Health Impact Attributable to Improvement of PM 2.5 Pollution from 2014–2018 and Its Potential Beneﬁts by 2030 in China

: With the advancement of urbanization and industrialization, air pollution has become one of the biggest challenges for sustainable development. In recent years, ambient PM 2.5 concentrations in China have declined substantially due to the combined effect of PM 2.5 control and meteorological conditions. To this end, it is critical to assess the health impact attributable to PM 2.5 pollution improvement and to explore the potential beneﬁts which may be obtained through the achievement of future PM 2.5 control targets. Based on PM 2.5 and population data with a 1 km resolution, premature mortality caused by exposure to PM 2.5 in China from 2014 to 2018 was estimated using the Global Exposure Mortality Model (GEMM). Then, the potential beneﬁts of achieving PM 2.5 control targets were estimated for 2030. The results show that premature mortality caused by PM 2.5 pollution decreased by 22.41%, from 2,361,880 in 2014 to 1,832,470 in 2018. Moreover, the reduction of premature mortality in six major regions of China accounted for 52.82% of the national total reduction. If the PM 2.5 control target can be achieved by 2030, PM 2.5 -related premature deaths will further decrease by 403,050, accounting for 21.99% of those in 2018. Among them, 87.02% of cities exhibited decreases in premature deaths. According to the potential beneﬁts in 2030, all cities were divided into three types, of which type III cities should set stricter PM 2.5 control targets and further strengthen the associated monitoring and governance. The results of this study provide a reference for the formulation of air pollution control policies based on regional differences.

The World Health Organization (WHO) has established an air quality guideline (AQG) for PM 2.5 concentration and has stipulated an annual average concentration limit of 10 µg/m 3 . Furthermore, three interim target values (IT-1, IT-2 and IT-3) have also been established, encompassing annual average concentration limits of 35, 25 and 15 µg/m 3 , respectively. In China, the guidelines for PM 2.5 in secondary environmental standards use the WHO IT-1 standard. The Ministry of Ecological Environment of the People's Republic of China has noted that 265 of 338 cities nationwide exceeded ambient air quality weighted regression algorithm integrating satellite observation data, a chemical transport model, and ground monitoring station data with 0.01-degree (~1 km) resolution [42]. These data were subsequently resampled at 1 km spatial resolution.
The population density data from 2014 to 2018 were obtained from the LandScan™ data set provided by Geographic Information Science and Technology (GIST), from Texas A&M University, with a 1 km resolution [43]. The spatial population distribution data for 2030 were obtained from Chen et al. [44], which also had a spatial resolution of 1 km.
The national disease-specific baseline mortality was obtained from the GBD database [45], and the baseline mortality rates of each province (municipality, autonomous region) were obtained from a survey of causes of death in various regions of China conducted by Zhou et al. [46]; however, this only provided data for 2013. Therefore, we combined these two databases to derive the provincial baseline mortality from 2014 to 2018.
In addition, administrative boundary data were obtained from the Resource and Environment Science and Data Center (RESDC), Chinese Academy of Sciences [47].

Global Exposure Mortality Model (GEMM)
In order to assess the health burden attributable to PM 2.5 in China from 2014 to 2018, we used the GEMM method to predict relative risk (RR). GEMM is the first exposureresponse model using Chinese cohort data, and has a large range of observed concentrations (2.4-84.0 µg/m 3 ) [41]. The formula for calculating the RR is as follows: where RR i is the RR of premature mortality due to PM 2.5 exposure in grid i; β is the exposure-response model coefficient; C i is the annual PM 2.5 concentration in grid i; C 0 indicates the theoretical minimum risk exposure level, which is 2.4 µg/m 3 ; and α, µ, and ν determine the curved form of the exposure-response relationship in GEMM. The values of the model parameters for this study are given in Table 1.

Estimation of Premature Mortality
Based on the GEMM, the premature mortality caused by PM 2.5 pollution can be estimated using Equation (2): where M i is the premature mortality caused by PM 2.5 pollution in grid i; RR i is the RR of IHD, stroke, COPD, LC, and LRI in grid i; B i is the baseline mortality rate of IHD, stroke, COPD, LC, and LRI in grid i; and P i is the exposed population in grid i.

Sensitivity Analysis
Changes in premature deaths were associated with the variations in PM 2.5 concentration, baseline mortality, and population density. In order to quantify the impact of each driving factor, we performed a sensitivity analysis where only a single factor was allowed to change from 2014 to 2018 (as a variable), while the other two factors were kept at their 2014 levels (as constants). Here, the difference between the results of the sensitivity analysis and the results of the 2014 baseline scenario (where 2014 values were used for all driving factors) was used to represent the relative contribution of each driving factor to premature deaths.

