A numerical method for model parameters identification is presented for a river model based on a finite volume discretization of the bidimensional shallow water equations. We use variational data assimilation to combine optimally physical information from the model and observation data of the physical system in order to identify the value of model inputs that correspond to a numerical simulation which is consistent with reality. Two numerical examples demonstrate the efficiency of the method for the identification of the inlet discharge and the bed elevation. An application to real data on the Pearl River for the identification of boundary conditions is presented.