Spatio-Temporal Analysis Of Multivariate Meteorological Data Based On Matrix Autoregressive Model

Meteorological data include spatiality and temporality,so it is often to establish a spatiotemporal model.However,the temporal and spatial research of meteorological data mostly focuses on the modeling of single meteorological index,but there is little research on the modeling of multiple joint meteorological indexes.Based on the idea of matrix autoregressive model,this paper made an innovative expansion of the model.Its advantage is that it not only makes full use of the temporal information and spatial information of each index,but also captures the correlation between each index,so as to improve the interpretability and reliability of the results.This paper creatively used the structural matrix autoregressive analysis method to model and analyzed the temporal and spatial data of meteorological indicators in Denmark.This paper referred to the characteristics of vector autoregressive model and structural vector autoregressive model and their correlation,changed the form of matrix autoregressive model and established structural matrix autoregressive model,so that the coefficient of the model includes not only the influence of the previous period,but also the influence of the current period.Through references of the projection method,least square method,maximum likelihood estimation method and full information maximum likelihood estimation method in the existing literature,this paper gave the basic ideas of these estimation methods and the sum of squares of residuals after fitting the model.The comparison results showed that the sum of squares of residuals obtained by estimating the parameters of the first-order matrix autoregressive model by the least square method is the smallest.After analyzing the residuals of the first-order matrix autoregressive model,it is found that the residuals have strong current structure.Therefore,based on the results of the first-order matrix autoregressive model,the structural matrix autoregressive model was established,and the projection method was selected to minimize the sum of squares of the model residuals.Finally,the parameter results of the two models are analyzed to obtain the current and lag effects between atmospheric indicators and regions.The study mainly found that the increase of humidity in the air can well reduce the concentration of air pollutants,and the meteorological conditions of the southeast coastal boundary of Denmark have a significant impact on the meteorological conditions of inland areas.