Health Benefits by 2030
In recent years, the Chinese government has taken several measures to promote PM 2.5 emissions reduction with the aim to reach the WTO IT-1 standard (35 µg/m 3 ) nationwide by 2030 [48]. In order to understand the reduction potential of health burden, the health benefits of achieving the PM 2.5 control target by 2030 were assessed. At the moment, there exist few evaluations of the health benefits associated with the current planning goals. Some studies have mainly used the future population data at the city or county level, which were obtained through the divisional statistics of the future population scenario data from the Shared Socioeconomic Pathways (SSPs) Public Database [49]. These data do not consider China's specific population policies (e.g., the two-child policy and population ceiling policy), or low spatial precision, as a recent study has found that the spatial resolution could significantly influence the results of air pollution on health impacts [50].
In this study, the spatial population distribution data for 2030 were gridded population data considering the specific population policy of China, with a spatial resolution of 1 km. Moreover, taking the annual mean PM 2.5 concentration in 2015 as the baseline concentration, regions with an annual mean PM 2.5 concentration more than 35 µg/m 3 will reach the PM 2.5 control target value by 2030, while regions in which the annual mean PM 2.5 concentration is less than or equal to 35 µg/m 3 will maintain the baseline concentration as the PM 2.5 control target [33]. The spatial distribution of population density and PM 2.5 concentration in 2030 are presented in Figure S1. As for the baseline mortality rate, according to the 'Healthy China 2030 Planning Outline', the mortality rate of diseases in China will drop by 30% in 2030 compared to 2015 [51].

Spatial Distribution of PM 2.5 Pollution and Its Relationship with Population Density
The spatial distribution of PM 2.5 in China was highly similar from 2014 to 2018, although the overall PM 2.5 concentration exhibited a decreasing trend (see Figure 1). Highvalue areas were mainly concentrated in the Beijing-Tianjin-Hebei region and surrounding areas (BTHS), the Yangtze River Delta (YRD), the Fenhe and Weihe Plain (FWP), the Chengdu-Chongqing region (CYR), and the Triangle of Central China (TCC) (see Table  S1 and Figure S2), as well as in Xinjiang. The lower-value areas were distributed in the western plateau and southeast coastal areas. However, the spatio-temporal changes of PM 2.5 significantly differed in China from 2014 to 2018. The PM 2.5 concentration in the eastern part of China decreased obviously, especially in the areas of high concentration ( Figure 2). The PM 2.5 concentration in the BTHS decreased from 73.55 µg/m 3 in 2014 to 49.49 µg/m 3 in 2018, accounting for a 32.71% reduction in five years. The PM 2.5 concentrations in YRD, FWP, CYR, and TCC decreased significantly, with a mean concentration of more than 10 µg/m 3 . Moreover, that in the Pearl River Delta (PRD) also decreased by 9.38 µg/m 3 , which was greater than the national mean (5.56 µg/m 3 ). While the western part of China was generally in a rising state, especially in the Taklimakan Desert in southern Xinjiang (with the maximum value of 20 µg/m 3 ; Figure 1f), more notably, the PM 2.5 concentration was slowly declining before 2018. This is because the Taklamakan Desert in Xinjiang is a major dust source [52]. The Aerosol Optical Depth (AOD) over Taklimakan showed a decreasing trend from 2014 to 2017, while the AOD value significantly increased in 2018 [53,54]. This may be also linked to natural soil characteristics, drought, and atmospheric dynamics in central Asia and West China [55].
Taklimakan Desert in southern Xinjiang (with the maximum value of 20 µg/m 3 ; Figure 1f), more notably, the PM 2.5 concentration was slowly declining before 2018. This is because the Taklamakan Desert in Xinjiang is a major dust source [52]. The Aerosol Optical Depth (AOD) over Taklimakan showed a decreasing trend from 2014 to 2017, while the AOD value significantly increased in 2018 [53,54]. This may be also linked to natural soil characteristics, drought, and atmospheric dynamics in central Asia and West China [55].   Taklimakan Desert in southern Xinjiang (with the maximum value of 20 µg/m 3 ; Figure 1f), more notably, the PM 2.5 concentration was slowly declining before 2018. This is because the Taklamakan Desert in Xinjiang is a major dust source [52]. The Aerosol Optical Depth (AOD) over Taklimakan showed a decreasing trend from 2014 to 2017, while the AOD value significantly increased in 2018 [53,54]. This may be also linked to natural soil characteristics, drought, and atmospheric dynamics in central Asia and West China [55].    (Table S2). At the city level, the PM 2.5 concentrations in almost all (95.87%) cities decreased, among which Shijiazhuang, Xingtai, Langfang, and Hengshui in Hebei Province experienced the most significant decreases (more than 30 µg/m 3 ). In addition, the PM 2.5 concentration in 4.13% of cities increased by less than 10 µg/m 3 , among which 12 cities exhibited an increase in PM 2.5 concentration of 0-5 µg/m 3 ( Figure S3). Changes in PM 2.5 concentrations are largely influenced by human activities [56,57] and, to analyze the relationship among these two factors, we fitted the PM 2.5 concentration with the population density ( Figure 3). Overall, there was a certain correlation between PM 2.5 concentrations and population density, where the value of the coefficient of determination (R 2 ) was about 0.1 from 2014-2018, meaning that most cities with severe pollution also had a concentrated population distribution. The average concentration of PM 2.5 from 2014-2018 was divided into low and high concentrations, for a more in-depth analysis of the effect of population density on PM 2.5 concentration (Figure 4). The correlation between PM 2.5 and population was more significant at high than at low concentration, with determination coefficients (R 2 ) of 0.095 and 0.053, respectively. The reason for this is that the cities with high population density were different: some of them were facing problems relating to PM 2.5 pollution, but some cities had lower PM 2.5 concentrations. In the early stage, population clustering further expanded the demand for necessities such as clothing, food, shelter and transportation, resulting in the use of more fossil fuels and the creation of more construction/domestic dust [58]. However, studies have pointed out that when socio-economic development reaches a higher stage the effect of human agglomeration on the environment can be mitigated through technological innovation and industrial transformation, as well as other approaches [59].
At the provincial level, only two out of 31 provinces exhibited increased PM2.5 concentrations from 2014 to 2018: Hainan and Xinjiang. Among the remaining 29 provinces, Beijing, Tianjin, Hebei, and Shandong in the North China Plain exhibited the most obvious decreases in PM2.5 concentrations (Table S2). At the city level, the PM2.5 concentrations in almost all (95.87%) cities decreased, among which Shijiazhuang, Xingtai, Langfang, and Hengshui in Hebei Province experienced the most significant decreases (more than 30 µg/m 3 ). In addition, the PM2.5 concentration in 4.13% of cities increased by less than 10 µg/m 3 , among which 12 cities exhibited an increase in PM2.5 concentration of 0-5 µg/m 3 ( Figure S3). Changes in PM2.5 concentrations are largely influenced by human activities [56,57] and, to analyze the relationship among these two factors, we fitted the PM2.5 concentration with the population density ( Figure 3). Overall, there was a certain correlation between PM2.5 concentrations and population density, where the value of the coefficient of determination (R 2 ) was about 0.1 from 2014-2018, meaning that most cities with severe pollution also had a concentrated population distribution. The average concentration of PM2.5 from 2014-2018 was divided into low and high concentrations, for a more in-depth analysis of the effect of population density on PM2.5 concentration (Figure 4). The correlation between PM2.5 and population was more significant at high than at low concentration, with determination coefficients (R 2 ) of 0.095 and 0.053, respectively. The reason for this is that the cities with high population density were different: some of them were facing problems relating to PM2.5 pollution, but some cities had lower PM2.5 concentrations. In the early stage, population clustering further expanded the demand for necessities such as clothing, food, shelter and transportation, resulting in the use of more fossil fuels and the creation of more construction/domestic dust [58]. However, studies have pointed out that when socio-economic development reaches a higher stage the effect of human agglomeration on the environment can be mitigated through technological innovation and industrial transformation, as well as other approaches [59].

