The use of appropriate hydrological models with real-time satellite rainfall estimates can help mitigate flood damage, provide support to contingency planning, and provide warning to people threatened by floods. In this study, the simulated streamflow from the Geospatial Stream Flow Model (GeoSFM) using NOAA Climate Prediction Centre Rainfall Estimates Version 2.0 (RFE) data was compared with observed streamflow in the Bagmati basin. The model showed poor performance when forced with 2002 monsoon RFE. However, the simulated streamflow showed a significant improvement when using improved RFE by incorporating local rain gauge data into the rainfall estimate. Simulated streamflow was much closer to the observed discharge, with a Nash–Sutcliffe efficiency coefficient of 0.91, flow ratio of 1.1, and coefficient of determination of 0.92. Thus, improved RFE can be considered for use in flood prediction. A flood hazard map prepared using the extreme rainfall event showed an inundation area of 315 km2, 8.4% of the basin.