One of the important products of MERRA-2 (Modern-Era Retrospective Analysis for Research and Applications) developed by NASA (National Aeronautics and Space Administration) is the long-term global records of surface PM2.5 mass concentration since 1980s, providing the ability of studying the interactions between air pollution, weather and climate changes. In this study, the PM2.5 mass concentrations of MERRA-2 are firstly validated across China mainland by independent surface measurements collected by Ministry of Ecology and Environment of People's Republic of China from 2014 to 2018. The results show that MERRA-2 well captures the spatial distribution and seasonal variation of PM2.5 mass concentration in China mainland. The spatial and temporal evolution of large scale persistent PM2.5 pollution event is also generally reflected by MERRA-2 by case study based on the target object match method. However significant underestimation of the PM2.5 mass concentration in MEERA-2 is revealed across China mainland, especially in BTH region by 34.6 μg m−3, followed by 19.8 μg m−3 in YRD and 9.1 μg m−3 in PRD region. Such underestimation is most substantial in winter and autumn seasons. In addition, the discrepancy between MERRA-2 and observations increases significantly with the enhanced PM2.5 level, for example, ranging from 29.9 μg m−3 in clean days, while 66.1 μg m−3 in polluted days in BTH region. We highlight the downward trends of PM2.5 from 2014 to 2018 in China mainland estimated by MERRA-2 which is basically consistent in the observations, but with ~50% underestimation, indicating the potential applications of MERRA-2 for the future aerosol climatological studies. We suggest that the underestimations of both magnitude and variability of PM2.5 in MERRA-2 probably result from the uncertainty of the magnitude of emission inventory used in GOES model (lower intensity and weaker variations), and the absence of nitrate in PM2.5 constitution. A parameterized method for nitrate is proposed and evaluated by the sensitive study to improve MERRA-2 PM2.5 underestimation by 19.2–23.6% in BTH region. However, the more comprehensive validations are still required in future studies, especially by the aerosol composition measurements. •The PM2.5 mass concentrations of MERRA-2 are firstly validated across China mainland from 2014 to 2018.•Significant underestimation of the PM2.5 concentration in MEERA-2 is revealed across China mainland.•A parameterized method for nitrate is proposed to improve MERRA-2 PM2.5 underestimation.