Air quality and health impacts from the updated industrial emission standards in China

The sectors in our research include all the industrial sectors that are discussed in the Handbook of Industrial Pollution Sources (Ministry of Environmental Protection 2011). Specifically, the power sector is also considered as an industrial sector in this study. Detailed information on each sector is given in table S1 which is available online at stacks.iop.org/ERL/14/124058/mmedia. We designed four emission standards based on specific emission targets which are stated in the emission standards (table 1). All the emission standards can be found at http://kjs.mee.gov.cn/hjbhbz/bzwb/dqhjbh/dqgdwrywrwpfbz/. The Old Emission Standard assumes that all industrial sectors follow the emission standards published before 2004. Under the Old Emission Standard, sector-specific emissions standards were established only for cement, coke, boiler and electricity sectors, and all the other sectors follow the emission limits established in 1996. Moreover, in the Old Emission Standard, there were no emission limits for NO_x emissions except for the nitric, cement and electricity sectors. In the updated emission standards that were established after 2004, there are two stages for the implementation of the new emission limits. The Temporal Emission Standard assumes that all industrial sectors follow the emission limits required in the first stage of the updated emission standards. In the Temporal Emission Standard, sector-specific emissions standards are established for most of the industrial sectors. Emission limits for smoke and dust are reduced by more than a factor in ferrous melting, non-ferrous melting and non-metal product sectors, and SO₂ emission standards are greatly improved in the ferrous melting and coke industry. More importantly, specific emission standards for NO_x are established. The New Emission Standard assumes that all industrial sectors follow the emission limits in the second stage, which are more stringent than those in the first stage. Compared with the Temporal Emission Standard, emission limits for SO₂ are reduced by half or more in most sectors, while smoke, dust and NO_x emission limits are reduced by about 40%. The Special Emission Standard assumes that all industrial sectors across China follow the lowest emission limits; this strategy was initially proposed for the North China Plain (NCP; Beijing, Tianjin, Hebei, Shanxi, Henan and Shandong). With the Special Emission Standard, nearly all the emission limits under the New Emission Standard are to be halved or more. In particular, the smoke emission limits for non-ferrous melting under the Special Emission Standard is one-eighth of that under the New Emission Standard. The emission limits for the implementation of the emission standards are shown in tables S2–S5. In this study, we calculated the quantities of SO₂, NO_x and primary PM_2.5 emissions when the emission standards are completely implemented by industrial sectors and assumed that other air pollutants (including VOCs) remain unchanged.

According to the classification of the emission standards, we designed three scenarios. The Old → Temporal scenario assumes that the emission standards were updated from the Old to the Temporal standards. The Temporal → New scenario assumes that the standards are improved from the Temporal to the New standards. The New → Special scenario assumes that the New Emission Standard was changed to reflect the Special Emission Standard.

We used equations (1)–(4) to calculate provincial SO₂, NO_x and primary PM_2.5 emissions in 2016 under different emission limits in the four emission standards. We did not calculate VOC emissions because only a few emission standards focused on VOCs, and VOC emissions from the industrial sector have continued to rise since 2010.

$$\begin{eqnarray}&&E{F}_{i,k}^{q}=\,\min \left\{{V}_{i}\times {C}_{i,k}^{q},E{F}_{i,k}^{unabated}\right\}\end{eqnarray} \tag{ 1 }$$

$$\begin{eqnarray}&&{E}_{i,k,p}^{q}={A}_{i,p}\times E{F}_{i,k}^{q}\end{eqnarray} \tag{ 2 }$$

$$\begin{eqnarray}&&{E}_{k,p}^{q}=\displaystyle \sum _{i}{A}_{i,p}\times E{F}_{i,k}^{q}\end{eqnarray} \tag{ 3 }$$

$$\begin{eqnarray}&&Standar{d}_{k,p,r}^{q}=MEI{C}_{k,p,r}^{industry}\times \displaystyle \frac{{E}_{k,p}^{q}}{MEI{C}_{k,p}^{industry}}\end{eqnarray} \tag{ 4 }$$

