Study On The Influencing Factors Of Air Pollutant Concentration Based On Semi-Parametric Regression Model

In the past two decades,with the rapid development of industrialization in China,energy consumption has also increased sharply,which has led to serious air pollution in most parts of China.Among them,smog and O₃pollution are typical problems of air pollution in recent years,and Beijing and Shanghai are typical cities of these two problems.Firstly,based on the data of air quality monitoring station of Beijing Olympic Sports Center from March 1,2013 to February 28,2017,random forest regression model,generalized additive model and generalized additive model based on location,scale and shape were established respectively.The results show that the random forest regression model has poor stability,and both GAM and GAMLSS can effectively predict the PM_2.5average daily concentration.However,in application,GAMLSS is more flexible than GAM,and can more fully exploit the characteristics and rules of distribution itself to dig out more information,which is conducive to improving the accuracy of prediction.Therefore,this paper discusses the influence of various pollutants and meteorological factors on PM_2.5average daily concentration in Beijing based on GAMLSS model,which provides decision-making basis for relevant departments to carry out early warning work.The results show that the PM_2.5concentration increases with the increase of WSPM and SO₂concentration,and the growth rate is stable.The higher the NO₂concentration and the higher the CO concentration,the higher the PM_2.5concentration,and the growth rate is first slow then fast and first fast then slow respectively.When O₃concentration has a certain value,PM_2.5concentration increases with the increase of O₃.The relationship between PM_2.5concentration and DEWP is complex.With the increase of DEWP,PM_2.5concentration first increases at a slow rate,then tends to a steady state,and then increases at a fast rate.Secondly,this paper proposes an additive tail index model,which can model both linear and nonlinear covariate effects and is more flexible than purely linear or nonlinear models.Furthermore,we approximate the unknown function by B-splines and construct an approximate log-likelihood function to estimate the coefficients of the linear covariates and the B-spline basis functions,and their asymptotic properties are also derived under some regularity conditions,the properties of the proposed estimation are shown by simulation.In addition,the model is applied to analyze the daily air pollution dataset in Shanghai.Finally,based on the results of empirical analysis and the current policy in our country,some suggestions are given.The results show that,NO₂、CO and PM_2.5have significant linear effects on O₃concentration,the relationship between NO₂、CO and O₃shows a linear negative correlation,while the relationship between PM_2.5and O₃shows a linear positive correlation.The influence of SO₂and PM₁₀on O₃concentration is nonlinear.The concentration of SO₂increases first and then decreases and then increases with the concentration of,while the concentration of PM₁₀increases first and then decreases with the concentration of.Finally,based on the empirical analysis results,in view of the haze pollution problem in Beijing and the O₃pollution problem in Shanghai,the following suggestions are put forward:for the country,reform the energy structure,adopt more pollution-free energy and low-pollution energy,improve combustion technology,encourage and support afforestation and so on;For the government,rational planning of industrial layout,designated air pollution prevention and control key cities and regions,actively develop urban central heating and so on;For individuals,travel as green as possible,join environmental organizations,cultivate children’s environmental awareness from an early age and so on.