Health Burden Attributable to PM2.5 and Factor Contribution Analysis
The total premature mortality caused by PM2.5 pollution was calculated using the PM2.5 concentration, population density, and baseline mortality in China from 2014 to 2018 ( Figure  5a). In general, PM2.5-related premature deaths decreased each year in China, with a de-

Health Burden Attributable to PM 2.5 and Factor Contribution Analysis
The total premature mortality caused by PM 2.5 pollution was calculated using the PM 2.5 concentration, population density, and baseline mortality in China from 2014 to 2018 ( Figure 5a). In general, PM 2.5 -related premature deaths decreased each year in China, with a decrease of 529,410 from 2014-2018. The continuous decline in PM 2.5 concentration was the main reason for the decrease in deaths [34]. Specifically, the premature deaths caused by IHD, stroke, COPD, LC, and LRI decreased by 145,840, 207,880, 104,770, 39,240 and 31,680, respectively. The relative percentage of premature deaths caused by PM 2.5 were calculated, as shown in Figure 5b. Among all PM 2.5 -related deaths in 2018, IHD and stroke were the two most important causes, accounting for 36.52 and 32.86%, respectively; followed by COPD, LC, and LRI, accounting for 14.41, 10.12 and 6.09%, respectively. Compared to 2014, the relative percentage of premature deaths by IHD, LC, and LRI increased by 5.82, 6.41 and 0.50%, while that caused by stroke and COPD decreased by 4.20 and 7.75%, respectively. This also demonstrated the improving PM 2.5 pollution at the present stage is more helpful to alleviate stroke and COPD.

