Spatial and temporal evolution of air pollution and verification of the environmental Kuznets curve in the Yangtze River Basin during 1980—2019

ABSTRACT Continuously high concentrations of haze pollution can hinder urban economic development. In order to improve the quality of the environment in the Yangtze River Economic Belt, it is necessary to investigate the spatio-temporal characteristics and impact factors of smog. This study, relying on multi-source remote sensing data, conducted a comprehensive study on the concentration of haze pollution based on long-term data, multiple spatial scales and pollution indicators. The results showed that the concentrations of seven air pollutants (PM2.5, SO4, SO2, BC, OC, SS and dust) in the Yangtze River Basin appeared to first increase and then decreased from 1980 to 2019. Dust pollution and sea salt pollution were concentrated in the upper reaches of the Yangtze River and the coastal areas of the Yangtze River Delta, while other pollutants were higher in the Sichuan Basin and northeast of the Yangtze River. Of the socioeconomic factors, the significance of different factors on pollutant concentration was obviously different. In addition, the environmental Kuznets curve relationship between economic gain and air pollution depended on the type of pollutant, and there were certain regional differences. This study provided a scientific basis for considering innovations in haze control in the urban agglomeration of the Yangtze River Economic Belt.


Introduction
Since the reform and opening up, the Chinese economy has always remained in continuous and rapid growth, with GDP growth of 579% in 1980-2000 and 1117.36% in 2000-2020.This is far higher than the global growth of 137.37%, and China has become the world's most dynamic economy.However, the problem of extensive development with "high inputs, high consumption and high pollution" behind the rapid economic growth is still prominent.The country has implemented a number of policies in hopes of improving the situation, such as the Chinese Ambient Air Quality Standards and the Air Pollution Prevention and Control Action Plan.However, the effect has not been obvious and air quality problems are still serious.The Chinese Environmental Status Bulletin states that: in 2013, testing according to the new standards was carried out in 74 cities in China, of which only three cities had air quality compliance; the percentage of days with excellent air quality in 2017 was 72.7% in the Yangtze River Delta and the Pearl River Delta, down 1.5% from 2016; in 2019, the ambient air quality in 180 out of 337 Chinese cities exceeded the standard, with heavy pollution increasing for 88 days compared to 2018.In the context of high-speed economic growth and increasingly deteriorating air quality, it has become a research hotspot to clarify the relationship between economic growth and atmospheric contamination, and to achieve coordinated economic and environmental development.
In the past few decades, by deepening the research on the relationship between economy and environment, the environmental Kuznets curve (EKC) has been proposed as an important tool for studying the relationship between the atmospheric environment and economic development (Grossman & Krueger, 1991).Then many boffins began to test the validity of the EKC hypothesis.Some studies have supported the EKC hypothesis.Tao (2008) used GDP per capita as an economic indicator to verify the inverted U-shaped relationship between economic gain and industrial waste.Using panel data analysis and quantile regression analysis, Wang (2013) showed that the relationship between carbon dioxide emissions and economic growth is consistent with the EKC hypothesis.Luo et al. (2017) found that EKC is valid for some cities in China for PM 2.5 , and its coefficient of GDP is negative.Similar results were obtained by Chakravarty & Mandal (2020) and Yang et al. (2020).In addition, there are different conclusions.For example, Du et al. (2012) investigated the driving forces and reduction potential of China's CO 2 emissions, and found that the inverted U-shaped relationship between per capita CO 2 emissions and the level of economic development was not strongly supported by the estimation results.Liu et al. (2015) studied the effects of human activities on industrial waste and argued that no strong evidence was found to support the EKC hypothesis for three industrial wastes in China.Therefore, further validation of the Chinese EKC hypothesis is required.
To date, most existing studies on air pollution have focused on only one pollutant such as O 3 (Tao et al., 2022;Tianzhen et al., 2022), SO 2 (Sinha et al., 2017;Zhou et al., 2017) and PM 2.5 (Fong et al., 2020;Gui et al., 2019).Furthermore, some studies used six atmospheric contaminants (SO 2 , NO 2 , PM 10 , PM 2.5 , CO and O 3 ) to construct the air quality indices (AQI).For instance, Pu et al. (2017) and Zhan et al. (2018) investigated the spatiotemporal variation of China's AQI and driving factors, providing a reference for the formulation of urban policies and the improvement of air quality in China.In these studies, spatial data analysis was a non-negligible means (Kaya et al., 2019).In the analysis of influencing factors on atmospheric contamination concentrations, numerous empirical studies have indicated a correlation between population, GDP, urbanization level, percentage of secondary industry, proportion of foreign investments and air pollution concentration (Fang et al., 2015;Shi et al., 2019;Yu & Ymla, 2016).In addition, studies have proven that factors such as industrial structure, technological innovation, and environmental regulation can promote environmental improvement (Rao et al., 2016;Yang et al., 2020;Zhang et al., 2021).In the case of research scales, they can range from national to regional to local scales, such as megacities and the Jing-Jin-Ji region (Li et al., 2021;Wang et al., 2017;Zhan et al., 2018).
Although many achievements have been made on the relationship between urbanization, industrial structure and ecological environment in the country and abroad, there are still some shortcomings.Studies show that social-economic factors and natural elements are the main factors influencing the concentration level and spatial distribution of atmospheric contamination.However, in the process of screening influencing factors and pollutant indexes, comprehensive analysis of multiple influencing factors based on long time series is still lacking.The characteristics of the spatial distribution of pollutant concentrations are usually quite different at different spatial scales.Therefore, it is difficult to reflect and apply research results at one spatial scale to results and evaluation at multiple spatial scales.The Yangtze River Economic Belt (YREB) is a key component of China's "T-shaped" development strategy.General Secretary Xi Jinping emphasized that the development of YREB must adhere to the strategic prioritization of ecological and green development.In view of this and in the context of the high-quality development of YREB, this study takes the three main urban agglomerations of YREB as the study.Furthermore, the study accurately estimated the concentration of air pollutants in the Yangtze River Belt from 1980 to 2019 based on the Merra-2 data set, and discussed the variation trend and spatio-temporal pattern of air pollutant concentrations in YRB.Based on the STIRPAT model, this study investigated the impact of social and economic development on the concentration of air pollutants and realized a comprehensive study of air quality in YREB based on long time series, multiple spatiotemporal scales and multiple impact indicators, thus providing a reference for air pollution control of the three major city clusters in the YREB, so as to promote the green and sustainable development of the study area.The structure of the study was organized as follows.Section 2 introduced the research area, data and methods.The results were presented in Section 3. Section 4 discussed the research results of the study.The conclusions were provided in Section 5.

