Abstract: The 2σ lightning jump algorithm is a method used to predict severe weather events, such as hail and tornadoes, by analyzing the rate of change in lightning frequency. However, this algorithm exhibits false alarm issues in practical applications, and its applicability varies in actual effectiveness across different regions. To address these limitations, the algorithm is improved by incorporating determination conditions based on the variation characteristics of the gradient of vertical integrated liquid (GVIL), which represents changes in radar vertically integrated liquid water (VIL) during hail processes. The performance of the improved algorithm in predicting hail is systematically evaluated using radar and lightning data from 116 hail events that occurred in Hubei Province between 2015 and 2023. The results indicate that the improved 2σ algorithm maintains a high probability of detection (POD=90.5%), while reducing the false alarm rate (FAR) from 41.4% to 20.1%, increaing the critical success index (CSI) from 56.4% to 73.7%, and providing an average warning lead time of approximately 21.8 minutes. By applying it to typical hail cases, the effectiveness and applicability of the improved 2σ algorithm are further validated. This study provides a new technical approach to enhance hail warning capabilities in the Hubei region.