Health Burden Attributable to PM2.5 and Factor Contribution Analysis
The total premature mortality caused by PM2.5 pollution was calculated using the PM2.5 concentration, population density, and baseline mortality in China from 2014 to 2018 ( Figure  5a). In general, PM 2.5-related premature deaths decreased each year in China, with a decrease of 529,410 from 2014-2018. The continuous decline in PM2.5 concentration was the main reason for the decrease in deaths [34]. Specifically, the premature deaths caused by IHD, stroke, COPD, LC, and LRI decreased by 145,840, 207,880, 104,770, 39,240 and 31,680, respectively. The relative percentage of premature deaths caused by PM2.5 were calculated, as shown in Figure 5b. Among all PM 2.5-related deaths in 2018, IHD and stroke were the two most important causes, accounting for 36.52 and 32.86%, respectively; followed by COPD, LC, and LRI, accounting for 14.41, 10.12 and 6.09%, respectively. Compared to 2014, the relative percentage of premature deaths by IHD, LC, and LRI increased by 5.82, 6.41 and 0.50%, while that caused by stroke and COPD decreased by 4.20 and 7.75%, respectively. This also demonstrated the improving PM 2.5 pollution at the present stage is more helpful to alleviate stroke and COPD.    Table 2 shows the PM 2.5 -related premature deaths in six major regions of China. The reduction in premature deaths in six major regions accounted for 52.82% of the total reduction in the whole country from 2014-2018. As the most polluted region in China, BTHS experienced the highest decrease in premature deaths (by 86,520). Similar to BTHS, the premature deaths in YRD, CYR, and TCC, along the Yangtze River, decreased by 68,360, 48,050 and 44,880, respectively. Compared to other regions, the decline in premature deaths of FWP and PRD was obviously lower, only 15,740 and 16,080. From 2014-2018, except for FWP, the premature deaths in BTHS, YRD, CYR, TCC, and PRD decreased by more than 20%, which also showed that PM 2.5 pollution improvement had achieved remarkable results within the five years. Meanwhile, the reduction of PM 2.5 -related premature deaths exhibited obvious differences among provinces ( Figure 6). Among the 31 provinces, Shandong exhibited the largest decrease in premature deaths due to the five diseases (47,040), while Tibet exhibited the smallest decrease (520). In the remaining provinces, those with the most significant decreases in deaths were Henan (40,250), Sichuan (37,630), Hebei (36,860), Hunan (30,160), Guangdong (28,480), and Hubei (25,360). In contrast, Hainan (620), Ningxia (1640), Qinghai (2440), Shanghai (4920), Tianjin (5590), and Xinjiang (6140)  Meanwhile, the reduction of PM 2.5 -related premature deaths exhibited obvious differences among provinces (Figure 6). Among the 31 provinces, Shandong exhibited the largest decrease in premature deaths due to the five diseases (47,040), while Tibet exhibited the smallest decrease (520). In the remaining provinces, those with the most significant decreases in deaths were Henan (40,250), Sichuan (37,630), Hebei (36,860), Hunan (30,160), Guangdong (28,480), and Hubei (25,36). In contrast, Hainan (620), Ningxia (1640), Qinghai (2440), Shanghai (4920), Tianjin (5590), and Xinjiang (6140)   The relative change of premature deaths attributable to PM 2.5 concentration, baseline mortality, and population density between 2014 and 2018 were quantified through the use of a sensitivity analysis (Figure 7). Overall, the PM 2.5 concentration reduction was the main reason for the decline in premature deaths; concretely, premature deaths declined by 393,460 (16.66%) due to PM 2.5 reduction from 2014-2018. The baseline mortality reduction contributed to the decline in premature deaths significantly, resulting in a 208,150 (8.81%) reduction in premature deaths. Conversely, changes in population density resulted in an increase The relative change of premature deaths attributable to PM 2.5 concentration, baseline mortality, and population density between 2014 and 2018 were quantified through the use of a sensitivity analysis (Figure 7). Overall, the PM 2.5 concentration reduction was the main reason for the decline in premature deaths; concretely, premature deaths declined by 393,460 (16.66%) due to PM 2.5 reduction from 2014-2018. The baseline mortality reduction contributed to the decline in premature deaths significantly, resulting in a 208,150 (8.81%) reduction in premature deaths. Conversely, changes in population density resulted in an increase by 49,240 (2.08%) in premature deaths. From 2014 to 2018, the total population of mainland China increased from 1367.82 million to 1395.38 million, an increase of only 2.01%, which greatly weakened the increase attributed to changes in the population density. At the provincial level, Shandong had the most significant decline in premature mortality due to PM 2.5 reduction, with a value of 36,170, while Sichuan, Hebei, Henan, Hunan, and Guangdong each exhibited decreases in premature mortality of more than 20,000. Meanwhile, the premature deaths caused by changes in population density also increased by more than 2500 in each of the aforementioned provinces. The impact of PM 2.5 reduction on premature deaths was greater than that of baseline mortality reduction in most provinces. However, the relative changes in premature deaths caused by baseline mortality reduction were greater than that caused by PM 2.5 reduction in Hainan and Xinjiang, with differences of 320 and 2190, respectively. each exhibited decreases in premature mortality of more than 20,000. Meanwhile, the premature deaths caused by changes in population density also increased by more than 2,500 in each of the aforementioned provinces. The impact of PM 2.5 reduction on premature deaths was greater than that of baseline mortality reduction in most provinces. However, the relative changes in premature deaths caused by baseline mortality reduction were greater than that caused by PM 2.5 reduction in Hainan and Xinjiang, with differences of 320 and 2,190, respectively.

