Abstract
The storage function method developed by Kimura is often used for short-term runoff simulation, and some on-line prediction methods coupling with the storage function method and Kalman filtering which aims to update forecasts, have been developed by several investigators including the authors. In this paper, the frameworks of a flood runoff prediction method and a treatment of uncertainty of model parameters in the storage function method are newly devised. In the new method, a number of filters with different parameters are run simultaneously, and the probability distribution assigned to these filters is updated by using Kalman filtering with real-time observation data. The application results show that the new treatment enhances the prediction accuracy.
