背景误差协方差矩阵不同求逆方案在高度计资料同化试验中的应用比较

The comparison of different inversion methods of a background error covariance matrix in assimilation experiments of altimeter data

针对海表高度计资料的同化,考查了背景误差协方差矩阵的不同求逆方案对同化效果的影响。所使用的求逆方案包括避免求逆的经验正交函数方案(EOF/EOF_var)、递归滤波方案(RF/RF_var)以及采用初等变换法直接求逆的方案(Inv)。基于上述方案开展了热带太平洋地区2002年1-7月的TOPEX/Poseidon高度计资料同化试验,并利用SODA再分析资料和TAO观测资料评估了各方案对温度场的同化效果,主要得到如下结果:与SODA相比较,Inv方案对模式温度场改进甚微,其余四种方案在100～300m深度之间对温度场改进较多,在其它深度范围内则改进较少;与TAO观测相比较,EOF_var、RF_var方案对模式温度场改进最多,EOF和RF方案次之,Inv方案则对温度场改进甚少。

For the assimilation of altimeter data,the influence of different analysis schemes on the assimilation effects is analyzed.Differing in the inversion methods of a background error covariance matrix,these schemes include a empirical orthogonal function scheme and its variant variance form（abbreviated as EOF and EOF_var）,a recursive filter scheme and its variant variance form（abbreviated as RF and RF_var）,and a matrix inversion using a elementary transformation method（abbreviated as Inv）.Using the above schemes,some assimilation experiments,based on TOPEX/Poseidon satellite altimeter data from January to July 2002,were carried out in tropical Pacific Ocean,and then the assimilation effects were assessed using SODA reanalysis data and TAO observations.On the one hand,the comparison with the SODA data shows that the Inv scheme has less improvement on the modeling temperature field,and the other four schemes improve the model fields more clearly in depth from 100 to 300 m and less in other range.On the other hand,the comparison with the TAO observations indicates that the EOF_var and RF_var are the most effective scheme to improve temperature field,next are the EOF and RF schemes,least is Inv scheme.