Abstract: Based on the data from an atmospheric electric field (AEF) meter at 500 m from the top of Canton Tower, dual-polarimetric radar in Guangzhou, Guangdong-Hong Kong-Macao lightning location system and ERA5 reanalysis data, this study comparatively investigates the characteristics of cloud-to-ground lightning (CG) activity and the relationship between CG activity and variation features of AEF in two thunderstorms in Guangzhou on August 15 and September 23, 2020. The results show that the"local Thunderstorm"caused by local thermal instability possesses is associated with less lightning frequency and weaker precipitation, while the"systematic Thunderstorm"caused by mesoscale system possesses is associated with higher lightning frequency and short-term heavy precipitation. The lightning frequency and variation of AEF in the two thunderstorm processes display the distributions with two peaks, and the peak of AEF variability and time-frequency domain energy correspond well to the lightning outbreak stage. In the decay of "systematic Thunderstorm", AEF fluctuates greatly and slowly between the positive and negative poles for a relatively long time, showing the characteristic of End of Storm-Oscillation (EOSO). However, such a characteristic of the oscillation of AEF does not occur in the decay of the"local Thunderstorm". The lightning often comes soon, when the AEF variability and time-frequency domain energy increase significantly. The thunderstorm is often dying out or moving away from the station, when the time-frequency domain energy decreases significantly and concentrates in the low frequency band. The variation characteristics of AEF in time domain and frequency domain play an indicative role in the occurrence and dissipation of thunderstorm. This can be considered as a reference for converged application of the atmospheric electric field data with other data in lightning risk early warning.