Flood monitoring in the Senegal River valley: First results based on SAR PRI data

This study focuses on the use of Synthetic Aperture Radar (SAR) data for mapping the extent of flooding events. A small test area in the middle Senegal River valley in the western part of Sahel is studied. Four ERS2 SAR PRI scenes have been used in the analysis for delineation of flooded areas. Change images have been produced by using the latest non-flood image (August 1999) as reference and dividing the backscatter values in the subsequent images by the values in this reference image. When plotting the temporal evolution in backscatter for different surface types an unambiguous picture emerges. Six different classes were found to have a distinct backscatter pattern trough time, and can therefore easily be identified in the images. For each change image in the flood period, an automatic standard classification algorithm has been applied, and the resulting classes have been assigned water or land. The classification result for the first SAR image in the flood period (September 22) has been compared to the outcome of a Landsat classification into water, land and vegetation for validation purposes. The agreement between the SARand the Landsat-based mappings is very high with an overall classification accuracy of 97.5%. Profiles of the inundated areas measured by differential GPS, have resulted in profiles of the surface elevation across the area of flood recession agriculture. It can be concluded that the high accuracy of the four profiles is within one decimeter and they are consequently suitable for the purpose of validating the SAR based delineation of flooded areas. The flooded/non-flooded areas in the classified image correspond very well to the profiles of topographic measurements carried out by differential GPS. This indicates that the water line estimated from the SAR data appears to be in agreement with the topography in the study area and the water level in the river. Consequently the duration and extent of the flood can be estimated from SAR images, and the results may thus serve useful to local decision-makers in crop yield estimation etc.