Changes in Health Burden at the City Level
As shown in Figure 8a, the spatial distribution of premature mortality caused by PM 2.5 pollution across Mainland China from 2014-2018 exhibited the pattern of "high in the south and low in the north, high in the east and low in the west". The PM 2.5 -related premature mortality northwest of the Heihe-Tengchong Line (as known as the Hu Line and is an imaginary empirical line that divides the population distribution in China) was lower, with less than 10 deaths per 100 km 2 in most areas. To the southeast of the Hu Line, BTHS, YRD, FWP, CYR, TCC and PRD were the main regions affected by high health burden attributed to PM 2.5 pollution. In particular, the premature mortality in cities with large population density, or high concentration of PM 2.5, was more than 100 deaths per 100 km 2 . Over the considered five years (i.e., 2014-2018), the health burden attributable to PM 2.5 pollution decreased in most areas (Figure 8b). In addition, note that the areas with increased premature mortality were scattered throughout the whole country, but the reasons in the southeast and northwest of the Hu Line were different. The main reason in the southeast was that the relative changes of premature mortality due to population increases were greater than those caused by the decrease in PM 2.5 concentration [34], such as in parts of Jiangsu and Henan. The main reason in the northwest was that the PM 2.5 concentration had experienced an increasing trend, such as in parts of Inner Mongolia and Xinjiang.

