| Nowadays,the problem of air pollution is becoming more and more serious,and my country often has large-scale haze weather.Research on the spatial distribution of PM2.5concentration plays an important role in human health and environmental prevention.Based on the analysis of the temporal and spatial characteristics of PM2.5 concentration and the exploration of the correlation between smog-related social text data and PM2.5 concentration,this paper focuses on the study of spatial interpolation methods,mainly including the study of univariate spatial interpolation methods and multivariate(Add auxiliary variables)spatial interpolation method research,the specific research content is as follows:(1)Analyzed the temporal and spatial characteristics of PM2.5 concentration,Shandong Province has a low average PM2.5 concentration in summer and a high average PM2.5concentration in winter;using the global Moran’s I and local Moran’s I methods,it is explored that there is a strong PM2.5 data In terms of spatial autocorrelation,the central and western parts of Shandong Province present a "high-high" agglomeration pattern,while the northeast presents a "low-low" agglomeration pattern.The PM2.5 concentration and the amount of smog-related microblog posts have similar changes in time.The sentiment value of the microblog text is calculated using the sentiment calculation model,and the average sentiment value of the microblog and the PM2.5 concentration in different radius buffers are found.There is a negative correlation between the data.When the PM2.5 concentration is high,the sentiment of the related Weibo text published by social users is more negative;as the buffer range increases,the absolute value of the correlation coefficient continues to increase.When it is 105 km,its absolute value reaches the maximum,about 0.78.(2)According to the trend effect of the spatial distribution of PM2.5 concentration data,based on the ordinary Kriging interpolation method,using the commonly used exponential,spherical and Gaussian function three variogram models,after removing the trend effect,the residual term is calculated The results show that the spherical function model that removes the first-order trend term has the best effect;according to the variation function,the spatial structural characteristics of PM2.5 concentration are further analyzed,the major and minor axis direction variation functions are calculated,and the The anisotropic variogram model,taking into account anisotropic spatial interpolation,fully considers the spatial structure and anisotropic characteristics of geographic phenomena,and its model evaluation results are better than isotropic methods.(3)The cross-variation function of microblog sentiment value and PM2.5 concentration data is analyzed,and the cokriging method that introduces the microblog sentiment value as a covariate factor improves the interpolation accuracy of PM2.5 concentration,and the accuracy of the model is better than that of ordinary grams.Based on ordinary Kriging method: In addition to the sentiment value of Weibo,influencing factors such as meteorological factors,road length and land use type are added,combined with correlation coefficients and geographic detectors to evaluate the contribution of potential influencing factors,and principal component analysis is used to eliminate multiple factors.A PCA-GWR model integrating multiple variables was constructed to simulate the spatial distribution of PM2.5concentration in Shandong Province.The results showed that the RMSE and MAE of the model were reduced.The multivariate spatial interpolation method is obviously better than the univariate spatial interpolation method,which can compensate for the uneven distribution of air quality monitoring stations,thereby improving the accuracy of spatial interpolation. |
PDF Full Text Request