Uncertainty assessment of radar-raingauge merged rainfall estimates in river discharge simulations

Abstract

Due to the huge potential of radar-raingauge merged rainfall products for improving quantitative rainfall estimates, the expectation of their applications in hydrological simulations has always been high. This study investigates the main strengths and limitations of radar-raingauge merged rainfall products over traditional radar-only and raingauge-only rainfall products for discharge simulations in both deterministic and stochastic ways. The Probability Distributed Moisture (PDM) model, was used to generate hourly discharge simulations over near-natural catchments in England. The generalized likelihood uncertainty estimation (GLUE) method was applied to quantify the uncertainty in river discharge simulated by using different rainfall estimates. The results suggest that the use of radar-raingauge merged rainfall estimates for hydrological simulations is beneficial even if the quality of radar rainfall is poor and there is a limited number of raingauges. Furthermore, The GLUE uncertainty analysis show that the posterior parameter distributions not only depend upon the choice of acceptability threshold values, but also on the quality of rainfall estimates. When increasing the GLUE threshold value, the parameter values are more constrained to a certain range to generate the behavioral runs. Rainfall estimators with good quality are more tolerant to the range of parameter sets, and leading to more “equifinality” and uncertainty in the parameters, particularly for a rainfall dependent parameter. This suggests that any factors affecting the quality of the rainfall estimates will have an impact on the parameter uncertainty and hence discharge simulations.