Abstract: An improvement on the trigger function of the shallow convection scheme in the GRAPES-Meso has been designed and tested referring the treatment suggested by Berg, et al (2005, 2013), in which the trigger function is related to the distribution of temperature and humidity in the convective surface layer determined by the Joint Probability Density Functions (JPDF). The results show that more shallow convection could be triggered and occur in an earlier simulation time than the original one, and the stronger feedbacks of temperature, moisture and water vapor condensate of cloud to model grid scale are correspondingly identified in the lower layers from several hundred meters to about 2-3 km above the ground; with an improvement in the shallow convection, there is a positive impact on the fraction of subgrid/grid scale precipitation to the total precipitation of GRAPES-Meso. The verification for the forecast results of 2-month experiment indicates that the improvement of the trigger function of shallow convection scheme could result in the better performances of 24 h forecasts of accumulative precipitation and surface air temperature.