Streamflow Forecasting using Satellite Products: A Benchmark Approach. Can We Reduce Uncertainty by using Multiple Products and Multiple Models?

Abstract:

Real-time satellite precipitation products can be used to drive hydrologic forecasts in downstream areas of poorly gauged basins. We present an improved approach to hydrologic modeling using satellite precipitation estimates to reduce uncertainty, consisting of: (1) bias-correction of satellite products, (2) re-calibration of hydrologic models using bias-corrected estimates, (3) bias-correction of streamflow outputs, and (4) plotting of uncertainty intervals. In addition, we evaluate the benefits of multi-product and model forecasts using four satellite precipitation products (CHIRPS, CMORPH, TMPA, and PERSIANN-CCS) to drive two hydrologic models (HYMOD and HBV-EDU), generating eight forecasts from different model-product-combinations following the approach described above. These probabilistic forecasts are then merged in an attempt to produce an improved forecast with higher accuracy and smaller uncertainty. These methods are applied in the Mara Basin in Kenya, facing serious water sustainability challenges, in an effort to support water management decisions balancing human and environmental needs, as part of the NASA SERVIR Applied Sciences Team.