We use PM_2.5 concentrations observed in 199 cities in provinces across China. The monthly PM_2.5 concentrations in 199 cities during 2014–2017 were obtained by averaging the hourly data of each city from the China National Environmental Monitoring Center (http://www.cnemc.cn/). Observed PM_2.5 concentrations in 2013–2017 in Japan were collected from the National Institute for Environmental Studies (http://www.nies.go.jp/igreen/), which provides PM_2.5 observations in 47 prefectures across Japan. The monthly mean PM_2.5 in a11 47 regions are available from this website. The observation of PM_2.5 in South Korea started in 2015. The observed PM_2.5 concentrations in 2015–2017 were downloaded from the official Air Korea website of the Ministry of Environment of Korea (http://www.airkorea.or.kr). We selected 30 cities from this website which had continuous and valid monthly averaged observation data (the percentage of observation days in a month is greater than 50%). The locations of cities in China, South Korea and Japan are shown in Figure 1.

For simulation of aerosols in East Asia, we used the 3-D Goddard Earth Observing System chemical transport model (GEOS-Chem) which includes fully coupled HO_x-NO_x-VOC-O₃-aerosol chemistry with more than 80 species and 300 reactions and online calculations of aerosols including sulfate (Park et al., 2004), nitrate (Pye et al., 2009), ammonium, BC and organic carbon (OC) (Park et al., 2003), mineral dust (Fairlie et al., 2007), and sea salt (Alexander et al., 2005). We simulated concentrations of aerosols using the nested-grid version of the GEOS-Chem model version11-01 (http://acmg.seas.harvard.edu/geos/) driven by Modern-Era Retrospective Analysis for Research and Applications Version (MERRA-2) assimilated meteorological fields from NASA’s Global Modeling and Assimilation Office (GMAO) (Gelaro et al., 2017). The nested domain for Asia (60–150°E, 11°S-55°N) has 47 vertical layers up to 0.01 hPa and a horizontal resolution of 0.5° latitude by 0.625° longitude. Both wet deposition of soluble aerosols and gases (Liu et al., 2001) and dry deposition (Wesely, 2007) are included in the model simulation. PM_2.5 concentration is calculated as the sum of sulfate, nitrate, ammonium, BC, and OC from the model. We do not consider sea salt and mineral dust aerosols in PM_2.5 in this study.

The NO_x emissions from soil and lighting were calculated online based on MERRA-2 meteorology. Biomass burning emissions followed the Global Fire Emissions database Version 4 (van der Werf et al., 2017), and biogenic emissions of volatile organic compounds (VOCs) were calculated online by Model of Emissions of Gases and Aerosols from Nature Version 2 (Guenther et al., 2006). We used anthropogenic emissions from the Community Emissions Data System (CEDS) (Hoesly et al., 2018) for the global domain during 2013–2017, overwritten by year-specific Multi-resolution Emission Inventory for China (MEIC) (Zheng et al., 2018) over China and Regional Emission inventory in Asia (REASv3.2) (Kurokawa and Ohara, 2020) over other regions of East Asia. The anthropogenic emissions from REAS in 2016 and 2017 were obtained by linear extrapolation of values over 2012–2015 for each grid. The anthropogenic emissions mainly come from five sectors of agriculture, power, transportation, industry and residential.

Table 1 shows the summary of annual anthropogenic emissions of aerosols and aerosol precursors in China from MEIC and in South Korea and Japan from REASv32 during 2013–2017. As the result of air pollution control policies, anthropogenic emissions of all species except non-methane volatile organic compounds (NMVOCs) decreased year by year over China from 2013 to 2017, with largest reductions in the North China Plain (NCP) and the Yangtze River Delta (YRD) (see Supplementary Figure S1). The reduction in emissions of SO₂ was the largest, with a decrease of 58.7% over 2013–2017. Over 2013–2017, the emissions of NO_x, CO, NH₃, BC and OC in China were reduced by 20.7, 22.9, 3.4, 28.2 and 32.4%, respectively. In South Korea, over 2013–2017, the annual anthropogenic emissions of BC and NO_x decreased, while other anthropogenic emissions increased, especially those of OC. In Japan, the anthropogenic emissions kept decreasing year by year during 2013–2017, the annual anthropogenic emissions of NO_x, CO, NMVOCs, SO₂, NH₃, BC and OC in Japan decreased by 20.0, 17.7, 12.5, 21.9, 4.9, 30.7 and 12.7%, respectively.

