Amélioration d'un modèle distribué d'érosion hydrique par la prise en compte spatiale de l'influence anthropique en milieu agricole
The following Ph.D. thesis is a contribution to the study of anthropic influences on soil erosion processes in an agricultural environment. Intensive exploitation of agricultural lands increases their fragility by amplifying their sensitivity to erosion and diffuse pollution. Agricultural areas are the focus of increased attention and are the object of numerous studies towards developing methodologies for their integrated management. At the center of the sustainable development issue, based on the compromise to be achieved between the controlled exploitation of resources and the preservation of areas in terms of soil and water quality, agricultural watershed management has become a priority in the fields of environmental research and organization. Modeling of physical functioning and the changes induced by man represents a solution to a better apprehension and understanding of the role of the different parameters involved in the dynamical hydrological and sedimentary processes and accordingly towards mastering more efficiently human effects on the different environments.The approach chosen is based on the spatial modeling of the anthropic influence on hydrological and sedimentary processes at the scale of the watershed combined with the use of remote sensing data for characterizing ground surface states. An existing erosion model was entirely reformulated for taking into consideration a certain number of these anthropic influences recognized as essential in soil erosion. These are namely the presence of tillage where orientation conditions flow directions as well as the spatial boundaries resulting from the organization of the landscape and which constitute obstacles to surface flows.The model was tested over a study site chosen for its high sensitivity to erosion, in Blosseville (Normandy, France).The site has been the object of multiple ground observations and hydrological monitoring over many years. Among the input parameters required for the model, certain are accessible through remote sensing. A methodology is proposed using satellite radar images for estimating ground moisture and surface roughness from the inversion of a radar backscattering model and permitting to take into consideration the spatial variability of these input parameters.