This talk reports on the use of satellite-based rainfall estimates (RFE) and distributed hydrologic models for flood forecasting purposes in the data sparse Himalayan region. The work addresses the need for bias-correction the RFE datasets before it is draw on for flood forecasting purposes. The study area is the watershed of the Narayani Basin, Nepal, upstream of the Devghat hydrometric station (32,000 km2). The rainfall-runoff model used is the Geospatial Stream Flow Model (GeoSFM), a spatially distributed, physically based hydrologic model. The RFE from the National Oceanic and Atmospheric Administration Climate Prediction Centre's rainfall estimates (CPC_RFE2.0) was compared with rainfall estimated from a dense network of rain gauges. The RFE rainfall probability of detection (POD) was 0.97 and the false alarm ratio (FAR) was 0.05. The RFE captures well the spatial trend of the rainfall but underestimates rainfall amounts on average by 50%. Two bias removal schemes were implemented for the RFE: a monthly bias adjustment by simple moving averages bias ratios, and an interpolation method that combines the RFE data with data from rain gauges. The hydrologic model was calibrated with RFE data from 2003. Simulated and observed flows were highly correlated during the calibration period—r2 was 0.88. Hindcast of the high flows for the spring season of 2004 were made with RFE and bias-adjusted RFE. When bias-adjusted monthly RFE was used, flood prediction the hydrologic model performance improved (r2 went from 0.69 to 0.76). When the local rain gauge bias-adjusted RFE was employed, simulated streamflow prediction improved further (r2 = 0.81). The results indicate a positive applicability of the RFE data for flood prediction purposes if an appropriate bias correction is used.