Analysis of Regional Differences in Energy-Related PM 2 . 5 Emissions in China : Influencing Factors and Mitigation Countermeasures

China’s rapid economic development has resulted in a series of serious environmental pollution problems, such as atmospheric particulate pollution. However, the socioeconomic factors affecting energy-related PM2.5 emissions are indistinct. Therefore, this study first explored the change in PM2.5 emissions over time in China from 1995 to 2012. Then the STIRPAT (Stochastic Impacts by Regression on Population, Affluence, and Technology) model was adopted for quantitatively revealing the mechanisms of various factors on energy-related PM2.5 emissions. Finally, the Environmental Kuznets Curve (EKC) hypothesis was adopted to examine whether an EKC relationship between affluence and energy-related PM2.5 emissions is present from a multiscale perspective. The results showed that energy-related PM2.5 emissions in most regions showed an increasing trend over the study period. The influences of the increase in population, energy intensity, and energy use mix on energy-related PM2.5 emissions were positive and heterogeneous, and population scale was the major driving force of energy-related PM2.5 emissions. The effects of the increase in the urbanization level and the proportion of tertiary industry increased value to GDP on energy-related PM2.5 emissions varied from area to area. An inverse U-shape EKC relationship for energy-related PM2.5 emissions was not verified except for eastern China. The conclusions are valuable for reducing PM2.5 emissions without affecting China’s economic development.


Introduction
The rapid advancement of China's industrialization has led to a deterioration of air quality in China.Severe air pollution has spread from developed regions, especially in the developed eastern provinces, to the whole country.The frequency of hazy weather showed a significant upward trend, and air pollution has increasingly become a core issue that constrains sustainable development and ecologically friendly urban development [1,2].Therefore, effective control of pollutant emissions and effective improvement of urban environmental air quality have become important targets for the Chinese government [3].Fine particulate matter (PM 2.5 ) can not only be suspended in the air, but also generates new pollutants through chemical reactions and reduces atmospheric visibility [4].In addition, PM 2.5 pollution can seriously affect human health [5][6][7].Many medical studies have shown that long-term exposure to air containing PM 2.5 can cause respiratory diseases and cardiovascular diseases, damage the body's immune system, and increase the risk of death in exposed populations [8][9][10][11].The number of premature deaths per year in China due prolonged exposure to polluted air exceeds 1.25 million [12].In the winter of 2013, the maximum concentration of PM 2.5 in Beijing exceeded 1000 µg/m 3 [13].Therefore, the PM 2.5 pollution problem has increasingly become the focus of China's air pollution prevention and control, and is also one of the hotspots of atmospheric environment research.
Many scholars have carried out a large number of studies on PM 2.5 pollution in recent years.The research has mainly focused on the transboundary diffusion of PM 2.5 [14][15][16], the health effects of PM 2.5 [17][18][19][20], source analysis of PM 2.5 pollutants [21][22][23][24], simulation of PM 2.5 concentration [25][26][27][28], spatiotemporal changes and patterns of PM 2.5 pollution [29][30][31], contributing factors analysis of PM 2.5 concentration [32][33][34][35], establishing a PM 2.5 emissions inventory [36,37], and determining the dispersion of PM [38][39][40].Total pollutant discharge control is a measure for environmental management, so reducing PM 2.5 emissions will be a commendable means of preventing PM 2.5 pollution.In 2012, new air quality standards were formulated, and PM 2.5 concentration was listed as a routine key environmental monitoring index for the first time.In 2013, the Chinese government published 10 air pollution control measures to cope with serious and persistent PM 2.5 pollution.The Chinese government is also trying to expand the scope of air quality monitoring.The monitoring sites for PM 2.5 concentration have increased from 612 in 2013 to 1436 in 2016.After that, a couple of studies tried to quantify the socioeconomic factors contributing to PM 2.5 emissions and reveal the dynamic relationship between these variables in recent years.For instance, Guan et al. [41] adopted the input-output method and the structural decomposition analysis (SDA) method to quantify influencing factors on China's primary PM 2.5 discharge from 1997 to 2010.Meng et al. [42] applied an input-output model for revealing the influence of trade contributing to primary PM discharge in Beijing.Lyu et al. [43] employed the Logarithmic Mean Divisia Index (LMDI) method for exploring the major driving force of primary PM 2.5 discharge from 1997 to 2012 in China.Xu et al. [44] applied the SDA method for identifying the socioeconomic factors contributing to China's primary air pollutant discharge.
The concept of the Environmental Kuznets Curve (EKC) hypothesis is that environmental quality initially deteriorates with economic growth, and then increases as economic development reaches a certain level.In recent years, this hypothesis has been used to empirically study the relationship between environmental quality and per capita income.For instance, the EKC hypothesis was applied by Brajer et al. [45] to test whether air pollution presented an inverted-U-type EKC relationship with economic growth.The EKC hypothesis for CO 2 and SO 2 emissions' relationship with economic growth in the United Kingdom was verified by Fosten et al. [46].In Malaysia, the EKC hypothesis for carbon dioxide emissions' correlation with increasing income was discovered by Saboori et al. [47].Shahbaz et al. [48] discovered evidence supporting an inverted-U-shaped nexus between financial development and carbon dioxide emissions.Hao et al. [49] verified an inverted-U-shaped EKC relationship for PM 2.5 concentrations in 73 Chinese cities with economic growth.The research by Wang et al. [50] confirmed that sulfur dioxide emissions had an inverted-U-shaped link with economic growth.The EKC hypothesis for three pollutant emissions (carbon dioxide, industrial wastewater, and industrial waste solid) in China was examined by Li et al. [51] and the results approved this hypothesis.Chen et al. [52] applied the EKC hypothesis to explore the nexus between affluence and Air Pollution Index (API) in China.
In summary, there are still some limitations in the existing literature.Firstly, the input-output method, the SDA method, and the LMDI method are often used in the existing research, but the econometric analysis method is seldom used.Secondly, there are few papers analyzing the factors contributing to PM 2.5 discharge, especially about the regional differences in China's energy-related PM 2.5 emissions influencing factors.Thirdly, although a lot of researchers have studied the EKC relationship between pollution and economic development, this paper is different because it examines primary PM 2.5 emissions for the first time.The main contributions of this study are: (1) filling the gap in that the econometric analysis method is seldom used to reveal the influencing factors of energy-related PM 2.5 emissions.It can also be regarded as an example of using the STIRPAT (Stochastic Impacts by Regression on Population, Affluence, and Technology) model to examine the influencing factors of energy-related PM 2.5 emissions.(2) The EKC hypothesis is used to study the relationship between energy-related PM 2.5 emissions and GDP for the first time, and this paper contributes to the empirical literature.(3) The study can improve the public's understanding of the changes of PM 2.5 emissions with economic growth, which is valuable for harmonizing economic growth and PM 2.5 emissions reduction.
This paper has three main objectives: (1) exploring the change in PM 2.5 emissions over time; (2) revealing quantitatively the influencing factors of energy-related PM 2.5 discharge from a multiscale perspective; (3) examining whether an EKC relationship between affluence and energy-related PM 2.5 emissions is present from a multiscale perspective.

