Ground-level PM2.5 concentrations—especially those during episodes of heavy pollution—are severely underestimated by mixed-effects models that ignore the effects of primary pollutant emissions and secondary pollutant conversion. In this work, meteorological parameters and NO2, SO2, CO, and O3 concentrations are introduced as predictors to a mixed-effects model to improve the estimated concentration of PM2.5, which is based on the Moderate Resolution Imaging Spectroradiometer (MODIS) aerosol optical depth (AOD). The Beijing-Tianjin-Hebei (JingJinJi) region is used as the study area. The model provides an overall cross-validation (CV) R2 of 0.84 and root-mean-square prediction error (RMSE) of 33.91 µg m–3. The CV R2 and RMSE of the proposed model are higher by 0.11 and lower by 9.16 µg m–3, respectively, than those of a model lacking gaseous pollutants as predictors. The R2 and RMSE of the proposed model increases and decreases by 0.14 and 13.37 µg m–3, respectively, when PM2.5 concentrations exceed 75 µg m–3. The high values predicted for the PM2.5 concentration indicate a drastic improvement in the estimation, and the spatial distribution generated by the model for periods of heavy pollution is highly consistent with that inferred from monitoring data. Thus, the proposed model can be used to generate highly accurate maps of the PM2.5 distribution for long-term and short-term exposure studies and to correctly classify exposure in heavily polluted areas.