Research On Air Quality In The Yangtze River Delta Region Based On Functional Data Analysis

The Yangtze River Delta urban agglomeration is the most economically developed region in China,but the environmental problems caused by the development cannot be ignored.In this paper,54 key cities in the Yangtze River Delta region were selected to study the air quality index(AQI)and the pollutants from 2017 to 2019 by using functional clustering and regression prediction model.Through analyzing the visualization of the results and the data variation characteristics,we give suggestions so as to improve the environment in the Yangtze River Delta region.For the purpose of improving clustering accuracy and robustness,this article first introduces traditional clustering models and functional clustering models,and summarizes existing improved methods,including the trimming process,functional principal component analysis,functional mixed model,etc.Then,by improving the coefficient of soft trimming method,the Functional clustering method based on Soft-trimming is proposed.In the simulation,the functional data with outliers was generated,and then these seven clustering methods were compared.The results showed that the functional clustering model based on soft trimming has the highest accuracy and accurately eliminates outline curves.The model was then applied to real data analysis.The data is the three-year air quality index of 54 cities in the Yangtze River Delta.The clustering results divided the research objects into three types and three outline curves,and the three types of air quality curves' patterns are different.From the visualization of air quality,we can see that the outline curves' pattern is obviously different.Through the visualization of the clustering results,we analysis the air quality curves' trend and the cause of the three types of cities from the perspective of geographic distribution;the seasonal division is performed and the characteristic analysis is performed from the perspective of seasonal fluctuations to provide a reference for pollution control.Secondly,to improve the accuracy of regression,we summarize the traditional regression methods and functional regression methods,and combine the outlier monitoring method for functional data into the regression model,and propose a pre-trimmed functional regression model.Using Nanjing PM2.5 data,and dividing the first 70% data as the training data and the last 30% data as the test data for simulation research,the results show that the pre-trimmed functional regression we proposed in this paper can improve the accuracy of regression.In the real data analysis study,in view of the lack of pollutant concentration data,the functional regression prediction model was used to refit.Therefore,the pre-trimmed functional regression prediction method is applied to the prediction of the six pollutants concentration in Wuhu City,and the prediction results are analyzed in the time dimension.The corresponding relations between individual air quality index(IAQI)and six pollutants are different.The prediction results are used to carry out pollution index statistics,we evaluate the degree of impact combined with the results,and the cause analysis of pollution characteristics is carried out.The real data analysis shows that the regional pollution of the Yangtze River Delta urban agglomerations presents a geographically stepped distribution,and has an obvious seasonality,especially the industrial areas are more serious.Therefore,in view of the pollution characteristics of urban agglomerations,we make specific recommendations based on implementation of large-scale joint prevention.In addition,through simulation and real data studies,we found that the functional data clustering and regression methods are well suited to the analysis of regional air quality characteristics in our country,and the trimming method can further improve its efficiency by eliminating outline curves,which is of great significance to the study of practical problems.