Research Area
In order to analyze the socioeconomic factors affecting energy-related PM 2.5 emissions from different scales, two research areas with different administrative scales were chosen in this paper.Mainland China was the first research area (Figure 1).With the start of Chinese economic reform in 1978, China's economy underwent rapid development that resulted in a series of serious environmental problems and massive energy-related PM 2.5 emissions.However, the influence mechanism of socioeconomic factors contributing to energy-related PM 2.5 emissions is still ill-defined.The eastern region, central region, and western region were together chosen as the second research area.Due to natural conditions and policy factors, eastern China took the lead in development and formed a relatively complete industrial economic system, while the development of central and western China was relatively slow, which makes China's regional economic development imbalanced.Regional differences in China are becoming more and more prominent, mainly in the following three aspects: (1) the difference in regional economic development level; (2) the difference in regional economic structure; (3) the difference in per capita income level.Different regions inevitably have different socioeconomic factors contributing to energy-related PM 2.5 emissions.
relationship between energy-related PM2.5 emissions and GDP for the first time, and this paper contributes to the empirical literature.(3) The study can improve the public's understanding of the changes of PM2.5 emissions with economic growth, which is valuable for harmonizing economic growth and PM2.5 emissions reduction.
This paper has three main objectives: (1) exploring the change in PM2.5 emissions over time; (2) revealing quantitatively the influencing factors of energy-related PM2.5 discharge from a multiscale perspective; (3) examining whether an EKC relationship between affluence and energy-related PM2.5 emissions is present from a multiscale perspective.