Changes in Health Burden at the City Level
As shown in Figure 8a, the spatial distribution of premature mortality caused by PM 2.5 pollution across Mainland China from 2014-2018 exhibited the pattern of "high in the south and low in the north, high in the east and low in the west". The PM 2.5 -related premature mortality northwest of the Heihe-Tengchong Line (as known as the Hu Line and is an imaginary empirical line that divides the population distribution in China) was lower, with less than 10 deaths per 100 km 2 in most areas. To the southeast of the Hu Line, BTHS, YRD, FWP, CYR, TCC and PRD were the main regions affected by high health burden attributed to PM 2.5 pollution. In particular, the premature mortality in cities with large population density, or high concentration of PM 2.5, was more than 100 deaths per 100 km 2 . Over the considered five years (i.e., 2014-2018), the health burden attributable to PM 2.5 pollution decreased in most areas (Figure 8b). In addition, note that the areas with increased premature mortality were scattered throughout the whole country, but the reasons in the southeast and northwest of the Hu Line were different. The main reason in the southeast was that the relative changes of premature mortality due to population increases were greater than those caused by the decrease in PM 2.5 concentration [34], such as in parts of Jiangsu and Henan. The main reason in the northwest was that the PM 2.5 concentration had experienced an increasing trend, such as in parts of Inner Mongolia and Xinjiang.
The changes of PM 2.5 -related premature mortality in Chinese cities from 2014 to 2018 were calculated; the results are presented in Figure 9 and Table S3. Western China had lower PM 2.5 concentrations and population density, and the cities had smaller decreases in premature mortality. Cities with a high population density and high decline in PM 2.5 concentration, such as those in BTHS, had a larger decrease of premature mortality (see Figure 9a). In BTHS, cities which experienced a large decrease of premature deaths were Beijing (8110), Baoding (6050), Tianjin (5590), and Shijiazhuang (5300). The premature deaths in the megacities Chongqing, Chengdu, Wuhan, and Shanghai decreased by 14,680, 7000, 5430, and 4920, respectively. In Northeast China, the large cities with large decreases in premature deaths were Harbin (7670), Changchun (6080), and Shenyang (5120). The cities of Linyi, Tangshan, Weifang, and Handan also had high decreases in premature deaths (4980, 4680, 4550 and 4490, respectively). In terms of the percentage change, cities in northeast and southwest parts of China exhibited the largest declines in premature mortality (Figure 9b). Specifically, the premature deaths in 33.92% of the cities decreased by 0 to 20%, while that in 59.59% of the cities decreased by 20 to 30%. Moreover, the premature deaths in other cities decreased by more than 30%. The changes of PM 2.5 -related premature mortality in Chinese cities from 2014 were calculated; the results are presented in Figure 9 and Table S3. Western China ha PM 2.5 concentrations and population density, and the cities had smaller decreases in ature mortality. Cities with a high population density and high decline in PM 2.5 con tion, such as those in BTHS, had a larger decrease of premature mortality (see Figure  BTHS, cities which experienced a large decrease of premature deaths were Beijing Baoding (6050), Tianjin (5590), and Shijiazhuang (5300). The premature deaths in th acities Chongqing, Chengdu, Wuhan, and Shanghai decreased by 14,680, 7000, 54 4920, respectively. In Northeast China, the large cities with large decreases in pre deaths were Harbin (7670), Changchun (6080), and Shenyang (5120). The cities o Tangshan, Weifang, and Handan also had high decreases in premature deaths (498 4550 and 4490, respectively). In terms of the percentage change, cities in northe southwest parts of China exhibited the largest declines in premature mortality ( Fig  Specifically, the premature deaths in 33.92% of the cities decreased by 0 to 20%, wh in 59.59% of the cities decreased by 20 to 30%. Moreover, the premature deaths i cities decreased by more than 30%.   Compared to 2018, the relative contribution of premature deaths caused by IHD, stroke, and LC decreased by 3.56, 1.67 and 10.97%, respectively, while those caused by COPD and LRI increased by 12.70 and 18.39%, respectively. At the provincial level (Figure 11), Henan exhibited the largest decreases in premature deaths (64,500), while Shandong, Hebei, Anhui, Jiangsu, Sichuan, and Hubei each decreased in premature deaths by more than 20,000 deaths, accounting for 67.96% of the national mortality decline. In contrast, the premature deaths in Guizhou and Yunnan increased by 11.81 and 1.33%, respectively. These were mainly because the PM 2.5 concentrations in 2015 had almost approached the PM 2.5 control target in 2030, such that the relative changes in premature mortality from the decrease in PM 2.5 were smaller than that caused by population increases. Figure 10 presents the premature mortality caused by PM 2.5 pollution in China, under the circumstance that the target of PM 2.5 concentration control in 2030 could be achieved. Compared to 2018, the premature deaths could be decreased by 403,050, accounting for 21.99% of those in 2018. In terms of PM 2.5 -related premature deaths in 2030, IHD and stroke were still the two most important causes, accounting for 35.22 and 32.31%, respectively; followed by COPD, LC, and LRI, accounting for 16.24, 9.01, and 7.21%, respectively. Compared to 2018, the relative contribution of premature deaths caused by IHD, stroke, and LC decreased by 3.56, 1.67 and 10.97%, respectively, while those caused by COPD and LRI increased by 12.70 and 18.39%, respectively. At the provincial level (Figure 11), Henan exhibited the largest decreases in premature deaths (64,500), while Shandong, Hebei, Anhui, Jiangsu, Sichuan, and Hubei each decreased in premature deaths by more than 20,000 deaths, accounting for 67.96% of the national mortality decline. In contrast, the premature deaths in Guizhou and Yunnan increased by 11.81 and 1.33%, respectively. These were mainly because the PM 2.5 concentrations in 2015 had almost approached the PM 2.5 control target in 2030, such that the relative changes in premature mortality from the decrease in PM 2.5 were smaller than that caused by population increases.  The changes of PM 2.5 -related premature mortality in Chinese cities from 2018 to 2030 were calculated; the results are presented in Figure 12 and Table S3. Among them, 87.02% of cities exhibited decreases in premature mortality caused by PM 2.5 . BTHS, YRD, FWP, CYR, and TCC would be the main contributors to the reduction in premature deaths. From 2018-2030, the premature mortality caused by PM 2.5 in BTHS, YRD, FWP, CYR, and TCC exhibited decreases of 106,440 (32.45% reduction), 68,360 (27.62% reduction), 23,490 (28.88% reduction), 28,620 (21.63% reduction), and 26,690 (19.44% reduction), respectively. The number of premature deaths in the megacities Beijing, Chongqing, Tianjin, and Shanghai decreased by 8,660, 6,570, 5,160 and 3,400 thousand, respectively. The cities Heze, Nanyang, Zhoukou, Baoding, and Cangzhou also had high decreases in premature deaths of 8,700, 8,690, 7,620, 7,560 and 7,160, respectively. However, note that 12.98% of cities exhibited increases in premature mortality caused by PM 2.5 pollution . These cities should further strengthen their monitoring and governance with the aim of reaching the PM 2.5 control target. Specifically, PM 2.5 -related premature deaths in the PRD experienced a 9,804 increase (15.95% growth), with Guangzhou and Shenzhen experiencing increases of 7,000 and 5,200, respectively. In addition, some provincial capital cities also experienced increases The changes of PM 2.5 -related premature mortality in Chinese cities from 2018 to 2030 were calculated; the results are presented in Figure 12 and Table S3. Among them, 87.02% of cities exhibited decreases in premature mortality caused by PM 2.5 . BTHS, YRD, FWP, CYR, and TCC would be the main contributors to the reduction in premature deaths. From 2018-2030, the premature mortality caused by PM 2.5 in BTHS, YRD, FWP, CYR, and TCC exhibited decreases of 106,440 (32.45% reduction), 68,360 (27.62% reduction), 23,490 (28.88% reduction), 28,620 (21.63% reduction), and 26,690 (19.44% reduction), respectively. The number of premature deaths in the megacities Beijing, Chongqing, Tianjin, and Shanghai decreased by 8660, 6570, 5160 and 3400 thousand, respectively. The cities Heze, Nanyang, Zhoukou, Baoding, and Cangzhou also had high decreases in premature deaths of 8700, 8690, 7620, 7560 and 7160, respectively. However, note that 12.98% of cities exhibited increases in premature mortality caused by PM 2.5 pollution. These cities should further strengthen their monitoring and governance with the aim of reaching the PM 2.5 control target. Specifically, PM 2.5 -related premature deaths in the PRD experienced a 9804 increase (15.95% growth), with Guangzhou and Shenzhen experiencing increases of 7000 and 5200, respectively. In addition, some provincial capital cities also experienced increases in premature deaths. Guiyang, Kunming, Nanning, and Nanchang were the top four cities, with the most obvious increases of 5260, 5060, 4030 and 2830, respectively.

