Development of Three‐Dimensional Variational Data Assimilation Method of Aerosol for the CMAQ Model: An Application for PM2.5 and PM10 Forecasts in the Sichuan Basin

Abstract

AbstractA three‐dimensional variational (3DVAR) data assimilation method for the aerosol variables of the community multiscale air quality (CMAQ) model was developed. This 3DVAR system uses PM2.5 and PM2.5‐10 (the difference between PM10 and PM2.5) as control variables and used the AERO6 aerosol chemical mechanism in the CMAQ model. Two parallel experiments (one with and one without data assimilation [DA]) were performed to evaluate the assimilating effects of surface PM2.5 and PM10 during a heavy haze episode from January 13 to 16, 2018 in the Sichuan Basin (SCB) region. The results show that simulations without DA clearly underestimated PM2.5 and PM10 concentrations, and the analysis field with aerosol DA is skillful at fitting the observations and effectively improving subsequent forecasts of PM2.5 and PM10. For the analysis fields of PM2.5 and PM10 after DA comparing with those without DA, the correlation coefficient (CORR) of PM2.5 and PM10 increased by 0.59 and 0.65, the bias (BIAS) increased by 82.29 and 125.41 μg/m3, and the root mean square error (RMSE) declined by 73.69 and 116.30 μg/m3, respectively. Improvement of subsequent 24‐h forecasts of PM2.5 and PM10 with DA is also significant. Statistical results of forecasting improvement with DA indicated that the CORR, BIAS, and RMSE for PM2.5 and PM10 at 78% and 89% of stations in the SCB region are improved, respectively. From the perspective of assimilation duration time, the improvement of PM2.5 and PM10 can be maintained for ∼24 h.