Research Area
In order to analyze the socioeconomic factors affecting energy-related PM2.5 emissions from different scales, two research areas with different administrative scales were chosen in this paper.Mainland China was the first research area (Figure 1).With the start of Chinese economic reform in 1978, China's economy underwent rapid development that resulted in a series of serious environmental problems and massive energy-related PM2.5 emissions.However, the influence mechanism of socioeconomic factors contributing to energy-related PM2.5 emissions is still ill-defined.The eastern region, central region, and western region were together chosen as the second research area.Due to natural conditions and policy factors, eastern China took the lead in development and formed a relatively complete industrial economic system, while the development of central and western China was relatively slow, which makes China's regional economic development imbalanced.Regional differences in China are becoming more and more prominent, mainly in the following three aspects: (1) the difference in regional economic development level; (2) the difference in regional economic structure; (3) the difference in per capita income level.Different regions inevitably have different socioeconomic factors contributing to energy-related PM2.5 emissions.

Data Sources
Based on data availability, this study selected the time series dataset of 30 provinces in China in 1995-2012, excluding Hong Kong, Macao, Taiwan, and Tibet.The provincial primary energy-related PM 2.5 emissions data in China were downloaded from the Multi-resolution Emission Inventory for China (MEIC).The data of the provincial gross domestic product (GDP), provincial total population, urban population, coal use, energy use, and the tertiary industry (primary industry refers to industry for the production of food and other biological materials, including planting, forestry, animal husbandry, aquaculture, etc.; secondary industry refers to industry that reprocesses the basic materials provided by the primary industry and nature, including mining, manufacturing, power, gas and water production and supply, construction, etc.; tertiary industry refers to the service industry, mainly including transportation, communications, commerce, catering, financial industry, education industry, etc.) increased value were all obtained from the National Bureau of Statistics of China.In order to eliminate inflation, GDP was calculated at a constant price in 1995.The detailed description of the data used in this study are shown in Table 1.

Time Variation Trend (Slope)
The simple linear regression model is a method frequently used to detect the time variation trends of observation data.The equation is as follows: where n denotes the time span and is equal to 18, and I a represents the energy-related PM 2.5 emissions in a year.

STIRPAT Model
The formula is as follows: where I denotes energy-related PM 2.5 emissions, P indicates the total population, A expresses the affluence, T represents the technology level, a denotes the model coefficients, b, c, and d represent the simulation coefficient of independent variables, and e means the error value sustainability-425450.
After taking natural logarithm, Equation ( 2) was transformed into: As the STIRPAT model gives consideration to the individual impact of different changes in P, A, and T on the pollutant emissions and can be extended by incorporating other factors and appropriately decomposing each factor, this method is widely applied to reveal the factors affecting pollutant discharge.For instance, Li et al. [53,54] utilized a STIRPAT model for exploring the influence of different factors on China's CO 2 emissions.A STIRPAT model was utilized for detecting the major factors affecting Beijing's carbon dioxide emissions [55].Shafiei et al. [56] explored the determinants of carbon emissions in the Organization for Economic Co-operation and Development (OECD) countries based on the STIRPAT model.Shahbaz et al. [57] explored the determinants of Malaysian energy use based on the STIRPAT model and found that urbanization was the major factor affecting Malaysian energy use growth.Laureti et al. [58] explored the determinants of Madrid's NOx discharge based on an augmented STIRPAT model.Zhang et al. [59] utilized a STIRPAT model for revealing the relationships between population aging and CO 2 emissions from multiscale perspective.Wang et al. [60] attempted to combine the STIRPAT model with a semi-parametric regression model for revealing the link between the income/urbanization and industrial CO 2 discharge.Xie et al. [61] employed an improved STIRPAT model for exploring the transportation infrastructure and urban carbon emissions nexus.Chai et al. [62] adopted a comprehensive LMDI-STIRPAT-PLSR model for analyzing influencing factors with regard to natural gas consumption.Wang et al. [63] quantified the major driving factors for Xinjiang's CO 2 emissions from different development stages.Zhang et al. [64] utilized an STIRPAT model for identifying the inherent relation between urbanization and CO 2 emissions.Yang et al. [65] utilized a STIRPAT model to quantify the influences of natural and socioeconomic parameters for CO 2 emissions.

