Multi-source uncertainties in streamflow modeling driven by TRMM satellite precipitation

Satellite Precipitation Estimate (SPE) with high spatio-temporal resolutions is frequently applied to hydrological modeling. However, uncertainty in hydrological modeling due to bias and errors has not been investigated. In the present work multi-source uncertainties in rainfall-runoff of two hydrological models (i.e., lumped GR model and distributed CREST model) forced by two precipitation inputs (Tropical Rainfall Measurement Mission post-processed 3B42v7 product and corresponding gauge rainfall) were quantified. The multi-source uncertainty components were partitioned by variance decomposition. To verify effect of proposed framework, streamflow at Waizhou outlet in the Ganjiang River Basin was simulated. Total uncertainty in CREST modeling driven by both SPE and gauge precipitation was found to be lower than in Ge´nie Rural (GR) modeling. Among 4 scenarios for 2 precipitation inputs combined with 2 hydrological models, input uncertainty of Coupled Routing and Excess Storage (CREST) modeling driven by SPE was found to be the lowest. These data indicate that distributed CREST model is better than lumped GR model to take advantage of the spatial information in TRMM satellite precipitation data.