Application Research Of Localization Method In Data Assimilation Error Processing

In the process of model driving,Data assimilation represents that utilize assimilation methods fuse the part or the whole of the observation information from different sources,and the process model and observation model are constantly adjusted by the observation information.so the prediction system of the reduced error model is obtained.However,most of the system models are non-linear.meanwhile,due to the false correlation between the long-distance observations and the assimilation-states during data assimilation,the localization method has attracted wide attention.In addition,the ensemble numbers fewer,the observation data is difficult to be used effectively,which makes the assimilation effect not good enough.In view of the above problems,the main research of this study is as follows:(1)In actual data assimilation,the accurate estimation of background error will affect the observation error,and the observation error mainly affects the final assimilation results.In addition,the main source of observation error is the fake correlation caused by the long-distance between observation value and state value.Therefore,the localization method is proposed to solve the problems of slow convergence speed and false correlation(observation value and assimilation state)caused by fewer ensemble numbers.(2)In the process of data assimilation of localization method,when the observation point is far from the state update point,the observation information can't fully utilize.Therefore,Fuzzy control-based localization method for data assimilation is proposed.The localization method mainly deals with the false correlation.The fuzzy control algorithms are used to judge the distance between the observation point and the update point of the state via fuzzy control construction of the weight of the observation position.The problem of inadequate utilization of observation information was solved during data assimilation.(3)In order to apply to high-dimensional/non-linear systems,New algorithm CF(covariance fuzzy)and FA(fuzzy analysis)is proposed.Firstly,covariance localization and localization analysis methods are introduced.Secondly,for the sake of improve the accuracy and robustness of filtering,covariance localization method and localization analysis method are combined with fuzzy control method is developed,and the coupling process has been discussed.A nonlinear Lorenz-96 model is used to compare the local analysis assimilation algorithm(LA),the covariance localization assimilation algorithm(CL),the local analysis algorithm coupled with the fuzzy control(FA)and the CL algorithm with fuzzy control(CF)...