Calibration of radar reflectivity measurements from the KMA operational radar network

A systematic procedure for calibrating system gain bias (so-called "calibration error") of radar reflectivity measurements from the Korea Meteorological Administration (KMA) operational radar network is presented. First, the RJNI radar located at Jindo Island is calibrated by comparing with radar reflectivities simulated theoretically by a scattering algorithm using drop spectra collected by a disdrometer from June 19 to 29, 2009. Once the RJNI radar is calibrated, the reflectivity measurements from nearby radars are compared with the RJNI radar reflectivities to determine existing gain biases of nearby radars. This radar-radar calibration procedure was repeated with the other radars within the network. For isolating a system gain bias, echoes affected by partial beam blockage due to ground clutter and by attenuation due to precipitation were removed. The system gain biases of the RJNI and RPSN radars were -3 and -4.2 dB, respectively, during the experimental period. The RBRI and RDNH radars revealed relatively large biases, above -8 dB. The other radars (RKSN, RGSN, RSSP, RKWK, RGDK, RIIA, and RMYN) revealed biases from -6 to -7 dB. Thus, the reflectivity measurements from all of the KMA radars were severely biased. New R-Z relations of R = 3.350 × 10^-2Z ^0.624 (Z = 231.1R ^1.6) for stratiform and R=1.546 × 10^-2 Z ^0.714 (Z = 342.4R ^1.4) for convective precipitations were derived using disdrometer data. Using these R-Z relations, the radar-derived total rainfall amounts from the reflectivity measurements without calibration produced significant underestimations, compared to gauge measurements at about 80 sites, with a normalized bias of about -56%. On the other hand, after calibrating the above system biases, the radar-derived rainfall amounts corresponded well with the gauge measurements, with a normalized bias of about -3%. In conclusions, the radar reflectivity measurements from the KMA radar network are severely biased and the procedure presented in this study can be used to resolve the system gain biases.