This study will be focused on aerosol concentrations in China, South Korea and Japan. To investigate the influences of anthropogenic emission reductions in China on air quality over South Korea and Japan, we conducted three 5-year simulations with the GEOS-Chem model with nested domain in Asia (Table 2):

1. BASE: The standard simulation with variations of anthropogenic emissions in Asia and meteorological fields over 2013–2017. Anthropogenic emissions outside Asia were fixed at 2013 levels.

The BASE simulation was used for model evaluation. The FIXEM_CKJ simulation shows the effects of changes in meteorological fields over 2013–2017 alone on aerosol levels in China, South Korea and Japan. The differences between FIXEM_KJ and FIXEM_CKJ (FIXEM_KJ minus FIXEM_CKJ) represent the impacts of changes in anthropogenic emissions in China over 2013–2017 on aerosol concentrations in China, South Korea, and Japan.

Figure 2 shows the simulated annual and seasonal mean surface layer PM_2.5 concentrations over East Asia from 2013 to 2017 from the BASE simulation. The simulated annual mean PM_2.5 concentrations in eastern China reached 80–120 μg m⁻³ during 2013–2017. Seasonally, PM_2.5 concentrations in China were the highest in DJF due to the strengthened anthropogenic emissions as well as the unfavorable weather conditions, and were the lowest in JJA because of the enhanced wet removal from the East Asian summer monsoon (Zhang et al., 2010). Both the annual and seasonal mean PM_2.5 concentrations in China exhibited clear decreasing trends over 2013–2017 because of the stringent emission control measures in China. PM_2.5 concentrations in South Korea were generally higher than those in Japan, the simulated annual mean PM_2.5 concentrations were in the range of 20–40 μg m⁻³ over South Korea and were 10–20 μg m⁻³ in Japan. PM_2.5 concentrations over South Korea and Japan had the same seasonal variations, with concentrations higher in DJF and MAM and lower in JJA and SON.

We use the ground-based observations to evaluate the model performance in simulating the surface-layer PM_2.5 for each year. Supplementary Figure S2–4 display the spatial distributions of simulated (from the BASE simulation) and observed annual mean PM_2.5 concentrations in China, South Korea, and Japan. The model can reproduce the spatial distributions of PM_2.5 concentrations in these three regions. The normalized mean biases (NMBs) between simulated and observed annual mean PM_2.5 concentrations in China are −0.5, 9.1, 8.9, and 15.1% in years of 2014–2017, respectively. The NMBs of annual mean PM_2.5 concentrations in South Korea are 25.2, 34.6, and 37.5% in years of 2015–2017, respectively, and in Japan are 5.5, 12.2, 14.7, 34.2, and 34.1% in years of 2013–2017, respectively. Thus the model has high biases in South Korea and Japan during 2016–2017. There might be two reasons for the increasing trend of biases. First, anthropogenic emissions outside Asia were kept at 2013 levels (see Table 2). Second, the REAS emission inventory is available only up to 2015, so anthropogenic emissions in South Korea and Japan in 2016 and 2017 were obtained by linear extrapolation of the REAS values over 2012–2015 for each grid. Figure 3 compares simulated and observed monthly mean PM_2.5 concentrations at 15 selected cities (5 cities in each of China, South Korea and Japan). Simulated concentrations of PM_2.5 in China agree fairly well with the observations, with high correlation coefficients of 0.81–0.89 and small NMBs of 2.1–24.0% in the five cities located in different polluted areas in China. In South Korea, the model has high biases with NMBs of 32.2–124.0% in the western part of South Korea (Seoul, Daejeon and Dae-gu locations, especially in the Seoul metropolitan area), and has small low biases with NMBs of −9.3% and −4.8% in remote areas like Busan and Jeju Island. The NMBs in five cities in Japan are in the range of −13.5–32.8%, but the correlation coefficients are relatively smaller (between 0.45 and 0.65) compared to the values in China and South Korea. The seasonal variations of the observed PM_2.5 in Japanese cities were small, but the simulated results show large seasonal variations. The model well captures the decreasing trends in PM_2.5 concentrations in all cities except for Seoul and the trends from the model were weaker than those in the observations in some cities.

