A Bayesian model for blending satellite precipitation estimates to enhance drought monitoring in poorly gauged regions

Droughts are among society's most damaging climatic disasters, representing a complex phenomenon in which precipitation plays a central role. In poorly gauged regions, precipitation estimates from remote sensing have been widely employed. However, these data often exhibit bias and may not accurately depict the spatiotemporal variability of the rainfall processes. To address these problems, this study aims to propose a novel prediction model to enhance the reliability of rainfall estimates in poorly gauged regions by utilizing a Bayesian approach to blend multiple Satellite Precipitation Estimates (SPEs). CHIRPSv2, IMERG V06, and CMORPH were the selected SPEs for building the blending model, which was applied to estimate monthly rainfall data in the Doce River catchment and the mid-lower portion of the São Francisco River catchment, both in Brazil. Comparative analyses between rainfall gauging data and the SPEs, as well as the blending model, unveiled the enhanced performance of the new precipitation estimates, demonstrating consistently lower values for RMSE, NRMSE, and PBIAS in both study areas. We also assessed the ability of the SPEs and the blending model to capture drought conditions across different aggregation scales of the Standardized Precipitation Index (1, 3, 6, 9, and 12 months). Regarding the number of drought months and magnitudes accurately identified, the blending model consistently achieved better performance from SPI-1 to SPI-9, but was outperformed at SPI-12 by different SPEs. Overall, despite its increased complexity, this approach can be effective for drought characterization in regions with sparse monitoring and further hydrological applications on a monthly scale. • A Bayesian model was applied for blending SPEs at the monthly scale. • The blending model generated more reliable precipitation estimates than selected SPEs. • CHIRPSv2, PERSIANN, IMERG V06, and CMORPH were assessed for drought characterization. • The blending product improves drought characterization and monitoring.