Role of Dust and Iron Solubility in Sulfate Formation during the Long-Range Transport in East Asia Evidenced by 17O-Excess Signatures.

Numerical models have been developed to elucidate air pollution caused by sulfate aerosols (SO42-). However, typical models generally underestimate SO42-, and oxidation processes have not been validated. This study improves the modeling of SO42- formation processes using the mass-independent oxygen isotopic composition [17O-excess; Δ17O(SO42-)], which reflects pathways from sulfur dioxide (SO2) to SO42-, at the background site in Japan throughout 2015. The standard setting in the Community Multiscale Air Quality (CMAQ) model captured SO42- concentration, whereas Δ17O(SO42-) was underestimated, suggesting that oxidation processes were not correctly represented. The dust inline calculation improved Δ17O(SO42-) because dust-derived increases in cloud-water pH promoted acidity-driven SO42- production, but Δ17O(SO42-) was still overestimated during winter as a result. Increasing solubilities of the transition-metal ions, such as iron, which are a highly uncertain modeling parameter, decreased the overestimated Δ17O(SO42-) in winter. Thus, dust and high metal solubility are essential factors for SO42- formation in the region downstream of China. It was estimated that the remaining mismatch of Δ17O(SO42-) between the observation and model can be explained by the proposed SO42- formation mechanisms in Chinese pollution. These accurately modeled SO42- formation mechanisms validated by Δ17O(SO42-) will contribute to emission regulation strategies required for better air quality and precise climate change predictions over East Asia.

Introduction

The formation of sulfate aerosols (SO42–) in East Asia is a concern because it causes severe haze and pollution events, particularly in China, resulting in low visibility[1] and public health problems.[2] SO42– is also one of the important short-lived climate forcers (SLCFs) related to regional and global climate changes.[3,4] Atmospheric SO42– is produced mainly via the oxidation of sulfur dioxide (SO2),[5] and SO2 emissions from China accounted for approximately a quarter of global total emissions.[6] The problem of SO42– pollution in East Asia is not confined to the emission sources in China because of the relatively long lifetime of SO42– and the westerlies in the mid-latitudes.[7] Thus, the long-range transport of SO42– from the source region on the Asian continent to the downstream region of the air masses (i.e., Republic of Korea and Japan) has been thoroughly investigated.[8−10]

To elucidate pollution caused by atmospheric SO42–, chemical transport models (CTMs) have been developed and used in various studies. In CTMs, the gas-phase oxidation by the hydroxyl (OH) radical and the aqueous-phase oxidations by hydrogen peroxide (H2O2), ozone (O3), and O2 catalyzed by transition-metal ions (TMIs) for SO42– formation are typically considered. However, typical CTMs often underestimate the burden of atmospheric SO42–, particularly in China,[11,12] suggesting that some SO42– formation in the atmosphere is missing. To date, heterogeneous SO42– production mechanisms particularly for aerosol surfaces[13−16] have been proposed to explain this missing formation such as the enhanced role of the oxidation by reactive nitrogen,[17−20] H2O2,[21] TMI-catalyzed O2,[22−24] and Mn-catalyzed oxidation.[25,26] Nonetheless, the observational evidence has not yet identified a specific mechanism. For example, while the importance of NO2 for SO42– formation within the aerosol surface was proposed,[17] follow-up studies have cast doubt on the impact of its reaction.[25,26] Thus, these proposed reactions have been highly controversial arguments.

These pollution and haze events over China and their impact on the downstream region should be mitigated, mainly via the reduction of anthropogenic SO2 emissions. However, according to reports in Western countries, atmospheric SO42– has declined less rapidly than would be expected from decreases in SO2 emissions.[27,28] This unknown response has been attributed to chemical feedback mechanisms of a weakening H2O2 limitation on the S(IV) + H2O2 pathway[29] and acidity-driven enhancement of the S(IV) + O3 pathway under low SO2 conditions[30,31] in Western countries. Therefore, accurate CTMs for both concentration and oxidation processes are required to establish an effective emission regulation strategy for improving air quality, and the accurate implementations of atmospheric SO42– formation mechanisms in CTMs are essential for elucidating present pollution and predicting future air quality. Yet, the atmospheric SO42– formation pathways implemented in CTMs have been simply evaluated by total SO42– masses and not validated by independent observational evidence.

