This research constructed an air quality ensemble forecasting model consisting of fifteen members using the China Meteorological Administration regional ensemble forecasting system (CMA_REPS) and Comprehensive Air Quality Model Extensions (CAMx) models to investigate the influence of atmospheric field uncertainty on air quality simulations. Focusing on the Beijing Winter Olympics in February 2022, this study examines the effects of both ground-level and vertical meteorological conditions on PM2.5 concentration distributions. The simulation accuracy of the model was validated, and its performance was analyzed. Results revealed that the ensemble mean simulations exhibit high correlation coefficients with observations for temperature (0.95), wind speed (0.80), relative humidity (0.83), and pressure (0.99). Both the control forecast and the ensemble mean for PM2.5 concentration aligned well with observations, with the ensemble mean demonstrating a strong correlation between the root mean square error and ensemble spread. In terms of reducing the false alarm rate (FAR) and improving the Bias Score (BS), the ensemble mean outperformed the control forecast. The control forecast for PM2.5 concentration was found to be more accurate at and around pollutant concentration inflection points, which may be attributed to simulation deviations in temperature and pressure that introduce uncertainty in atmospheric stability simulations. The correlation between PM2.5 and various meteorological elements varied during different periods. The vertical distribution of meteorological factors also significantly affected simulation outcomes, particularly uncertainties in simulating wind speed and inversion temperature processes, which further contributed to the uncertainty in pollutant simulations.