Study area
The YRB has a total length of 6,300 km and a total basin area of 1.8 million km2 .It covers three major economic zones in eastern, central and western China, including Qinghai, Tibet, Sichuan, Yunnan, Chongqing, Hubei, Hunan, Jiangxi, Anhui, Jiangsu and Shanghai.Since the reform and opening up, the development of the area along the Yangtze River has been included in the national development strategy.Now, YREB covers nine provinces and two cities in eastern, central and western China, with an area of 2.05 million km 2 .It has become an important densely populated region and industrial area in China, with a strategic supporting role in China's economic development.However, rapid economic development has also brought many environmental problems, such as air pollution, soil erosion, and water ecology.Therefore, this study takes YREB (Figure 1) as the research area to explore the variations in the concentration of atmospheric pollutants in this region and the factors affecting them.

Air pollution data
Merra-2 is a new generation of atmospheric analysis data released by NASA in 2016.It has long time series and global coverage, with a spatial resolution of 0.5°×0.625°and a temporal resolution of 1 hour.This dataset provides annual, monthly, diurnal, and time series data for coordinate points and annual averages for regions.In addition, atmospheric aerosols, temperature, precipitation and other products provided by Merra-2 have high accuracy in real-time analysis.Merra-2 data sets are widely used in remote sensing analysis.Significant changes in human emissions caused by the epidemic emergency in 2020 may cause errors in the air pollution concentration trend.Considering the long time series characteristics of this study and the availability of PM 2.5 data.In this study, the Merra-2 product was used to invert the concentrations of PM 2.5 components SO 4 , SO2, black carbon (BC), organic carbon (OC), sea salt (SS) and dust in the YRB during 1980-2019 (Table 1).From this, the concentration of PM 2.5 is calculated.The formula is as follows: where [SO 4 ] is the concentration of sulfate particulates.The diameter of all particles is less than or equal to 2.5 μm.

Socioeconomic data
The development of the economy inevitably leads to many environmental problems.In order to explore the influencing factors of air pollution, this study relies on the existing literature and data availability and selects a set of social and economic data from 1980 to 2019 (Table 1).The considered data include the economic scale of each city, GDP per capita, population density, proportion of industry, urban green space coverage, scientific career expenses, fiscal expenditure of local government (excluding tech spending), and local economic density.