Classification of Potential Benefits and Policy Recommendations
A recent study has shown that age structure change may still increase premature mortality by about 30%, under the condition of reaching the PM 2.5 control target by 2030 [60]. As the population distribution data used in this study had no age structure, cities with less than 30% reduction in health burden were associated with uncertainty. In order to further explore the health benefits of achieving PM 2.5 control targets by 2030, taking PM 2.5 -related premature deaths in 2018 as the horizontal axis and the change of premature deaths from 2018 to 2030 as the longitudinal axis, all cities were divided into three types, as shown in Figure 13a. Type I represents areas where the reduction in health burden was more than 30%; Type II represents areas where the reduction in health burden was less than 30%; and Type III represents areas where the health burden exhibits increased compared that in 2018. For a more intuitive observation and analysis, the classification results of potential benefits are shown in Figure 13b.  Type I consisted of 148 cities, accounting for 43.66% of the total number of cities, and was the most widely distributed across China. Under the circumstance that the target of PM 2.5 concentration control in 2030 could be achieved, type I cities are expected to experience reductions in health burden by more than 30%. Even considering the increase of health burden caused by age structure change, the potential benefits of type I cities are still positive. Therefore, the PM 2.5 control target of these cities should not need to be changed and implementing the current emission reduction policies would be sufficient.
Type II accounted for 43.36% of the total number of cities and had a relatively concentrated distribution across China. Type II cities are expected to experience reductions in health burden by less than 30% under the circumstance that the target of PM 2.5 concentration control in 2030 is achieved. Considering the age structure change, it is difficult to determine whether the potential benefits of type II cities are positive or negative benefits. Thus, these cities should adapt to local conditions and adjust the PM 2.5 control target rationally, based on their own age structure characteristics.
Type III accounted for 12.98% of the total number of cities, having a scattered distribution across China and being relatively small in number: mainly provincial capitals, subprovincial cities, and other cities with good socio-economic development. Under the circumstance that the target of PM 2.5 concentration control in 2030 could be achieved, type III cities are expected to experience increases in health burden in varying degrees. In order to make the potential benefits of these cities positive in 2030, type III cities should set stricter PM 2.5 control targets and further strengthen monitoring and governance.

Discussion
In this study, the health burden attributed to PM 2.5 pollution was assessed. Compared with existing studies on the mortality burden of long-term exposure to PM 2.5 , one advantage of this study is the high spatial resolution of the data used. Hu et al. [61] pointed out that using coarse-scale data may affect the accuracy of assessment with respect to PM 2.5 pollution, and that using fine-scale data can provide more comprehensive information. However, previous studies mostly evaluated the premature mortality caused by PM 2.5 pollution using coarse-scale PM 2.5 and population data [31][32][33][34][35]. A recent study also noted the limitations of using county-level population data to estimate premature deaths [62]. To solve this problem, we used PM 2.5 and population data with finer resolution (1 km), in order to assess the health burden attributed to PM 2.5 pollution. This considered the spatial heterogeneity of PM 2.5 concentration and population distribution, and greatly improved the accuracy of estimation.
On the other hand, the latest exposure-response model (GEMM) was used to assess the relative risk (RR) of five diseases (including IHD, stroke, COPD, LC, and LRI) under different concentrations of PM 2.5 , as shown in Figure 14. Overall, the RR of PM 2.5 to different diseases increased with the PM 2.5 concentration. The curves of stroke, COPD, and LC were roughly straight lines, indicating that the RR increased evenly with the PM 2.5 concentration. As for the IHD curve, the RR increased rapidly when the concentration of PM 2.5 was lower than 20 µg/m 3 , then increased steadily in a straight line. For the LRI curve, the RR increased rapidly when the concentration of PM 2.5 was below 40 µg/m 3 then the growth rate slowed down. In this study, the estimated premature mortality caused by PM 2.5 pollution was higher than that in past studies [32][33][34]63]. The main reason may be that the exposure-response model used was different, whereas previous studies mostly used the integrated exposure-response (IER) model for estimation. The results of the study by Burnett et al. [41] showed that the global estimates of mortality attributable to ambient fine particulate air pollution by the GEMM method were 120% larger than that estimated by the IER model. Meanwhile, another study pointed out that the estimation results of the GEMM method may be more applicable to high pollution areas such as China [64]. concentration. However, PM 2.5 concentrations might increase in low-concentration regions due to economic development by 2030. Meanwhile, the PM 2.5 concentration in low-concentration regions may also be considerably lower than 35 µg/m 3 when emissions in high-concentration regions are further controlled. Third, the GBD database does not provide baseline mortality data for future years, and no data are available for the future baseline mortality of each disease (at the country or provincial level). According to the 'Healthy China 2030 Planning Outline', we assumed that the baseline mortality in China will drop by 30% in 2030 compared to 2015. However, substantial spatial disparities exist in the baseline mortality of each disease due to the differences in socio-economic contexts, environments, and health services. Fourth, we assumed that all chemical species in PM 2.5 have equivalent toxicity, whereas the effects of different chemical species on human health are different [65]. It is generally believed that black carbon and organic carbon in PM 2.5 are more harmful to humans [66], which also leads to uncertainty in the health burden assessment.

