The future pollutant emission outcome is a cumulative result of a range of variables including socio-economic development, technological change, efficiency improvements and policies directed at pollution control as well as alternative concerns including climate change, energy access and agricultural production (Rao et al., 2017). The shared socioeconomic pathways (SSPs) include five distinctly different pathways about how the future might unfold in terms of major socioeconomic, demographic, technological, lifestyle, policy and institutional trends (O'Neill et al., 2014; van Vuuren et al., 2017). We select three scenarios to represent low-, medium- and high-emission cases: SSP1 – sustainability, SSP2 – middle of the road and SSP3 – regional rivalry. An assumption about the degree of air pollution control (strong, medium or weak) is included on top of the baseline pathway. Weak air pollution controls occur in SSP3, with medium controls in SSP2 and strong controls in SSP1 (Turnock et al., 2020). However, SSP scenarios do not include explicit climate policies (O'Neill et al., 2020). Instead, the RCPs generate climate projections targeting at a range of climate forcing levels, such as 1.9 W m2 (1.5 ∘C warming), 4.5 W m2 (3 ∘C) and 7.0 W m2 (4 ∘C) in 2100. Thus, the combination of SSP–RCP scenario framework depicts societal and climate futures in parallel and explores plausible futures of human activities, the changing climate and emissions (O'Neill et al., 2020).

Here, we use anthropogenic aerosol emissions projected under SSP–RCP scenarios for foreign countries. We select three scenarios to represent low, middle and high air pollutant emissions in plausible futures: SSP119 (a sustainable development pathway targeting at a rise of the global mean surface temperature below 1.5 ∘C from the pre-industrial levels by the end of the century), SSP245 (a business-as-usual development pathway with 3 ∘C warming) and SSP370 (a regional-rivalry development pathway with 4 ∘C warming). Table 1 summarizes the scenario settings in more detail.

The original SSP–RCP anthropogenic emission future projection starts from 2015. Here, we update the start year to 2019, which is the latest year that anthropogenic emissions from the Community Emissions Data System (CEDS; Hoesly et al., 2018) are available. Because CEDS historical emissions and SSP–RCP future emissions were developed upon different assumptions and methods, it is important to harmonize the two datasets to ensure smooth transitions between the two sets of emission trajectories (Gidden et al., 2019). The update of the harmonization year from 2015 to 2019 in this study could better represent the actual emission trajectory between 2015 and 2019. We also use the most recently developed CEDS version 2 (https://data.pnnl.gov/dataset/CEDS-4-21-21, last access: 6 September 2023; Smith et al., 2023​​​​​​​) to harmonize the future SSP–RCP emission projections as the new emissions can better represent the historical trend of pollutant emissions (https://github.com/JGCRI/CEDS/blob/master/documentation/Version_comparison_figures_v_2021_04_21_vs_v_2016_07_16(CMIP6).pdf, last access: 6 September 2023). The final future SSP–RCP emissions for foreign countries used in this study are presented in Fig. S1a in the Supplement.

Future scenarios of anthropogenic pollutant emissions for China were developed by Tong et al. (2020) and Cheng et al. (2021b). We adopt these emissions instead of SSP–RCP emissions because Tong et al. (2020) and Cheng et al. (2021b) developed the emissions specifically to meet China's recent climate goals (i.e., 2030 carbon peak and 2060 carbon neutrality) and used parameters (e.g., emission factors, energy use) to reflect the most up-to-date pollutant control policies (i.e., the Clean Air Action since 2013) and technologies in China. Their emissions have been suggested to better capture China's PM2.5 concentration decline during 2015–2019 than those driven by SSP–RCP emissions (Cheng et al., 2021a). We select two plausible scenarios: the current-policy scenario and the carbon-neutral scenario. The current-policy scenario seeks to achieve China's Nationally Determined Contribution (NDC) pledges and the national PM2.5 air quality goal (i.e., 35 µg m-3) by 2030, elucidating China's future air pollution mitigation pathway towards all the released and determined upcoming clean air policies since 2015. The carbon-neutral scenario pursues China's carbon-neutral commitment and the WHO's old PM2.5 guideline (10 µg m-3) by 2060. It implements the best-available end-of-pipe technologies and more stringent pollution control policies than the current-policy scenario. Future anthropogenic pollutant emissions for China under these scenarios are developed by firstly simulating China's future energy and socioeconomic evolution using the Global Change Assessment Model (GCAM-China) and then translating it into pollutant emissions by the Dynamic Projection model for Emissions in China (DPEC; Tong et al., 2020). More details are summarized in Table 1. The actual future air pollutant emissions for China used in this study are presented in Fig. S1b.

