Foreign emissions exacerbate PM_2.5 pollution in China through nitrate chemistry

Fine particulate matter (PM_2.5) pollution is a severe problem in China. Research on the sources of Chinese PM_2.5 pollution has focused on the contributions of China's domestic emissions. However, the impact of foreign anthropogenic emissions has typically been simplified or neglected, partly due to the perception that the short lifetime of PM_2.5 (a few days) does not allow long-distance transport. Here we explore the role of foreign anthropogenic emissions in Chinese PM_2.5 pollution in 2015 using the GEOS-Chem chemical transport model. We validate the model simulations with a comprehensive set of observations of PM_2.5 and its composition, including sulfate, nitrate, ammonium, black carbon, and primary organic aerosols, over China and its surrounding regions. We find that 8 % of PM_2.5 (5 µg m⁻³) and 19 % of nitrate (2.6 µg m⁻³) over eastern China in 2015 was contributed by foreign anthropogenic emissions. The contributions were the highest in January (6.9 µg m⁻³ PM_2.5, with 68 % nitrate) and the lowest in July (2.7 µg m⁻³ PM_2.5, with 11 % nitrate). Yet, only 30 % of such foreign contributions in January were through direct atmospheric transport. The majority (70 %) were instead through chemical interactions between foreign-transported aerosol precursors and China's domestic emissions of pollutants. Specifically, the transport of non-methane volatile organic compounds (NMVOCs) from foreign countries enhanced the atmospheric oxidizing capacity and facilitated the oxidation of Chinese nitrogen oxides (NO_x) to form nitric acid (HNO₃) over eastern China. The abundance of Chinese ammonia (NH₃) further partitioned nearly all HNO₃ gas to particulate nitrate, leading to considerable foreign contributions of nitrate and PM_2.5 to eastern China. Over southwestern China, foreign anthropogenic emissions contributed 4.9 µg m⁻³ PM_2.5 concentrations (18 % of total PM_2.5 mass) to Yunnan Province, with 37 % as organics and 27 % as sulfate. Our findings suggest that foreign anthropogenic emissions play an important role in Chinese PM_2.5 pollution because of direct aerosol transport and, more importantly, chemical interactions between transported pollutants and China's local emissions. Thus, foreign emission reductions will be very beneficial for improving Chinese air quality.

1 Introduction

China has been severely affected by fine particulate matter (PM_2.5, particulate matter smaller than 2.5 µm in aerodynamic diameter) pollution over the past decades from processes of industrialization and urbanization (Geng et al., 2021; West et al., 2016). Over 1 million premature deaths associated with PM_2.5 pollution occur in China every year (Cohen et al., 2017; Yue et al., 2020; Zhang et al., 2017). In response, the Chinese government imposed stringent emission controls on primary particles and precursor gases in the 5-year Clean Air Action in 2013 (China State Council, 2013), leading to a nationwide emission reduction of 59 % for sulfur dioxide (SO₂) and 21 % for nitrogen oxides (NO_x≡ NO + NO₂) from 2013 to 2017 (Zhang et al., 2019; Zheng et al., 2018). Correspondingly, annual-mean PM_2.5 concentrations in China decreased by 30 %–50 % from 2013 to 2017 (Ding et al., 2019; Geng et al., 2021; Li et al., 2019), avoiding 64 000 (6.8 %) premature deaths. Despite these remarkable achievements, population-weighted mean PM_2.5 concentration in China was still as high as 42.1 µg m⁻³ in 2017 (with 2.1 million associated premature deaths; Geng et al., 2021a), far exceeding the newly revised threshold of 5 µg m⁻³ in the World Health Organization (WHO) Air Quality Guidelines (WHO, 2021). In 2020, the Chinese government further launched the “Beautiful China” strategy, which requires an annual-mean PM_2.5 concentration of ≤ 35 µg m⁻³ in all cities by 2035. Yet, nearly 30 % of cities in China exceeded that standard based on the 2021 national observation data (Ministry of Ecology and Environment, MEE, 2021). Thus, further improvement of air quality is pressing. However, air quality management has been progressively challenging with the abundance of end-of-pipe control technologies in dominant sources (industrial and power sectors; Xing et al., 2020) and the exhausting benefits of such technologies (Geng et al., 2021). Hence, more comprehensive and in-depth understanding of Chinese PM_2.5 pollution sources is urgently needed to help prioritize increasingly limited resources for accurate and effective mitigation action.

