The Space-time Distribution,Prediction And Diffusion Analysis Of PM2.5 In Chengdu

As the development of economy and the rapid advance of urbanization in China,the air pollution has become more and more serious and has become a social problem.Particulate matter especially PM2.5 is primary pollutants in most of the cities and it will bring bad influence for all the citizens.The pollution of PM2.5 mass concentration is a good way for citizens to avoid damage.Therefore,the research for PM2.5 has become a very important topic.This paper focuses on the research of pollutant data of six monitoring points in Chengdu.Firstly,this paper uses the daily average concentration of PM2.5 to draw the concentration time series,and finds that PM2.5 in Chengdu occurs mostly in spring and winter,and the frequency of occurrence in autumn and summer is lower.Then the Moran index of six monitoring points was calculated,and it was found that there was a significant positive spatial correlation between Chengdu PM2.5 in different months,and 9 months passed the significance test,showing obvious aggregation pattern;in spatial regression analysis The model passed the correlation test,and the coefficients of O3 and SO2 and air humidity were not significant,and the regression coefficients of other variables were significant,indicating that air quality index variables,wind speed and temperature all had significant effects on PM2.5 concentration.Then using support vector regression and neural network to predict the concentration of pollutants,the model prediction results have certain errors,and the model's estimation of concentration tends to be conservative.For some larger or smaller concentration values,the prediction error is larger.However,after the spatial factors were added to both models,the effect of the model was significantly improved,and it was confirmed from the side that the pollutants in Chengdu did have spatial interactions.The prediction error of the radial basis neural network is smaller than that of the support vector machine,and the mean square error of the pollution concentration prediction in the urban area of Chengdu is about 10.Finally,spatial interpolation and Gaussian diffusion model are used to spatially predict the pollutant concentration.Since only the pollution data of six monitoring points are collected in this paper,we assume other pollution source points,use the inverse distance weight method for spatial interpolation,and predict the initial pollutants of other pollution source points.Then,a Gaussian diffusion model was used to simulate the spread of pollutants in the entire space after a sudden one hour of pollution.The results show that the concentration of pollutants along the downwind is increasing,and at the same time,due to the topography,the pollutants are difficult to spread in the central part of the city,thus forming a high-pollution area.