We use the GEOS-Chem model to simulate PM2.5 concentrations in China and other Asian countries under each emission scenario and year. Detailed descriptions of the model configurations can be found in Xu et al. (2023). Briefly, we use the flex-grid capability of the GEOS-Chem classic model v13.2.1 to simulate PM2.5 concentrations over Asia and its adjacent regions (11∘ S–60∘ N, 30–150∘ E; covering the entire Asia, eastern Africa and eastern Europe regions). We use MERRA-2-assimilated meteorological data provided by the Global Modeling and Assimilation Office (GMAO) at the NASA Goddard Space Flight Center to drive the model. Our simulation is conducted at a horizontal resolution of 0.5∘ latitude × 0.625∘ longitude with 47 vertical levels between the surface and ∼ 0.01 hPa. Chemical boundary conditions are taken from corresponding global simulations under each emission scenario in Table 2 at a resolution of 2∘ latitude × 2.5∘ longitude. The regeneration of boundary conditions under each emission scenario could ensure the inclusion of pollution transported from countries both within and outside the flex-grid domain as transboundary pollution to China. We spin up every simulation for 1 month to remove the effects of initial conditions.

Anthropogenic emissions for the base year (2015) for China are taken from the Multi-resolution Emission Inventory (MEIC) for 2015 (Zheng et al., 2018b), and for the rest of the world they are taken from the Community Emissions Data System (CEDS) version 2 for 2015 (https://data.pnnl.gov/dataset/CEDS-4-21-21). For future simulations, anthropogenic emissions for China and foreign countries for each scenario are described above and are specified in Table 2. Other emissions are default in GEOS-Chem following Xu et al. (2023) and are fixed in all present and future scenarios.

We conduct simulations for January, April, July and October and treat the mean of the four months as an annual mean. Our meteorological fields are fixed to 2015 for all scenarios and years to exclude the influence of climate on the results. More details of our model configurations can be found in Table 2. We conduct two types of simulations: (1) baseline simulations (simulations with “Base_” prefix in Table 2) that include complete anthropogenic emissions for both China and foreign countries and (2) sensitivity simulations (simulations with “China_” prefix in Table 2) that exclude anthropogenic emissions for foreign countries from the baseline simulation. Baseline simulations calculate PM2.5 concentrations in China that are driven by both Chinese and foreign emissions, while sensitivity simulations calculate PM2.5 concentrations in China that are driven merely by China's domestic emissions. The impacts of transboundary pollution on China's PM2.5 are calculated as the difference in China's PM2.5 between a baseline simulation and a sensitivity simulation in a specified year and under a specified Chinese emission scenario.

We correct our simulated PM2.5 concentrations under each scenario, based on a large set of ground-based observations of PM2.5, because our simulated PM2.5 concentrations are biased high by roughly 15.7 % (Fig. S2). Descriptions and data screening method of our ground-based observations can be found in Xu et al. (2023). Further evaluations of PM2.5 composition concentrations in China and PM2.5 concentrations in other Asian countries with ground-based observations are presented in Xu et al. (2023). We correct simulated PM2.5 concentrations in anthropogenic concentration-dominated grid cells (where anthropogenic emissions exceed natural emissions) by observations located in anthropogenic concentration-dominated grid cells. Similarly, we correct simulated concentrations in natural concentration-dominated grid cells (where natural emissions exceed anthropogenic emissions) by observations located in natural concentration-dominated grid cells. Natural and anthropogenic concentration-dominated grid cells in the model are shown in Fig. S3. To isolate anthropogenic concentration-dominated grid cells, we conduct a sensitivity that excludes natural emissions in China and in foreign countries. PM2.5 concentrations from this sensitivity simulation are referred to as natural PM2.5 concentrations. The difference between concentrations from the Base_2015 simulation (with complete natural and anthropogenic emissions) in Table 2 and the natural concentrations is the anthropogenic emission-contributed concentrations (anthropogenic concentrations). We scale anthropogenic concentrations by the average ratio of observed concentrations and simulated concentrations at each observation site that falls into an anthropogenically dominated grid cell. Concentrations before and after the correction for anthropogenic and all grid cells (including natural grid cells) are shown in Fig. S2. The overestimation in both anthropogenic and total PM2.5 concentrations is removed after the correction.