Recent research on the sources of PM_2.5 pollution in China has focused mostly on China's domestic anthropogenic emissions (An et al., 2019; Cheng et al., 2021b; Meng et al., 2019; Tang et al., 2022; Tong et al., 2018; Xing et al., 2020). A number of studies have explored how China could further reduce its own emissions of air pollutants through a wide range of energy transformation scenarios to achieve co-benefits of air quality improvement and climate mitigation (Cheng et al., 2021b; Peng et al., 2018; Tong et al., 2018, 2020; Xing et al., 2020). Studies have also investigated factors hindering the effectiveness of emission reductions on air quality improvement in China, such as excess ammonia emissions (Bai et al., 2019; Gu et al., 2021; Yan et al., 2021a) and enhanced atmospheric oxidizing capacity associated with NO_x emission reductions in recent years (Huang et al., 2021; Le et al., 2020; Ren et al., 2021; Zang et al., 2022). A few works have studied the inter-provincial transport of pollution across China and found that the contribution of inter-provincial transport to PM_2.5 concentrations in the most severely polluted regions (such as Beijing) might have exceeded that of local emissions (Li et al., 2015). In addition, the range of inter-provincial transport of pollution was not confined within city clusters, such as Beijing–Tianjin–Hebei, the Yangtze River Delta, and the Pearl River Delta, but also extended over a long distance across city clusters (Wang et al., 2022). Yan et al. (2021b) further found that transboundary transport from Asian regions (18.5 %–19.2 %, including Chinese regions outside the Wuhan city cluster) contributed much more to ozone concentrations in the Wuhan city cluster in Central China than the transport within the city cluster (2.5 %–3.1 %), highlighting the importance of transboundary transport of pollutants to China.

However, the influence of transboundary transport to China air quality has been hardly investigated for PM_2.5. This is likely due to the perception that the relatively short lifetime of PM_2.5 (a few days) does not permit long-distance transport (Wang et al., 2019). Only a few have investigated the influence of pollutant emissions from neighboring countries (Jiang et al., 2013; Koplitz et al., 2017) on China; however they have typically focused on one particular sector, such as biomass burning emissions from South Asia (Jiang et al., 2013) or coal emissions from Southeast Asia (Koplitz et al., 2017). A comprehensive assessment of transboundary PM_2.5 pollution in China from foreign sources is lacking. In contrast, studies on the transboundary PM_2.5 pollution from China to neighboring countries have received considerable attention (Choi et al., 2019; Jiang et al., 2013; Kurokawa and Ohara, 2020; Park et al., 2014). This contrast is likely due to another perception that transboundary pollution from foreign countries to China is minor since China's domestic emissions far exceeded those from neighboring countries, such as Korea, Japan, India, and Southeast Asia (Kurokawa and Ohara, 2020; McDuffie et al., 2020). However, the pollutant emission pattern in China and neighboring countries may shift in the future. Emissions in China have decreased considerably (Zheng et al., 2018), and the trend is expected to continue with the launch of ambitious policies on air pollution (the 2035 Beautiful China strategy) and climate change (the 2060 carbon neutrality). In contrast, emissions from India and Southeast Asian countries have been estimated to increase in the future by various projections, given their fast economic growth and a lack of clear commitments on either air quality or climate mitigation (IEA, 2021). For example, Koplitz et al. (2017) revealed that the projected increase in coal emissions in Southeast Asian countries will lead to 49 780 excess deaths per year associated with PM_2.5 pollution in 2030, with 9000 (18 %) of these excess deaths occurring in China. The transboundary pollution from neighboring countries to China may become increasingly prominent in the future. Thus, an effective air quality management action for the achievement of the Beautiful China target requires a clear understanding of the current contribution of foreign anthropogenic emissions to Chinese PM_2.5 pollution.

A comprehensive assessment of foreign contributions to PM_2.5 pollution in China relies on a complex representation of aerosol emissions and chemical reactions across a large spatial domain. The GEOS-Chem global chemical transport model has been widely applied to PM_2.5 studies over Asia (i.e., China, India, Southeast Asia, Korea, and Japan; Choi et al., 2019; Koplitz et al., 2017; Miao et al., 2020; Venkataraman et al., 2018; Wang et al., 2004; Zhang et al., 2015) and is thereby applicable to such research. Although the model has been extensively validated for total PM_2.5 mass concentrations over China using observational data, compositional PM_2.5 across China and total PM_2.5 for other Asian countries are far less evaluated due to scarce observations (Cheng et al., 2021a; Koplitz et al., 2017; Miao et al., 2020). This limits the credibility of the model's representation of aerosol emission and chemical reactions across a large domain. Thus, a more comprehensive evaluation of the GEOS-Chem simulation is needed to support model estimates of the influence of transboundary pollution on air quality in China.

In this study, we use the GEOS-Chem model to quantify the contributions of foreign anthropogenic emissions to total and compositional PM_2.5 mass concentrations over China in 2015. We first evaluate our model simulations with comprehensive observations of total and compositional PM_2.5 concentrations across China and other Asian countries. Then, we quantify the contributions of foreign anthropogenic emissions to Chinese PM_2.5 and compositional concentrations in 2015. Finally, we reveal the physical and chemical pathways leading to such contributions.

2 GEOS-Chem simulations

We conducted a series of simulations using the GEOS-Chem chemical transport model (v13.2.1; http://www.geos-chem.org, last access: 4 April 2023) to (1) represent 2015 PM_2.5 and composition concentrations over Asia, (2) quantify the contributions of foreign anthropogenic emissions to total and compositional PM_2.5 concentrations over China, and (3) understand the role and the mechanisms of direct transport and chemical interactions in transboundary pollution in China. Simulation configurations are summarized in Table 1 and are elaborated as follows.

