为了验证风云三号D星MERSI传感器的气溶胶光学厚度（AOD）数据对地面PM_2.5的污染过程预报的效果，本文基于WRF-Chem（Weather Research and Forecasting model coupled with Chemistry）大气化学模式和三维变分同化方法，针对2020-02-10—2020-02-12中国北方地区的一次PM_2.5重污染过程，进行了同化和预报试验研究。同化数据来自常规地面站点的PM_2.5浓度数据和风云三号D星MERSI传感器的气溶胶光学厚度（AOD）数据。控制试验不同化任何资料，3组同化试验分别为仅同化地面PM_2.5，仅同化卫星AOD，以及同时同化PM_2.5和卫星AOD两种资料。结果表明，3组同化试验都可以有效提高初始场准确率，以地面PM_2.5作为检验标准，仅同化PM_2.5、仅同化AOD、同时同化两种资料相对于控制试验，初始场的平均偏差分别降低54.9%、21.9%和49.0%，平均相关系数分别提升51.4%、16.0%和34.0%，平均均方根误差分别降低50.6%、17.2%和42.3%。以卫星AOD作为检验标准，3组同化试验相对于控制试验，初始场的平均偏差分别降低37.6%、78.4%和83%，平均均方根误差分别降低31.6%、62.2%和65.2%。同化后的初始场对预报有显著的改进，改进持续时间达24 h，以地面PM_2.5作为检验标准，同时同化两种资料的试验对24 h预报的平均偏差减少19.7%，相关系数提升8.8%，均方根误差减少17.2%；以卫星AOD作为检验标准，24 h预报的平均偏差减少40.1%，相关系数提升25.9%，均方根误差降低34.7%。试验结论为，相对于仅同化地面PM_2.5资料，同化风云卫星AOD资料可以提升后期预报效果。

This study aims to verify the effect of the aerosol optical thickness data of the Fengyun-3D satellite MERSI sensor on the pollution process prediction of PM_2.5.This study was based on WRF-Chem (Weather Research and Forecasting Model Coupled with) Atmospheric Chemistry model and three-dimensional variational assimilation method, which were used to study the assimilation and prediction of a PM_2.5 pollution process in northern China from February 10 to 13, 2020.The assimilation data were derived from PM_2.5 concentration data from conventional ground stations and Aerosol Optical Depth (AOD) data from the MERSI sensor on the FY-3D satellite. The control experiment did not assimilate any data. The three groups of assimilation experiments were to assimilate ground PM_2.5, satellite AOD, and PM_2.5 and AOD data at the same time.Results show that the three groups of assimilation experiments can effectively improve the accuracy of the initial field. With ground PM_2.5 as the test standard, compared with the control experiment, assimilating PM_2.5 data, AOD data, and PM_2.5 and AOD data at the same time, the average mean deviation of the initial field was decreased by 54.9%, 21.9%, and 49.0%, the average correlation coefficient was increased by 51.4%, 16.0%, and 34.0%, and the average root mean square error was decreased by 50.6%, 17.2%, and 42.3%. With AOD as the test standard, compared with the control experiment, the average mean deviation of the initial field in three assimilation experiments was decreased by 37.6%, 78.4%, and 83%, and the average root mean square error was decreased by 31.6%, 62.2%, and 65.2%. The initial field after assimilation can significantly improve the prediction, and the improvement lasted for more than 24 h. In general, the experiment of assimilating two kinds of data at the same time had the best improvement effect on the 24 h prediction. With ground PM_2.5 as the test standard, the average mean deviation of the 24 h forecast was decreased by 19.7%, the correlation coefficient was increased by 8.8%, and the root mean square error was decreased by 17.2. With AOD as the test standard, the average mean deviation of 24 h forecast was decreased by 40.1%, the correlation coefficient was increased by 25.9%, and the root mean square error was decreased by 34.7%.The experiment also found that the assimilation of FY-3D satellite AOD data had a better lasting effect on the late prediction than only assimilating ground PM_2.5 data.