Multicollinearity Testing
Multicollinearity represents a situation in which there is a strong and linear correlation between independent variables.If multicollinearity is present among explanatory variables in the linear regression model, it will lead to: (1) an OLS estimator with large variance and low precision; (2) the individual effects of variables not being judged; (3) a nonsense significance test; and (4) an unstable regression model [69].These unstable alterations will bring about an unreasonable regression model, thus providing unreliable results for variables.
There are many methods to assess "strong and linear correlations" between factors, such as the variance inflation factor (VIF) and correlation matrix.VIF, calculated by the ordinary least square (OLS) regression method, is the most commonly used method to evaluate whether there is multicollinearity between independent variables in regression models.The larger the VIF, the more serious the multicollinearity.It is generally believed that there is serious multicollinearity if a VIF is greater than 10 [70-72].

Ridge Regression
Ridge regression is an improved OLS estimation method because it abandons its unbiased property and can effectively decrease the standard error.Ridge regression is a more practical and reliable way to obtain fitting coefficients at the cost of precision loss.If there is multicollinearity between independent variables, the value of determinant of X'X matrix is approximately 0 (X is an n×b matrix (rank b) of independent variables).Therefore, matrix (X'X) −1 will be hypersensitive to tiny changes in the data.If X'X is added to the constant matrix KI under the condition K ≥ 0, the sensitivity of (X'X + KI) −1 will be improved (K is the constant).Therefore, the coefficients estimation of ridge regression is more stable than the OLS regression method.Its estimator is as follows: If K = 0, its estimator is the result of OLS regression; if K→∞, its sustainability-425450 is close to 0, so K should not be too large.

The Variation Trend of Energy-Related PM 2.5 Emissions
The energy-related PM 2.5 emissions in 1995-2012 in China and the three economic zones are displayed in Figure 2.During the period from 1995 to 2006, the energy-related PM 2.5 emissions in the three economic zones showed an increasing trend.The rise of energy-related PM 2.5 emissions in western China was the fastest.The growth rate of energy-related PM 2.5 emissions in central China was the second highest.Energy-related PM 2.5 emissions in the eastern regions grew slowly and remained basically unchanged.During the period from 2006 to 2012, there were obvious differences in the changes of energy-related PM 2.5 emissions.Energy-related PM 2.5 emissions in eastern China featured a downward trend, while energy-related PM 2.5 emissions in central and western China showed an upward trend.A simple linear regression model was used to calculate the slopes of energy-related PM 2.5 emissions from 1995 to 2012.The slope was 5.29 × 10 4 tons/year in China, and the energy-related PM 2.5 emissions were fluctuating upward.The slope was −4.21 × 10 4 tons/year in the eastern region, indicating a downward trend from 1995 to 2012.The slope of central and western China were 3.61 × 10 4 tons/year and 5.89 × 10 4 tons/year respectively, which showed that energy-related PM 2.5 emissions had an upward trend in 1995-2012.

Results of Multicollinearity Inspection
According to the data collected, OLS regression was adopted in SPSS to determine whether multicollinearity is present (OLS results are displayed in Table 3).Several VIFs were much higher than 10, which implied that multicollinearity between independent variables was present.Therefore, the coefficients fitted by the OLS regression method cannot be guaranteed and the OLS results cannot be applied to quantify the factors affecting energy-related PM2.5 emissions.Obviously, in order to get reliable regression results, the multicollinearity between independent variables must be eliminated.

Results of Multicollinearity Inspection
According to the data collected, OLS regression was adopted in SPSS to determine whether multicollinearity is present (OLS results are displayed in Table 3).Several VIFs were much higher than 10, which implied that multicollinearity between independent variables was present.Therefore, the coefficients fitted by the OLS regression method cannot be guaranteed and the OLS results cannot be applied to quantify the factors affecting energy-related PM 2.5 emissions.Obviously, in order to get reliable regression results, the multicollinearity between independent variables must be eliminated.

Ridge Regression Estimation
The ridge regression method was applied to fit Equation (4) for reducing the impact of multicollinearity between independent variables; the simulation coefficients are shown in Table 4. Its estimation coefficients are selected according to the ridge trace.When K = 0.1 (whole of China), K = 0.05 (eastern region), K = 0.08 (central region), K = 0.1 (western region), the ridge trace is almost stable.The specific coefficients are displayed in Table 4.

