扩展卡尔曼滤波数据同化试验--基于Lorenz(1960)系统的研究

本文采用Lorenz(1960)系统，在只考虑初始误差及观测误差而不考虑模式误差的情况下，利用扩展卡尔曼滤波(Extended Kalman Filter)数据同化方法进行了数值模拟试验。数值试验的结果表明：扩展卡尔曼滤波数据同化方法对系统状态的估计有较好的改善作用，能有效的抑制估计误差的增长；加大观测频率，可以进一步改善数据同化的效果，使估计误差进一步减小；由于模式误差的存在，系统的不稳定能量会不断的累积，出现了估计误差的异常增长和计算的不连续现象，在模式预报方程中的均值演变方程加人二阶偏差纠错项，对控制估计误差的异常增长，进一步改善数据同化的效果有较明显作用。

DATA ASSIMILATION TEST ON THEEXTENDED KALMAN FILTER——THE STUDY BASED ON LORENZ(1960) MODEL

In this paper, based on the Lorenz (1960) model, without thinking of the model error ,the authors had a numerical simulated test on the extended Kalman filter data assimilation method. The results indicate that EKF data assimilation method has a preferable function on improving the estimate of model state and restraining effectively the unbounded increase of the error of estimation. When the frequency of observation is increasing ,there will be a better influenceon improving the effect of data assimilation and decreasing the error of estimation ;and at the same time ,because of model error ,the unstable energy of the system will cumulate continuously ,there has been a phenomenon of abnormal increase of the error of estimation and the incontinuity in calculation ;including the second bias correction term in the mean evolution equation of model propagation ,it is more effective on controlling the indefinite growth of error and improving the impact of data assimilation .