Abstract: With the global climate change and rapid urbanization in China, extreme rainfall response has become one of the priorities of urban flood management. Dynamic rolling flood risk analysis under extreme rainfall events can provide reference for forecasting and early warning, enhancing the scientific basis for decision-making, and is an effective means of preventing flood disasters. Although traditional hydrological-hydrodynamic models can accurately calculate flood routing and retreat in fine grids, they often spended extensive computation time due to the big grid scale, and the real-time rolling water depth prediction often cannot meet the requirements. This paper proposes an efficient interpolation technique based on static-dynamic variable decoupling by establishing a functional mapping relationship between flood inundation depth and rainfall. Combined with distributed databases and parallel computing techniques, this method can provide results consistent with those of hydrological-hydrodynamic models in a very short period of time. The study aims to provide effective support for real-time flood risk analysis.