An Integrated Approach to Error Correction for Real-Time Radar-Rainfall Estimation

Abstract

Abstract A procedure for estimating radar rainfall in real time consists of three main steps: 1) the measurement of reflectivity and removal of known sources of errors, 2) the conversion of the reflectivity to a rainfall rate (Z–R conversion), and 3) the adjustment of the mean field bias as assessed using a rain gauge network. Error correction is associated with the first two steps and incorporates removing erroneous measurements and correcting biases in the Z–R conversion. This paper investigates the relative importance of error correction and the mean field bias–adjustment processes. In addition to the correction for ground clutter, the bright band, and hail, the two error correction strategies considered here are 1) a scale transformation function to remove range-dependent bias in measured reflectivity resulting from an increase in observation volume with range, and 2) the classification of storm types to account for the variation in Z–R relationships for convective and stratiform rainfall. The mean field bias is removed using two alternatives: 1) estimation of the bias at each time step based on the sample of observations available, and 2) use of a Kalman filter to estimate the bias under assumptions of a Markovian dependence structure. A 7-month record of radar and rain gauge rainfall for Sydney, Australia, were used in this study. The results show a stepwise decrease in the root-mean-square error (rmse) of radar rainfall with added levels of error correction using either of the two mean field bias–adjustment methods considered in our study. It was found that although the effects of the two error correction strategies were small compared to bias adjustment, they do form an important step of radar-rainfall estimation.