Cities are suffering from extremely heavy rain and urban flood disaster in recent years,due to the change in the natural hydrological processes impacted by global climate change and rapid urbanization. In order to predict floods and mitigate urban flood disasters effectively,hydro-hydraulic models are used to simulate the sub-cycle of urban water cycle. However, the uncertainty of forecast data of single rainstorm is a major challenge to forecast flood in hydrological model. From the perspective of urban rain flood model input uncertainty,the development process of the rainfall observation and prediction technology is reviewed. The characteristics,applicability and limitations of different prediction methods are also summarized. Thus,the basic framework of multi-source storm forecast data fusion is proposed,and the development trend and prospect of urban flood forecasting is illustrated. That is to say,in the future,in order to realize high-precision and longterm forecast of flood forecast,it is necessary to improve the ability of rainfall observation and quantitative precipitation forecasting, and develop multi-source information coupling technology,which will provide the basis for urban flood control and disaster reduction.