Coarse particulate matter air quality in East Asia: implications for fine particulate nitrate

Air quality network data in China and South Korea show very high year-round mass concentrations of coarse particulate matter (PM) between 2.5 μm and 10 μm aerodynamic diameter, as inferred by the difference between PM 10 and PM 2.5 observations. This coarse PM averages 47 μg m -3 in the North China Plain (NCP) region in 2015-2019 and 21 μg m -3 in the Seoul Metropolitan Area (SMA). It is dominantly contributed by urban fugitive dust, rather than by natural dust as is often presumed. Concentrations decreased by 21% from 2015 to 2019 and further dropped abruptly in 2020 due to COVID-19 reductions in construction and vehicle traffic. This anthropogenic coarse PM is generally not included in air quality models but acts as a sink of nitric acid (HNO 3 ), thus affecting fine particulate nitrate which is a major air quality concern in China and South Korea. GEOS-Chem model simulation of surface and aircraft observations from the KORUS-AQ campaign over the SMA in May-June 2016 shows that consideration of anthropogenic coarse PM largely resolves the previous model overestimate of PM 1 nitrate. Anthropogenic coarse PM in the model increases the sensitivity of PM 2.5 nitrate to ammonia (NH 3 ) emission in winter. In summer, anthropogenic coarse PM directly affects PM 2.5 nitrate by HNO 3 uptake, and we find that the decrease of anthropogenic coarse PM over 2015-2019 offset the PM 2.5 nitrate reductions expected from decreasing NO x and NH 3 emissions.


Introduction
Coarse particulate matter (coarse PM; particulate matter between 2.5 µm and 10 µm in aerodynamic diameter) poses a severe air pollution problem in East Asia, constituting a particle mass comparable to fine particulate matter (PM2. 5) and thus about half of PM10 (1)(2)(3)(4). That coarse PM is mainly fugitive mineral dust, with contributions from both natural desert dust and human activity including on-road traffic, construction, and agriculture (5)(6)(7)(8). Atmospheric chemistry models used in air quality applications generally do not include anthropogenic fugitive dust, due to the lack of available emission inventories except for a few urban areas (9)(10)(11). Aside from its interest as an air quality issue, coarse PM can also affect PM2.5 by heterogeneously taking up acids (HNO3, SO2, and H2SO4). This uptake has been observed for natural dust events (12)(13)(14)(15)(16), but the more ubiquitous effect from anthropogenic dust has received little study (17,18). With increasingly stringent control measures to decrease fugitive dust pollution in East Asia (5,(19)(20)(21), it is pressing to better understand its impacts on PM2.5 air quality.
A specific issue is the effect of anthropogenic dust on PM2.5 nitrate. PM2.5 in winter is dominated by the nitrate component (22,23), and both winter and summer haze pollution events are driven by nitrate (24,25). PM2.5 over North China in winter has not been effectively decreasing despite reductions in emissions of the precursor nitrogen oxides (NOx ≡ NO + NO2) (23,24). Simulation with the GEOS-Chem model of fine particulate nitrate and HNO3 observations from the KORUS-AQ ground and aircraft campaign over South Korea in May-June 2016 previously showed a factor of two overestimate of observed PM1 nitrate (26,27). At the same time, coarse PM at KORUS-AQ surface sites averaged 21 μg m -3 while the GEOS-Chem model simulated only 3.5 μg m -3 from natural dust and sea salt sources (26). Ref (27) postulated a five-fold increase in the HNO3 dry deposition to explain the fine particulate nitrate overestimate in KORUS-AQ. Uptake of HNO3 by coarse PM could provide an alternative explanation. Ammonia (NH3) emissions have previously been found to be the most effective lever for PM2.5 nitrate control over North China in winter by altering the gas-particle partitioning of total nitrate (particulate nitrate + HNO3) and affecting the lifetime of total nitrate against deposition (23). NH3 is also important in summer in driving the partitioning of HNO3 to PM2.5 nitrate (25). Coarse PM would affect the HNO3-NH3 thermodynamics by providing a sink of HNO3, thus potentially affecting the sensitivity of PM2.5 nitrate to NH3 emissions. Better understanding this sensitivity is of crucial importance because of recent efforts by the Chinese government to decrease NH3 emissions (28).
