中文摘要：

在客观分析中,背景误差协方差对观测信息的传播和平滑、反映不同变量之间的关系有着非常重要的作用.构造合理的背景误差协方差矩阵对于同化系统至关重要,甚至会决定同化分析的好坏.作者主要利用观测余差方法,用T213预报资料和无线电探空观测资料统计我国区域的背景位势高度误差协方差样本,分析背景误差协方差场的结构特征和拟合误差场的空间分布.

英文摘要：

Background error covariance is very important to govern the amount of smoothing and spreading of the observed information and to decide the relationships between different variables in variational data assimilation. Because of the existence of a balance in the reality and in the model state, there is a version of the balance that exists in the background error covariances. Background error covariances depend on the uncertainty of the previous analysis and forecast. To a large extent, the form of this background error covariance governs the resulting objective analysis. With the development of data assimilation, the methods to estimate the forecast error correlation structure have been reported in many literatures. However there is a little work about background error covariance in our country and the work is needed in the operational data assimilation system and GRAPES （Global and Regional Assimilation and PrEdiction System） 3D Var （three-Dimensional Variational data assimilation） research. So the statistical struc- ture of background error covariance is studied in this paper. It is difficult to directly get error covariances, which can only be estimated in a statistical sense. In order to get the height background error covariance, the innovation vector method is used in this paper. The data consist of innovation data （12 h and 24 h predicted height of T213 model minus radiosonde measurements） at 0000 UTC and 1200 UTC. Horizontal characteristic length, prediction error variance and observation error variance are obtained using Gauss correlation function approximation in a particular level. The straightforward way and the empirical thickness method are used to get the approximate function in interlevel values. In the vertical direction, vertical covariance approximation is obtained by the second-order autoregressive （SOAR） correlation function and distance transformation method, The resulting three-dimensional approximation function is partially separable, which is the product of the horizon