Trend analysis
The Mann-Kendall test is a non-parametric statistical test with many advantages.It is often applied to test the variation tendency of variables.It was first proposed by Mann and Kendall (Mann, 1945) in 1945, and the formula is as follows: ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi where S is the test statistic of the normal distribution; x i and x j are two series with different distributions in the same sample (1≤j<i≤n), σ s is the standard variance, n is the total number of samples; and Z MK is the test value.When the absolute value of Z MK is greater than 2.32, 1.64 and 1.28, it means that the 1%, 5% and 10% significance level is passed, respectively.In this study, the non-parametric Mann-Kendall method was used to examine the variation tendency of atmospheric pollution from 1980 to 2019.The obtained results are described in Section 3.1.

STIRPAT model
IPAT is a quantitative relationship model that represents the impact of anthropogenic activities on the environment.It was first proposed by Ehrlich & Holdren (1971).The model is defined as follows: I stands for the anthropogenic impact on the environment, and P stands for population.The average affluence (A) is usually measured by GDP per capita.T indicates the technology level.Combined with previous research, this study takes the proportion of industry and the scientific career expenses to represent the technological level.However, the IPAT model is flawed in that its problem analysis tends to have a proportional effect on the dependent variable.To overcome this shortcoming, Dietz & Rosa (1994) improved IPAT into an extensible stochastic environmental impact assessment model (STIRPAT).The model takes the following form: After taking logarithms, the equation is as follows: where, suffix i refers to the city, while b, c and d represent the coefficients of P, A and T, respectively.ε is the error term and α is the constant term.Based on existing studies, this study improved the STIRPAT model by introducing factors such as proportion of industry, urban green space coverage, local economic density and local government fiscal expenditures.
where, I it represents the concentration of air pollution.The coefficients (b 1 , b 2 . . .b 9 ) were calculated by the second order least squares method, which represent the factor coefficients; gdp is the regional GDP, which represents the economic scale of each city; pgdp is real GDP per capita; popdens is population density, industry is proportion of industry; gre is urban green space coverage; tec is scientific career expenses; gov is local government fiscal expenditure, excluding science and technology expenditure; and ecodens represents the local economic density.
EKC is an important hypothesis that explains the relationship between economic gain and environmental quality.It was first proposed by American scholars in 1991 (Grossman & Krueger, 1991).The EKC hypothesis holds that there is an inverted U-shaped relationship between economic growth and environmental pollution.According to the EKC hypothesis, the environment deteriorates with economic gain in the early stage and starts to improve with economic gain after reaching a critical value.According to Formula (9), we can also verify the EKC hypothesis.If the coefficient b 2 is greater than 0 and the coefficient b 3 is less than 0 (i.e.b 3 <0 and b 2 >0), then the EKC hypothesis is true, and the turning point of EKC is calculated by the Formula (−0.5b 2 /b 3 ).