In this work, we show that coarse PM decreased by 21% over the North China Plain (NCP) region and the Seoul Metropolitan Area (SMA) over the 2015-2019 period as a result of measures to control fugitive dust emissions. We also show from the abrupt drop of coarse PM in China observed during the COVID-19 lockdown and the strong daily correlations between coarse PM and carbon monoxide (CO) that coarse PM in urban China and South Korea is mainly anthropogenic. We find that coarse PM largely resolves the previous GEOS-Chem model overestimate of PM1 nitrate in the SMA during the KORUS-AQ campaign. We infer from model simulations that coarse PM has a major and highly seasonal effect on PM2.5 nitrate and on the sensitivity of PM2.5 nitrate to NH3 emission, with implications for the impact of fugitive dust emission controls on PM2.5 nitrate levels.

Results and Discussion
Coarse PM in China and South Korea. Concentrations of coarse PM are available in China and South Korea from the dense air quality monitoring networks reporting PM2.5 and PM10 (29,30). Fig. 1 shows the annual mean concentrations in 2015, 2019, and 2020. Concentrations in each country are highest in the NCP and in the SMA, indicating a dominant urban anthropogenic origin. Concentrations decreased by 21% in both regions from 2015 to 2019, reflecting efforts to control fugitive dust pollution (5,20,31,32). They further decreased sharply in 2020 because of the COVID-19 slowdown that curtailed traffic and construction. Concentrations before and during the sharp COVID-19 lockdown in China starting on January 24, 2020, show a 40-50% drop (SI Appendix, Fig. S1).
The spring maxima in the NCP and SMA are due to natural desert dust events superimposed on a high anthropogenic baseline. It is well known that the influence of natural dust in East Asia peaks in spring (33). Fig. 2 shows the daily correlations of regional mean coarse PM and carbon monoxide (CO, a tracer of incomplete combustion) over the NCP and SMA in 2015 separated by seasons. Correlations are strong except for spring, and then mainly because of high coarse PM outliers attributable to events of long-range transport of natural dust. Taking CO as a tracer of urban air, this further implies that coarse PM is dominantly urban in origin. Similar correlations are found in other years (SI Appendix, Fig. S2). Observed coarse PM decreases from 2015 to 2019 in the NCP are 32% in winter and 40% in summer, and in the SMA are 38% in winter and 40% in summer, when the natural contribution is minimum (SI Appendix, Fig. S3).
Effect of anthropogenic coarse PM on fine particulate nitrate. Coarse PM in East Asia is alkaline (34,35) and can therefore readily take up HNO3 to affect PM2.5 nitrate. We included anthropogenic coarse PM in the GEOS-Chem model by using the air quality network observations as boundary conditions in the lowest model level (Materials and Methods). The resulting model provides a simulation of coarse PM vertical profiles consistent with KORUS-AQ aircraft observations over the SMA (SI Appendix, Fig. S4). We simulate alkalinity and the uptake of acids (HNO3, SO2, and H2SO4) by coarse PM on the basis of laboratory data and ambient dust composition observations in East Asia (34)(35)(36)(37) as detailed in Materials and Methods. We find that dust uptake of SO2 and H2SO4 does not affect sulfate significantly during KORUS-AQ (SI Appendix, Fig. S5). We therefore focus our analysis on the effect on nitrate.
We start by applying GEOS-Chem to the simulation of observations during the KORUS-AQ campaign over South Korea (May 1 to June 10, 2016), including continuous gas and aerosol concentrations at surface sites and vertical profiles from 20 flights all in daytime (38). Nitrate was measured at the Korea Institute of Science and Technology (KIST) surface site and on the aircraft by Aerosol Mass Spectrometers (AMSs) with size cut of 1 μm diameter (PM1 nitrate) (39,40), and also on the aircraft by the Soluble Acidic Gases and Aerosol (SAGA) instrument with size cut of 4 μm diameter (PM4 nitrate) (41,42). The PM1 nitrate measured by the AMS during KORUS-AQ was mainly inorganic (SI Appendix, Fig. S6). Here we take ammonium nitrate in the model for comparison to PM1 observations, and ammonium nitrate plus size-resolved dust nitrate for comparison to PM4 observations. Fig. 3 shows SMA observations during KORUS-AQ and their simulation by GEOS-Chem including median PM1 nitrate diurnal variations at the KIST surface site and median vertical profiles of PM1 nitrate, PM1-4 nitrate (between 1 and 4 μm), PM4 Ca 2+ (a tracer of dust), and HNO3. PM1 nitrate at the KIST site increases over the course of the night, which is attributed in the model to reactive uptake of NO2 and N2O5 by aqueous-phase particles and decreases in the morning due to dilution by mixed layer growth (27). Here we adjusted the diurnal variation of NH3 emission in the model to match the NH3 observations made at the Olympic Park site, 7 km southeast of KIST (Materials and Methods, SI Appendix, Fig. S7). We find that the standard GEOS-Chem simulation without anthropogenic coarse PM overestimates the observed PM1 nitrate and HNO3 concentrations by about a factor of two (by about a factor of 3-4 for PM1 nitrate during the night), and underestimates PM1-4 nitrate and PM4 Ca 2+ by about a factor of two. Adding anthropogenic coarse PM to the model largely corrects model biases for daytime PM1 nitrate, PM1-4 nitrate, and PM4 Ca 2+ and corrects half of the overestimate of HNO3 and nighttime PM1 nitrate. One reason for the remaining model overestimate of PM1 nitrate at the KIST site could be that a large fraction of ammonium nitrate has diameter larger than 1 µm and is not detected by AMS (40). We find that coarse PM takes up HNO3 three times faster than dry deposition and that this uptake is not limited by alkalinity (only 60% of the dust alkalinity in surface air is neutralized on average).