We calculate sectoral contributions of foreign anthropogenic emissions to China's PM2.5 concentrations under China's current policy and foreign countries' SSP370 scenario at a resolution of 2∘ × 2.5∘ for January in 2030 and 2060. Sectoral contributions are calculated by taking the difference of a simulation that includes one sector of SSP370 foreign anthropogenic emissions (agriculture, industry, energy, traffic, residential combustion, solvent use, waste burning) at a time and a simulation without foreign anthropogenic emissions (“China_current_2030” and “China_current_2060” runs in Table 2).

We also conduct sensitivity simulations at a resolution of 2∘ × 2.5∘ to separate transboundary contributions to China's future PM2.5 concentrations from anthropogenic emissions within or outside the flex-grid domain of the model under the current-policy scenario and SSP370 scenario. Contributions of foreign anthropogenic emissions from within the flex-grid domain are calculated as the difference of a simulation without foreign anthropogenic emissions outside the flex-grid domain and a simulation without foreign anthropogenic emissions over the globe. Contributions of foreign anthropogenic emissions from outside the flex-grid domain are calculated as the difference between the total foreign contributions and foreign contributions from flex-grid countries.

We use the GEMM (Burnett et al., 2018) to estimate premature deaths attributable to ambient PM2.5 exposure for noncommunicable diseases (NCDs) and lower respiratory infections (LRIs) in China under each scenario. GEMM NCD + LRI calculates premature deaths associated with ambient PM2.5 exposure (M) for each population subgroup s (by age and sex) in grid cell g as 1Ms,g(Cg)=Bs×AFs(Cg)×Pg, where Bs is the national baseline mortality rate of NCD + LRI for the exposed population subgroup s. AFs(Cg) is the attributable fraction of NCD + LRI to PM2.5 exposure at level Cg for population subgroup s. Pg represents the total exposed population in grid cell g. In particular AF was calculated as AF = (RR - 1) / RR, where RR is the relative risk of NCD + LRI attributable to ambient PM2.5 exposure. The dependence of RR of NCD + LRI on PM2.5 concentrations is calculated as 2RR=e1+eθ×ln⁡zα+1-z-μν,wherez=max⁡(0,Cg-2.4), where θ, α, μ and ν are fitted parameters of PM2.5–mortality relationships. According to the GEMM, the RR of NCD + LRI is calculated by age for adults aged from 25 to greater than 85 years in 5-year intervals.

In this study, we use the national baseline mortality data by age group and disease type from the Global Burden of Disease Results Tool, 2017 version (GBD 2017; Institute for Health Metrics and Evaluation, 2018). For future baseline mortality, we use the age-specific baseline mortality rates projected by the International Futures (IFs) model v7.89 (Hughes et al., 2011). Population and age structure data for China for 2015 and future years are obtained from the SSP dataset (Samir and Lutz, 2017; Riahi et al., 2017) because China's current-policy scenario is built upon the SSP2 scenario, and the carbon-neutral scenario is built upon the SSP1 scenario (Cheng et al., 2021b; Tong et al., 2020). The gridded population data for China on a 0.5∘ × 0.5∘ spatial resolution for 2015 and future years under each SSP scenario are developed by Huang et al. (2019). The future cause-and-effect relationship between PM2.5 exposure and mortality follows the PM2.5–mortality hazard ratio function in the GEMM, which has been widely used in the calculation of PM2.5-associated mortality in the future (Hong et al., 2019; Liu et al., 2021; Yang et al., 2022).

