Beginning with the mechanism of rainstorm floods, and taking the Dongjin River Basin as a case, this paper carries out the study on the rainstorm flood risk regionalization of medium and small river basins. Using the meteorological data, hydrological data, geographic information data, social and economic statistical data, historic disaster data as well as the hydrological model of TOPMODEL and statistical method, we determine the flood critical area rainfall and also reconstruct the data sequence of regional meteorological stations by the method of stepwise regression. Then by using the generalized extreme value distribution function, we work out the floodcausing area rainfall in different return periods. According to the distribution of hourly precipitation rainfallpattern in the river basin, we use the FloodArea model to simulate the flood with the data of the floodcausing area rainfall in different return periods, the DEM added with the dykedam information, and the coefficient of manning etc., obtaining the flood inundation maps for different return periods. Finally the risk regionalization atlas for different return periods are drawn based on the data of population, GDP and landuse in the river basin. The technique established in this paper for rainstorm flood risk regionalization of medium and small river basins is convenient and feasible, and the results of risk regionalization are of high precision and strong practicability. Therefore, it is meaningful for dynamic disaster risk management which faces the realtime disaster prevention and mitigation.