where i represents the industrial process (or fuel consumption of industrial boilers and power plants), k is the type of air pollutant, p is the province in China and q is the emission standard scenario. $E{F}_{i,k}^{q}$ represents the abated emission factor for product i following the emission limit of air pollutant k in scenario $q.\,{V}_{i}$ (m³ t^–1) is the volume of smoke emitted from industrial process i. ${C}_{i,k}^{q}$ is the concentration limit (kg m^–3) of air pollutant k in scenario q for the smoke emitted from process $i.E{F}_{i,k}^{unabated}$ is the unabated emission factor for air pollutant k of process i. ${A}_{i,p}$ is the activity rate of industrial process i in province p. ${E}_{i,k,p}^{q}$ is the total emissions of air pollutant k from process i in province p. ${E}_{k,p}^{q}$ represents the total emissions of air pollutant k in province p. $MEI{C}_{k,p,r}^{industry}$ represents the industrial emissions of air pollutant k in MEIC grid r within the boundary of province p. $MEI{C}_{k,p}^{industry}$ represents the industrial emissions of air pollutant k in province p. $Standar{d}_{k,p,r}^{q}$ represents the emissions of air pollutant k in grid r within the boundary of province p in scenario q. To avoid the influence of activity rates on emissions, we assumed that the production and fuel consumption in the industrial sectors remained the same as those in the year 2016.

The emission standards only regulate the emission limits for total suspended particulates (TSPs); therefore, equations (5)–(8) were developed to estimate the PM_2.5, black carbon (BC) and organic carbon (OC) emissions.

$$\begin{eqnarray}&&{R}_{i,TSP}^{q}=1-E{F}_{i,TSP}^{q}/E{F}_{i,TSP}^{unbated}\end{eqnarray} \tag{ 5 }$$

$$\begin{eqnarray}&&O{P}_{i,j,PM}^{q}=\left\{\displaystyle \frac{{R}_{i,TSP}^{q}}{{\eta }_{i,1,TSP}},\cdots ,\displaystyle \frac{{R}_{i,TSP}^{q}}{{\eta }_{i,m,TSP}},\cdots \right\}\,,\displaystyle \frac{{R}_{i,m,TSP}}{{\eta }_{i,m,TSP}}\lt 1\end{eqnarray} \tag{ 6 }$$

$$\begin{eqnarray}&&\begin{array}{l}E{F}_{i,P{M}_{2.5}}^{q}=E{F}_{i,TSP}^{q}\times {f}_{i,P{M}_{2.5}}\times \left[O{P}_{i,j,PM}^{q}\right.\\ \,\,\,\,\left.\times \,(1-{\eta }_{i,j,PM2.5})+(1-O{P}_{i,j,PM}^{q})\right]\end{array}\end{eqnarray} \tag{ 7 }$$

$$\begin{eqnarray}&&E{F}_{i,BC/OC}^{q}=E{F}_{i,P{M}_{2.5}}^{q}\times {f}_{i,BC/OC}\end{eqnarray} \tag{ 8 }$$

where m is the available removal devices for PM, j is the type of removal device used to remove PM in scenario q, ${R}_{i,TSP}^{q}$ is the rate of removal of TSPs during industrial process i under scenario q, ${\eta }_{i,m,TSP}$ represents the rate of removal of TSPs for device m during process i, $O{P}_{i,j,PM}^{q}$ is the rate of operation of PM removal device j during industrial process i in scenario q and approaches ${R}_{i,TSP}^{q},\,{\eta }_{i,j,PM2.5}$ is the rate of removal of PM_2.5 for device j during process i, ${f}_{i,P{M}_{2.5}}$ represents the mass ratio of PM_2.5 to TSPs during the industrial process of product i, and ${f}_{i,BC/OC}$ is the BC/OC mass ratio in PM_2.5.

The concentration limits for SO₂, NO_x and TSP were based on the emission standards published from 1996 to 2016 (http://kjs.mee.gov.cn/hjbhbz/bzwb/dqhjbh/dqgdwrywrwpfbz/). The unabated emission factors and the volume of smoke were adopted from the Handbook of Industrial Pollution Sources (Ministry of Environmental Protection 2011). We collected provincial activity rates for industrial processes and fuel consumption in 2016 from the China Industrial Yearbook and China Energy Yearbook, respectively. The rates of removal of TSP and PM_2.5 as well as the mass ratios of PM_2.5, BC and OC were compiled from GAINS (Lei et al 2011, Klimont et al 2017). The rate of operation was based on the value used in the Asia-Pacific Integrated Model (Hibino et al 2003), and the modification of the emission inventory followed the method proposed by Peng et al (2017) and Qin et al (2017). The MEIC inventory (http://meicmodel.org) for 2016 was used to determine the spatial allocation of air pollutants. A validation and uncertainty analysis of the estimation is presented in table S7.