With the baseline mortality, the population data and the age-structure data, we calculate, grid cell by grid cell, the age-specific baseline mortality rate under present and future scenarios. The impacts of transboundary pollution on mortality in China are calculated as the difference between mortality associated with PM2.5 simulated by the full anthropogenic emissions for China and foreign countries and mortality associated with PM2.5 simulated by excluding foreign emissions from the full anthropogenic emissions.

In 2015, the national mean population-weighted PM2.5 over China is about 48 µg m-3 (Fig. 1) after the observation-based correction, consistent with previous studies (48–55 µg m-3) that used various models (Burnett et al., 2018; Cheng et al., 2021a, b; Tang et al., 2022; Zhang and Cao, 2015). The correction reduces the overestimation in the model by roughly 16 % (Fig. S2); details about the correction approach are described in the Method section. From 2015–2030 and 2030–2060, there is a remarkable decreasing trend in China's annual mean population-weighted PM2.5 concentrations under plausible futures (Fig. 1). In 2030, achieving the 35 µg m-3 goal on a national average level would be feasible even under the fossil-fuel-intensive pathways in China (current-policy scenario) and foreign countries (SSP370), yet the most polluted provinces might not be able to achieve the goal under such scenarios (upper whiskers in Fig. 1). In 2060, achieving the WHO AQG goal of 5 µg m-3 would be highly unlikely under any emission pathway analyzed in this study (Fig. 1).

Considering air quality goals at the city level in China (35 µg m-3 in 2015 and 2030; 5 µg m-3 in 2060), the fraction of cities achieving air quality goals increases considerably as China and foreign countries transition from the fossil-fuel-intensive to low-carbon pathways, yet the achievement is not fully attainable in all cities. In 2015, only 30 % of 365 cities in the national PM2.5 monitoring network has an annual mean population-weighted PM2.5 below the 35 µg m-3 threshold (Fig. 2a). In 2030, the percentage increases considerably (Fig. 2b). Under the current-policy emission pathway in China (the current clean-air policies and nationally determined contribution pledges), roughly 65 % of Chinese cities (average of the three foreign scenarios) are able to achieve the goal, doubling the percentage in 2015. Under the carbon-neutral emission pathway in China (stringent clean-air policies and carbon neutrality commitments), the percentage further increases to about 92 % (average of the three foreign scenarios). However, even the cleanest emission pathway in both China and foreign countries cannot allow all cities to attain the 35 µg m-3 goal. In 2060 (Fig. 2c), the WHO AQG goal of 5 µg m-3 is not achievable for the majority of cities (≥ 75 %) in China. Emission pathways adopted by foreign countries can affect up to 6 % of cities achieving the AQG goal, and the influence could be even larger over border regions (as shown in Fig. 3).

In 2015, transboundary pollution contributes about 3.8 µg m-3 population-weighted PM2.5 to China (Fig. 3a), accounting for roughly 8 % of the total population-weighted PM2.5 (Fig. 4). In the future, transboundary pollution becomes increasingly important in China, as the share of transboundary pollution in China's total population-weighted PM2.5 increases to 12 %–22 % in 2060 (Fig. 4).

For future PM2.5 concentrations, the contribution of transboundary pollution to PM2.5 in China decreases as foreign countries and China undergo the low-carbon pathways (Fig. 3a). In 2030, under the current-policy scenario in China, transboundary contributions to PM2.5 in China would be reduced by 1.2 µg m-3 (29 %) as foreign countries transition from the fossil-fuel-intensive (SSP370) to the low-carbon (SSP119) scenario. By 2060, the difference would be increased to 1.8 µg m-3 (45 %). The transboundary pollution will also depend on Chinese domestic emissions because of their chemical interactions with foreign-transported pollution (Xu et al., 2023). In 2030, under the SSP370 scenario in foreign countries, transboundary contributions to PM2.5 in China could be reduced by 0.6 µg m-3 (14 %) as China transitions from the current-policy to the carbon-neutral scenario. In 2060, the PM2.5 reduction could be increased to 1.8 µg m-3 (45 %).

