Evaluation of Quantitative Precipitation based on X-Band Phased Array Radar

Urban flash flooding has been one of main natural calamities that threaten human’s lives and property in megacities. High resolution rapid-fresh and accurate rainfall information play important role in prediction and timely warning of high impact weather event. Thanks to the rapid development of the sensing capabilities in the past decades of years, weather radar has great advantages to realize large-scale and finer space-time observation of precipitation storms, being an important part of operational weather observation. Apart from operational S-band polarimetric radar, Guangzhou has built the first network of 4 X-band polarimetric phase array weather radars (XPPARs) around the world in order to response to local-form and rapid-strengthen weather phenomena. Nowadays, four XPPARs have been under trial run for weather observation for more than one year. XPPAR has fast update rate of roughly 1 minute, subtle spatial resolution about 30 meters, adding vertical scanning mode, contributing to strengthening small–scale convection monitoring capacity, in collaborative coverage particularly. The real-time distributions of observation parameters from RHI/PPI scan mode are frequently used by forecasters to track high-impact heavy storms, which are regarded as a very effective instrument. XPPAR-based QPE has not been widely employed so far, let alone quantitative assessment during the experiment. It is worth conducting research to carry forward this task for the sake of deepening understanding of precipitation storm process. In this paper, we employed three formulas with various combination of Z_H, K_DP and Z_DR to compute the XPPAR-based QPE for a locally generated and rapidly intensifying storm occurred on 3rd, May, 2018, during pre-summer period. Then the XPPAR-based QPE was compared to dense gauge observations in terms of accumulated precipitation and hourly rainfall rate. Results show that: 1) both radar and gauge demonstrate similar spatial QPE distributions, but the intensity derived with R(Z_H) is much lower than both R(Z_H, Z_DR) and R(K_DP), which can be quantitative estimated by negative RB(-0.67) in R(Z_H), but positive one in the latter two. 2) In terms of temporal evolution characteristics, both R(Z_DR, Z_H) and R(K_DP) produce significant overestimation of precipitation, but R(Z_DR, Z_H) can generally capture the temporal variation while R(K_DP) failed to capture the trend. Among the three relationships, R(Z_H) has the best CC ( ~0.6910 ) and the smallest RMSE ( ~4.7453 ) but shows pronounced underestimation with RB about -58.09%. 3) As the distance increases, R(K_DP) has the minimal absolute values of RB and R(Z_H) owns the largest one, while R(Z_DR, Z_H) changes pronouncedly. 4) All results witness a dramatically decreasing CC, while R(Z_DR, Z_H) performs best, followed by R(K_DP), and R(Z_H) is the worst. 5) R(Z_H) has poorest performances in detecting heavy rain in this case with low probability of detection(POD).