Implications for the trends of fine particulate nitrate and the response to emission controls. PM2.5 nitrate observations in the North China Plain in winter show either no or marginal decreases since 2015 (24,43). No observations are available outside of winter. The Multiresolution Emission Inventory for China (MEIC) reports that NCP emissions of NOx decreased by 16% from 2015 to 2019, while SO2 emissions decreased by 58% (44). Efforts to control agricultural NH3 emissions in China began in 2015 (28). The MEIC reports a 15% decrease in NH3 emission over China from 2015 to 2019 (22% for the NCP), while the PKU-NH3 emission inventory reports a 6% decrease from 2015 to 2018 (28). Emissions of primary PM2.5 from combustion also decreased by 35% in winter and by 20% in summer from 2015 to 2019. As previously mentioned, coarse PM decreased by 32% in winter and 40% in summer.
Here we use the GEOS-Chem model driven by the MEIC emission inventory and observed coarse PM to investigate the factors controlling the PM2.5 nitrate trend in the NCP over 2015-2019. The model simulates a 27% decrease of January-February-March mean fine particulate nitrate in downtown Beijing from 2015 to 2019, comparable to the 22% decrease in concurrent observations (43). By contrast, the mean 2015-2019 January-February-March PM2.5 nitrate over the NCP in GEOS-Chem increased by 5%, with changes in meteorology as an important driver (SI Appendix, Fig. S8).  Table S1). The weak trend in winter reflects the effect of decreasing SO2 emissions offsetting the benefits of decreasing NOx and NH3 emissions. The 16% reduction of NOx emission decreases PM2.5 nitrate only by 3%, consistent with our previous finding (23). The 22% reduction of NH3 emission decreases PM2.5 nitrate by 6% if not accounting for coarse PM, again consistent with our previous finding. Accounting for coarse PM, the 22% reduction of NH3 emission decreases PM2.5 nitrate by 9%. The 35% decrease of primary PM2.5 from combustion decreases PM2.5 nitrate by 6.6%, mainly due to less NO2 hydrolysis on aerosol surfaces. Coarse PM has little direct effect on PM2.5 nitrate because the thermodynamics of ammonium nitrate formation drives HNO3 into fine particles in winter (SI Appendix, Fig. S9), but it increases the sensitivity of PM2.5 nitrate to NH3 and SO2 emissions because it accelerates the loss rate of HNO3 and thus amplifies the effect of NH3 on total nitrate lifetime (23).
In summer, we find in the model that the 40% decrease of coarse PM over 2015-2019 increases summertime PM2.5 nitrate by 20%, offsetting the effectiveness of NOx and NH3 emission controls. This is because a large fraction of total nitrate in summer remains in the gas phase as HNO3 under the warm temperatures, and with the decrease of coarse PM, HNO3 scavenged by coarse PM is reduced (SI Appendix, Fig. S9). The 58% decrease in SO2 emission reduces PM2.5 nitrate by 12% because less sulfate makes HNO3 less likely to partition to the aerosol phase (45).
In summary, we find that coarse PM (PM10 -PM2. 5 shows that anthropogenic coarse PM largely resolves the previous large model overestimates of PM1 nitrate. We find in the model that the summertime decrease of coarse PM in the NCP from 2015 to 2019 increased fine particulate nitrate by 20%, offsetting the effectiveness of NOx and NH3 emission reductions. In winter, the decrease of coarse PM has little direct effect on fine particulate nitrate trends but increases the effectiveness of NH3 emission reductions. As coarse PM continues to decrease in response to fugitive dust pollution control, there is a greater need to reduce NH3 and NOx emissions in order to decrease fine particulate nitrate air pollution in East Asia.