The Whole of China
For the whole of China, population scale, energy intensity, urbanization, and energy use mix all had a positive and statistically effects on energy-related PM 2.5 emissions, whereas the negative effects of industrial mix on energy-related PM 2.5 emissions was significant.The fit coefficients of energy-related PM 2.5 emissions to population scale, energy intensity, urbanization, and energy use mix were 0.864, 0.174, 0.213, and 0.526 respectively, which showed that energy-related PM 2.5 emissions increased with the increase of total population, technology level, urbanization level, and ratio of coal use to total energy use.The fit coefficients of energy-related PM 2.5 emissions to industrial mix was −0.679, indicating that China's energy-related PM 2.5 emissions declined with the increase of the proportion of added value of tertiary industry to GDP.Moreover, the fit coefficients of energy-related PM 2.5 emissions to per capita GDP and its square were all positive and significant at the level of 5%.This indicated that an EKC relationship between wealth and energy-related PM 2.5 emissions was not present in China during this study period.

The Eastern Region of China
For the eastern region, energy-related PM 2.5 emissions were positively and statistically significantly correlated with the population scale, energy intensity, and energy use mix.The fit coefficients of energy-related PM 2.5 emissions to population scale, energy intensity, and energy use mix are 0.936, 0.288, and 0.453, respectively.This implied that the influences of the increase the increase in the total population, technology level, and ratio of coal use increased the energy-related PM 2.5 emissions in the eastern region.The proportion of tertiary industry increased value to GDP showed a negative correlation with the energy-related PM 2.5 emissions, and was statistically significantly under the level of 5%.The proportion of increased value of tertiary industry to GDP increased by 1%, which resulted in a 0.242% decrease in the energy-related PM 2.5 emissions.Moreover, the sensitivity of energy-related PM 2.5 emissions production in eastern China to urbanization was minus but statistically not significant.Interestingly, the fit coefficients of energy-related PM 2.5 emissions to per capita GDP and its square value were found to be statistically significant with a positive and negative correlation at the level α = 0.1, respectively.Energy-related PM 2.5 emissions first increased sharply and then decreased with wealth growth, indicating that an EKC relationship between economic level and energy-related PM 2.5 emissions was present in the eastern region during this study period.

The Central Region of China
For the central region, the simulated elasticities were all statistically significant under the level α = 0.1 or lower.Similar to the national scale, population size, energy intensity, urbanization, and energy use mix all positively influenced energy-related PM 2.5 emissions, whereas industrial mix negatively affected energy-related PM 2.5 emissions.The proportion of increased value of tertiary industry to GDP increased by 1%, which resulted in a 0.073% decrease in the energy-related PM 2.5 emissions.The elasticities of energy-related PM 2.5 emissions to population scale, energy intensity, urbanization and energy use mix were 0.839, 0.420, 0.052, and 0.511, respectively, indicating that the impacts of the decrease in the total population, technology level, urbanization level and ratio of coal use to total energy use all decreased the energy-related PM 2.5 emissions in the central China.Moreover, the elasticities of energy-related PM 2.5 emissions to per capita GDP and its square value were significant at the level α = 0.1 and showed a positive correlation.This demonstrated that the existence of an inverted-U-type EKC relationship between wealth and energy-related PM 2.5 emissions in central China was not validated during this study period.

The Western Region of China
For the western region, the effect of the total population, technology level, and percentages of coal use to total energy use on energy-related PM 2.5 emissions were positive.The positive sensitivity of total population, energy use per unit GDP, and percentages of coal use on energy-related PM 2.5 emissions were 0.828, 0.419, and 0.286, respectively, which implied that the effects of the decrease in total population, energy use per GDP and percentages of coal use fell by energy-related PM 2.5 emissions.The effects of the urbanization level and the ratio of tertiary industry increased value to GDP on energy-related PM 2.5 emissions were all not significant even at the 10% level.Besides, the simulated elasticities of energy-related PM 2.5 emissions to per capita GDP and its square value were all statistically significant with a positive correlation under the level of 1%, demonstrating no EKC relationship between affluence and energy-related PM 2.5 emissions in western China during the study period.

