Research On Data Assimilation Method Based On OpenDA

Data assimilation originates from meteorology.It is a data fusion algorithm that integrates different sources of observation information into natural numerical models.The previous research and implementation of data assimilation system is based on specific models or specific methods,and the development and implementation of time consuming,high cost,low efficiency and poor universality.Therefore,data assimilation research based on general toolbox is the main trend nowadays.OpenDA(Open Data Assimilation)is a universal toolbox developed for data assimilation.It has been widely used abroad,but seldom applied in China.This paper will take OpenDA as the platform to carry out data assimilation research.In view of the lack of literature in the domestic OpenDA,First,the structure of OpenDA platform was introduced,including its application framework,model encapsulation,black box coupling,configuration file,localization and parallelization.Secondly,we introduces the assimilation and calibration algorithms and models of existing integrated OpenDA,and the installation and operation methods of OpenDA.Aiming at the uncertainty of particle trajectories and the divergence of ensemble filter in damped harmonic oscillator,the main research work in this paper is as follows:(1)aiming at the uncertainty of the particle trajectory of the damping harmonic Oscillator model,using the Ensemble Kalman Filter algorithm(EnKF)in OpenDA,we take the damping harmonic Oscillator model(Oscill)as the platform,take the different aggregation numbers,explore the influence of the change of the set number on the assimilation effect,and realize the real-time prediction of the position of the model particle trajectory.Experiments show that EnKF algorithm can achieve real-time prediction of particle location of the model.(2)to solve the problem of filtering divergence caused by set number restriction,the background error covariance amplification technique and analysis error covariance amplification technology are proposed to eliminate false correlation and improve the performance of filtering.the Ensemble Transform Kalman Filter algorithm(ETKF)in OpenDA is amplified by covariance,and the Lorenz-96 model is used as the platform to take different sets of sets.The influence of the magnifying factor on the filtering assimilation performance is explored when the number of sets is changed.The experiment shows that the assimilation effect can be achieved by effective selection of amplification factor and small set number.(3)in order to solve the problem of filter divergences caused by the set class filtering and the forced parameter restriction,the Ensemble Time Local Robust Filter algorithm(EnTLHF)in OpenDA is amplified by covariance,and the different forcing parameters are taken on the Lorenz-96 model as the platform,and the influence of the magnifying factor on the filtering assimilation performance is explored when the forced parameter changes.Different amplification factors are taken and compared with ETKF,the experiment shows that the effect of large forcing parameters can be achieved by selecting the magnifying factor and the smaller force parameters effectively.Compared with the ETKF,the robustness of EnTLHF is always better than that of ETKF.In this paper,the Open DA data assimilation toolbox is used as a platform to study the real-time prediction of the position of the particle trajectories of the oscill model.It is verified that the amplification factor can solve the problem of filtering divergence caused by the number of set filter sets and the constraint of forced parameter.Considering the many advantages of OpenDA data assimilation toolbox,it is expected to be further promoted and applied in China.