Assimilation of Remote Sensing Data for River Flows

Abstract. We address the problem of parameters identification and data assimilation for river flows modeled by the 2D St-Venant equations. In practice, available observations are very sparse especially during flood events (very few measurements of elevation at gauging stations in the main channel). We assume we have in addition either surface trajectories extracted from video images (lagrangian data) or space distributed water levels extracted from one satellite image. Then we identify parameters such as the inflow discharge or the topography and/or the initial state (depending on the configuration and the observations available). Numerical twin experiments demonstrate the efficiency of the present method for toy test cases