A method for improving the performance of the 2σ lightning jump algorithm for nowcasting hail

Abstract

Studies have shown that the 2σ lightning jump (LJ) algorithm exhibits good performance for severe weather warnings but continues to produce a high false alarm rate (FAR). A method is proposed in this study for improving the 2σ LJ hail warning performance by combining the hydrometeor classification results of dual-polarization radar with the 2σ LJ algorithm. A total of 17 hail events occurred across Jiangsu, China on April 29th -30th, 2021. Dual-polarization radar and hail observation data were used to calculate five variables, including the rate, first derivative (FD), rate of rate (rate2), second derivative (SD) and the rate of FD (FD_rate), for large hail (HA), graupel and small hail (G/SH), rain and hail (RH), and ice crystal (IC) grid point numbers. The evolution of these variables was compared and analyzed. The results indicate that the positive local peaks of rate2 of the HA and G/SH grid point numbers can be used to effectively identify valid and invalid LJs. Removing the identified invalid LJs reduces the FAR without affecting the probability of detection (POD), thus improving the hail warning performance of the 2σ LJ algorithm. This method was tested for the aforementioned 17 hail events. The rejection rate of invalid LJs reached 69.6%, and the false identification rate of valid LJs was only 9/29 (31%). The proposed method for hail warning had a POD equal to that of the conventional 2σ LJ algorithm (100%) and a significantly lower FAR (29.2% versus 58.5%). Moreover, the critical success index (CSI) of the proposed method (70.8%) was considerably higher than that (41.5%) of the conventional 2σ LJ algorithm. However, the average early warning lead time of the proposed method (35.1 min) was slightly lower than that (37.9 min) of the conventional 2σ LJ algorithm. The proposed method is superior to the conventional 2σ LJ algorithm in terms of the FAR and CSI for hail warning and therefore considerably improves the ability of the 2σ LJ algorithm to produce refined early warnings of hail or even severe weather for meteorological services.