Figure 4 shows the yearly changes in PM_2.5 in East Asia compared to the previous year. In China, the simulated annual mean PM_2.5 concentrations in the BASE simulation generally decreased in all years compared to the previous year, except for a partial increase in North China caused by the meteorological conditions in 2015. In the downwind regions of China, with the changes in both meteorology and anthropogenic emissions, simulated annual mean PM_2.5 concentrations in the BASE simulation showed overall increases in 2014 and 2016 compared to the previous year whereas decreases in 2015 and 2017. The patterns of simulated year-to-year changes in PM_2.5 concentrations over South Korea and Japan in BASE simulation were similar to those in FIXEM_CKJ simulation, indicating the dominant influence of meteorology. Changes in anthropogenic emissions over China are estimated to reduce PM_2.5 levels in China and its downwind regions year by year during 2013–2017 (FIXEM_KJ minus FIXEM_CKJ).

We further analyze the 2013–2017 changes in PM_2.5 concentrations in eastern China (25–40°N, 110–120°E), South Korea (34.3–38°N, 126.2–129.5°E), and Japan (38–45°N, 139–145°E and 31–38°N, 130–141°E) and attribute these changes to the changes in anthropogenic emissions (FIXEM_KJ minus FIXEM_CKJ) or the changes in meteorological fields (FIXEM_CKJ) (Figure 5).

Figure 6 shows the simulated spatial distributions of the trends of concentrations of PM_2.5 components (sulfate, nitrate, ammonium, BC and OC) over 2013–2017 in East Asia. In the BASE simulation, the concentrations of all aerosol species had downward trends in eastern China, in which sulfate and nitrate had the largest decreases. In the downwind regions of China, sulfate, ammonium, and BC exhibited decreasing trends, nitrate decreased in South Korea and increased in Japan, and OC concentrations had increasing trends in both South Korea and Japan in the BASE simulation. In the FIXEM_CKJ simulation, the trends of all components caused by changes in meteorological conditions had similar spatial distributions, with negative trends in northeastern and southeastern China and South Korea whereas positive trends in central eastern China and Japan. The changes in anthropogenic emissions in China led to reductions in sulfate, ammonium, BC, and OC in the whole of East Asia. However, nitrate aerosol had an increasing trend in the downwind regions of China.

The regional trends of aerosol components are summarized in Figure 7. The trends of sulfate, nitrate, ammonium, BC and OC in eastern China in the BASE simulation were −0.86, −0.75, −0.53, −0.24, and −0.53 μg m⁻³ year⁻¹, respectively, which were much higher than the trends of decreases in South Korea and Japan (Figure 7). These decreasing trends in eastern China were mainly caused by domestic changes in anthropogenic emissions.

In South Korea, sulfate had the largest decreasing trend (−0.32 μg m⁻³ year⁻¹), followed by nitrate (−0.16 μg m⁻³ year⁻¹) and ammonium (−0.15 μg m⁻³ year⁻¹), in the BASE simulation. The changes in anthropogenic emissions in China led to large reductions in sulfate and ammonium whereas an increase in nitrate. Changes in meteorological conditions led to a relatively large decreasing trend of nitrate in South Korea.

The trends of aerosol components in the BASE simulation in Japan were the smallest among the three countries. The trends of aerosol concentrations influenced by meteorology were positive for all aerosol species. The changes in anthropogenic emissions over China had similar impacts on aerosol components in Japan as compared to South Korea. Nitrate concentrations increased in Japan from 2013 to 2017 by changes both in anthropogenic emissions in China and in meteorological fields.

To summarize, the changes in anthropogenic emissions over China in 2013–2017 had the largest impacts on sulfate concentrations in South Korea and Japan, leading to large reductions in sulfate. On the contrary, the changes in meteorology had almost no effect on sulfate in South Korea and Japan. Over 2013–2017, anthropogenic emissions of nitrate precursors of NO_x and NH₃ were reduced by 20.7 and 3.4% in China, respectively, and the concentrations of nitrate were reduced in China. However, the changes in anthropogenic emissions over China led to an increasing trend of nitrate in South Korea and Japan, which can be explained by that the reductions in SO₂ emissions changed the balance of the ammonia–nitric acid–sulfuric acid–water system, creating free NH₃ that reacted with HNO₃ to form NH₄NO₃ (Bae et al., 2021; Uno et al., 2020).

