传统三维大气模式在京津冀的模拟存在较大不确定性。结合观测的三维变分资料同化方法，通过纳入观测修正模拟，同时产生三维再分析场，有效提高气溶胶模拟能力。NCAR团队曾基于GSI-WRF/Chem系统，开发了面向GOCART沙尘气溶胶的同化系统；但是该机制并不适宜表达中国重污染地区气溶胶组分特点。本研究拟基于GSI-WRF/Chem同化系统，建立面向MOSAIC多物种多粒径段气溶胶方案的同化模块，并将该系统应用于京津冀地区。主要研究工作包括：（1）建立面向MOSAIC方案的分析变量系统和同化框架；（2）重建观测算子，构建针对环保部国控网的PM2.5观测误差模块；（3）建立面向MOSAIC方案的背景误差协方差模块。拟利用该系统构建纳入观测的2015-2017冬季逐小时PM2.5三维再分析场，分析京津冀地区冬季污染变化特征（时空变化、年际变化及驱动要素等），揭示该区域气溶胶污染的气象和化学机制。

The winter-time aerosol pollution in Beijing-Tianjin-Hebei Region fluctuate during 2015-2017. Simulations by traditional 3-D regional air quality models have large uncertainties for that region. 3DVAR data assimilation combining the observational data and regional models, can correct concentration fields and also generate reanalysis fields, thus improve model performances. NCAR group developed the GSI-WRF/Chem data assimilation system for the GOCART dust aerosol scheme. However this scheme is not appropriate for the aerosol characteristics in the heavy polluted regions in China. In this proposal, we will build the GSI-WRF/Chem system for the MOSAIC aerosol scheme and apply the system to Beijing-Tianjin-Hebei Region. The development includes three parts: (1) the framework of the DA system and the control variable interface, (2) Observation operator and observation error modules to use the MEP PM2.5 observation network, (3) Background error covariance module for the MOSAIC scheme. By using the new DA system, we will also generate 2015/2016/2017 winter-time PM2.5 reanalysis fields and investigate the winter-pollutions in Beijing-Tianjin-Hebei region, focusing on the temporal/spatial changes, annual changes and driving factors.