| As a global hot issue,air pollution has attracted more and more attention.Compared with the air quality of other countries,the air pollution in China is relatively serious,which poses a certain threat to people's health.At the meeting of the 19 th CPC National Congress report,General Secretary Xi Jinping clearly proposed to solve the prominent environmental problems which are harmful to the health of the people.Therefore,it is necessary to study the urban air pollution data in China.So far,the quantile regression method has been applied relatively rarely in the field of environment.In the field of environmental pollution,the real spatio-temporal data generation process does not satisfy the hypothesis of independent repeated test in classical statistics,and the asymptotic validity of estimators in traditional quantile regression is based on the assumption that the number of individuals and the number of periods tend to be infinite.However,in practical application,the observed values of air pollution data are limited,so it is difficult to ensure the validity of the traditional quantile regression estimator.In view of these shortcomings in the traditional quantile regression method,this paper decides to choose Bayesian parameter quantile regression method,which takes full account of the uncertainty of parameters and is also applicable in the case of small sample and non-convex or unoptimal objective function.Under the framework of Bayesian theory,Bayesian estimation of quantile regression model can be realized by combining asymmetric Laplace distribution and quantile regression theory.Considering that PM2.5 is affected by adjacent cities in space.At the same time,it is also influenced by its internal structure.Therefore,the conditional autoregressive model is introduced on the basis of the general quantile regression model,the temporal and spatial pattern of PM2.5 is reflected by setting virtual variables.In this paper,the PM2.5 hourly concentration data of 1497 stations in China from January 1st 2015 to December 31 st 2017 were collected by using the Python crawler tool in the national air quality real-time publishing platform of China Environmental Monitoring General Station.According to these data,based on Moran index and Gini coefficient,the spatial autocorrelation and geographical concentration degree of urban air pollution in China are briefly explained.It is found that the spatial distribution of PM2.5 in Chinese cities is not completely random,and the spatial distribution is concentrated in winter.But in summer,the spatial distribution is relatively dispersed,and the spatial distribution is more balanced.Combined with Bayesian spatio-temporal quantile regression model,using annual and quarterly models as research objects,the results show that the temporal and spatial distribution of PM2.5 at the level of 0.3?0.5?0.7 quartile is approximately similar,whether the year was used as the research object or the quarter as the study object.On the whole,the air quality in March and December is close to that in heavy pollution,in summer and autumn,the spatial effects of Gansu and Guizhou are relatively large,some eastern cities are also relatively large in winter.Under mild and moderate pollution,air quality gradually improved in January and June and deteriorated gradually after September,and the spatial effect of some coastal cities in winter is relatively large under the condition of light pollution.The fluctuation periodicity of each season is about 4 weeks.In this paper,with regard to the spatial-temporal pattern of urban PM2.5 pollution in China,it is suggested that for better joint governance,the government can control the combustion of fossil fuels,increase the level of economic development in the western region,and change the economic growth mode in the eastern region. |
PDF Full Text Request