In order to have a clearer understanding of the impacts of reductions in anthropogenic emissions in China as well as the changes in meteorological conditions on PM_2.5 in South Korea and Japan, we analyzed the changes in mass outflow from eastern China during 2013–2017. Figure 8 shows the simulated annual and seasonal mass fluxes of PM_2.5 and zonal winds across the selected plane from the surface to 100 hPa altitude and from 25°N to 55°N along 120°E in the BASE simulation. The maximum mass fluxes of PM_2.5 were located at 700–1,000 hPa. On an annual mean basis, a maximum flux was located over 40°N throughout 2013–2017, and another maximum flux occurred over 46°N in 2016 and 2017 due to the strengthening of the westerlies in this location in MAM and DJF of 2016 and in SON and DJF of 2017.

As summarized in Table 3, the seasonal outflow flux of PM_2.5 peaked in MAM of every year as a result of the strong westerlies, although PM_2.5 levels in MAM in China were not the highest (Figure 2). The mass outflow of PM_2.5 was the lowest in JJA, which was caused by the lowest concentrations of PM_2.5 in China as well as the weakest westerlies in this season (Figure 8). From 2013 to 2017, the annual flux of PM_2.5 was the highest in 2013 and the lowest in 2015 (Table 3). The annual mean westerlies were the weakest over 25°-55°N in 2015 (Figure 8). Compared to 2013, the flux of PM_2.5 decreased by 15.1% in MAM of 2015, but increased by 28.0% in DJF of 2015. These yearly changes in simulated mass fluxes of PM_2.5 led to corresponding changes in simulated PM_2.5 concentrations in South Korea and Japan (Relative to 2013, the PM_2.5 concentrations in South Korea and Japan from the BASE simulation decreased significantly in MAM of 2015 and increased largely in DJF of 2015, Figure 5). The interannual and seasonal variations in mass flux are mainly driven by the changes in meteorological parameters (Holzer et al., 2005; Yang et al., 2015). With the changes in both emissions and meteorology, from 2013 to 2017, the annual mass fluxes of PM_2.5 from eastern China had a linear increasing trend of 13.1 × 10⁶ kg yr⁻¹ in the BASE simulation. However, the trend of flux was −99.4 × 10⁶ kg yr⁻¹ over 2013–2017 with changes in anthropogenic emissions in China alone (FIXEM_KJ minus FIXEM_CKJ).

Over 2013–2017, the changes in meteorological conditions have different effects on PM_2.5 concentrations in South Korea and Japan. The changes in meteorological conditions led to a decreasing trend of 0.30 μg m⁻³ year⁻¹ in PM_2.5 concentration in South Korea, whereas an increasing trend of 0.08 μg m⁻³ year⁻¹ in Japan (Figure 5B). Therefore, to further understand the impacts of meteorological fields on PM_2.5 in South Korea and Japan, we calculated the trends of annual mean temperature, relative humidity (RH) and planetary boundary layer height (PBLH) as well as the trend of annual total precipitation over 2013–2017 (Figure 9).

Temperature and RH can affect the thermodynamics of ammonium nitrate formation. In South Korea, the combined effect of increases in temperature and decreases in RH had an effect of reducing nitrate concentrations from 2013 to 2017 (Zhai et al., 2021) (Figure 7B). In South Korea, from 2013 to 2017, PBLH had an increasing trend, which increased the diffusion of pollutants and also reduced PM_2.5 concentrations. Although precipitation decreased in South Korea, the overall impact of meteorological conditions in South Korea was to reduce PM_2.5 over 2013–2017. In Japan, temperature and RH increased whereas PBLH decreased over 2013–2017. The increases in temperature could increase sulfate but reduce nitrate (Aw and Kleeman 2003). Higher RH was reported to increase the formation of secondary aerosols through enhanced chemical reactions (Yin et al., 2016). The decreases in PBLH in Japan restrained vertical mixing and lead to the accumulation of particles (Gao et al., 2015). The spatial distributions of changes in precipitation in Japan were uneven. As a result, the meteorological conditions in Japan led to increases of PM_2.5.