One proven method to validate atmospheric SO42– formation is the mass-independent oxygen isotopic composition (Δ17O)[32] of SO42–, which reflects the formation pathway from SO2 to SO42–. The comparison of observed and modeled Δ17O(SO42–) has enabled the recognition and quantification of SO42– formation mechanisms, including TMI-catalyzed S(IV) + O2,[33] SO42– formation by O3 oxidation on sea salt aerosol,[34] S(IV) oxidation by hypohalous acids,[35,36] and acidity-driven changes in SO42– formation pathways.[31,37] For extreme pollution events in China, the importance of the heterogeneous chemistry of SO42– production, which is not considered in typical CTMs, has been discussed based on Δ17O(SO42–);[13,14] however, verification has been limited in China. Studies to reveal the SO42– formation process in the long-range transport over East Asia using Δ17O(SO42–) have been limited.

SO2 emissions in East Asia have recently begun to decrease.[38] Without the validation of SO42– formation processes in East Asia, studies using current typical CTMs may not adequately predict air quality and climate change, given the possible cause of chemical feedback mechanisms, which have been studied in Western countries.[27,28] Therefore, the observation of Δ17O(SO42–) in regions downstream from intense emission sources, such as China, is critical for research into SO42–. The present study describes the annual observation of SO42– and Δ17O(SO42–) in 2015 at the background site in Japan (NOTO Ground-based Research Observatory [NOTOGRO] located over the downstream region of China) (Figure ). Regional CTM analyses over East Asia are conducted with the Community Multiscale Air Quality (CMAQ) model with three different configurations (Exps. A, B, and C). The role of mineral dust as a supply of alkaline material,[20,39,40] which is a unique aspect of the East Asian environment, is investigated because dust-driven higher pH alters acidity-dependent aqueous-phase SO42– formation, such as TMI-catalyzed oxidation by O2 and oxidation by O3. In addition, the impact of TMIs on SO42– formation is examined because these are uncertain parameters for emissions and solubilities in the model.[41] Through comparisons of observed and modeled SO42– and Δ17O(SO42–) at NOTOGRO in Japan, this study clarifies the key factors controlling the formation process of long-range-transported SO42– in East Asia.

Figure 1

SO2 emissions compiled in this study plotted over the modeling domain, and the location of the NOTOGRO observation site.

Methods

Observation of Aerosol and Oxygen Isotopic Composition in Japan

The atmospheric observations were performed at NOTOGRO at 37.45°N, 137.36°E (Figure ). The Noto peninsula extends from the west coast of mainland Japan approximately 150 km into the sea of Japan, and NOTOGRO is located on the tip of this peninsula. The geographical location of NOTOGRO is ideal for capturing the atmospheric variation in East Asia because it is surrounded by the sea and isolated from major pollution sources in Japan.[42]

The aerosol samples were collected by a high-volume air sampler (MODEL-120 SL, Kimoto Co., Ltd., Japan) mounted on the rooftop (∼10 m above sea level) of NOTOGRO. Fine (<2.5 μm) and coarse (>2.5 μm) samples were collected on prebaked (450 °C for 4 h) quartz filters (2500QAT-UP, Pall Co., Ltd.; TE-230-QZ, Tisch Environmental Inc.). Sampling was performed at a flow rate of ∼1.05 m3/min, and the sampling interval was usually 1 or 2 weeks. After sampling, the filters were wrapped in aluminum foil, sealed in polyethylene bags, and stored in a clean freezer at −20 °C prior to the measurement at the Tokyo Institute of Technology, Japan. In the laboratory, half of each filter was soaked in ultrapure water (30 mL) in a 50 mL centrifuge tube (Centricon plus-70, Millipore). The sample solution was separated from the insoluble materials and the filter by centrifuging in a centrifugal filter unit for 10 min. This method can recover >98% of the initial water volume. The major anions were quantified using an ion chromatograph (Dionex ICS-2100, Thermo Fisher Scientific) with a guard column (Dionex IonPac AG19, Thermo Fisher Scientific) and a separation column (Dionex IonPac AS19, Thermo Fisher Scientific). The major cations were quantified using an ion chromatograph (881 Compact IC Pro, Metrohm, Switzerland) with a guard column (Metrosep C4 S-Guard/4.0, Metrohm) and a separation column (Metrosep C4-150/4.0, Metrohm). The uncertainties of measurement errors were estimated by analyzing five different concentration standards at intervals of every ∼20 sample measurements, resulting in approximately 4% for both cation and anion measurements.

