观测误差协方差估计下的集合鲁棒滤波数据同化方法

在数据同化方法中，观测误差协方差矩阵是相关的，且与时间和状态有一定的依赖性。针对这种相关特性，将鲁棒滤波方法与观测误差协方差估计方法相结合，得到随状态时间变化的观测误差协方差，提出一种带有观测误差估计的鲁棒数据同化新方法，更新观测误差协方差，改善估计效果。从分析误差协方差，转移矩阵特征值放大等角度优化同化方法。利用非线性Lorenz-96混沌系统，对三种不同优化角度下带有观测误差估计的鲁棒滤波和原鲁棒滤波方法的鲁棒性和同化精度进行评估，并比较分析了两种方法在模型误差、观测数目和性能水平系数变化时的性能。结果表明：观测误差估计技术能够提高状态估计的精确性，带有观测误差估计的鲁棒滤波对系统参数变化具有较好的鲁棒性。

Ensemble Robust Filtering Data Assimilation Method with Estimation of Observation Error Covariance

In the data assimilation method， the observation error covariance matrix is correlated and dependent on time and state. in view of such correlation characteristics， the robust filtering method is combined with the estimation of observation error covariance to obtain the time-varying covariance of observation error， and the robust data assimilation method with observation error estimation is proposed to update the observation error covariance and improve the estimation performance. In this work， nonlinear lorenz-96 chaotic system is used to evaluate the robustness and assimilation accuracy of robust filtering with observation error estimation and original robust filtering under three different optimization methods. The performance of the two methods is compared and analyzed when the model error， the number of observations and the performance level coefficient change. The results show that the observation error estimation technique can improve the accuracy of the state estimation， and the robust data assimilation with the observation error estimation is more robust on the change of system parameters.