WRF-Chem is a meteorological model coupled to a chemical mechanism that enables the simulation of atmospheric phenomena across scales ranging from meters to thousands of kilometers (Grell et al 2005, Fast et al 2006). In this study, we conducted simulations covering East Asia at a 30 km × 30 km horizontal resolution and with 36 vertical layers. The Carbon Bond Mechanism version Z (CBMZ) (Zaveri and Peters 1999) and Model for Simulating Aerosol Interactions and Chemistry (MOSAIC) (Zaveri et al 2008) were used to simulate the gas phase and aerosol chemical mechanisms, respectively. The initial conditions for meteorological variables were obtained from the Global Forecast System (GFS) of the National Center for Environmental Prediction (NCEP) analyses that are updated every 6 h. The chemical boundary and initial conditions were based on the archived simulations of the Model for Ozone and Related chemical Tracers (MOZART; www.acom.ucar.edu/wrf-chem/mozart.shtml) model. To improve the simulation of secondary inorganic aerosols, a new heterogeneous reaction was added to the CBMZ–MOSAIC chemical mechanism following the approach of Chen et al (2016) and Li et al (2017). The emission inventory developed for the four emission scenarios was based on anthropogenic emissions. The rest of the information describing the WRF-Chem model (e.g. the photolysis scheme) is presented in table S8.

Due to the high computational demands of these simulations, we simulated January, April, July, and October in 2016 as representative months for each season, with 1 week at the end of December, March, June and September used for model spin-up. In each season, the meteorological condition is different for the formation of PM_2.5 and O₃. Additionally, the contribution of industrial emissions to total emissions also varies with season. Therefore, we used the representative month of each season to reflect the seasonal change of meteorological conditions and emissions. The differences in air quality impacts due to different emission scenarios were evaluated by computing the changes in air pollutants.

We compared the monthly average of PM_2.5 in our results with the 2016 PM_2.5 observations (measurement stations are shown in figure S1, obtained from China National Environmental Monitoring Centre, http://106.37.208.233:20035/), and also compared the monthly average of the simulated daily maximum 1 h O₃ with the 2016 O₃ observations in figure S2. The normalized mean bias (NMB) and normalized mean error (NME) were used to evaluate the performance of simulation, which summarizes the discrepancy between the observation and the simulation. Figure S2 shows that NMB ranged between 0.1 and 0.4 for PM_2.5 and 0.3 and 0.6 for O₃, while NME was around 0.5 for PM_2.5 and 0.7 for O₃, which reasonably reproduce the observed PM_2.5 and O₃ concentrations fields compared with other research (Gao et al 2015, Chen et al 2016). Although there are underestimations at certain areas in some months, our model reasonably captured the general trend of observation. We also ran present-day simulations using 2016 MEIC data, and compared these with our simulations using the Temporal Emission Standard and New Emission Standard. We found that the temporal and spatial characteristics of PM_2.5 and O₃ in our results shared a pattern with those of the present-day simulations (figures S5 and S6).

We estimated the cause-specific premature deaths in each WRF-Chem grid using the relevant attributable fractions (Ezzati et al 2003), as recommended by GBD (GBD 2016 Risk Factors Collaborators 2017) and Environment Protection Agency (Cohan et al 2007):

$$\begin{eqnarray}&&{\rm{\Delta }}Mor{t}_{r}=Po{p}_{r}\times AF\times {y}_{0}\end{eqnarray} \tag{ 9 }$$

$$\begin{eqnarray}&&AF=1-1/RR\end{eqnarray} \tag{ 10 }$$

where ${\rm{\Delta }}Mor{t}_{r}$ is the number of premature deaths due to PM_2.5 pollution or O₃ pollution, ${y}_{0}$ is the baseline mortality rate for the exposed population, AF is the attributable fraction, RR represents the relative risk of death due to a change in the air pollutant concentration and $Po{p}_{r}$ is the exposed population in grid r (adults ≥ 25 years old).

We calculated the cause-specific premature mortality values attributed to long-term PM_2.5 exposure following the IER function developed by Burnett et al (2014):

$$\begin{eqnarray}&&RR\left(z\right)=\left\{\begin{array}{c}1,\,if\,z\lt {z}_{cf}\\ 1+\alpha \left\{1-\exp \left[\gamma {(z-{z}_{cf})}^{\delta }\right]\right\},\,if\,z\geqslant {z}_{cf}\end{array}\right.\,\end{eqnarray} \tag{ 11 }$$

where z is the annual average PM_2.5 concentration and ${z}_{cf}$ is the counterfactual concentration below which we assume that there is no health risk attributed to PM_2.5. For PM_2.5, we estimated the age-specific RR for mortality due to ischemic heart disease (IHD), stroke, chronic obstructive pulmonary disease (COPD) and lung cancer (LC). The values of parameters α, γ, δ and ${z}_{cf}$ were given by the Global Burden of Disease Collaborative Network (2013).

