Research Of A Localized Equal Weight Particle Filter Method In Ocean-atmosphere Coupled Model

Ocean is an indispensable resource in human production and life,which plays an important role in promoting economic development,safeguarding national sovereignty,enhancing national defense and security capacity building.The study of Oceanography mainly depends on numerical simulation and direct observation of the ocean,which have their own advantages.The data assimilation method is an important means to combine the multi-source atmospheric ocean observation and numerical prediction model,combining the two research methods to obtain better application results..This paper studies the data assimilation method of equal weight particle filter,analyzes the dependence of traditional particle filter method on future observation,and gives an equal weight particle filter assimilation method based on the density proposed by Kalman;at the same time,in order to meet the application in the actual complex mode,on the basis of the new method,gives out the local application scheme,and compares it with other mainstream localities Comparison and analysis of chemical methods.First of all,the paper introduces the principle of equal weight particle filter,and expounds the current research status of many mainstream data assimilation methods,especially the hybrid data assimilation method and particle filter method.Combined with the simple experimental results of particle filter hybrid method,the advantages and disadvantages of each method are analyzed,and the feasibility of proposing density improvement for the traditional equal weight particle filter method is analyzed.Secondly,this paper analyzes the data assimilation framework of traditional equal weight particle filter,and proposes an equal weight particle filter method based on the density proposed by Kalman.In this method,Kalman filter is introduced into the proposed density calculation to improve the dependence of traditional methods on future observation,so that the method can better meet the assimilation requirements in real-time observation.At the same time,in the simple lorenz-63 mode,it is compared with the traditional equal weight particle filter method.Thirdly,due to the fact that the interaction between the atmosphere and the ocean should be considered in the actual application of ocean assimilation,the method of particle filter with equal weight based on the density proposed by Kalman is applied to the ocean atmosphere coupled 5vccm model to further verify the improvement of the assimilation accuracy of this method.Experimental results show that the accuracy of this method is significantly higher than that of the traditional equal weight particle filter,and the overall assimilation results are more stable.Finally,considering the gridding model used in the actual data assimilation application,the localization improvement based on the density average weight particle filter proposed by Kalman is proposed.In this method,the localization scheme is introduced into the average weight particle filter assimilation method,which makes the method more suitable for the application of grid sparse observation.In lorenz-96 mode,we choose to use the "twin" experimental framework to use the nonlinear observation operator and the current mainstream localized particle filter and local Ensemble Kalman filter for comparative study.The experimental results show that in the case of non-linear non Gaussian observation,the density average weight localization particle filter method based on the Karl Mann proposal can use the least number of particles to get the optimal Improve the accuracy of assimilation.