基于大气化学模式的分析预报已成为有效应对PM2.5、PM10等气溶胶污染重要的手段之一。资料同化通过融合观测资料，能够有效改进大气化学模式的初始场，提高模式预报的准确性。卫星AOD资料由于覆盖面广、分辨率高，是气溶胶资料同化最重要的资料源，但由于卫星AOD与气溶胶的物种（化学成分）、粒径分布以及相对湿度之间存在复杂的非线性关系，难以直接进行同化。本项目将基于WRF-Chem模式中MOSAIC气溶胶方案的多物种和多粒径段变量，研究卫星AOD的直接同化方法；根据气溶胶光学特性、水平和垂直相关特征对气溶胶变量进行优化组合，设计区分物种及粒径大小的背景误差协方差，基于米散射理论构建复杂的AOD观测算子及其伴随算子；建立一个准确、高效的卫星AOD资料直接同化系统，并通过数值试验验证同化系统的合理性和可靠性，为提高气溶胶模拟和预报水平提供理论和技术支撑。

Modeling and forecasting based on atmospheric chemistry model has become an important ways to effectively deal with aerosol pollution such as PM2.5 and PM10. Data assimilation can improve the initial field of atmospheric chemistry model by integrating the observed data, so as to improve the accuracy of the model prediction. Because of its wide coverage and high resolution, satellite AOD data is the most important source of aerosol data assimilation. However, due to the complex non-linear relationship among satellite AOD and aerosol species (chemical composition), particle size distribution and relative humidity, it is difficult to directly assimilate. This project will focus on developing direct assimilation of satellite AODs based on the multi-species and multi-particle size bins of the MOSAIC aerosol program in the WRF-Chem model. Then, aerosol variables are optimally combined based on the aerosol optical properties, horizontal and vertical correlations to construct background error covariance. Developing Complex AOD observation operator and its adjoint operator based on the Mie theory. An efficient satellite AOD data assimilation system was developed and validated by numerical experiments. The system will provide theoretical and technical support for improving aerosol simulation and forecasting.