Abstract. Lightning is a major threat to people's daily activities. Therefore, to forecast its occurrence is very important to ensure the safety of life and property. Compared to a series of signals from instruments like the Doppler radar, the lightning locator and the meteorological satellite, the electromagnetic field signal is easy to be detected by the atmosphere electronic field instrument and strongly correlated with lightning. We analyze the defections of traditional warning methods through setting threshold on the atmospheric electronic field value and propose a new method which uses the first-order derivative to model the persistent and low-intensive fluctuation characteristic of atmospheric electronic field before the lightning. The experiment on real data proves that our method could improve the lightning warning accuracy.