Table 1 Configuration summary of GEOS-Chem simulations in this study. Emission abbreviations are elaborated in Sect. 2.2.

^* OTR refers to other species including SO₂, NH₃, BC, OC, and CO.

2.1 The GEOS-Chem simulation of ground-level PM_2.5

We used the flex-grid capability of the GEOS-Chem classic model v13.2.1 to simulate aerosol concentrations over Asia and the adjacent area (11^∘ S–60^∘ N, 30–150^∘ E; Fig. 2a) at a horizontal resolution of 0.5^∘ × 0.625^∘ and at 47 vertical levels between the surface and ∼ 0.01 hPa. The lowest vertical layer has a thickness of about 130 m. We regard the pollutant concentrations in this layer as “ground-level”. Detailed descriptions of the flex-grid setup can be found at http://wiki.seas.harvard.edu/geos-chem/index.php/FlexGrid (last access: 4 April 2023). Our flex-grid domain extended the traditionally defined nested Asia domain (11^∘ S–55^∘ N, 60–150^∘ E) in the model (Fig. 2a) to better represent the transport of anthropogenic pollutants from Central Asia to China that has not been studied yet. Our simulations were driven by assimilated meteorological data from MERRA-2 provided by the Global Modeling and Assimilation Office (GMAO) at NASA Goddard Space Flight Center. Convective transport in the model was computed from the convective mass fluxes in the meteorological archive as described by Wu et al. (2007). A non-local scheme was used to represent vertical mixing within the planetary boundary layer (PBL), as it accounts for different states of mixing based on the static instability (Lin and McElroy, 2010). Boundary conditions were archived from global simulations at a resolution of 2^∘ × 2.5^∘. We spun up every simulation for 1 month to remove the effects of initial conditions.

GEOS-Chem simulates PM_2.5 concentrations as the sum of sulfate (SO${}_{\mathrm{4}}^{\mathrm{2}-}$), nitrate (NO${}_{\mathrm{3}}^{-}$), ammonium (NH${}_{\mathrm{4}}^{+}$), organic aerosol (OA ≡ primary OA + secondary OA), black carbon (BC), fine dust, and fine sea salt component concentrations. The sulfate–nitrate–ammonium (SNA) aerosol system was simulated following Fountoukis and Nenes (2007) and Park et al. (2004), including heterogeneous chemistry with dinitrogen pentoxide (N₂O₅) uptake by aerosol and hydroperoxyl radical (HO₂) uptake by aerosol. Gas−aerosol partitioning of SNA was simulated by the ISORROPIA II thermodynamic equilibrium scheme (Pye et al., 2009). We used a simple scheme to represent secondary organic aerosol formation (Heald et al., 2012) and used a spatially resolved ratio to calculate organic mass from organic aerosol concentrations (Philip et al., 2014). Natural dust simulation followed the Mineral Dust Entrainment and Deposition (DEAD) scheme (Duncan Fairlie et al., 2007). Sea salt aerosol simulation was described in Jaeglé et al. (2011). Dry deposition of gases and particles followed a standard resistance-in-series scheme, with updates from Jaeglé et al. (2018). Wet deposition was described in Liu et al. (2001) and Wang et al. (2011, 2014), with updates from Luo et al. (2020) that included a faster below-cloud scavenging of HNO₃. We calculated the simulated PM_2.5 and composition concentrations at 35 % relative humidity (RH) for consistency with ground-based measurements.

2.2 Emissions for baseline simulation

We conducted a baseline simulation (“Base” run in Table 1) for January, April, July, and October 2015 and treated the mean of the 4 months as the annual mean. Our simulations were all at a resolution of 0.5^∘ × 0.625^∘, unless otherwise specified. The baseline simulation used emissions as described below.

Anthropogenic emissions for China were taken from the Multi-resolution Emission Inventory (MEIC) for 2015 (Zheng et al., 2018), and for the rest of the world they were taken from the Community Emissions Data System (CEDS) version 2 for 2015 (https://data.pnnl.gov/dataset/CEDS-4-21-21, last access: 4 April 2023). Other emissions were default in GEOS-Chem. Fine anthropogenic fugitive dust emissions from combustion and industrial sources for countries except China (FR_AFCID) were taken from Philip et al. (2017) and from the MEIC inventory for China (CH_AFCID). Aircraft emissions were from the Aviation Emissions Inventory Code (AEIC) inventory (Stettler et al., 2011). Natural emissions include lightning NO_x from Murray et al. (2012); soil NO_x, biogenic non-methane volatile organic carbons (NMVOCs), and sea salt from offline emissions developed by Weng et al. (2020); biomass burning emissions from the Global Fire Emissions Database version 4 (GFED4; Randerson et al., 2015); volcano emissions from Fisher et al. (2011); marine dimethyl sulfide (DMS) emissions from Breider et al. (2017); and dust emissions using the Mineral Dust Entrainment and Deposition (DEAD) scheme (Zender et al., 2003).