The measurement procedures for Δ17O(SO42–) are described in our previous studies.[31,37,43] Briefly, 1 or 2 μmol of H2SO4 separated by ion chromatography was chemically converted to Na2SO4, and 30% of H2O2 solution (1 mL) was added, and the mixture was dried. The Na2SO4 was converted to silver sulfate (Ag2SO4) using an ion-exchange resin. This Ag2SO4 powder was transported in a custom-made quartz cup, which was dropped into the furnace of a high-temperature conversion elemental analyzer (TC/EA, Thermo Fisher Scientific) at 1000 °C and thermally decomposed into O2. The O2 gas was introduced separately into an isotope ratio mass spectrometer to measure m/z = 32, 33, and 34. The Δ17O(SO42–) measurements were corrected using our working standard B (Δ17O(SO42–) = 2.4‰) with the same procedure described previously.[31,37,43] In this correction for isotopic analysis, SD (1σ) for the corrected values for standard B was 0.11‰ based on measurements for the samples collected in 2015, and this 1σ uncertainty was used for the error of the isotopic measurements in the present study. The raw observation data are presented in the Supporting Information (Table S1). All data for SO42– concentration and Δ17O(SO42–) are corrected using Na+ concentration to non-sea salt fraction of SO42– (nss-SO42–) in a similar manner reported previously.[31]

Regional Chemical Transport Modeling over East Asia

Model Description

The regional air quality modeling was conducted with the CMAQ model version 5.3.1.[44,45] In this study, the simulation domain covered the entirety of East Asia with a horizontal resolution of 36 km (Figure ) and 44 nonuniform layers from the surface to 50 hPa. One gas-phase reaction and five aqueous-phase reactions in cloud are involved in SO2 oxidation (i.e., SO42– formation) in the original CMAQ. The one gas-phase reaction is SO2 oxidation by OH (GAS), and the five oxidants in the aqueous-phase reactions in cloud are H2O2 (AQ(H2O2)), O3 (AQ(O3)), O2 catalyzed by TMIs (AQ(O2)), peroxyacetic acid (PAA) (AQ(PAA)), and methyl hydrogen peroxide (MHP) (AQ(MHP)). In addition to these five oxidants in the original CMAQ, taking into account the elevated NO2 concentration in Chinese pollutions,[17−19] the aqueous-phase pathway in cloud via NO2 was added in this study. This inclusion partly improved the model underestimation issue during winter in our previous study.[46] Note that the production of SO42– on the aerosol surface is not considered in this study because the specific Chinese haze events that occurred in limited areas are out of scope and our focus is on capturing SO42– pollution over East Asia from the viewpoint of the background site in Japan.

This version 5.3.1 of CMAQ has several features to achieve the purpose of this study. At first, Fe and Mn emissions can be treated as independent variables, although these emissions had been previously dependent on the total PM2.5 emission. Thus, we implemented the Transition Metal Inventory-Asia version 1.0[47] to calculate the emissions of Fe and Mn. Furthermore, although the default setting of CMAQ does not possess, we implemented the pH dependency rate constant for the calculation of AQ(O2) using the synergistic relationship between Fe and Mn developed in our previous study.[46] These explicit treatments of Fe and Mn concentrations are major advantages of CMAQ and our study because other CTMs, such as GEOS-Chem, treat TMI concentrations as a fraction of the PM2.5 concentration. Because of these improvements for Fe and Mn emissions and the rate constant for the calculation for AQ(O2), we tested the remaining uncertainty of solubilities of Fe and Mn, as described in Section .