The premature deaths due to long-term exposure to O₃ were estimated following the method given by Silva et al (2016), where the daily maximum 1 h O₃ concentrations is used in the health impact function:

$$\begin{eqnarray}&&RR\left(X\right)=\exp \left(\beta {\rm{\Delta }}X\right)\end{eqnarray} \tag{ 12 }$$

$$\begin{eqnarray}&&{\rm{\Delta }}X=\left\{\begin{array}{c}\,\,0,\,if\,X\lt LCC\,\\ X-LCC,if\,X\geqslant LCC\end{array}\right.\end{eqnarray} \tag{ 13 }$$

where X is the average daily maximum 1 h O₃ concentration, ${\rm{\Delta }}X$ is the change in O₃ pollution and LCC (low-concentration cutoff) represents the threshold exposure concentration below which we assume there is no effect of O₃ exposure on mortality. β is the concentration–response factor, which reflects the percentage change in mortality due to a unitary change in O₃. According to Jerrett et al (2009), RR = 1.040 for a 10 ppb increase in the O₃ concentration, and the average daily maximum 1 h O₃ concentration was calculated in the consecutive 6 month period with the highest average. The value of LCC, derived from Malley et al (2017), is 33.3 ppb. In our research, the 2 months with the highest averages among January, April, July and October were used to estimate the average daily maximum 1 h O₃ concentration. Additionally, we estimated premature deaths due to all chronic respiratory diseases.

We used the 30'' × 30'' global population distribution of the Center for International Earth Science Information Network (CIESIN), Columbia University (2016) from 2015 and regridded the population data to the WRF-Chem resolution (30 km × 30 km). The distribution of age groups was obtained from the 2010 Population Census of China. The baseline mortality rates for IHD, stroke, COPD, LC and all chronic respiratory diseases were collected from the Global Burden of Disease 2016 report (GBD 2016; http://ghdx.healthdata.org/gbd-results-tool).

To analyze the uncertainties associated with the RRs, baseline mortality rates and function parameters, we conducted 1000 Monte Carlo simulations. We assumed that the RRs of O₃ follow a normal distribution with 95% confidential intervals (CIs), as reported by Jerrett et al (2009) and the GBD. For PM_2.5, the parameter values used for the 1000 simulations were reported by the Global Burden of Disease Collaborative Network (2013). For baseline mortality rates, we assumed lognormal distributions with 95% CIs, as adopted from GBD 2016. The 95% CIs of the provincial mortality rates are shown in tables S9 and S10. The uncertainty ranges of the mortality shown in this study all refer to the uncertainty caused by the scientific uncertainty of the health function, and do not incorporate uncertainty from emissions or the air quality modeling.

Figure 1 shows the reductions in emissions under different scenarios. In total, SO₂, NO_x and PM_2.5 emissions in the Old scenario would be reduced by 83.6%, 65.5% and 83.8% with the Special Emission Standards, respectively (table S7). The Temporal → New scenario contributed to the largest reduction in SO₂ (7.6 Tg). Of the 7.6 Tg SO₂ reduction, 45.8% was associated with the power sector, followed by the contributions of industrial boilers (23.5%) and the ferrous metal sector (16.0%). The Old → Temporal and New → Special scenarios both reduced SO₂ emissions by approximately 4.5 Tg. The ferrous metal sector, industrial boilers and non-metal sectors contributed nearly 80% of the 4.3 Tg SO₂ emission reduction in the Old → Temporal scenario, and the power sector accounted for over half of the emission reduction in the New → Special scenario.

Zoom In Zoom Out Reset image size

For NO_x emissions, the power sector accounted for 96.0% of the 5.9 Tg reduction in the Temporal → New scenario, and the power sector dominated the reduction in NO_x emissions in the Old → Temporal scenario. In contrast, for the New → Special scenario, the reduction in NO_x emissions in the power sector contributed little to this emission reduction, and the non-metal product sector accounted for over half of this reduction.