2.3 Sensitivity simulations for the contributions of foreign anthropogenic emissions to Chinese PM_2.5 and composition concentrations

We quantified contributions of foreign anthropogenic emissions to China's total and compositional PM_2.5 concentrations by taking the difference in the baseline simulation (“Base” run in Table 1) and a sensitivity simulation that excluded foreign anthropogenic emissions from the baseline simulation (“CHAnth” run in Table 1). Such a foreign contribution is referred to as “FR_total”. We also conducted simulations to quantify the sectoral contributions of foreign anthropogenic emissions to China's PM_2.5 concentrations. Sectoral contributions were calculated by taking the difference in a simulation that included one sector of foreign anthropogenic emissions (agriculture, industry, energy, traffic, residential combustion, solvent use, waste burning) at a time and a simulation without foreign anthropogenic emissions (“CHAnth” in Table 1).

We further conducted sensitivity simulations to attribute the transboundary pollution to (1) direct transport of foreign PM_2.5 to China and (2) chemical interactions between transported foreign pollutants and Chinese emissions. We quantified the contribution of direct transport in transboundary pollution (referred to as “FR_transport”) by taking the difference in a sensitivity simulation that excluded Chinese anthropogenic emissions (“FRAnth” run in Table 1) and another sensitivity simulation that excluded both Chinese and foreign anthropogenic emissions (“NoAnth” run in Table 1). Transboundary pollution through chemical interactions with China's local emissions (referred to as “FR_chemistry”) was calculated as the differences between total foreign anthropogenic contributions (FR_total) and direct transport contributions (FR_transport).

We conducted sensitivity simulations to understand main pollutants driving the chemical interactions of transboundary pollution with Chinese emissions. Specifically, we quantified the contributions of foreign anthropogenic emissions of NMVOCs and NO_x to O₃, NO₃, N₂O₅, HNO₃, and NO${}_{\mathrm{3}}^{-}$ concentrations in China. Such contributions include both the direct transport of foreign pollutants and chemical interactions between foreign-transported and Chinese domestic emissions of pollutants. We quantified the total contributions of foreign anthropogenic emissions of NMVOCs as the difference between a simulation with full emissions and a simulation that excluded foreign anthropogenic emissions of NMVOCs (“Base” – “No_FRAnthNMVOCs” runs in Table 1). To quantify the direct transport share of the total contributions, we excluded China's domestic emissions to avoid interactions with foreign-transported pollutants and calculated them as the difference between simulations that included and excluded foreign anthropogenic emissions (“FRAnthNMVOCs” – “NoAnth” runs in Table 1). The contributions of chemical interactions between foreign NMVOCs and China's domestic emissions were quantified as the difference between the total contributions and the direct transport share of the total contributions. Similarly, the total contributions by foreign anthropogenic emissions of NO_x were calculated as “Base” – “No_FRAnthNO_x” runs in Table 1, with the corresponding direct transport share of the total contributions calculated as “FRAnthNO_x” – “NoAnth” runs in Table 1 and the chemistry share calculated as the difference between the total and the direct transport share.

To reduce computational costs, we conducted NMVOC-related sensitivity simulations at a resolution of 2^∘ × 2.5^∘ (Table 1). The difference in PM_2.5 over eastern China between resolutions of 2^∘ × 2.5^∘ and 0.5^∘ × 0.625^∘ is about 5 % (3 µg m⁻³). Compositional differences are within 20 %, with the largest difference in nitrate (2.6 µg m⁻³, 18 %) and the lowest in black carbon (0.1 µg m⁻³, 4 %). These differences are within the reasonable range associated with spatial resolutions.

3 Ground-level observations of PM_2.5 and composition in China and other Asian countries

We evaluated the modeled PM_2.5 and composition concentrations in the base year (2015) using ground-level PM_2.5 observations from the network of the China National Environmental Monitoring Center (CNEMC), a literature search, and the World Health Organization (WHO) database.

Ground-level PM_2.5 observations were obtained from the CNEMC network (http://106.37.208.228:8082/, last access: 4 April 2023). We used hourly measurements for January, April, July, and October 2015. PM_2.5 mass concentrations were measured using the micro-oscillating balance method or the β absorption method (Zhang and Cao, 2015). We further applied quality controls to hourly CNEMC data. Specifically, we removed a day if there were <14 valid data points within the day and removed a month if there were <25 d of valid data within the month. These in whole removed 12 % of the PM_2.5 hourly data and finally retained observations for 1179 sites in 314 cities for model evaluation (Fig. S1). To compare with the GEOS-Chem-simulated PM_2.5 concentrations, we calculated the grid-averaged and monthly averaged PM_2.5 concentrations from the CNEMC to match spatially and temporally with the model.