Direct atmospheric transport and chemical interactions play different roles in transboundary pollution over different regions in China. Over northern China (outlined in Fig. 3c–o), the influence of chemical interactions on transboundary pollution is prominent. Xu et al. (2023) found that foreign-transported NMVOCs could enhance the atmospheric oxidizing capacity and facilitate the oxidation of Chinese nitrogen oxides (NOx) to form nitrate over northern China. This feature persists in the future. Due to chemical interactions, northern China is the region that is the most strongly affected by transboundary pollution in most scenarios. In 2030, transboundary pollution contributes 3 to 10 µg m-3 PM2.5 concentrations to northern China, even under China's carbon-neutral scenario (Fig. 3b and i–l), greatly increasing the difficulty for this region to achieve the 35 µg m-3 goal. Further reduction of transboundary pollution on northern China requires the reduction of not only foreign anthropogenic emissions but also China's domestic emission. For example, in 2060, if China adopts the current-policy pathway (Fig. 3f–h), achieving the WHO AQG goal of 5 µg m-3 would be highly unlikely for the majority of northern China, as transboundary pollution alone would contribute roughly 3–10 µg m-3 of PM2.5 concentration (Fig. 3b and f–h). Alternatively, adopting the carbon-neutral rather than the current-policy pathway in China reduces roughly 41 %–58 % of transboundary PM2.5 over northern China in 2060 (Fig. 3b), making it possible for the region to achieve the 5 µg m-3 goal. Thus, a low-emission pathway for China has a large co-benefit on reducing transboundary pollution exerted upon its populous northern area by reducing the aforementioned chemical interactions.

The western border provinces of Yunnan and Xinjiang (denoted in Fig. 3c–o) are also influenced substantially by transboundary pollution, with the magnitude of transboundary pollution determined predominantly by foreign emission pathways. For example, in 2060, under China's carbon-neutral emission pathway, transboundary PM2.5 over Yunnan decreases from 6 to 3 µg m-3 (a 50 % reduction) as the foreign pathway switches from a fossil-fuel-intensive (SSP370; Fig. 3m) to low-carbon (SSP119; Fig. 3o) pathway, reflecting the considerable impact of South Asian pollution to China (Jiang et al., 2013). Over Xinjiang, the transboundary pollution driven predominantly by direct atmospheric transport can reach 6 µg m-3 in many scenarios (i.e., current-policy plus SSP370 for 2030 and 2060, carbon-neutral plus SSP245 for 2030), indicating the important influence of anthropogenic emissions from Central Asia in the future that has hardly been investigated in previous studies.

As transboundary contributions are the largest under China's current-policy and foreign countries' SSP370 emission pathways (Fig. 3a), we further investigate major emission sectors in foreign countries that contribute to China's future PM2.5 concentrations. We take January as an example because Xu et al. (2023) found that transboundary contributions to China's PM2.5 were the largest in January. Figure S4 reveals that agriculture, energy, industry, transportation and residential combustion emissions in foreign countries are major sources of transboundary pollution to China's national average future PM2.5 concentrations in 2030 and 2060 January, and these sectors contribute roughly evenly (15 %–23 % for each source). There is pronounced spatial heterogeneity in the source attribution. Solvent use emissions in foreign countries make roughly 3 times larger contributions to northern China (18 % in 2030 and 2060) than to the national average (6 %–7 % in 2030 and 2060). This is in line with the result of Xu et al. (2023) that the NMVOCs are the primary drivers of transboundary PM2.5 over northern China. Residential combustion in foreign countries contributes about 16 % in 2030 and 10 % 2060 more PM2.5 to the province of Yunnan than to the entirety of China, as Yunnan is mostly affected by emissions from South Asia (Jiang et al., 2013), where residential combustion is intensive (McDuffie et al., 2020).