GEOS-Chem model description.
We use the GEOS-Chem global atmospheric chemistry model version 13.0.2 (https://zenodo.org/record/4681204) in a nested-grid simulation over East Asia (100 -150° E, 20 -50° N) with a horizontal resolution of 0.5°× 0.625°. The GEOS-Chem model simulates detailed oxidant-aerosol chemistry (46)(47)(48)(49)(50) and is driven by meteorological data from the NASA Modern-Era Retrospective Analysis for Research and Applications, Version 2 (MERRA-2). Dry deposition of gases and particles follows a standard resistance-in-series scheme (51). Wet deposition of gases and particles includes contributions from rainout, washout, and scavenging in convective updrafts (52)(53)(54)(55), with recent updates featuring faster below-cloud scavenging of HNO3 (56). We conducted a number of sensitivity simulations as listed in SI Appendix, Table S1 to investigate the model sensitivity of 2015-2019 PM2.5 nitrate trends to decreases in individual emissions of NOx, SO2, NH3, primary PM2.5, and coarse PM. PM1 nitrate is diagnosed in the model by the ammonium nitrate component, while PM2.5 nitrate is diagnosed as the sum of the ammonium nitrate and fine (PM2.5) dust nitrate components.

Coarse PM simulation.
We simulate coarse PM in GEOS-Chem by using 24-hour average coarse PM observations from the national networks in China and South Korea as boundary conditions at the lowest model level. For this purpose, we linearly interpolate the daily network data to the GEOS-Chem model grid and apply them to the coarse dust GEOS-Chem model component (57) with an effective diameter of 4.8 µm. Dust alkalinity is included in these boundary conditions with 7.1% Ca 2+ and 1.1% Mg 2+ as carbonates by dry mass (34,35,58). We include reactive uptake of HNO3 and SO2 on dust limited by dust alkalinity and mass transfer, and uptake of H2SO4 limited by competition with other aerosol surfaces (59). The relative humidity (RH)dependent reactive uptake coefficients (γ) of HNO3 and SO2 are based on laboratory studies (36,37) and constrained by observations of coarse and fine particulate nitrate and sulfate observations during natural dust events in Beijing (12,60). γ(HNO3) increases from 0.06 to 0.21 as RH increases from 40% to 80%, and γ(SO2) increases from 7.0x10 -5 to 4x10 -4 .

Emissions. Monthly anthropogenic emissions for China are from the Multi-resolution Emission
Inventory for China (MEIC) (44,61,62), and emissions for other Asian countries including South Korea are from the KORUSv5 inventory (63). We adjusted the diurnal scaling factors of the NH3 emissions in both inventories based on our simulation of the Olympic Park NH3 observations during KORUS-AQ (SI Appendix, Fig. S7). MEIC and KORUSv5 including primary anthropogenic PM2.5 emissions but solely from combustion. According to MEIC, NOx emissions in the North China Plain decreased by 16% from 2015 to 2019, SO2 emission decreased by 58%, NH3 emission decreased by 22%, and primary PM2.5 emission from combustion decreased by 35% in winter and by 20% in summer (44). To this we add here a 32% decrease in coarse PM in winter and 40% decrease in summer in the North China Plain region.   The correlation coefficients in spring are not shown because they are insignificant. Here spring is defined as February-March-April-May, summer as June-July-August, Autumn as September-October-November, and winter as December and January. We include February in spring to cover the season of natural dust events (34). We have excluded observations from two process-directed flights (RF7 and RF8) and the Daesan power plant plume following ref. (30). PM1 nitrate refers to particles smaller than 1 μm as measured by the Aerosol Mass Spectrometer (AMS), and PM1-4 nitrate refers to particles between 1 and 4 μm diameter as measured by the difference between the Soluble Acidic Gases and Aerosol (SAGA) instrument and the AMS. GEOS-Chem simulates PM1 nitrate by ammoniumassociated nitrate, PM1-4 nitrate by dust-associated nitrate with a diameter less than 4 µm, and PM4 Ca 2+ by 7.1% mass of dust with a diameter less than 4 µm.  Table S1) and are shown as seasonal means for winter and summer. Results from simulations not accounting for the effect of HNO3 uptake by anthropogenic coarse PM are shown as "+" symbols.      averaged over the NCP. Particulate nitrate is partitioned to nitrate that is respectively associated with ammonium, fine dust (PM2.5), and coarse dust.