The Error Propagation of Global Precipitation Estimation through Distributed Hydrological Modelling

Abstract:

This study characterizes precipitation error propagation through distributed hydrological modelling based on an extensive number of basins across the CONUS, aiming to promote the understanding of the relationship between the errors in precipitation inputs and simulated discharge (P-Q error relationship). Relying on 9,322 streamflow gauges over the CONUS, we first evaluate the simulated discharge with a distributed hydrological model driven by NLDAS-2 precipitation. Then, using the NLDAS-2 precipitation and the simulated discharge as the reference at the continental scale, the relative errors in multiple conventional precipitation products and their simulated discharges for the period of 2002-2013. The results reveal positive linear P-Q error relationships at annual and monthly timescale, and stronger linearity for larger temporal accumulations. The stronger amplification effects on precipitation error propagation were observed at larger temporal accumulations, while stronger dampening effects for peak discharges. The study also highlights the role of climate regimes, basin size and topography on the patterns of P-Q error relationship and error propagation. Specifically, the strong seasonality in precipitation and streamflow for arid/semi-arid basins were found in western CONUS. The precipitation errors are strongly dampened for basins with temperate and continental climate regimes, particularly for peak discharges, showing highly nonlinear relationships. This study suggests that under the circumstance of lacking real-time validation (e.g., for applications of flood forecasting and monitoring), using high-resolution precipitation products and incorporating an evaluation of spatial variability in precipitation will improve the predictability of errors in simulated discharge.