Abstract:To effectively utilize the characteristics of atmospheric electric field signals and improve the precision of lightning risk warning, using data from an atmospheric electric field instrument installed on a 500 m-height platform at the Guangzhou Tower within a 20 km (radius) range, this paper analyzes the full day (00:00 BT-24:00 BT) data from 83 days with thunderstorm processes and 123 days with non-thunderstorm processes corresponding to the instrument in 2021 and 2022, and then proposes a lightning risk warning method combining time-frequency domain features and one-dimensional Morpho based on the fusion of enhanced empirical wavelet transform and adaptive Savitzky-Golay (EEWT-ASG). This proposed method uses the spectral width first-order backward difference and mean square error of time-frequency domain features as warning and judgment features for lightning and non-lightning processes, and energy difference is used as a feature for judging the release of warnings. Through selecting samples for effectiveness testing in this article, the proposed lightning risk warning method achieves an accuracy (POD) of 77.11% during thunderstorms, and also has the lowest false alarm ratio (40.00%) and highest critical success index (0.51) performance. Besides, an average warning lead time of 22.27 min is achieved. During non-thunderstorm processes, POD reaches 90.24%. Most warnings have a delay time of 0-40 min, with an average delay time of 32 min. According to the comparison with previous algorithm models, the method proposed in this paper for warning and relief can meet the needs of lightning risk warning in industries with high lightning impact.