Abstract: The physics package of the NCEP Global Forecast System (GFS) and its single column version have been implemented in the GRAPES-MESO model. Using the data collected at the ARM (Atmospheric Radiation Measurement Program) Southern Great Plain site during the intensive observing period in the summer of 1997, several experiments of single-column model (SCM) are conducted to test the performance of a set of original physical schemes of GRAPES-TCM and the GFS physics package implemented. Temperature, moisture, radiation, surface heat flux, surface air temperature and precipitation are evaluated. It is found that the potential temperature and water vapor mixing ratio simulated by the GFS experiment is more accurate than that simulated by the original physical schemes in GRAPESTCM. Errors of surface downward solar and long-wave radiation fluxes simulated by the GFS experiment are less than that by the original physical schemes and upward latent and sensible heat fluxes also agree better with observations. The maximum and minimum 2-m air temperatures of the GFS experiment are close to observations compared with that simulated with the original physical schemes. Analysis of precipitation simulation shows that both sets of physical schemes can well reproduce heavy rainfall events. The failure and delay of moderate rainfall events and overestimation of drizzle events are commonly found for the two sets of experiments. For the case studies of three rainfall events, the errors of precipitation rate simulated by the GFS experiment are smaller than that simulated by the original physical schemes. This study also analyzes cases in which rainfall events are either missed or delayed in forecasts of both experiments and reveals that the modification of the trigger method in the cumulus parameterization scheme can significantly improve the simulation of precipitation, which subsequently improves the simulation of large scale fields via more accurate feedback.