Satellite-Based Rainfall Estimates for Flood Prediction in the Brahmaputra Basin

The Brahmaputra basin. with a catchment area of 540,000 km2, shared by four countries, China, India, Bhutan and Bangladesh are adversely affected by floods on an annual basis. In this transboundary basin there is inadequate number and uneven distribution of hydro-meteorological observation stations and limited sharing of data and information amongst the countries in this transboundary basin making it challenging for flood forecasting. The knowledge of variation and amount of rainfall is crucial in hydrological modeling to implement a reliable and timely early warning system. As the spatial distribution of the rain gauge in the basin is not sufficient to provide a detail perspective on highly varied spatial nature of rainfall estimates using satellite-enhanced rainfall estimation provides an opportunity for application.

In this study we investigated the use of satellite-based rainfall datasets for flood prediction purposes. We utilized the United States Geological Survey rainfall runoff model (GeoSFM) with the NOAA CPC_RFE2.0 product to predict the discharge of the Brahmaputra River at Bahadurabad in Bangladesh. We have evaluated the model by comparing the simulated discharge with observed discharge at Bahadurabad. In general the simulated discharges follow the trend of the observed values quite well however, there is a significant difference in the magnitude of the flows. With the CPC_RFE2.0 data from 2002 and 2003 the Nash Sutcliff Coefficient of Efficiency was 0.23 with correlation (r) of 0.59. With the modified CPC_RFE2.0 data with ingestion of local rain gauge data, improvement in the results were obtained. The calibration and validation of the model is still in progress. However, our preliminary results suggest that CPC_RFE2.0 could provide a useful rainfall datasets for flood forecasting purposes in poorly gauged basin scale catchments but needs to be adjusted before application.