Second, it is possible to implement the developed physics-based inline dust calculation after CMAQ version 5.2. Before this version, the effect of soil dust for neutralization and altering cloud drop pH could not be considered, even though one of the characteristics in East Asia is the role of mineral dust originating from the Taklamakan and Gobi Deserts, the Loess Plateau, and Inner Mongolia. However, dust is not implemented as the default setting of CMAQ, which hampers precise calculation for the pH-dependent SO42– formation pathways, especially for AQ(O3) and AQ(O2). The impact on SO42– formation caused by the difference with or without implementation of inline dust calculation was tested in this study, as explained in Section . The details of the modeling description and chemical configurations are, respectively, given in the Supporting Information (Sections S1 and S2).

Model Experiments

In this study, we compared the following three experiments (Exp. A, Exp. B, and Exp. C) to investigate the role of mineral dust and solubilities of Fe and Mn for SO42– formation, as summarized in Table . In Exp. A, which is the original settings in the CMAQ model version 5.3.1, the solubilities of anthropogenic Fe and Mn were set as 10 and 50% and dust inline calculation was not implemented (Table ).

Table 1

Modeling Experimental Design and Settings of Solubility for Anthropogenic and Dust TMIsa

		Fe		Mn
modeling experiments	design	anthropogenic (%)	dust (%)	anthropogenic (%)	dust (%)
Exp. A	the standard simulation	10	-	50	-
Exp. B	incorporation of dust inline calculation	10	1	50	50
Exp. C	same as Exp. B, but increasing TMI solubilities	54	3	97	50

Note: dash means dust simulation was not implemented in Exp. A. In the aqueous-phase reaction of O2 catalyzed by TMIs (AQ(O2)), Fe(III), and Mn(II) are related. Fe(III) was assumed to be 10% of the total dissolved Fe during the day and 90% at night as the diurnal variation. Mn(II) was assumed to be the same for all dissolved Mn. These diurnal variations of Fe(III) and Mn(II) were the same in all three experiments.

In Exp. B, the modulation of pH by mineral dust was considered, and thus the newly developed dust inline calculation scheme in CMAQ[48] was applied. In addition to pH changes by mineral dust, the dust-derived Fe and Mn were considered in Exp. B. The solubility of dust Fe and Mn was set as 1 and 50%, respectively[33] (Table ). The details of the dust inline calculation are given in the Supporting Information (Section S3).

In Exp. C, increased TMI solubilities were considered because one of the uncertainties in modeling settings still under debate is TMI solubilities. Based on a literature review, the range of TMI solubility for Fe and Mn are 0.03–54 and 1.2–97%, respectively.[14] The solubilities of anthropogenic Fe and Mn were set at 54 and 97%, respectively, for the maximum possible production through TMI processes (Table ). Based on the integrated massively parallel atmospheric chemical transport (IMPACT) global aerosol model[41] (Figures S1 and S2 in the Supporting Information), the solubility of dust Fe was increased from 1% in Exp. B to 3% in Exp. C (Table ). The result from the IMPACT global aerosol model also suggested the verification of increasing anthropogenic Fe solubility (Figures S3 and S4 in the Supporting Information). The discussion of TMI solubilities is also given in the Supporting Information (Section S4).

