Research On Related Problems Of Errors In Data Assimilation

With the rapid development of the Earth's surface study such as the atmosphere and oceans,data assimilation,as a "bridge" between observations and models,has gradually become a popular method in earth science research.No matter what data assimilation method is,the error in DA system has a great influence on the final assimilation effect.Therefore,the research on all kinds of errors in data assimilation has become one of the important issues in data assimilation.In data assimilation,constructing a reasonable error covariance matrix is very important for the assimilation system.Considering the correlation problem in the error matrix,there will be a new method for estimating the error covariance matrix and more realistic and effective analysis values.Therefore,in-depth exploration of the generation mechanism,estimation and processing methods of error covariance will ultimately promote data assimilation and can be better applied to actual forecasting.Through the study of related theoretical knowledge of data assimilation and error processing,and related literature research.In view of the difficulty in determining the background error covariance matrix and the prediction of pollutant diffusion in the atmosphere,the research contents of this thesis are mainly developed from the following aspects:(1)Starting from the basic theory and research status of data assimilation,this thesis analyzes and introduces two types of data assimilation concepts and methods:sequential data assimilation and continuous data assimilation.The error sources and processing methods in DA are summarized.(2)This thesis introduces the background error covariance correlation theory and several common methods for estimating the B matrix in variational data assimilation.Lorentz 63 is used as a prediction model.The usability of the NMC method is verified by a three-dimensional variational data assimilation algorithm.It is further obtained through numerical experimental analysis.The statistical characteristics of the background error covariance in the assimilation of three-dimensional variational data.At the same time,the influence of the two on the background error covariance is discussed for the dimension of the variable and the width of the assimilation window.(3)Taking the problem of the diffusion of atmospheric pollutants as the research background,a two-dimensional advection-diffusion equation and Ensemble KalmanFilter(EnKF)were used as prediction models and data assimilation algorithms to establish a simulation assimilation system,and the number of sets in assimilation was studied.2.The influence of observation point position and observation distance on the assimilation effect.The thesis focuses on the difficulty of determining the B matrix and the diffusion of pollutants in the atmosphere.It verifies the feasibility of the NMC method to calculate the background error covariance,and discusses the influence of the variable dimension and the assimilation window on the assimilation performance.In this thesis,the influence of the number of collections,the position and distance of observation points on the assimilation effect is discussed.