We collected compositional PM_2.5 observations from publicly available studies, as shown in Table S1. We selected observations that spun at least 1 year or seasonal/monthly measurements centered at January, April, July, or October to match our model simulations. A total of 56 observation data points from 17 cities in 16 provinces for 2014–2016 were collected for major PM_2.5 chemical composition, including sulfate, nitrate, ammonium, organic aerosol (OA), and black carbon. We sampled the GEOS-Chem-simulated concentrations from locations and periods (monthly or annual) of observations for evaluation.

To evaluate the modeled PM_2.5 concentrations outside China, we collected PM_2.5 measurement data for 2013–2016 from the WHO's database on ambient air pollution in cities (https://www.who.int/data/gho/data/themes/air-pollution, last access: 4 April 2023). This database provides annual-average PM_2.5 concentrations for more than 500 cities globally. We retained PM_2.5 measurements that were directly measured and excluded data inferred from PM₁₀ concentrations. We also retained the sites within our simulation domain and the grid cells where anthropogenic emissions exceeded natural emissions (by comparing PM_2.5 concentrations in “FRAnth” and “NoAnth” runs in Table 1). A total of 83 sites covering 10 Asian countries (Bangladesh, Indonesia, India, Japan, Korea, Malaysia, Myanmar, the Philippines, Thailand, and Vietnam) were finally selected in our study as shown in Fig. 2a. Because of the differences in measurement approaches between jurisdictions and the absence of details regarding measurement data (Brauer et al., 2016), we compared annual-average PM_2.5 concentrations in the WHO dataset with those estimated from a hybrid of satellite observations, a chemical transport model, and ground-based measurements (van Donkelaar et al., 2021). We removed the sites if the differences were more than 40 %. The final sites retained in our study are shown in Fig. 2b.

4 PM_2.5 and composition concentrations across China and other Asian countries

Figure 1 shows modeled and observed annual-mean concentrations of PM_2.5 and composition across China. The modeled total and compositional PM_2.5 are for 2015. Observations of total PM_2.5 are for 2015, and those of composition are for 2014–2016. The spatial distribution of modeled PM_2.5 exhibits a broad high across eastern China, driven primarily by sulfate (SO${}_{\mathrm{4}}^{\mathrm{2}-}$), nitrate (NO${}_{\mathrm{3}}^{-}$), ammonium (NH${}_{\mathrm{4}}^{+}$), and organics (OA). PM_2.5 over the Sichuan Basin is also high, contributed mostly by sulfate, ammonium, and organics. The enhanced PM_2.5 concentrations over the west are dominated by mineral dust. Compared to observations, the modeled PM_2.5 reproduces the spatial variation in PM_2.5 well across China, with a correlation coefficient (r) of 0.73 and a normalized mean bias (NMB) of 15.7 %. The slight overestimation is primarily contributed by sites in the Sichuan Basin (Fig. S1a), where local emissions (anthropogenic and vegetation emissions; Wang et al., 2018) and meteorological conditions (humid and stagnant; Chen et al., 2014; Liao et al., 2017), combined with the special terrain (plain surrounded by hills; Chen et al., 2014; Wang et al., 2017), make it hard for the model to represent aerosol processes, especially the secondary aerosol formation process (Liao et al., 2017; Tao et al., 2017; Wang et al., 2018). The seasonal evaluation of simulated PM_2.5 is shown in Fig. S2 and exhibits a good consistency with observations across seasons (r = 0.61–0.77), demonstrating the model's capability in capturing the seasonal pattern of aerosol processes.

Figure 1 Total and compositional PM_2.5 concentrations in China. Spatial distributions of PM_2.5 and composition concentrations are annual-mean concentrations simulated by the GEOS-Chem model for January, April, July, and October 2015 with a resolution of 0.5^∘ × 0.625^∘. We regard the mean of the 4 months as the annual mean. Thick black lines outline eastern China, as discussed in this work. Text in the bottom left corner of each map refers to mean concentrations (µg m⁻³) over eastern China. The scatterplot of PM_2.5 compares the simulated annual-mean concentrations with collocated and coincident observations for 2015 from the CNEMC network. Scatterplots of composition compare both annual and monthly concentrations of the observed and the coincident simulated concentrations according to the availability of observations from the literature for 2014–2016.

We further evaluate our modeled compositional PM_2.5 with observations from 56 sites across 16 provinces and municipalities in China to better understand the performance of the simulation. The scatterplots of composition comparison in Fig. 1 show a mixture of annual and monthly data depending on the availability of observation data from the literature. We find that the modeled compositional PM_2.5 reasonably reproduces the vast spatial variation in PM_2.5 composition from observations. Sulfate and nitrate simulations are particularly improved over previous model studies where sulfate was significantly underestimated (NMB of −40 %) and nitrate was substantially overestimated (NMB – 80 %; Gao et al., 2018; Miao et al., 2020). The better representation of nitrate is due to the faster nitrate removal in the Luo et al. (2020) deposition scheme. Organics are also well reproduced by the model (NMB = 4.7 %), suggesting the effectiveness of the simple secondary organic aerosol (SOA) scheme in representing total SOA mass. Ammonium and black carbon show relatively large discrepancies (NMB = 24.9 % and −12.8 %, respectively), potentially reflecting model biases in chemical reactions or gas–particle partitioning for ammonium formation (Miao et al., 2020) and emission inventories for BC (Zhang et al., 2019). In addition, outstanding differences in observation approaches (i.e., thermal, optical, or incandescence measurements for black carbon and different relative humidity in measurements; Bond et al., 2013; Snider et al., 2016) across the literature are another major reasons for the discrepancies.

