Statistical analysis of error propagation from radar rainfall to hydrological models

摘要

This study attempts to characterise the manner with which inherent error inradar rainfall estimates input influence the character of the stream flowsimulation uncertainty in validated hydrological modelling. An artificialstatistical error model described by Gaussian distribution was developed togenerate realisations of possible combinations of normalised errors andnormalised bias to reflect the identified radar error and temporaldependence. These realisations were embedded in the 5 km/15 min UK Nimrodradar rainfall data and used to generate ensembles of stream flowsimulations using three different hydrological models with varying degreesof complexity, which consists of a fully distributed physically-based modelMIKE SHE, a semi-distributed, lumped model TOPMODEL and the unit hydrographmodel PRTF. Thesemodels were built for this purpose and applied to the Upper Medway Catchment(220 km2) in South-East England. The results show that the normalisedbias of the radar rainfall estimates was enhanced in the simulated streamflow and also the dominate factor that had a significant impact on streamflow simulations. This preliminary radar-error-generation model could bedeveloped more rigorously and comprehensively for the error characteristicsof weather radars for quantitative measurement of rainfall.