Abstract: A three-level index system for urban flood risk assessment is put forward, in which higher level indexes of the system can be calculated from lower level indexes following a bottom-to-up deducing rule. The possibility estimation of risk zoning can be conducted in two ways. The first way is based on historical data. 100 extreme storms in study areas are selected from historical storm data, and then the precipitation of these rainstorms in the area are interpolated for 100 times, followed by summing up risk values of these storm events for each grid cell. The second way is based on observation and forecasting precipitation data. Sensitive indexes are determined by evaluating the effects of land surface topography, impervious surface area and river network on flood formation. The sensitive index of topography is obtained by considering land relief and elevation in each grid cell, and the measurement transforms land topographical characteristics into flood risk sensitivity. Impervious surface also plays an important role in flood formation since it greatly reduces water infiltration, quickens stream flow peaks, and increases storm flow. On the other hand, population density and GDP per square kilometer as well as distribution of critical areas for flood controlling are indicators for exposure in the flood risk analysis. As a result, the comprehensive flood risk indexes have been deduced by integrating the estimated second-level indexes (the precipitation index, the sensitive index and the exposure index). The application of the flood risk zoning with the bottom-to-up assessment method indicates that the risk zoning areas can be promptly accomplished by this way, estimating the distribution of actual risk areas properly. Meanwhile, it is found that the flood risk in urban is higher than that of rural and suburb, because the urban is flat in topography with much larger areas with impervious surface, and its population and economic entities are more heavily aggregated, and the rainstorm possibility is greater in urban with the analysis of the rainstorm historical data. Referring to the historical loss data of rainstorm floods occurring from 2004 to 2008 in Beijing, it has also been demonstrated that the result is consistent. Application of the model for risk warning is further validated with a storm event happened on 23 June 2011 in Beijing. In this case, quantity precipitation estimation (QPE) is used to estimate flood risk index, combined with the bottom-to-up assessment framework to estimate potential flood risk. Results of this application suggest that the bottom-to-up quick risk assessment fits the actual risk condition very well. Besides risk zoning and assessing, it can also be used to provide quickly-processed products for flash flood risk warning.