Spatiotemporal trends in air pollution concentrations
Temporal trends of 40-year average air pollution concentrations across the YRB were illustrated in Figure 2. Similar trends were seen between SO 2 , SO 4 , PM 2.5, BC, and OC.The concentrations of these air pollutants tended to increase gradually from 1980 to 2000.For BC and OC there were an inflection point in 1986, i.e. unlike the rapid increase in concentrations before 1986, the increase in concentrations has slowed down until 2000.It is clear from Figure 2 that there were five steep peaks in concentration growth rates for each air pollutant between 2000 and 2008.The increase in these air pollutants during 2000-2008 may be related to rapid economic growth and urbanization.After 2008, concentrations of SO 4 , BC, OC and PM 2.5 decreased with fluctuations.This may be in response to highlighted reports of atmospheric pollution during the 2008 Beijing Olympic Games, and therefore a number of mitigation options have been implemented.Also, the global economic crisis of 2008 may be partly the answer to the reduction.
Unlike the pollution trends mentioned above, the trend of sea salt pollutants was less obvious between 1980 and 2000.The change was in the range of 0.5-0.7 μg•m −3 and 2000 was the pivot point where the magnitude of the increase increases swiftly.The peak concentration was reached in 2005 at a value of around 1.5 μg•m −3 , and then the concentration fluctuates with a general downward trend.While the change of sand dust concentration can be divided into four parts (1980-1988, 1989-1996, 1997-2005 and 2006-2019), the concentration trend showed first increase and then decrease in each period.The highest concentration appeared in 2000 with a value about 24 μg•m −3 , while a valley value of 14 μg•m −3 appeared in 1988.
Figure 3 showed the spatial distribution of 40-year (1980-2019) average concentrations of different pollutants.The concentrations of PM 2.5 , SO 2 and SO 4 were at a low level across the YRB in 1980.Unlike these three pollutants, BC and OC concentrations showed spatial differentiation with high concentrations (above 3 μg•m −3 and 6 μg•m −3 , respectively) in the Sichuan Basin.For dust, high values (>20 μg•m −3 ) were observed in the north of YRB and at the source of the Yangtze River, which is adjacent to the Loess Plateau, a place with severe soil erosion and large wind and sand.Due to its proximity to the ocean, the Yangtze River Delta was a high-value area for sea salt pollution.The upper reaches of the YRB have low values of pollution, except for the dust pollution in 1980.
In 2019, a similar spatial pattern occurred for PM 2.5 , SO 2 , SO 4 , BC and OC, with high loading areas in the Sichuan Basin and the eastern part of the YRB, especially in the Yangtze River Delta.For example, the concentration of PM 2.5 in the Sichuan Basin was over 45 μg•m −3 while for the east of the YRB, the concentration is over 35 μg•m −3 .As for SO 2 , a high loading over 30 μg•m −3 occurred in the Sichuan Basin (SB) and the Yangtze River Delta.The change in BC was also obvious.In 1980, the highest loading area was in SB with a value range of 2-4 μg•m −3 .The situation changed in 2019.Both in SB and in the lower reaches of the Yangtze River, the concentration was above 4 μg•m −3 .The main areas polluted by dust and sea salt have changed less than in 1980, but the pollution range was greater, and the pollution situation was more serious.Take OC for example, in regions with strong anthropogenic activities, such as the SB and the lower stretch of the Yangtze, the concentration was at least 8 μg•m −3 .
The right column in Figure 3 described the average annual growth trend of air pollutants.A grid with black spots indicates that the P < 0.01 confidence test is passed.The significant trends of PM 2.5 , SO 2, SO 4 , OC and SS were identified over wide areas of the YRB, and only a small part of the study area showed small changes.For PM 2.5 , the changes in the whole region were significant, except for the unapparent trend in the source of the Yangtze and a small part of the northern and southern YRB.The concentration increases by at least 0.2 μg•m −3 and 0.3 μg•m −3 per year in the upper and middle reaches, respectively, in the Sichuan Basin.The lower reaches of the YRB, the increase in concentration is greater than 0.6 μg•m −3 per year.Unlike the lower and middle reaches of the YRB, the increase in BC was not as strong and obvious in the upper reaches of the Yangtze.For dust trend, only the source of the Yangtze and the southeast part of the YRB have significant upward growth.

Seasonal variations in air pollution concentrations
Figure 4a showed monthly and quarterly data on each pollutant concentration for 40 years.Among them, the monthly concentration distributions of SO 2 , SO 4 , BC and OC showed the "U" type.For SO 2 , SO 4 and BC, the minimum concentration appeared in July, while OC reached the minimum value in June.From the perspective of seasonal differentiation, the concentration of the four pollutants peaked in winter, followed by spring.Anthropogenic emissions due to fuel combustion or heating in winter are the response to the peak value, with both reaching the valley value in summer, then in autumn.This may be related to windy days and abundant precipitation in summer and autumn.This is also in accordance with the research of (Shen et al., 2020).
For PM 2.5 , the monthly concentration showed a wavy distribution.The maximum of about 28 μg•m −3 occurs in March, and the minimum of about 18 μg•m −3 occurs in July. Figure 4b described the quarterly variation and spatial distribution of each pollutant concentration.After calculating the average concentration in each season, it was noticed that spring was the season with the highest PM 2.5 concentration, followed by winter, autumn, and summer.There are two reasons for seasonal variations in reaches of the Yangtze River.All these factors have led to an increase in PM 2.5 concentration.As for natural factors, seasonal precipitation and wind in summer and autumn can reduce the loading of PM 2.5 .These reasons are discussed in detail in the next section.
Spring was the season with the highest dust concentration, especially in March (Xu et al., 2019).It is mainly distributed in the upper reaches of the Yangtze River.The reason is obvious, i.e. multiple sandstorms occurred in the spring, and sandstorms from the Taklimakan Desert have the main responsibility for this.There were no major differences in dust concentration in other seasons.The change in sea salt concentration was similar, and sea salt is mainly concentrated in the Yangtze River Delta.This shows that these two pollutants are mainly influenced by natural factors.