The evaluation of modeled PM_2.5 concentrations for other Asian regions has been rarely conducted due to limited observations (Koplitz et al., 2017). Here, we compare our modeled PM_2.5 concentrations with observations from 10 Asian countries around China to understand the model performance in regions whose pollution could influence China through transboundary transport. Figure 2 shows good agreement between modeled and observed total PM_2.5 mass across countries, with a correlation coefficient of 0.76 and a NMB of 3.5 %, despite large uncertainties in the measurements collected by different countries. There is an underestimate at coastal sites (e.g., in the Philippines; Fig. 2a), where sea salt aerosols potentially make larger contributions than our simulations. Specifically, our simulated PM_2.5 is very consistent with observations in Southeast Asia (NMB = 2.7 %; including Indonesia, Myanmar, the Philippines, Thailand, and Vietnam), India (3.8 %), and other countries in South Asia (NMB = 7.9 %; including Bangladesh, Bhutan, the Maldives, Nepal, Pakistan, and Sri Lanka). Simulations over Japan and South Korea show relatively larger bias (NMB = 17.4 % for Japan and 40 % for South Korea). Zhai et al. (2021) attributed the 43 % bias in the GEOS-Chem simulation of surface PM_2.5 over South Korea in their study to nighttime nitrate formation, although their updates of faster below-cloud scavenging of HNO₃ (Luo et al., 2020) corrected the overall nitrate bias in East Asia.

Figure 2 Annual-mean PM_2.5 concentrations at anthropogenic-emission-dominated sites in the simulation domain outside China. (a) The spatial distribution of simulated and observed PM_2.5 concentration ratios. The simulated concentration at each measurement site represents the 0.5^∘ × 0.625^∘ grid cell covering that site. Dashed lines represent the default nested Asia domain (11^∘ S–55^∘ N, 60–150^∘ E) in the model. The flex-grid domain in our study is shown as the entire domain of the map. (b) Scatterplot comparing simulated concentrations for 2015 with collocated observations from the WHO for 2013–2016.

5 Contributions of foreign anthropogenic emissions to total and compositional PM_2.5 in China

Figure 3 shows the contributions of foreign anthropogenic emissions to total and compositional PM_2.5 concentrations over China in 2015. On the national level, foreign anthropogenic emissions contribute about 2.4 µg m⁻³ PM_2.5 to China in 2015, accounting for 6.2 % of the national-average PM_2.5 concentration. The foreign influence exhibits prominent spatial heterogeneity, with the largest contribution of 5.0 µg m⁻³ PM_2.5 (8 %) to eastern China (outlined in Fig. 3; including Anhui, Hebei, Henan, Jiangsu, Liaoning, Shandong, Beijing, and Tianjin). Considering the WHO's newly revised guideline for PM_2.5 annual exposure level (≤ 5 µg m⁻³), foreign anthropogenic emissions alone would make PM_2.5 concentrations over eastern China reach the WHO standard, threatening the health of nearly 500 million residents there. Transboundary pollution is also outstanding along the southwestern border, contributing 4.9 µg m⁻³ PM_2.5 (18 %) to Yunnan Province, mostly driven by anthropogenic emissions from South Asia (i.e., India).

Figure 3 Simulated annual-mean contributions of foreign anthropogenic emissions to China's total and compositional PM_2.5 concentrations in 2015. Thick black lines outline eastern China, as discussed in this work. Text in the bottom left corner of each panel refers to mean concentrations (µg m⁻³) over eastern China contributed by foreign anthropogenic emissions. The foreign impact on China's anthropogenic dust concentrations is shown in Fig. S3. YN in the PM_2.5 subplot refers to the location of Yunnan Province.

