| With the rapid development of China’s economy and technology level,the air pollution problem has become increasingly prominent.Since the strategy for the rise of central China has been implemented,cities have made significant progress in terms of their economic development potential.But with the rapid development of these cities,air pollution caused by heavy industries and human activities has become more and more serious,and the air quality problems in the six central provinces have become more urgent.Therefore,studying the air pollution situation is important for promoting the environment-friendly and high-quality development of cities in the six central provinces.This thesis takes six cities in central provinces as the research scope,and develops the methodological research and empirical analysis around the air quality index(AQI),data of six pollutant concentrations and data of six meteorological influencing factors from 2016 to 2021.Firstly,the spatial and temporal distribution characteristics,clustering and unevenness of AQI were derived using Kriging interpolation method of Arc GIS software and exploratory data analysis(ESDA)method of Geo Da software.Secondly,grey correlation analysis model and geographically weighted regression model(GWR)were used to study the regional classification of air pollution control in key cities in six central provinces,the analysis and identification of major influencing factors and the spatial heterogeneity of each influencing factor.Finally,using Hefei,the capital of Anhui Province,as a representative city,the MATLAB software was used to build a momentum multi-hidden layer adaptive BP neural network and a double-hidden layer Bayesian(BNN)neural network model to predict the air quality index and compared their fitting effects.The main findings of this thesis are as follows.(1)The air pollution conditions of 82 cities in the six central provinces improved year by year from 2016 to 2021.The quarterly averages of AQI in the six central provinces reached the lowest value in summer,the highest value in winter,and intermediate values in spring and autumn.The monthly average of AQI had a"U"shape distribution,reaching a peak in January or December,then fluctuating downward,reaching a trough in July and then oscillating upward.(2)Most of the six central provinces had excellent annual air quality,and the AQI values generally showed a pattern of"high in the north and low in the south".The air pollution areas were mainly located in the central and southern parts of Shanxi Province,the western and northern parts of Henan Province,the northern part of Anhui Province and the central part of Hubei Province.(3)During the study period,the air quality in the six central provinces showed a strong positive spatial correlation,and the spatial clustering was first weak and then strong.The high-high clustering area was located in the northern areas of Henan Province,Shanxi Province and some cities in Anhui Province,where cities with high AQI values were surrounded by neighboring cities with high AQI values.The low-low agglomeration area was mainly distributed in the southern areas such as Hunan Province,Jiangxi Province and some cities in the south of Anhui Province,which were closely related to the conditions such as developed agriculture,large forest coverage and underdeveloped industry in the city.(4)The 26 key cities in the six central provinces could be divided into 7 regions running north and south,forming a distribution structure of"two points,one line and four sides",indicating the regionality and transmission of air pollution.The identification and analysis results of meteorological influencing factors in the region showed that the influence of different meteorological factors changes with time,and there were obvious differences in the role and direction of meteorological factors.However,in most regions,there was a negative correlation between precipitation,average temperature,relative humidity and air quality index,and there was a positive correlation between average air pressure and sunshine duration.(5)The prediction accuracy and prediction error of the double hidden layer Bayesian network(BNN)model are better than the momentum multi-hidden layer adaptive BP neural network model,the R-Squared coefficient(R~2)is as high as 99%,and the prediction curve and the true value curve almost coincide. |
