Abstract
The model to predict water level from precipitation directly was built by the storage function method and an H-Q equation. Because this model does not use flow quantity in calculation process, this model is available in a flood prediction at the water level station not to get adjustment of flow quantity and the precipitation, and not to observe flow quantity. The constant number of this model is expressed in a selfregression model and Unscented Kalman Filter is applied to data assimilation technique. As a result of having applied this model to the Doki River basin, we obtained an almost proper result by setting selfregression coefficient and a system noise appropriately. In addition, it was revealed that the flow quantity counted backward accorded with flow quantity estimated by the storage function method. Therefore, it was understood that the built model reflected a rain-runoff process.