Transboundary pollution to China could arise from emissions of countries within and outside the flex-grid domain of the model. Thus, we conduct a sensitivity simulation driven by China's current-policy and foreign countries' SSP370 emissions in 2030 and 2060 as an example to explore the relative importance of anthropogenic emissions from countries within and outside the flex-grid domain for transboundary pollution to China. Figure S5 shows that anthropogenic emissions from countries within the flex-grid domain dominate the transboundary pollution to China in all seasons (> 80 %), and their contributions remain in 2030 and 2060. Contributions of anthropogenic emissions from countries outside the flex-grid domain are larger in April and October, when westerly winds prevail and wet deposition is not as strong as in July (Jiang et al., 2013; Leibensperger et al., 2011; Ni et al., 2018). Northern China is more influenced by emissions outside the flex-grid domain (13 % in April and 24 % in October) than China on average is (7 % in April and 13 % in October), indicating that northern China is more influenced by emissions outside Asia, such as Europe and North America. Leibensperger et al. (2011) have also found that NOx emission in the US could lead to an enhancement of PM2.5 over northern China.

Figure 5 shows our estimated PM2.5-associated premature deaths in 2015 and the future. Our estimated PM2.5-associated premature deaths in China in 2015 (2.03 million) are comparable to other studies (2 to 2.4 million; Burnett et al., 2018; Geng et al., 2021; Tang et al., 2022). Our estimated increasing trend of premature deaths in China from SSP370 to SSP119 is also consistent with previous works (Tang et al., 2022; Yang et al., 2022), which is driven primarily by population aging (Fig. S6). Our estimated premature deaths in 2030 and 2060 under the current-policy emission scenario (2.68 to 2.82 million for 2030; 1.8 to 2.05 million for 2060) are substantially lower than those in Tang et al. (2022) (3.5 to 4 million for 2030; 6.5 to 7.5 million for 2060). The difference is primarily because Tang et al. (2022) fixed the baseline mortality rates in future years at the 2015 level, while our future baseline mortality rates from the IFs are projected on the basis of income, education and technology advancement, and other factors (Hughes et al., 2011).

The PM2.5 concentrations contributed by transboundary pollution can lead to an extra number of people (i.e., excess population) exposed to PM2.5 concentrations above the targeted air quality levels in China (35 µg m-3 in 2015 and 2030; 5 µg m-3 in 2060), which may lead to potential health threats. In the future, the amount of excess population due to transboundary pollution depends on both foreign and China's emission pathways (Fig. 6a). In 2030, with Chinese emissions following the carbon-neutral scenario, adopting the low-carbon (SSP119) rather than the fossil-fuel-intensive pathway (SSP370) in foreign countries could prevent 10 million Chinese people from being exposed to PM2.5 concentrations above 35 µg m-3. In 2060, 5 million people could be prevented from being exposed to PM2.5 concentrations above 5 µg m-3 if the foreign scenario switches from SSP370 to SSP119. These results indicate remarkable health benefits that the low-carbon emission pathway in foreign countries would bring to China.

For a given future year and foreign emission scenario, the excess population due to transboundary pollution tends to be larger when China adopts the carbon-neutral pathway than when China adopts the current-policy pathway (Fig. 6a). This reflects the increasing influence of transboundary pollution on air quality in China as China's overall PM2.5 concentrations drop sharply towards the air quality goals under the carbon-neutral pathway (Fig. 1).

Another measure of health threat that transboundary pollution could exert upon China is PM2.5-related premature deaths. As shown in Fig. 6b, there is an obvious decreasing trend of transboundary-contributed PM2.5-related premature deaths in China as foreign countries and China transition from the fossil-fuel-intensive to the low-carbon pathway. In 2030, adopting the low-carbon pathway (SSP119) in foreign countries would prevent 41 % (178 000 under China's carbon-neutral pathway) to 45 % (207 000 under China's current-policy pathway) of premature deaths that would occur under the fossil-fuel-intensive pathway (SSP370) in foreign countries. In 2060, the prevention would be as large as 76 % (270 000 under China's current-policy pathway) to 91 % (63 000 under China's carbon-neutral pathway). In addition, China's low-carbon emission pathway could also bring considerable health benefits through reducing the chemical interaction-related transboundary pollution and associated premature deaths in China. In 2030, adopting the carbon-neutral pathway in China would prevent 99 000 (SSP245) to 211 000 (SSP119) people from transboundary pollution-associated premature deaths relative to adopting the current-policy emission pathway. In 2060, the prevented deaths would be 76 000 (SSP119) to 283 000 (SSP370). These findings highlight the considerable health benefits to China if foreign countries and China could adopt the low-carbon emission pathways coincidently.