Abstract: It is well acknowledged that the accuracy of stream flow predictions from a hydrologic model is heavily dependent on the accuracy of the precipitation inputs. Particularly, high variability of rainfall exists both in time and space, and mountainous basins have in general fast response time. Therefore, hydrological models taking into account the rainfall variability should play an important role in flood alert systems in mountainous basins. In this sense, weather radar inform ation is a key element in flood forecasting. The studied closed basin, Foziling (1813 km²), has a basin-to-radar distance of 100 km from southwest of the Hefei CINRAD/SA radar (31.866°N, 117.257°E). Comparisons of raingage-based and radar-gage-based simulated discharges using TOPMODEL (TOPography based hydrological MODEL) are performed for the Foziling basin and its 6 subcatchments. Rainfall observations of the basin are available from 12 raingauges operated by Bureau of Hydrology, Huaihe River Commission, Ministry of Water Resources PRC. In order to utilize true independent data sources for verification purposes, 6 gauges are withheld from the estimation scheme and used for verification instead. The radar data are selected from an S-band Doppler weather radar located at Hefei, Anhui Province throughout the period from June 20 to July 12, 2003. The TOPMODEL rainfall-runoff model used in this study is a semi-distributed watershed model that simulates the variable-source-area mechanism of storm runoff generation and incorporates the effects of topography on flow paths. For the application of the TOPMODEL, the topographic index is computed based on 1:250000 DEM (Digital Elevation Model). To some extent, whether the merged radar-gage estimates are better than the gage-only estimates or not is relative to the density and representativeness of the raingage network; even when the raingage-only and radargage mean areal rainfall estimates show nearly the same value for the Foziling basin, there is discrepancy between raingage-only and radar-gage estimates of mean areal rainfall for each subcatchment; for certain subcatchment, the greater the discrepancy between the raingage-only and radar-gage mean areal rainfall estim ates, the bigger the divergence between the corresponding simulated runoff depths from TOPMODEL. Therefore, even when a relatively dense raingauge network exists, the rain gauge data alone do not provide an initial rainfall state that is detailed enough for accurate hydrologic simulation, and radar information is essential to provide accurate flow estimates using a rainfall-runoff model. In fact, a density of about 1 raingauge per 300 km², which is the case of the Foziling basin after the 6 verification raingauges are withheld, is insufficient to reproduce the spatial precipitation pattern of the event studied. Conclusions from the study may be specific to the target case, which is based on the characteristics of the QPE (Quantitative Precipitation Estimation) inputs and the TOPMODEL hydrologic model, or specific hydrologic characteristics of the Foziling basin. In the future, additional events, other hydrological models and some more robust radar-raingauge correction procedures will be investigated.