Abstract:Previous researches showed that the application of artificial intelligence (AI) techniques could substantially improve severe convective weather nowcasting capabilities. Cooperating with Baidu Inc., Chongqing Meteorological Service preliminarily implemented the intelligent nowcasting model of three dimensional radar echo using TrajGRU technique based on the radar data of 2008-2018. Together with the ground observation data and the U net technique, the intelligent identification model of thunderstorm gale and hail were established. The AI techniques above were used to operational thunderstorm gale, hail and precipitation nowcasting. Compared to the operational products of the Severe Weather Automatic Nowcast (SWAN) system, AI products showed obvious improvement in the prediction of radar echo evolution. The intelligent prediction ability of radar echo decreased with time at a lower speed than that of the traditional extrapolation method, and the prediction for longer lead times was improved. Verification showed that AI precipitation forecast had a higher skill than SWAN to a certain extent. During the 2019 trial operational application period, the lead times of intelligent forecast model for thunderstorm gale and hail forecast were 77 min and 37 min, respectively. The research and application of AI techniques in the early warning of severe convective weather in Chongqing have achieved initial results.