Résumé
Variational data assimilation has been used to assimilate water surface elevation, slope and river top width, outputs of the forthcoming Surface Water and Ocean Topography (SWOT) simulator, for simultaneous estimation of inflow discharge, river bathymetry and/or friction. This method has been applied in two benchmarks of the SWOT Discharge Algorithm Working Group (DAWG) using the 1D full Saint-Venant hydraulic model SIC (super 2) , developed at Irstea-Montpellier (France). Discharge was first estimated assuming known river bathymetry and roughness coefficient, underlying the influence of the spatial and temporal frequencies of the SWOT observations. We emphasize that the temporal sampling frequency should be consistent with the characteristic (travel) time of the dynamical system, in which case discharge is successfully estimated (rRMSE = 2.1%). Larger temporal sampling intervals require complementary information given by the a-priori information of discharge or a larger spatial scale dynamical system. Second, we investigate discharge estimation under uncertainty in bathymetry or/and friction. Simultaneous estimation of discharge together with (i) the roughness coefficient, (ii) the bed level, and (iii) both the roughness coefficient and the bed level, have been performed. The algorithm provide successful estimators in cases (i) and (ii) with respective relative errors on discharge of 2.6% and 3.8% which represent an improvement of 10% and 47.1% over the estimation of discharge solely. Whereas in configuration (iii) the optimal solution is not unique, this is typical of the equifinality issue. In the latter case, discharge is estimated with an rRMSE of 7.1% with an improvement of 33.4%.