The transboundary pollution over eastern and southwestern China is contributed by different chemical components of PM_2.5, as shown in Fig. 3. Eastern China is mainly driven by nitrate and ammonium, explaining 70 % of transboundary PM_2.5. Particularly, 18 % of nitrate and 12 % of ammonium concentrations over eastern China in 2015 are driven by transboundary pollution. However, such an enhancement over eastern China is not observed for organic matter, partly due to the simplified SOA formation scheme that is not able to fully represent the chemical formation of SOA, which is a common issue in chemical transport models (Pennington et al., 2021; Shrivastava et al., 2017). Leibensperger et al. (2011) and Koplitz et al. (2017) found similar nitrate enhancement, yet to a much lesser extent (<0.2 µg m⁻³). They attributed the nitrate enhancement to the increase in ozone that speeded up the rate at which Chinese local NO_x emissions were converted to nitrate. We discuss the mechanism in more detail in the next section. The sulfate contribution is very small (<0.5 µg m⁻³). Leibensperger et al. (2011) proposed a reason for why H₂O₂ (the key oxidant for sulfate formation) is abundant most of the year over eastern China and is thereby insensitive to additional transboundary sources. When H₂O₂-limited conditions prevail in winter, cloud cover is infrequent, limiting the in-cloud oxidation of SO₂. Thus, the influence of transboundary pollution to sulfate over eastern China is weak. The transboundary PM_2.5 over southwestern China is primarily contributed by organics (1.8 µg m⁻³, 37 % of transboundary PM_2.5), which is consistent with previous studies that found that massive biomass burning emissions in South Asia contributed considerable organics to southern China (Jiang et al., 2013). Sulfate contributes 27 % of transboundary PM_2.5 over southwestern China, where the inflow of hot and humid atmosphere from South Asia facilitates the in-cloud formation of sulfate (Jiang et al., 2013). In addition, anthropogenic fugitive dust emissions from foreign countries have an influence on Chinese PM_2.5, accounting for 14 % of the PM_2.5 increase in both eastern and southwestern regions, as shown in Fig. S3.

We further investigate the seasonal variation in transboundary pollution in China to understand potential sources of foreign contributions. Figure 4 presents the seasonal enhancement of PM_2.5, nitrate, and ammonium over China in 2015 driven by foreign anthropogenic emissions. Transboundary pollution of PM_2.5 over eastern China is the largest in January (6.9 µg m⁻³) and gradually decreases to a minimum in July (2.7 µg m⁻³). Affected regions also change prominently with seasons. In January, the boundary of transboundary PM_2.5 larger than 9 µg m⁻³ extends to the south of the eastern China domain outlined in Fig. 4, whereas in July, that boundary shrinks to a much smaller region along the east coast. Source attribution (Fig. S4) reveals that all of the seven major anthropogenic emission source sectors in foreign countries contributed similarly (10 %–17 % for each sector) to Chinese PM_2.5 in January, yet their relative importance exhibits spatial heterogeneity. Over eastern China, industry and solvent use in foreign countries are the largest sources (likely because of their considerable amount of NMVOCs emissions that increase the atmospheric oxidizing capacity over eastern China, as is discussed in Sect. 6), whereas over Yunnan Province, residential combustion in foreign countries makes the largest contribution.

Figure 4 Same as Fig. 3, but for January, April, July, and October as denoted by text in the top left corner of each row.

These seasonal characteristics of transboundary PM_2.5 are similar to those of nitrate and ammonium. In January, 68 % of transboundary PM_2.5 over eastern China is contributed by nitrate (4.7 µg m⁻³) and 19 % by ammonium (1.3 µg m⁻³), yet these fractions decrease to 11 % for both nitrate and ammonium in July. The transboundary nitrate for the majority of eastern China exceeds 5 µg m⁻³ in January yet decreases to less than 1 µg m⁻³ and is even negative in July. These prominent seasonal variations in transboundary PM_2.5, nitrate, and ammonium reflect different processes controlling the transboundary pollution in winter and summer in China.

6 Physical and chemical mechanisms of foreign anthropogenic contributions to PM_2.5 over eastern China

We investigate the relative importance of direct transport and chemical interactions in driving the considerable transboundary pollution over eastern China in January and July, as shown in Fig. 5. In January, the transboundary PM_2.5, nitrate, and ammonium are predominantly (71 %–97 %) driven by chemical interactions, suggesting that the transboundary pollution in winter over eastern China is not through direct transport of nitrate and ammonium but through chemical interactions between directly transported precursors from foreign countries and local emissions in China. In July, however, nearly all transboundary PM_2.5 over eastern China is driven by direct transport, with 30 % of the direct transport contributed by anthropogenic fugitive dust from foreign countries (Fig. S3). The transboundary nitrate is still primarily driven by chemical interactions in July (89 %), yet the magnitude is too small to substantially affect total PM_2.5.

Figure 5 Foreign anthropogenic contributions to PM_2.5, nitrate, and ammonium concentrations in January and July over eastern China. Total concentration contributions of foreign anthropogenic emissions are split into contributions from direct transport and chemical interactions according to the legend. The contributions to other PM_2.5 components are shown in Fig. S5.

Figure 6 Contributions of foreign anthropogenic emissions of NMVOCs (a, b) and NO_x (c, d) to atmospheric oxidizing capacity and oxidation products over eastern China as simulated by the GEOS-Chem model for January (a, c) and July (b, d) 2015. Total foreign anthropogenic contributions are split into contributions from direct transport and chemical interactions according to the legend. O₃ concentrations are 24 h average concentrations. Nitrogen-related species are presented in micrograms of nitrogen per cubic meter for the convenience of nitrogen budget calculation.