Calculation of Δ17O(SO42–) from Model Outputs

The diagnostic tool of the sulfur tracking method in CMAQ was used to output each oxidation process involved in SO42– formation. From these outputs in CMAQ, the modeled Δ17O(SO42–) was derived based on eqs and 2.In these equations, [SO42–]i is the SO42– concentration from each process and Fi represents the fractional contribution for each process, where i indicates the SO42– formation process. The contributions of PAA and MHP were negligible (around 0.1% contribution throughout the year) and hence were omitted from this calculation. In this study, Δ17O(SO42–) of 0‰ was set for the GAS and emission oxidation pathways,[49] and the end members and their uncertainties for Δ17O(SO42–) in the other oxidation pathways were set to 0.8 ± 0.2‰ for AQ(H2O2), 6.4 ± 0.3‰ for AQ(O3), and −0.1‰ for AQ(O2) (see our previous studies[31,37,43] for the determination of the end members). The value of AQ(NO2) was set as 0‰ according to He et al.[13] and references therein. Δ17O(SO42–) = 0‰ for SO42– produced by AQ(NO2) is expected based on the following three mechanisms: radical chain mechanism,[50] oxygen-atom transfer from OH–,[51] or from O2.[52] Additionally, a detailed discussion of the uncertainties of the AQ(O3) end member is given in the Supporting Information (Section S5 and Figure S5). The SO42– derived from the boundary conditions is considered the background existing SO42–, and thus monthly values observed at Alert, Canada, ranging from 0.5 to 1.3‰[33] were used.

Statistical Analysis

The model performance for SO42– concentration and calculated Δ17O(SO42–) was statistically evaluated. The metrics were correlation coefficient (R), normalized mean bias (NMB), and normalized mean error (NME).Here, N is the total number of paired observations (O) and models (M), and these averages are denoted as O̅ and M̅, respectively. The recommended metrics for SO42– concentration are model performance goals for best performance of R > 0.7, NMB < ±10%, and NME < +35% and model performance criteria for acceptable performance of R > 0.4, NMB < ±30%, and NME < +50%.[53]

Results and Discussion

Observed and Simulated SO42– Characteristics over East Asia

The simulated spatial distribution of SO42– over East Asia was divided into three seasons, late winter to spring (February–May), summer (June–August), and autumn to winter (September–December), to characterize seasonal behaviors (Figure ). The highest SO42– concentrations were found over mainland China, and the higher-concentration regions extended into the downstream region over the Korean Peninsula and Japan. This feature was dominant from spring to summer; therefore, the observations in the downstream region of Japan detected the polluted air mass resulting from the long-range transport over East Asia. In contrast, from autumn to winter when the strong northwesterly wind field by the Asian monsoon was dominated, the SO42– concentration was low and was characterized by clean background conditions. The weekly and monthly average variations of SO42– concentration observed at NOTOGRO also showed higher concentrations from spring to summer and lower concentrations during autumn to winter (Figure a). As the background site in Japan, Chichijima island located in the western North Pacific (i.e., south of Tokyo) showed higher concentration in winter and lower concentration in summer.[54] The difference in the seasonal variation of SO42– concentration seen in NOTGRO is caused by the outflow pattern in East Asia, as found in Figure . Overall, throughout the year, the statistical analyses showed that all three modeling experiments (Exps. A, B, and C) generally captured the SO42– concentration (Table ).

Figure 2

Spatial distribution of simulated SO42– with wind field over East Asia during late winter to spring (February–May), summer (June–August), and autumn to winter (September–December).

Figure 3

Observed and simulated (a) SO42– concentration and (b) Δ17O(SO42–) from February to December 2015. Simulated SO42– formation in (c) Exp. A, (d) Exp. B, and (e) Exp. C. Each oxidation process is normalized to the simulated SO42– concentration and is shown as relative percentages. The values for this figure are listed in the Supporting Information (Tables S2–S4).

Table 2

Statistical Analysis of Model Performance for the Monthly Average of SO42– and the Monthly Weighted Average of Δ17O(SO42–)a

components	SO42–			Δ17O(SO42–)
metrics	Exp. A	Exp. B	Exp. C	Exp. A	Exp. B	Exp. C
mean (obs.)	3.80			1.20
mean (model)	3.43	3.26	3.30	0.57	1.21	1.03
R	0.67	0.79	0.79	0.23	0.47	0.48
NMB (%)	–12.42	–15.08	–13.86	–47.70	+7.13	–9.69
NME (%)	+28.63	+24.34	+23.59	+49.24	+34.38	+24.95

Note: units of mean are μg/m3 for SO42– and ‰ for Δ17O(SO42–), respectively. The score improvement compared with Exp. A is shown in bold font.