Shifting from the Old Standard to the Temporal Standard (Old → Temporal scenario) resulted in the largest reduction in most primary PM_2.5 emissions (1.4 Tg), followed by those for the Temporal → New (1.0 Tg) and New → Spec (0.5 Tg) shifts. The ferrous and non-ferrous metal sectors contributed to 70% of the 1.4 Tg primary PM_2.5 emission reduction in the Old → Temporal scenario. In the Temporal → New scenario, the improved emission limits for the industrial boiler, non-metal product and ferrous metal sectors accounted for nearly 90% of the reduction. For the New → Special scenario, the power sector accounted for approximately half of the reduced emissions.

Assuming that the emission standards are updated from the Old to the Special cases, figure 2 shows that the grid-level monthly average PM_2.5 would decrease by 20 μg m⁻³ in eastern China and by less than 5 μg m⁻³ in western China (the geographical distribution of provinces is shown in figure S3). The largest percentage decrease in PM_2.5 occurred in July, as more than 35% of the PM_2.5 was reduced in eastern China, with peaks of more than 50% in northeastern (Inner Mongolia, Heilongjiang and Jilin; the distribution of provinces can be seen in figure S3) and southern (Jiangsu, Zhejiang, Jiangxi, Fujian and Guangdong) China. The smallest reduction in PM_2.5 in eastern China occurred in January, when the monthly average PM_2.5 was reduced by 10%–20% in northern China and by 20%–35% in southern China. Specifically, the NCP region experienced the most serious PM_2.5 pollution (figure S5), and the monthly average PM_2.5 was reduced by over 60–100 μg m⁻³ (35%–50%) in July, 40–80 μg m⁻³ (25%–35%) in April, 20–60 μg m⁻³ (20%–35%) in October and 10–40 μg m⁻³ (10%–15%) in January.

Zoom In Zoom Out Reset image size

The improvements in the emission standards also led to an increase in O₃ pollution. Except in July, the monthly average of daily maximum 1 h O₃ concentrations generally increased by over 20 μg m⁻³ in northern China. In January, most areas in eastern China experienced increases of over 50% in O₃. In April and October, the O₃ concentration generally increased by 25%–50% in northern China and decreased by 0%–15% in southern China. In contrast, the O₃ concentrations decreased across China in July. Specifically, O₃ decreased by 0%–10% in mid China, and other areas commonly observed decreases of 10%–20%.

Figures S7 and S8 show the changes in PM_2.5 and O₃ in each of the emission reduction scenarios. The Temporal → New scenario would result in the largest reduction in PM_2.5, followed by the Old → Temporal and New → Special scenarios. In the Temporal → New scenario, a reduction of over 40 μg m⁻³ PM_2.5 was achieved in the NCP region in July, with reductions of 20–30 μg m⁻³ in April, 10–20 μg m⁻³ in October and 2–10 μg m⁻³ in January. In the Old → Temp scenario, the reduction in PM_2.5 was approximately 10 μg m⁻³ less than that in the Temp → New scenario, except in the NCP region. In January, PM_2.5 in the NCP decreased by 10–20 μg m⁻³ in January in the Old → Temp scenario and decreased by 2–10 μg m⁻³ in the Temporal → New scenario.

The largest increase and decrease in O₃ both occurred in the Temporal → New scenario. In the Temporal → New scenario, the increase in surface O₃ exceeded 40 μg m⁻³ in the NCP in October and YRD in January and October. The largest decrease in O₃ occurred in July in the Temporal → New scenario, as O₃ decreased by 10–20 μg m⁻³ in the NCP and 20–30 μg m⁻³ in the YRD and southern China. In the Old → Temporal scenario, the increase in O₃ occurred over a more widespread area. In July in the Old → Temporal scenario, O₃ increased by 2–20 μg m⁻³ in Jiangsu, Anhui, Shandong, Henan, Tianjin and Hebei, and the O₃ concentrations in these provinces generally decreased in the Temporal → New scenario. Furthermore, in April, the increase in O₃ covered larger areas of the NCP and YRD in the Old → Temporal scenario than that in the Temporal → New scenario. Additionally, in January in the Old → Temporal scenario, O₃ increased in the southern provinces (e.g. Fujian, Sichuan and Yunnan), which was not the case in the Temporal → New scenario. For the New → Special scenario, the increase in O₃ was less than or equal to 10 μg m⁻³ in January and April and less than or equal to 20 μg m⁻³ in October. In particular, in April, Anhui and Jiangsu provinces exhibited decreasing O₃ concentrations in the New → Special scenario, and O₃ increased in these provinces in the Old → Temporal and Temporal → New scenarios.