Drivers of air pollution in the period 1980-2019
(a) The relationship between economic scale (Lngdp) and air pollution.Table 2 showed that all estimated economic scale coefficients exerted a significant positive effect on air pollution indicators, except for the effect on BC and OC.Specifically, a 1% expansion in economic scale increased SO 2 and SS by 0.152% and 0.231% in the YREB, respectively.Additionally, the impact of economic scale on PM 2.5 has passed the 5% significance level, the confidence measures of the check data, effected on SO 4 and dust, are over 90%.As shown in Table 2, there was a significant positive relationship between the selected air pollution indicators (PM 2.5 , SO 4 , BC, dust, OC, SO 2 and SS) and the economic scale characterized by regional GDP.That is, the greater the economic scale, the more serious the air pollution.In other words, the effect of the scale of economic development cannot effectively reduce the concentration of the atmospheric environmental indicators.Therefore, in the future development of the YREB, the Chinese government should not excessively pursue the scale of GDP, but gradually establish a new mode of modern, intensive and ecological urban development in order to achieve intensive, efficient and green development of the urban economy.
(b) The relationship between the economic development level (Lnpgdp/ln2pgdp) and air pollution.Through the STIRPAT model, we tested the impact of economic development level (Lnpgdp/ln2pgdp) on air pollution in the period 1980-2019.The results showed that the coefficients of income per capita and its squared term on PM 2.5 , SO 2 and SS were statistically significant at the level of 10% or lower with positive and negative signs, respectively, indicating an inverse U-shaped relationship.In contrast, the level of economic development was found to play a key role in reducing the concentration of the other air pollutants at 5% level or lower, except for SO 4 , and also a U-shaped relationship was obtained.This does not simply mean that the concentration of air pollutants (BC, dust, OC) will increase with the improvement of economic development and the explanation will be detailed in the urban zoning comparison table (Tables 3-9).(c) The relationship between population density (Lnpopdens) and air pollution.As described in Table 2, the correlation coefficients between population density and air pollution (excluding SS) were statistically significant and positive at the 1% level.This indicated that the population, excessively concentrated in the city, exerted serious pressure on the urban environment (Dan et al., 2021).Although the purpose of population concentration in cities is to seek good job opportunities and enjoy social welfare such as education, medical care and nursing, the population pressure in the main urban areas of some big cities is too great, and the overall carrying capacity is insufficient, which may lead to lower quality of life.Therefore, with further acceleration of urbanization, the government and relevant administrative departments should appropriately control the urban population density, pay attention to  (d) The relationship between industrial structure (Lnindustry) and air pollution.Similar to the effect of population density on atmospheric contamination, it was found that all estimated coefficients of industrial structure have a significant positive effect on all selected indicators of atmospheric contamination at the 1% level.The results showed that the industrial structure, represented by the ratio of industry, exerts an inflationary influence in relation to air pollution.That is, the higher the proportion of secondary industry, the more serious the degree of air pollution (G.Li et al., 2020;Luo et al., 2018;Zhou et al., 2019).Future development of major cities in the YRB should coordinate the development of urban space, economic scale and industrial structure, especially the adjustment of the industrial structure of the three major economic zones (Shanghai, Chongqing, and Wuhan) in the YREB.Cities in the YREB will continue to reduce the ratio of high-energy, highpollution and high-emission industries, increase the proportion of tertiary industries such as green industries and high-tech industries, and accelerate the transformation and improvement of traditional industries in order to decrease urban air pollution (Liu, 2021;Ren & Matsumoto, 2020).(e) The relationship between urban greening, technology, government investment and economic density (Lngre, Lntec, Lngov and Lnecoden) and air pollution.All the estimated coefficients of greening, technology, government investment and economic density were statistically significant and negatively correlated with air pollutants concentration in the YREB at the level of 1% or higher (Table 2).Negative correlations between greening, technology, government investment and economic density and air pollution were described in the following points.First, improving the level of urban greening, represented by the urban green area coverage, simultaneously increases the possibility of vegetation to adsorb toxic substances such as PM 2.5 , SO 4 , SO 2 , and dust, and enhances the city's self-ecological resilience.Second, technological progress is a long-term determinant of air pollution management, and investment in research and development of technology largely determines the direction of influence on China's air pollution control.Table 2 showed that the selected cities of the YREB have begun to take research and technological progress as the entry points for reducing pollution and improving the air pollution treatment in China.Third, China adopts policies and measures in the fields of finance, taxation, pricing, and government procurement to support air pollution management, thus reflecting the national support for air pollution control in terms of financial input (Li et al., 2020).There is no doubt that the more government investments, the greater the degree of intervention will be, and the concentration of atmospheric pollution will decrease.Finally, the results reveal that economic density had negative, inhibitory effects on air pollutant concentrations.We can argue that economically intensive development will be an effective way to improve the situation of air pollution.