Conclusions
In summary, with the aid of high spatial resolution PM 2.5 concentration and population density data, we assessed the health burden attributed to PM 2.5 pollution in China. The results show that premature mortality caused by PM 2.5 pollution in China decreased by 529,410 from 2014-2018, including decreases of 145,840, 207,880, 104,770, 39,240, and 31,680 in the number of premature deaths caused by IHD, stroke, COPD, LC, and LRI, respectively. In addition, the reduction of premature mortality in six major regions (BTHS, YRD, FWP, CYR, TCC, and PRD) of China accounted for 52.82% of the total national reduction from 2014-2018. If the PM 2.5 control target can be achieved by 2030, PM 2.5 -related premature deaths will decrease by 403,050, accounting for 21.99% of those in 2018. Among them, 87.02% of cities exhibited decreases in premature deaths. According to the potential benefits in 2030, all cities were divided into three types, namely type I, type II, and type III. For type I cities, achieving the current PM 2.5 control target will be sufficient. Although the health burden in Type II cities experienced reductions, they still need to adjust the PM 2.5 control target according to local conditions, especially in cities with an aging age structure. Type III cities, with increases in health burden, should set stricter PM 2.5 control targets and further strengthen the associated monitoring and governance. We suggest that the relevant Chinese government departments pay attention to regional differences, as well as formulate differentiated air pollution control policies according to the specific circumstances. However, this study had some limitations. First, as baseline mortality rate and population density data with a 1 km resolution for different age ranges are difficult to obtain, all-age population density and baseline mortality were used to estimate premature mortality. Related studies have shown that the RR of PM 2.5 pollution on cardio-cerebro-vascular diseases in different age ranges may significantly differ [41], causing uncertainty in the health burden assessment. Second, for the low-concentration regions (PM 2.5 concentrations lower than or equal to 35 µg/m 3 ), we assumed that the air quality was maintained at the baseline concentration. However, PM 2.5 concentrations might increase in low-concentration regions due to economic development by 2030. Meanwhile, the PM 2.5 concentration in low-concentration regions may also be considerably lower than 35 µg/m 3 when emissions in high-concentration regions are further controlled. Third, the GBD database does not provide baseline mortality data for future years, and no data are available for the future baseline mortality of each disease (at the country or provincial level). According to the 'Healthy China 2030 Planning Outline', we assumed that the baseline mortality in China will drop by 30% in 2030 compared to 2015. However, substantial spatial disparities exist in the baseline mortality of each disease due to the differences in socio-economic contexts, environments, and health services. Fourth, we assumed that all chemical species in PM 2.5 have equivalent toxicity, whereas the effects of different chemical species on human health are different [65]. It is generally believed that black carbon and organic carbon in PM 2.5 are more harmful to humans [66], which also leads to uncertainty in the health burden assessment.

Conclusions
In summary, with the aid of high spatial resolution PM 2.5 concentration and population density data, we assessed the health burden attributed to PM 2.5 pollution in China. The results show that premature mortality caused by PM 2.5 pollution in China decreased by 529,410 from 2014-2018, including decreases of 145,840, 207,880, 104,770, 39,240, and 31,680 in the number of premature deaths caused by IHD, stroke, COPD, LC, and LRI, respectively. In addition, the reduction of premature mortality in six major regions (BTHS, YRD, FWP, CYR, TCC, and PRD) of China accounted for 52.82% of the total national reduction from 2014-2018. If the PM 2.5 control target can be achieved by 2030, PM 2.5 -related premature deaths will decrease by 403,050, accounting for 21.99% of those in 2018. Among them, 87.02% of cities exhibited decreases in premature deaths. According to the potential benefits in 2030, all cities were divided into three types, namely type I, type II, and type III. For type I cities, achieving the current PM 2.5 control target will be sufficient. Although the health burden in Type II cities experienced reductions, they still need to adjust the PM 2.5 control target according to local conditions, especially in cities with an aging age structure. Type III cities, with increases in health burden, should set stricter PM 2.5 control targets and further strengthen the associated monitoring and governance. We suggest that the relevant Chinese government departments pay attention to regional differences, as well as formulate differentiated air pollution control policies according to the specific circumstances.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/su13179690/s1. Figure S1: Spatial distribution of population density (a) and PM 2.5 concentration (b) across China in 2030. Table S1: List of the cities and division basis of six major regions in China. Figure S2: Geographical locations of Chinese provinces and the six major regions. Table S2