[Objective] It is necessary to study on the spatial and temporal process of urban overland runoff that could provide support decisions for waterlogging prevention and drainage emergency.[Methods] Because of the fragmentation of underlying surface and complexity of micro topology in urban environment, urban hydrology process is too complicated to present exactly characteristics, such as changeability of flow routes, spatial and temporal variation of flooding. Based on cellular automata, an urban surface flooding model(USFM) is built to simulate surface hydrologic process by defining cellular's state and its evolution rules. In this model, several processes were considered, including rainfall, vegetation intercept, depression storage, runoff generation, infiltration and inlet discharging. The runoff process was handled by Manning's equations and rules, and runoff volume and flow path were calculated respectively.[Results] Because the hydrologic response units(HRU) are treated as grids, USFM can be defined as a distributed model, which can deal with the spatial variability of urban space very well based on DEM and land use data. Moreover, by combining the cell rules with mathematical formula for urban space flow uncertainty, USFM has the capacity of determining the flow routing and volume of runoff automatically. The basic input data of USFM are DEM data, land use data and gutter inlet data, which are raster data format. The USFM output data is a spatial and temporal dataset, which contains surface runoff information of any given time and space. A basin in Guangzhou Panyu District was chosen as the experimental area to test the model.[Conclusion] Through parameter calibration and model validation, the proposed approach well reflects the urban surface flooding hydrological details with a small amount of input data and simple modeling working. The USFM is useful to forecast the flooded areas in urban areas.