Drivers of air pollution in different areas during the period 1980-2019
Taking into account the regression coefficients shown in Table 2, this study analyzed the impact of socioeconomic data on atmospheric contamination in the YREB during 1980-2019.However, the YREB is a vast area with significant differences between different regions.Given the regional differences, it is necessary to further investigate the relation between different socioeconomic factors and air pollution.Therefore, urban areas with a population ≥ 100,000 in 2019 were selected for analysis, which included 95 urban areas.
The urban areas were stratified by population size into the large city region (1,000,000 ~ 5,000,000), the megacity region (5,000,000 ~ 10,000,000), and the super city region (>10,000,000) (Larkin et al., 2016).The super city region includes Shanghai, Wuhan and Chongqing, and the megacity region includes Hangzhou, Chengdu and Nanjing.Tables 3-9 listed the estimates of the impact of population and other variables on air pollution indicators for the entire YREB and its three regions, respectively.The findings were discussed as follows.
Similar to the entire YREB, all estimated coefficients of economic scale, population density and industrial structure were positively correlated with seven types of air pollution in the three regions.On the contrary, the correlations between greening, technology, government investment, economic density and air pollutants were negative in the three regions.It should be noted that the effect of different socioeconomic variables on air pollution was also heterogeneous across regions, especially in the super city regions and the megacity regions.For example, Table 3 provided estimates of the impact of human activities on PM 2.5 emissions for the YREB and for the super city, mega city, and large city regions.With a 1% increase in population density, the PM 2.5 concentration in the YREB, megacity, and large city regions increased by 0.352%, 1.675% and 0.323%, respectively.When other variables were kept constant, a 1% increase in greening reduced PM 2.5 concentrations by 0.218%, 1.748% and 0.207% in the YREB, megacity, and large city regions, respectively.However, the correlations between population density, greening, industrial structure and PM 2.5 concentration in the super city region were insignificant during the study period.In addition, the estimated coefficients of correlation between greening, local economic density and BC concentration were all statistically significant at the 5% level or lower in the YREB, mega city and large city regions, except for the super city region.Although the socioeconomic variables selected in this study can be reflected in the relationship with air quality to a certain extent, the impact on air pollution will not become dominant for the super city regions in China.There are also cases where socio-economic indicators have a crucial influence on the air pollution in the YREB and its three regions.For example, the estimated coefficients of population density, industrial structure, greening, and government investment were all statistically significant and positively or negatively correlated with SO 2 and SS at the 10% level or lower for the four regions, respectively (Tables 8 and 9).

Environmental kuznets curve
The aforementioned extended STIRPAT model suggests that there are differences in the EKC relation between different regions and different types of pollutants (Table 10).This result was supported by the research of Liu et al. (2015) and Managi & Kaneko (2009), who suggested that the relationship between economic gain and environmental pollution is complex.According to the empirical results described in Table 10, the three Kuznets curves will be discussed as follows: (a) Inverted U-shaped.With the quadratic EKC specification, the estimated coefficients on income per capita and its squared value in the entire YREB and its three parts were significant at the 10% level with positive and negative signs, respectively (Table 3).This result suggested that there is an inverse U-shaped relationship between PM 2.5 and income per capita in the study area.After the inflection point, the PM 2.5 concentration will decrease with the increase in income per capita (Liu et al., 2015).Although the current GDP per capita in the super city and mega city regions surpassed their inflection points of 51,021 yuan, and 29,143 yuan, the PM 2.5 concentration did not decrease significantly with economic growth.The reason for this difference may be that this study considered other factors, such as urban green space coverage and economic scale.Similarly, with the exception of the megacity region, the coefficient of income per capita and its squared term for SS concentration were statistically significant at the 1% level with positive and negative signs, respectively, indicating an inverted U-shaped relationship.(b) U-shaped.Both income per capita and its squared value for BC and OC pollutants in the three regions (YREB, megacity and large city) were statistically significant with negative and positive signs, respectively, indicating a U-shaped relationship.Interestingly, an invert U-shaped relationship was found for SO 2 in the YREB and large city region.However, a U-shaped relationship was found for SO 2 in the mega city region.Furthermore, the linkages between income per capita and SO 2 in the super city region were not obvious.
Considering that the current income per capita for the YREB and large city regions did not exceed these inflection points, but for the mega city region the current income per capita has exceeded its inflection point (42,617 yuan), the per capita SO 2 concentration increases with the increase in GDP per capita in the short term.Therefore, it is necessary to combine specific urban governance measures and policies to ensure a healthy relationship between economic development and SO 2 pollution.(c) Not obvious.For SO 4 , there was no statistical significance at the 10% level or higher, indicating that SO 4 pollution was inconsistent with the EKC hypothesis.For example, in the super city region, the linkages between income per capita and BC, dust, OC and SO 2 concentrations were not obvious.Similarly, the EKC is not significant for SS in the mega city region.

