Improving Polarimetric C-Band Radar Rainfall Estimation with Two-Dimensional Video Disdrometer Observations in Eastern China

Abstract

Abstract In this study, the capability of using a C-band polarimetric Doppler radar and a two-dimensional video disdrometer (2DVD) to estimate monsoon-influenced summer rainfall during the Observation, Prediction and Analysis of Severe Convection of China (OPACC) field campaign in 2014 and 2015 in eastern China is investigated. Three different rainfall R estimators, for reflectivity at horizontal polarization [R(Zh)], for reflectivity at horizontal polarization and differential reflectivity factor [R(Zh, Zdr)], and for specific differential phase [R(KDP)], are derived from 2-yr 2DVD observations of summer precipitation systems. The radar-estimated rainfall is compared to gauge observations from eight rainfall episodes. Results show that the two polarimetric estimators, R(Zh, Zdr) and R(KDP), perform better than the traditional Zh–R relation [i.e., R(Zh)]. The KDP-based estimator [i.e., R(KDP)] produces the best rainfall accumulations. The radar rainfall estimators perform differently across the three organized convective systems (mei-yu rainband, typhoon rainband, and squall line). Estimator R(Zh) overestimates rainfall in the mei-yu rainband and squall line, and R(Zh, Zdr) mitigates the overestimation in the mei-yu rainband but has a large bias in the squall line. QPE from R(KDP) is the most accurate among the three estimators, but it possesses a relatively large bias for the squall line compared to the mei-yu case. The high variability of drop size distribution (DSD) related to the precipitation microphysics in different types of rain is largely responsible for the case-dependent QPE performance using any single radar rainfall estimator. The squall line has a distinct ice-phase process with a large mean size of raindrops, while the mei-yu rainband and typhoon rainband are composed of smaller raindrops. Based on the statistical QPE error in the ZH–ZDR space, a new composite rainfall estimator is constructed by combining R(Zh), R(Zh, Zdr), and R(KDP) and is proven to outperform any single rainfall estimator.