A new real-time flood forecasting model based on robust kalman filter method

The telemetry inevitably brings some outliers in observation data. The real-time flood forecasting based on conventional Kalman filter is sensitive to the outliers, that cannot guarantee the reliable forecasting results. In this paper a new real-time flood forecasting model based on a robust Kalman filter is described. The proposed robust filter assigns the less weight to the outliers by inserting the weight factors, and the unity weight to the normal data. That makes the robust Kalman filter method is insensitive the outliers, meanwhile keeps the properties of the conventional Kalman filer. The feasibility of the robust approach is demonstrated with unbiased and biased data.