Policy implications
Currently, China has not completed its historic task of industrialization and urbanization, and air pollutants are predicted to increase in the coming years.The outlook for China's air pollutant levels is not optimistic, which led to international and domestic pressure.
Minimizing air pollutants while maintaining economic growth and social progress is a major challenge facing China today.Our empirical results will be of particular interest to Chinese policymakers and urban planners, especially to local governments in the YREB.
Optimizing urban form and traffic through sound The results obtained from the STIRPAT model described above are based on the estimated coefficients of the per capita income and their squared values.Not obvious a means that the estimated coefficients for the per capita income and their squared terms are not statistically significant at the 10% level or higher.
urban planning decisions is crucial for the treatment of most air pollutants.Future urban development planning in China must therefore begin to take into account issues of spatial optimization in addition to the socioeconomic considerations.In the context of such rapid industrialization and urbanization, policymakers also require more detailed information about the complex links between human activities and environmental impacts.Furthermore, urban planning and management policymakers should take into account regional pollutant reduction priorities.Considering the strongest impacts of population, income and technology on pollutant concentration, the super city and mega city regions should continuously introduce and develop technologies for industrial upgrading and develop green industries.Large urban areas should seek the best mode and reasonable path to undertake industrial transfer, according to their different geographical locations, resources, industrial bases and capabilities of industrial enterprises.Despite the abundance of resources, residents in large urban areas should be guided to form energy-saving and ecological consumption patterns.However, due to the distinct spatial clustering properties of air pollutants, while prioritizing regional emission reductions, it is also necessary to pay attention to regional interactions and spatial heterogeneity.In decision-making, the development status and advantages of various regions should be combined to promote coordinated emission reduction among regions and common development.In addition, energy consumption has always been an important source of air pollution, promote renewable energy, optimize the energy structure, according to regional advantages, improve the dual control path of energy consumption, such as the economic development of solar energy and other renewable energy in the middle reaches of the Yangtze River, the construction of offshore power generation facilities in coastal areas, to achieve the goal of controlling regional air pollution.

