Abstract: In order to synchronously correct river roughness parameters and hydraulic state variables (including discharge, stage or water depth) of river network hydrodynamic model, extended Kalman filter method is used to develop a data assimilation model. This model takes roughness parameters and hydraulic state variables as model state variables for synchronization correction. In a simulation example, it systematically examined and analyzed the factors including roughness dynamic noise level, water level dynamic noise level, initial value of roughness and station number, which have important influence on model correction capability. The results reveal that the present model is able to carry out data assimilation of dynamic river system efficaciously. Corrected roughness values near stations tend to be true values, and those far away from stations tend to be initial values. By adjusting roughness dynamic noise level, it can effectively control roughness correction degree, avoiding calculating failure.