Abstract：This paper evaluated the performance of National Aeronautics and Space Administration （NASA）’s Modern-Era Retrospective Analysis for Research and Applications， Version 2 Reanalysis （MERRA-2） over Guangdong by validating its PM2.5 concentrations against that of the hourly observations from 2015 to 2021. The results show that PM2.5 of MERRA-2 were consistent with those of measured ones when the concentrations were less than 50 μg?m-3， but were obviously underestimated when the pollution was serious. Moreover， the simulation performance of MERRA-2 had obvious seasonal dependence， which performed better in the wet seasons than in dry seasons. The difference between MERRA-2 and observed values increased significantly with increasing pollution levels， mainly due to the lack of nitrate simulation and the errors of the emission inventory. Comprehensively considering the impact of meteorological conditions， a parameterized method for nitrate and ammonium was proposed based on the 2016 datasets of aerosol chemical compositions in Guangzhou Panyu atmospheric composition site， which was used to correct the aerosol composition of PM2.5 under unfavorable diffusion conditions. The correction results showed that the parameterization can effectively reduce the deviation between MERRA-2 and observations.