Research On Ocean-atmosphere Coupled Model Data Assimilation Based On Particle Filter

Data assimilation as an effective method to intergrate multi-source atmospheric and ocean observation information and numerical model,it is able to use the limited observation data to best estimate the ocean state,improve model estimate accuracy,and effectively reduce the uncertainty from the initial conditions.Particle Filter(PF)is a kind of the advanced data assimilation methods in the atmospheric and oceanic researches.Theoretically,PF overcomes the Gaussian assumption and nonlinear constraint.It plays an important role in the development of atmospheric ocean business.This paper deeply analyzes PF data assimilation problem.The PF often suffers from a well-known problems,i.e.,the degeneracy phenomenon and particle impoverishment.Based on the equal-weight particle filter,then it proposes an equal-weight particle filter with observation window that can be better use observation adjust the proposed density and an equal-weight particle filter with improve gain matrix K.The feasibility and assimilative quality of the method was verified by different models.Firstly,this paper gives the definition of data assimailation,then the constitution,characters,deficiencies of data assimailation.In addition,this paper reviews the Particle Filter data assimailation research status at abroad and analyzes the feasibility of using it in oceanatmosphere coupled model.Secondly,this paper overviews principle and characters of PF's well known problems.Considering of the ocean-atmosphere coupled model,Based on the Lorenz-63 model designed experiments to analyze and compare two different particle filterThirdly,this paper expounds the fundamental of PF and propse two improve methods for equal-weight particle filter.One is using observation window,another is improve gain matrix K.Based on simple model designed experiments to analyze the feasibility of two methods.And the detailed technological process is given.Finally,this paper based on couple model designed experiments to analyze and compare PF with the equal-weight particle filter which use two different improve methods.The result shows that the RMSEs of two improve methods are lower.Comparing this method with EAKF,particle filter's RMSEs are similar to EAKF but particle filter is more stable.