Modélisation de l'érosion hydrique en milieu semi-aride de forte énergie de relief à partir de données de télédétection application à la Bolivie
In this study, we propose a model of water erosion risks for such environments using remote sensing and morphometric data: MEH-SAFER (Modèle d'Érosion Hydrique en milieu Semi-Aride de Forte Energie de Relief). MEH-SAFER is based on the Lamachère and Guillet model (MLG) (Burkina Faso) to the Lake Laka-Laka drainage basin (Bolivia). While preserving the same principle calculations of runoff potential, we have improved the acquisition method of the biophysical data in order to avoid errors related to topography and to the ground data. We have conceived a method based on multisource satellite images (RADARSAT-1, Landsat-7 and SPOT-4). We replaced the geomorphological graphic models used by Lamachère and Guillet with a topographic vulnerability map originating from the MVT (topographic vulnerability model). This, in turn, was derived from a DEM. Several combinations of multisource image and texture bands give classification accuracies greater than 80% for the land use classes. These include, among others, the combinations of entropy-ETM+ 2-ETM+ 4, correlation-ETM+ 2-ETM+ 4, homogeneity-ETM + 2-ETM+ 4, mean-ETM+ 2-ETM + 4, standard deviation-ETM+ 2-ETM+ 4, original radar image-ETM+ 2-ETM+ 4, dissimilarity-ETM + 2-ETM+ 4, angular second moment-ETM+ 2-ETM + 4, XS1-ETM+ 3-ETM+ 4 and XS2-ETM + 3-ETM+ 4. In 83% of the basin, the potential runoff is superior to 0,50 on a maximum of 1. In the remaining 17% of the basin, the potential varies of 0 to 0,42, which is explained by the resistant hydrodynamical characteristics of the geoecological units. The results reported here reveal that in general the study area is not particularly vulnerable to erosion and that as a consequence the perceived rate of sedimentation is a natural consequence of the morphoclimatic conditions of the drainage basin. The main contribution of this study is the development of the MEH-SAFER. It includes several original ingredients including the numerical processing of multisource satellite data and morphometric topographical and auxiliary data. It also incorporate statistically identified biophysical parameters to be included in the modeling"--Résumé abrégé par UMI.