Multiple Aerosol Data Assimilation Based On IMPROVE Observation Operator

A three-dimensional variational assimilation system for the multiple aerosol observation data in the Weather Research and Forecasting model coupled to Chemistry?WRF-Chem?model with Model for Simulating Aerosol Interactions and Chemistry?MOSAIC?aerosol scheme is established,using which the assimilation of multi-source observations including lidar-derived aerosol extinction coefficients,satellite-derived aerosol optical depth?AOD;or aerosol optical thickness,AOT?and aerosol mass concentration can be implemented.The observation operator scheme in the assimilation system is based on IMPROVE?The Interagency Monitoring of Protected Visual Environments?equation.The background error covariance of 16 control variables within the MOSAIC was statistically analyzed,of which the structure characteristics were analyzed.It is found that the background error standard deviation of different control variables is quite different with the OIN_2.5-10.5-10 has the maximum value and the NA_2.5-10.5-10 has the minimum value near the ground;the background error standard deviation decreases with height and the rate of decrease is fastest in the upper part of the boundary layer;the attenuation scale of the background error horizontal correlation coefficient is different among control variables,with the NH4_2.5 has the maximum value of 25.5km,and the OIN_2.5-10.5-10 has the minimum value of 13.2km.The ideal assimilation test shows that the assimilation incremental field corresponds well with the structural of the background error covariance.Assimilation experiments for multi-observation data during March 8-15,2018,including mass concentrations of PM_2.5 and PM₁₀ in air quality monitoring stations,lidar aerosol extinction coefficients,and Himawari-8 satellite AOD,were designed to test the improvement effect of assimilation on aerosol simulation and prediction.The RMSE of the PM data assimilation,the extinction coefficient data assimilation and the combined assimilation in the third domain?D03?were reduced by 140.6?gm^-3?79.0%?,28.1?gm^-3?15.8%?and 140.2?gm^-3?78.8%?compared with that without data assimilation?DA?,respectively.The duration of DA positive impact is about 17 hours.The RMSE of the PM data assimilatiot,the AOD data assimilation and the combined assimilation in the second domain?D02?were reduced by 50.3?gm^-3?70.7%?,22.9?gm^-3?32.2%?,and50.9?gm^-3?71.3%?.The duration of DA positive impact is more than 24 hours.