The Analysis Of Air Quality Index Based On Quantile Regression

The problem of the environment is more and more serious due to the rapid economic development of different countries, especially the suspended particles PM2.5 and PM10 pollution to the environment produced by industrial production, fossil fuel consumption, vehicle emissions and sandstorm invasion. Therefore, the exploration of effective measures to control environmental pollution is imperative. The goal of environmental protection is to improve air quality, reduce the emissions of toxic gas and pollutants. Among which, the monitoring, forecasting and controlling of air quality are the most important. This thesis devotes to evaluate and predict the air quality index(AQI) of three cities in Gansu province by quantile regression method, and provide some suggestions for the relevant department to enact policies.In this thesis, the concentration of six major pollutants SO2, NO2, PM 10, PM2.5, O3, CO and three environment variables temperature, humidity, wind scale in Lanzhou, Tianshui and Jinchang are chosen as alternative factors affecting AQI. In order to analyze the correlation and impact extent between influential factors and AQI, and select the main factors to air quality, we calculate Pearson correlation coefficients within these elements and AQI, then establish a linear quantile regression model using above 9 factors as independent variables and AQI as dependent variable. Adopting the quantile autoregressive and quantile autoregressive distributed lag model which are based on the Adaptive-Lasso variable selection, we carry out a point and interval prediction on AQI. Empirical analysis shows that: PM10 is the dominant factor of air quality in Gansu Province in spring; the performance of models can be promoted by adding other exogenous variables in predicting AQI; the complexity and redundancy of original models is reduced after variable selection under the premise of keeping the accuracy of prediction; interval prediction can reflect the variation range and fluctuation of AQI more reasonably, which provides a more effective reference tool for on-line reporting of air quality index.