Figure 3(a) shows the change in the number of annual premature deaths due to PM_2.5 and O₃ in the 30 provinces in China (except Tibet and Taiwan) after shifting from the Old Emission Standard to the Special Emission Standard. For the PM_2.5-related premature deaths, eastern, central and southern China generally experienced higher reductions in mortality than other geographical regions. The largest reduction in PM_2.5-related premature deaths occurred in Guangdong [15.2 × 10³, 95% CI (12.5–17.7) × 10³], followed by Jiangsu, Shandong, Hebei and Henan (over 10⁴ in each province). Considering the population, the highest decrease in premature deaths per capita occurred in Shanghai (143 per million population, 95% CI 125–162), followed by Guangdong and Jiangsu (figure S9(a)). In total, by improving the emission limits from the Old Standard to the Special Standard, the 1.4 × 10⁶ [95% CI (1.1–1.6) × 10⁶] PM_2.5-related premature deaths under the Old Standard could be reduced by 0.15 × 10⁶ [95% CI (0.13–0.17) × 10⁶]. Due to the nonlinearity of the health assessment model, the health burden was only reduced by 10.9%, although PM_2.5 concentrations were reduced by more than 25% in most areas.

Zoom In Zoom Out Reset image size

The changes in premature deaths related to O₃ varied by geographical region in China. VOC-limited regimes were found in northern and eastern provinces, while most southern provinces were NO_x-limited (Jin and Holloway 2015). This geographical difference means that reducing NO_x emissions would increase O₃ concentrations in northern and eastern provinces and decrease O₃ concentrations in southern provinces. As a result, the O₃-related premature deaths in northern China and eastern China generally increased, with a maximum change in Hebei Province (1100, 95% CI 500–1500). However, after considering the effect of population, Beijing and Tianjin experienced the largest increase in O₃-related premature deaths per million population. The increase in O₃ partly reduced the health benefits due to the decrease in PM_2.5 in several provinces. The increase in premature deaths due to O₃ offset approximately 20% of the decrease in PM_2.5-related premature deaths in Beijing and Tianjin and 10% of those deaths in Hebei and Shandong. In contrast, deaths due to O₃ decreased in southern China and southwestern China. The largest reduction in O₃-related premature deaths (1000, 95% CI 400–1400) occurred in Sichuan Province. Following Sichuan Province were Jiangxi and Hunan, which experienced reductions of 990 (95% CI 400–1300) and 930 (95% CI 380–1300) O₃-related premature deaths. As for the reduction in premature death per capita, the largest decrease occurred in Jiangxi and Guangxi provinces. Based on the total change in premature deaths in all provinces, the premature deaths attributed to O₃ decreased by 1800 (95% CI 500–3000) after shifting from the Old Emission Standard to the Special Emission Standard.

Figures S12 and S13 show that the changes in premature deaths due to O₃ are strongly related to the reduction scenario. Specifically, O₃-related premature deaths increased by 3200 (95% CI 1400–4200) in the Old → Temporal scenario, and the largest increases occurred in Shandong (820), Jiangsu (790), Henan (540) and Hebei (530). Conversely, O₃-related premature deaths decreased by 2400 (95% CI 920–3500) and 2700 (95% CI 1100–3700) in the Temporal → New and New → Special scenarios, respectively. The changes in PM_2.5-related premature deaths are consistent with the changes in PM_2.5 concentrations (figure S10). Premature deaths attributed to PM_2.5 decreased the most in the Temporal → New scenario [69.8 × 10³, 95% CI (60.7–78.3) × 10³], followed by the changes in the Old → Temporal [45.9 × 10³, 95% CI (40.4–51.2) × 10³] and the New → Special scenarios [29.3 × 10³, 95% CI (29.3–38.3) × 10³].

Figure 3(b) shows the total change in premature deaths (by combining the change in PM_2.5-related and O₃-related premature deaths) and the initial premature deaths (by combining PM_2.5-related and O₃-related premature deaths) in the Old Emission Standard. There is a strong linearity between the total reduction in premature deaths and the initial premature deaths. Larger health benefits commonly occurred in provinces with a larger number of initial premature deaths. Figure S9(b) reveals there is less linearity between the total reduction of premature deaths per capita and the initial premature deaths per capita. Northern and central provinces generally experienced higher initial premature deaths per capita, but larger reductions per capita occurred in southern and eastern China.