classification maps of flooded and emerged cover types.