The weekly observed Δ17O(SO42–) at NOTOGRO ranged from 0.46 to 1.98‰, and the monthly weighted average ranged from 1.0 to 1.5‰, except for a value of 0.70‰ in August (Figure b). These observed Δ17O(SO42–) values ranging from 1.0 to 1.5‰ were higher than those reported in the polluted region in China, where the observations from Beijing varied from 0.1 to 1.6‰ with a mean of 0.9 ± 0.3‰[13] and the data from Wuhan varied from 0.14 to 1.02‰ with a time-weighted average of 0.53‰.[23] The observed higher Δ17O(SO42–) values at NOTOGRO are explained by either the greater importance of AQ(O3) (Δ17O(SO42–) = 6.4 ± 0.3‰) or the lesser importance of pathways such as GAS (Δ17O(SO42–) = 0‰) and AQ(O2) (Δ17O(SO42–) = −0.1‰). We note that the low Δ17O(SO42–) values with high SO42– concentration in August probably originated from volcanic eruptions in western Japan, which is discussed in Section .

Disagreement between Observed and Modeled Δ17O in the Standard Model

In the standard CMAQ model experiment (Exp. A), although the modeled SO42– concentrations agreed with the observations (Figure a, black line), the modeled Δ17O(SO42–) ranged from 0.21 to 1.24‰ and underestimated the observations over the period except in winter (Figure b, black line). In Exp. A, GAS was the dominant oxidation process, contributing more than 30% of SO42– formation during the period and reaching up to 80% during spring and summer. Following the GAS process, AQ(H2O2) was the next most important process, contributing 10–30% of SO42– formation (Figure c). In the model, a higher SO42– concentration with lower Δ17O(SO42–) was found in August, and the low Δ17O(SO42–) was due mainly to the high contribution of the GAS process (Δ17O(SO42–) = 0‰). The importance of GAS in volcano-originated SO42– was suggested in this study, whereas the importance of AQ(O2) was suggested to be the dominant process in volcanic plumes.[55] This difference indicates that the oxidation processes contributing to SO42– formation inside a plume are different from those contributing to SO2 oxidation after diffusion into the atmosphere. The discussion of this volcanic impact is given in the Supporting Information (Section S6 and Figure S6). Overall, the underestimation of Δ17O(SO42–) in Exp. A indicates that the standard model parameterization in CMAQ is missing oxidation pathways that increase Δ17O(SO42–).

Importance of Dust-Derived pH Increase in SO42– Formation

Given the importance of dust in SO42– formation and neutralization,[22,37,56] the dust inline calculation was implemented in Exp. B. The supply of alkaline dust to the atmosphere increases the pH and thus increases the acidity-dependent reaction rates for AQ(O2) and AQ(O3).[5] In Exp. B, the dust-derived Fe and Mn were calculated in addition to anthropogenic Fe and Mn. Although Exp. B did not improve the estimation of SO42– concentration substantially compared with Exp. A (Figure a and Table ), Δ17O(SO42–) for Exp. B ranged from 0.56 to 2.36‰ (Figure b, brown line; Table ). For the relative contributions of the SO2 oxidation processes in the model, AQ(O3) increased, whereas GAS decreased (Figure d). This switch can be explained by the neutralization of atmospheric acidity by dust-derived CaCO3, and the increased pH obtained by including the inline dust calculation is consistent with previous works.[22,40] The results in Exp. B showed better agreement with the observations of Δ17O(SO42–) (Table ), and this improvement in Δ17O(SO42–) in Exp. B suggests that mineral dust supply is crucial in increasing the acidity-driven SO42– formation in East Asia. However, the overestimation of Δ17O(SO42–) from February to March and from October to December remained.

