Abstract
A simple real-time system for forecasting flood water level at a retardation pond of a drainage pump station is presented.In this system, a lumped runoff model composed of two tanks is adopted.One of the tanks is a non?linear storage representing runoff from urban areas, while the other tank is a linear storage used for runoff from rural areas.This system is called the Lumped Tank Model system (LTM system).The runoff coefficients and storage heights for both tanks are estimated iteratively, using Kalman filtering and the difference between estimated and observed water levels of the retardation pond as the criterion.
The performance of the LTM system is compared with that of the system adopting a simplified version of the Diffusive Tank Model (DTM) called SDTM system.Comparison of the two systems was performed using the floods of July, 1972 and July, 1986 in the Ogura basin.The results show that the LTM system has almost the same level of accuracy in forecasting flood level as the SDTM system and requiring much less computing time.Because of the equal level of accuracy and computational efficiency, the application of LTM system should be preferred.