Conclusions and Policy Recommendations
This study explored the change in PM 2.5 emissions over time in China.Then, an extended STIRPAT model was adopted for quantitatively revealing the various socioeconomic factors on energy-related PM 2.5 emissions from multiscale perspective in 1995-2012.Finally, the EKC hypothesis was adopted to test whether there is an EKC relationship between affluence and energy-related PM 2.5 emissions from multiscale perspective and some conclusions have been drawn.
The results showed that energy-related PM 2.5 emissions in most regions showed an increasing trend over the study period.The effects of the increase in population size, energy intensity and energy use mix on energy-related PM 2.5 emissions were positive and heterogeneous in China, population size was the major driving force of energy-related PM 2.5 emissions.Higher urbanization increased energy-related PM 2.5 emissions for the whole of China and central China, whereas the effects of urbanization level on energy-related PM 2.5 emissions in eastern and western China were statistically insignificant even at the level α = 0.1.The proportion of tertiary industry increased value to GDP had negative influences on energy-related PM 2.5 emissions for the whole of China and eastern and central China, whereas the effects of the proportion of tertiary industry increased value to GDP on energy-related PM 2.5 emissions of western China was statistically insignificant even at the level α = 0.1.Moreover, the EKC relationship for energy-related PM 2.5 emissions has not been verified, excluding eastern China.
Population size had a significant positive effect on PM 2.5 emissions.In general, an increase in population size can affect PM 2.5 emissions in two ways: (1) the agglomeration effect-increasing population size often produces an agglomeration effect, which will improve technological level, public transport sharing efficiency, and energy efficiency to reduce PM 2.5 emissions; (2) the scale effect-the increase in population size will directly or indirectly lead to an increase in energy consumption, which will result in the increase of energy-related PM 2.5 emissions.The results showed that the scale effect of population size was significantly higher than its aggregation effect during the study period.Therefore, in the future, the government should pay more attention to the role of the population agglomeration effect for mitigating its scale effect on primary PM 2.5 emissions.For example, the government should use the media for publicizing the concept of a green life, striving to raise public awareness of low energy use, advocating a low-energy lifestyle, and promoting sustainable consumption patterns for households.
The eastern region demonstrated an EKC relationship as GDP increased.Most regions of China have levels of PM 2.5 emissions that are still increasing, that is to say, the PM 2.5 emissions will continue to be positively correlated with economic growth for some time.The "decoupling" stage between PM 2.5 emissions and economic growth has not yet arrived, which again illustrates the urgency of energy conservation and emissions reduction.Therefore, the government should emphasize sustainable development and further guide residents' green travel and consumption in order to achieve a win-win situation of stable economic growth and continuous decline of PM 2.5 emissions.For example, advocating "low-energy transportation" is also an effective measure to reduce PM 2.5 emissions.
The effect of the increase in energy intensity on energy-related PM 2.5 emissions was positive.Technological progress can often affect PM 2.5 emissions through production technology and emissions reduction technology, that is to say, technological progress can reduce PM 2.5 emissions, but also can promote economic growth, resulting in an increase in PM 2.5 emissions.The results showed that the effect of production technology is significantly higher than that of emissions reduction technology during the research period.Therefore, the government should accelerate research and development of energy conservation technology, develop new energy-efficient products, implement incentive policies, and guide enterprises to improve energy efficiency.
The effects of the increase in energy use mix on energy-related PM 2.5 emissions were positive, indicating that the increase of coal proportion increased primary PM 2.5 emissions.Therefore, it is necessary to adjust the energy use mix for mitigating its positive impact on energy-related PM 2.5 emissions.The Chinese government should actively develop clean, green energy to reduce the proportion of coal use.
Higher urbanization increased energy-related PM 2.5 emissions for the whole of China and the central region, whereas the impact of urbanization on energy-related PM 2.5 emissions in the eastern and western region was not significant.In the process of urbanization, building new buildings will consume a large amount of energy.Therefore, the government should advocate the use of more environmentally friendly materials instead of traditional cement and improve the quality of buildings for reducing energy use.
The proportion of tertiary industry increased value to GDP had negative influences on energy-related PM 2.5 emissions, excluding the western region.This is because the leading industry in the western region is secondary industry, and the tertiary industry accounts for a relatively low proportion and develops slowly.Therefore, the government should promote the optimization, transformation, and upgrading of industrial structures, and develop tertiary industry.
Our conclusions have improved the public's understanding of the changes of PM 2.5 emissions with economic growth, and are valuable for harmonizing economic growth and PM 2.5 emissions

Figure 1 .
Figure 1.The spatial distribution of the research areas.Figure 1.The spatial distribution of the research areas.

Figure 1 .
Figure 1.The spatial distribution of the research areas.Figure 1.The spatial distribution of the research areas.

Figure 2 .
Figure 2. The variation trend of energy-related PM 2.5 emissions.

Table 1 .
Description of the data used in this study.

Table 2 .
Description of variables.