We further explore the changes in key chemical species for nitrate formation to understand the chemical mechanism driving the considerable nitrate enhancement over eastern China. Figure 6 shows the contributions of transboundary pollution to concentrations of precursor gases (NO_x), oxidants (O₃, N₂O₅, and NO₃), and oxidized products (total inorganic nitrate including gas-phase HNO₃ and particulate NO${}_{\mathrm{3}}^{-}$) in nitrate chemistry over eastern China in January and July. Figure 6a shows that, in January, chemical interactions between foreign-emitted NMVOCs and China's local aerosols promote the O₃ production over eastern China by 2.6 ppb, leading to a drop in NO_x (−1.8 µg N m⁻³) and an increase in N₂O₅+ NO₃ and HNO₃+ NO${}_{\mathrm{3}}^{-}$ (0.88 µg N m⁻³) concentrations over eastern China. The presence of excess ammonia over eastern China due to the reduction in SO₂ (Liu et al., 2018, 2019) further partitions nearly all HNO₃ gas into particulate nitrate (nitrate ammonium, NH₄NO₃), leading to a nitrate increase of about 1 µg N m⁻³ (or 2.6 µg m⁻³) there. These results reveal that, in January, when the O₃ production is in a NMVOCs-limited regime, the additional NMVOCs from transboundary sources promote the O₃ production and the subsequent nitrate formation over eastern China, which otherwise would be limited by the lack of O₃ (Jin and Holloway, 2015; Li et al., 2018; Wang et al., 2017).

In addition to NMVOCs, NO_x is another important precursor of O₃. We therefore conducted sensitivity simulations to understand the role that foreign anthropogenic emissions of NO_x play in the nitrate enhancement in eastern China. As shown in Fig. 6c for January, the promotion of nitrate formation is very small, with the additional influence of NO_x emitted from foreign sources. This is due to the high local emissions of NO_x over eastern China in January that suppress the O₃ formation under a NO_x-saturated regime.

In July, although foreign NMVOCs contribute about 4 ppb O₃ to eastern China through the chemistry process, they hardly lead to much difference in nitrate concentrations (Fig. 6b, d). This is because a lack of excess aerosols (compared to winter) limits the transformation of N₂O₅ and NO₃ to HNO₃ on the aerosol surface. In addition, the oxidation of NO_x by OH is sufficiently fast in summer, and the abundance of OH over eastern China in summer makes the NO_x oxidation process insensitive to the added O₃ from foreign countries. Therefore, a lack of excess aerosols and the abundance of OH in summer make the transboundary transport of ozone precursors minor in contributions to nitrate concentrations over eastern China.

7 Conclusions

An effective air quality improvement action requires an accurate understanding of PM_2.5 sources. This work complements our understanding of PM_2.5 sources in China by investigating the influence of foreign transboundary transport through the GEOS-Chem simulation. Our extensive and comprehensive evaluation of the GEOS-Chem model for PM_2.5 total and compositional mass concentrations in China and 10 additional Asian countries showed a reasonable consistency with observations. Based on model simulations, we found that foreign anthropogenic emissions played an important role in Chinese PM_2.5 pollution because of direct aerosol transport and, more importantly, chemical interactions between transboundary pollutants and China's local emissions. Over eastern China, the transport of NMVOCs from foreign anthropogenic emissions increased O₃ concentrations by 2.6 ppb through chemical interactions with China's local aerosols. The additional O₃ combined with high local emissions of NO_x and ammonia led to a nitrate enhancement of 2.6 µg m⁻³ in January. Over southwestern China, transboundary transport contributed 18 % of the PM_2.5 to Yunnan Province in 2015, mostly driven by the direct transport of aerosols from anthropogenic emissions in South Asia. There are a few sources of uncertainty in this study, for example the wet deposition of nitrate and the simplified secondary organic aerosol formation scheme, but they do not manifest themselves as systematic biases.

In light of the physical and chemical mechanisms of transboundary pollution in China, further improvements of air quality for the Beautiful China target require different emission reduction strategies for different regions. Over eastern China, reductions in both foreign and domestic anthropogenic emissions can reduce transboundary PM_2.5 pollution to China, since a considerable amount of transboundary-transported PM_2.5 is formed through the chemical interactions between pollutants from both sources. To reduce the transboundary influence driven by chemical interactions, source sectors with considerable emissions of NMVOCs in foreign countries, such as industry and solvent use, are of particular importance. Over southwestern China, foreign emission reductions will be necessary for improving air quality there since the transboundary pollution was primarily through the direct transport of aerosols from foreign countries. Given the declining trend of Chinese anthropogenic emissions in the present and the future, the likely rising emissions from adjacent countries in the future could become an increasingly important problem for Chinese air quality protection if no action is taken to avoid emissions increases in adjacent countries. Our study points to the need to carefully consider the potential influence of transboundary pollution when making long-term air quality improvement strategies in China. Future studies could further investigate the impact of transboundary pollution transport to China under future emission scenarios or extend this study to other regions in the world where transboundary pollution could potentially increase domestic PM_2.5 pollution through nitrate chemistry. Investigation into the bi-directional PM_2.5 contributions between countries would also be helpful for developing practical policies for regional cooperation on emission reductions.