Importance of TMI Solubilities in SO42– Formation

To account for the remaining gap between the model and observed values of Δ17O(SO42–), the increases in TMI solubilities in the polluted air were considered in Exp. C. In the standard model of Exp. A, although the anthropogenic emission of TMIs is accurately considered based on the latest emission data set, the modeled estimates of solubilities of anthropogenic and dust Fe and Mn remain highly uncertain in cloud water, as pointed out previously.[14,29,31,33] The increased Fe solubilities have been indicated based on global model comparisons with multiple field campaigns over the Northwest Pacific[57] and also observed toward the downstream region of East Asia,[58,59] mainly because fine mineral aerosols can be acidified due to air pollution. Thus, for Exp. C, we implemented the higher solubilities for both anthropogenic and dust Fe. Compared to the anthropogenic Fe solubility (10%) and Mn solubility (50%) considered in Exps. A and B, the maximum solubility for anthropogenic Fe (54%) and Mn (97%) over the literature[14] was taken in Exp. C (Table ). Note that the higher solubility of Fe in dust (3%) considered in Exp. C is generally consistent with that simulated by the IMPACT global aerosol model[57] (Figures S1 and S2 in the Supporting Information).

The SO42– concentration in Exp. C did not increase substantially compared with that in Exps. A and B (Figure a and Table ), but the overestimation in modeled Δ17O(SO42–) from February to March was clearly decreased in Exp. C (Figure b, orange line; Table ) and ranged from 0.54 to 1.94‰. The improvement in Δ17O(SO42–) values was explained by the increased contribution of AQ(O2) owing to the higher solubilities of Fe and Mn (Figure e), especially from October to December. Regarding the importance of the AQ(O2) process, TMI concentration levels have been discussed in China. In Beijing, TMI-catalyzed oxidation showed a clear distinguishment of higher/lower contribution during polluted/clean periods.[14] In Wuhan, the enhanced role of TMI-catalyzed oxidation in winter was suggested due to higher PM2.5 concentration.[23] In contrast, this study highlights the role of TMI-catalyzed AQ(O2) through the consideration of solubilities over the downstream region of East Asia, as evidenced by Δ17O(SO42–). The importance of TMI solubilities, which led to the improvement of Δ17O(SO42–), was found during February and March when SO42– concentration was higher and from October to December when SO42– concentration was lower. Therefore, it is concluded that the role of TMI-catalyzed AQ(O2) by enhancing solubilities does not depend on the pollution level in this case. Overall, the increased pH obtained by including dust and the increase in TMI solubilities in Exp. C showed the best match for SO42– concentration and Δ17O(SO42–) among three experiments conducted in this study. The series of results strongly indicate the importance of dust and TMI solubilities for SO42– formation via changes in oxidation processes in downstream regions of East Asia, which was not discovered in previous studies that considered only SO42– mass.

Toward Closer Agreement between Observed and Modeled SO42– Formation

However, SO42– concentration was still underestimated from February to March (observed values: 3.58 ± 0.26 and 5.64 ± 1.71 μg/m3; modeled values in Exp. C: 2.61 and 3.66 μg/m3; Table S2 in the Supporting Information) and Δ17O(SO42–) values were overestimated, even in Exp. C (observed values: 1.24 ± 0.20 and 1.26 ± 0.33‰; modeled values in Exp. C: 1.94 ± 0.09 and 1.57 ± 0.08‰; Table S3 in the Supporting Information). Since the domestic contribution to SO42– in Japan is estimated to be small except in summer,[8,9] the most plausible reason for missing processes is attributed to inadequate SO42– formation in polluted areas over China. Here, Δ17O(SO42–) values of possible missing processes (defined as Δ17O(SO42–)missing) are calculated from the following mass-balance calculationwhere [SO42–]obs., [SO42–]model, and [SO42–]missing are SO42– concentrations for observation, model, and missing processes (i.e., difference between the observation and model), respectively, and Δ17O(SO42–)obs., Δ17O(SO42–)model, and Δ17O(SO42–)missing are Δ17O(SO42–) for the observation, model, and missing processes, respectively. Given that [SO42–]missing is calculated by [SO42–]obs. minus [SO42–]model, Δ17O(SO42–)missing can be obtained from eq . The calculated Δ17O(SO42–)missing was −0.63 ± 0.52‰ during February and 0.70 ± 1.01‰ during March, respectively.

