The present paper describes practical issues and their resolutions for increasing predictive capabilities of real-time flood forecasting methods. Special attention is directed to explicit incorporations of the water level-discharge (H-Q) relationship and rainfall measurements by radar into the filter prediction formulation.It is well recognized from hydrologic practices that measurement errors involved in the H-Q relationship are quite significant in flow and water stage forecastiong problems. The current approach proves extremely helpful in assessing the relative importance of modeling and measurement errors, and inaccuracy in rainfall forecasts for flood prediction applications.For the behavior of watershed response, the present study uses a nonlinear storage routing model converted from the kinematic wave equations where accurate linkages between the parameters for the two approaches are maintained. The proposed algorithms. were applied to the river stage forecasts of the 1988 Flood data in the Uryu River, which is a tributary of the Ishikari River.