Radar‐derived parameters in hail‐producing storms and the estimation of hail occurrence in Romania using a logistic regression approach

Abstract

This study investigated the radar‐derived parameters in hail‐producing storms and compared two hail‐detection approaches in order to select the optimal method for real‐time hail detection for the southern part of Romania. A statistical analysis of radar parameters was made, and the thresholds for radar‐deriver parameters in hail‐producing storms were determined. In addition, the hail storms’ life cycle and diurnal and seasonal cycles of hail occurrence were investigated. The first algorithm developed in this study was based on the logistic regression model (LRM); the second one on the vertically integrated liquid (VIL) density algorithm. These algorithms were tested for the convective seasons (May–September) of 2014 and 2015. For the construction of the LRM, the composite reflectivity (CR), VIL and echo‐top high (ET) were used as predictors for the probability of hail occurrence determination for 43 hail events observed at the synoptic station located in the radar‐coverage area. The obtained statistics showed a marked diurnal cycle of hail occurrence with a maximum in the afternoon hours (1500–1800 LT), and the highest occurrence frequency of hail falls was in June. The results revealed that 93% of hail events occurred with 60 dBZ CR, 87% with 13 km ET, 100% with 55 kg/m2 VIL and 100% with 4.5 g/m3 VIL density. For the two hail‐detection methods’ performance comparison, a two‐by‐two contingency table was constructed and several statistical scores calculated. The LRM algorithm performed better than the VIL density algorithm.