一个用于气候模式的三维变分资料同化系统——方案设计及检验

介绍了一种适合格点模式的三维变分同化方案,并应用于中国科学院大气物理研究所的9层大气环流格点模式（IAP9L2°×2.5°-AGCM）中。同化方案考虑质量场和风场的动力约束关系,并通过变量变换将模式变量转化为互不相关的控制变量：流函数、非平衡速度势、非平衡位势和比湿;同时采用共轭梯度法求解代价函数避免了直接计算背景误差协方差矩阵逆的困难。通过理想试验、模式连续同化积分试验以及对东亚地区特殊寒潮天气的同化试验,结果表明：该系统能够正确反应多变量之间的相互关系,并且长时间运行稳定,更重要的是同化不仅可以明显改善直接参与同化的物理量,而且经过模式的动力调整也改善了未参与同化的物理量（如海平面气压、近地面气温、降水等）,这为大气的持续性信息存储于陆面物理量提供了可能,从而可为IAP9L2°×2.5°-AGCM提供陆—气相协调的初始场。

A Three-Dimensional Variational Data Assimilation System for a Climate Model(AGCM)——Basic Scheme and Tests

A scheme of three-dimensional variational assimilation（3D-Var） is used in a grid-point At- mospheric General Circulation Model with 9 vertical levels, developed by Institute of Atmospheric Phys- ics, Chinese Academy of Sciences （IAP9L2°× 2.5°-AGCM）. In the scheme, model variables are converted to control variables which are stream function, unbalanced potential velocity, unbalanced geopotential height and specific humidity by considering relationships among multi-variables and using variable transformation. In order to avoid the complexity of background error covariance and reduce the cost of minimization, Conjugate-Gradient Method is used. Many experiments are performed for testing the validity of the system and dynamical constraints between mass field and wind field,including single observation, continuous assimilation added in the model and special cold wave event in East Asia. Results show that the relationships among multi-variables are reasonable, and the system is stable for a long time, more importantly, it is corresponded to dynamical model. The variables which participate in data assimilation are improved obviously, and some other variables （for example, sea level pressure, surface air temperature and precipitation） become better too. Therefore ,it is possible for the atmospheric information to be reserved in land surface variables through data assimilation for some time. Thus it may be provide harmonious initial conditions including land and atmosphere for the IAP9L2° ×2.5°-AGCM model.