迭代集合卡尔曼滤波方法的性能比较研究

针对数据同化过程中模型的非线性问题,通过分析对比得出了一种适合强非线性系统的迭代集合Kalman滤波(IEnKF)。在Lorenz|63模型的框架内,比较分析集合Kalman滤波(EnKF)、迭代集合Kalman滤波(IEnKF)和迭代扩展卡Kalman滤波(IEKF)在集合数、观测误差方差、放大因子和模型步长不同时同化性能差异,由此探讨这3种方法的优劣。研究结果表明:随着集合数的增加,3种算法的同化性能都得到了一定的改善;放大因子的增大,使其同化性能变差且EnKF呈现出多重波峰波谷的现象;3种方法的均方误差(RMSE）随观测误差方差和模型步长的增大而增大,其同化精度都变差;而IEnKF同化性能最优,更具有鲁棒性。

Comparative Studies on Iterative Ensemble Kalman Filter Methods

With regard to model non-linear problems in data assimilation process,an Iterative Ensemble Kalman Filter (IEnKF) is derived by thoroughly analysis and comparison.Within the framework of Lorenz-63model,this paper compared the different performances among the following three methods,Ensemble Kalman Filter (EnKF) Iterative Ensemble Kalman Filter (IEnKF) and Iterative extended Kalman Filter (IEKF),by changing ensemble numbers,observation error variance,the inflation factors and the model steps.The final comparative studies show that the assimilation accuracy of all three algorithms can be improved when ensemble numbers increase.When we change the inflation factors,the assimilation results are becoming worse and the EnKF presents obvious multihill and multivalley phenomena.The RMSE of all three algorithms increase when observation error variance and the model steps increase,and the results of algorithms get worse as well.The results show that the IEnKF is the most optimal algorithms with a much better robust performance.