The impact of radar and raingauge sampling errors when calibrating a weather radar

This paper investigates the impact of sampling errors on the estimation of the parameters of the relation Z = aRb used in the measurement of rainfall (R) from radar reflectivity (Z). Simulation results show that, at levels of sampling errors likely to be observed in real data sets, of the order 100 gauge and radar observation pairs are required to estimate a with an accuracy better than 15% and b better than 10%. Errors in radar rainfall estimates for rain rates less than 15 mm h−1 are insensitive to the Z-R parameters. High-resolution measurements from a vertically pointing radar (300 to 400 m above ground level, 1-minute averages) are compared with raingauge measurements taken at the radar site during a period of widespread rainfall. The radar provided surprisingly poor estimates of rain rates, with mean standard errors of 80% for rain rates less than 2 mm h−1. The auto-correlations of the residuals in the radar estimates were found to be very low after only 2 minutes, leading to the conclusion that radar estimates of rainfall intensities should be made at a high space/time resolution and then averaged so as to reduce measurement errors arising from sampling problems.