The model of 1D unsteady channel flow combined with the Kalmanfilter for real-time channel flood forecasting was attempted in this study. The suitable upstream and downstream boundary conditions were suggested. The system equation was given by the linearization of the finitedifference equations of the mass conservation and momentum equations as well as the boundary conditions. In the Kalman filter updating model, because the number of measurement variable is less then that of state-space variables, the measurement error covariance matrix could be estimated in real time through the innovation sequence, and the system error covariance matrix needs to be estimated preliminarily. A real example of flood forecasting in the Huaihe River was given to explain how the method works. The results show that the model is reasonable and effective.