Conclusions
In this study, data from three major city clusters in the YREB from 1980 to 2019 were used as research samples.PM 2.5 , SO 4, SO 2 , BC, dust, OC and SS were used as air pollution indicators in order to verify the EKC hypothesis using the improved STIRPAT model.All air pollution data were from the Merra-2 data set in this paper.Several papers have compared Merra-2 PM 2.5 with ground-based monitoring values from air quality monitoring stations, and the results showed that Merra-2 PM 2.5 is in good agreement with groundbased observations, demonstrating the availability of Merra-2 data for this study (Song et al., 2018).In addition, there were also papers using Merra-2 data to study the temporal and spatial characteristics of each component of PM 2.5 in the Yangtze River Basin and the influence of meteorological factors (He et al., 2019), which was similar to the research results in this paper, proving the reliability of the research results in this study.The main conclusions were as follows: (a) In the past 40 years, the concentration of all pollutants in the YRB first increased and then decreased.The concentration of dust increased after a sudden decline in 2005, and the concentration of all pollutants began to decline in 2008.
Obviously, the prevention and control measures of air pollution have achieved certain results.The pollution in the YRB presented obvious spatial aggregation characteristics.PM 2.5 , SO 4 , BC, OC and SO 2 pollution was noticed throughout the Yangtze River Delta, with high concentration areas in the Sichuan Basin and the eastern YRB, especially in the Yangtze River Delta, which are densely populated and economically developed areas.As far as dust was concerned, the areas of high pollutant concentration were mainly located in the upper reaches of the YRB, adjacent to the Loess Plateau.In contrast, the concentration of other pollutants was low in this area.Due to its proximity to the ocean, the sea salt pollution was concentrated in the Yangtze River Delta.In addition, dust and sea salt pollution have gradually expanded the scope of the trend, and the pollution is more serious.(b) After adding variables such as economic scale, population density and scientific research, PM 2.5 , SO 2 and SS in the YREB showed an inverted U-shaped relationship with economic development, which supported the EKC hypothesis.
According to the year-end population, cities are divided into the super cities, megalopolis and large cities.Only the research results of the large cities were the same as those for the YREB.In megacities, only PM 2.5 pollution showed an inverted U-shaped relationship.PM 2.5 and SS have an inverted U-shaped relationship with super cities.Only the relationship between PM 2.5 and economic development in megacities supported the EKC hypothesis.In addition to the positive U-shaped relationship between megalopolis economic development and SO 2 , it was also discovered that the relationship between other atmospheric pollutants and economic development has an U-shaped relationship.It was determined that some cities, especially large cities, implemented the policies, and that there was an inflection point when the air quality improved.It can be seen that China's air pollution control has achieved phased results.Meanwhile, SO 4 , BC, dust and OC failed the significance test except for SO 4 , the regression coefficients of the other three indices were all significant above 5% level, showing a positive "U" curve relationship with economic development.This did not simply mean that the concentration of the three indices will increase with the improvement of economic development level.Other factors affecting pollutant concentrations (such as population change, scientific and technological progress, green coverage rate, etc.) should also be considered.(c) Among many control variables, economic scale, industrial structure and population density contributed to the increase of air pollution, while scientific and technological progress, economic density and increase of green coverage contributed to the reduction of air pollution, and especially the increase of government investments was an effective measure to reduce air pollution.Despite the current atmospheric pollution, management has achieved certain achievements, but it should clearly recognize that atmospheric pollution governance is a long-term process.In future economic development, more attention should be paid to accelerating the transformation of economic development mode, adjusting and optimizing the industrial structure, increasing contributions in scientific and technological innovations, promoting the application of scientific and technological achievements, and adhering to the concept of green development, in order to achieve coordinated economic and environmental development.

Figure 1 .
Figure 1.The study areas of the Yangtze River economic Belt.

Figure 2 .
Figure 2. Long-term trends of different atmospheric pollutants from 1980 to 2019.

Figure 3 .
Figure 3.The distribution of air pollutants in 1980 and 2019, and the average annual growth trend of air pollutants (the grid with black dots passes a confidence test of less than 0.01).
Figure 4a.(a)monthly and quarterly changes in concentrations of seven pollutants.(a) monthly and quarterly concentrations.

Table 1 .
The variables used in the study span the period 1980-2019.

Table 2 .
Estimation results for air pollution by STIRPAT model.The estimated method is STIRPAT.Ln () is natural logarithms; gdp, which is the regional GDP, represents the economic scale of each city; pgdp is real GDP per capital; popdens is the population density; industry is the proportion of industry; gre is the urban green space coverage; tec is the scientific career expenses; gov is the local government fiscal expenditure, excluding science and technology expenditure; ecodens represents the local economic density.Obs is observations.*Indicate statistical significance at the 10% level.**Indicate statistical significance at the 5% level.***Indicate statistical significance at the 1% level.

Table 3 .
Multivariate relationship of socioeconomic metrics with the near-surface PM 2.5 concentrations calculated by the improved STIRPAT model from 1980-2019.

Table 4 .
Similar with Table3, but for the near-surface SO 4 concentrations.

Table 5 .
Similar with Table3, but for the near-surface BC concentrations.

Table 6 .
Similar with Table3, but for the near-surface dust concentrations.

Table 7 .
Similar with Table3, but for the near-surface OC concentrations.

Table 8 .
Similar with Table3, but for the near-surface SO 2 concentrations.

Table 9 .
Similar with Table3, but for the near-surface SS concentrations.

Table 10 .
Environmental Kuznets curve (EKC) hypothesis between economic growth and seven types of air pollutant emissions in whole YREB sample and its three parts stratified by urban population from 1980-2019.