局地误差子空间变换卡尔曼滤波方法的最优参数选取  

The choice of the optimal parameters in a Local Error Subspace Transform Kalman Filter

作　　者：王昊运 王辉[1,3] 张宇[1,3] 万莉颖[1,3] WANG Hao-yun;WANG Hui;ZHANG Yu;WAN Li-ying(National Marine Environmental Forecasting Center,Beijing 100081 China;Laboratory of Marine Environment and Ecology,Ocean University of China,Qingdao 266100 China;Key Laboratory of Research on Marine Hazards Forecasting,National Marine Environmental Forecasting Center,Beijing 100081 China)

机构地区：[1]国家海洋环境预报中心,北京100081 [2]中国海洋大学海洋环境与生态教育部重点实验室,山东青岛266100 [3]国家海洋局海洋灾害预报技术研究重点实验室,北京100081

出　　处：《海洋预报》2020年第5期42-51,共10页Marine Forecasts

基　　金：国家重点研发计划(2016YFC1401409)。

摘　　要：基于PDAF同化框架,通过Lorenz96模型的孪生实验,探讨了LESTKF同化方案中的两个重要参数局地化半径和遗忘因子对于同化结果的影响。实验结果表明:局地化半径对分析结果的空间分布影响明显。局地化半径过大,不能很好地滤去背景误差协方差矩阵中的虚假相关;局地化半径过小则分析太细节化使得物理量场不符合实际。遗忘因子的选取对于同化效果影响显著:孪生实验证明遗忘因子(取值为0~1)选取适当能够明显提高同化效果,但如果选得太小则会使同化结果过于接近模式,观测信息对模式的调整将减弱。当二者同时作为自变量影响Lorenz96模式的同化效果时,存在一个最优的参数选择区域,但该最优区域紧邻滤波发散的区域,因此在实际同化应用中应格外重视。Based on the ensemble assimilation method LESTKF within PDAF,this paper analyzes the impact of local radius and forgetting factor on assimilation results using twin experiments with Lorenz96 model.The results show that the localization radius significantly impacts the spatial distribution of the analysis results.The spurious correlations in the background error covariance can’t be well filtered if the localization radius is too large,while the physical quantities field doesn’t conform with the reality due to over-detailed analysis if the localization radius is too small.The twin experiments reveal that the assimilation performance could be significantly improved by choosing proper forget factor(0~1).The smaller the forget factor is,the closer the assimilation results will resemble the model simulation,which indicates the weakening impacts of observations in adjusting the model.Therefore,special attention should be made in choosing the localization radius and forget factor in data assimilation applications.

关 键 词：Lorenz96模型 LESTKF同化方法 PDAF同化框架 

分 类 号：P456.7[天文地球—大气科学及气象学]