Hydrological data and modeling to combine and validate precipitation datasets relevant to hydrological applications

Study Region: Forty-one river basins in Brazil and neighboring countries in South America.Study Focus: In large river basins, on countrywide or continental scales, it is often difficult to have consistent and accurate long time series of spatially distributed precipitation data available. However, these are needed to calibrate hydrological models and to run hydrological simulations continuously in real-time streamflow forecasting. In this study, we assess two real-time precipi-tation products based on rain gauges and satellite data (TRMM-MERGE and CPC-NOAA) for their use in streamflow forecasting in the hydropower sector in Brazil. To take advantage of each precipitation data source and derive a unique dataset, a methodology is proposed to combine, extend, and validate the datasets. We consider the discharges at the river basin outlets as an independent and robust reference for hydrological applications. Observed discharges are used to quantify precipitation uncertainties and to weight the blending, while discharges obtained from hydrological modeling are used to validate the final precipitation product.New Hydrological Insights for the Region: The proposed blending method, which uses the uncer-tainty of the original datasets to define the weighting factors, was efficient in generating a pre-cipitation product that performs better than each dataset separately when used to force a hydrological model. The use of the double-mass curve correlation to extend the time series of the datasets beyond their common period allowed us to produce long time series of precipitation for South American basins and hydrological applications. The study shows that it is possible to rely on river discharge data and hydrological modeling to select and combine different precipitation products in the region and presents a step-by-step methodology to do so.