For the calculated Δ17O(SO42–)missing of −0.63 ± 0.52‰ during February, the result can be explained by SO42– formation via TMI-catalyzed oxidation by O2, Mn-catalyzed oxidation, and other reactions (i.e., oxidation by NO2) having Δ17O(SO42–) close to 0‰. This interpretation is consistent with previous studies, which proposed these mechanisms as missing processes in Chinese haze.[17,18,24−26] It is worthy to conclude that AQ(H2O2) does not explain this missing SO42– formation during February.

Conversely, for the calculated Δ17O(SO42–)missing of 0.70 ± 1.01‰ during March, this value is close to Δ17O(SO42–) = 0.8 ± 0.2‰ for AQ(H2O2), although we could not fully exclude the possibilities of other reactions that have lower Δ17O(SO42–). In terms of the SO42– formation process, faster H2O2 oxidation of SO42– formation in high solute strength was suggested in Chinese haze events.[21] The current CTMs are based on kinetics research in dilute aqueous solutions and may miss such strengthened features in the atmosphere. The contribution of AQ(H2O2) was declined in Exp. C compared to that in Exp. A (Figure e,c) through increased dust-derived pH and enhanced TMI solubilities in this study. Because AQ(H2O2) does not depend on pH, faster oxidation by H2O2 in Chinese pollution would also affect the downstream region over East Asia even in the application of Exp. C. In this study, although we do not include reactions on the aerosol surface because our focus is not on the Chinese haze itself, the accurate modeling to capture the enhanced SO42– concentration in haze events will also improve SO42– behaviors in the downstream region if the long-range transport occurs.

To date, SO42– production mechanisms have been proposed to explain this missing formation in Chinese pollution and haze events, as introduced. Although this study cannot identify a single mechanism for this missing formation pathway, our results imply that the oxidation pathways for inadequate SO42– formation in polluted areas over China are not always identical. Based on our Δ17O(SO42–) approach, these estimations for missing processes from the perspective of the downstream region in East Asia are the first supportable information for unexplained SO42– formation over China. A recent study[26] proposed the dominance of Mn-catalyzed oxidation of SO2 and negligible formation pathways by gas- and aqueous-phase reactions in Chinese haze because the meteorological condition of the stable boundary layer with weak turbulence prohibits their formations. Under such conditions dominated by the Mn-catalyzed oxidation pathway, the value of Δ17O(SO42–) is expected to be close to 0‰; however, the reported value of Δ17O(SO42–) in Beijing haze from October 2014 to January 2015 ranged from 0.1 to 1.6‰ with a mean of 0.9 ± 0.3‰.[13] Therefore, the dominance of Mn-catalyzed oxidation and the negligible contributions from other processes will be required to be carefully examined in terms of the validation with Δ17O(SO42–) in the future study. To validate SO42– formation over East Asia covering from the haze above intense emission sources to a background condition over the downstream region, furthermore studies including laboratory experiments, measurements of both SO42– concentration and Δ17O(SO42–), and numerical modeling are needed.

Implications for Future Air Quality and Climate Studies

We demonstrated the effect of the dust-derived increase in pH on SO42– production and the need to increase TMI solubilities in the modeling by constraining both SO42– concentration and Δ17O(SO42–) at the background site in Japan over the downstream region. On the current typical CTMs, TMI solubilities are fixed as constant, which does not account for spatial (both horizontal and vertical) and temporal variations particularly found in upwind and downstream differences over East Asia (e.g., Figures S1–S4 in the Supporting Information). This study highlights the importance of TMI solubilities to determine the role of the SO42– formation process. Compared to the parameterization in this study, the assumptions of solubilities for dust Fe (0.45%) and Mn (5%) for the previous study conducting the modeling of Δ17O(SO42–) in Beijing haze[14] were significantly lower. This study improved SO42– formation in the model by implementing detailed emissions of Fe and Mn, pH-dependent rate constants for AQ(O2) catalyzed by TMIs, and increasing in solubilities of Fe and Mn; however, solubilities were taken as fixed parameters in time and space; hence, the detailed spatiotemporal variations of solubilities have not been accurately investigated. Thus, we propose considering spatiotemporal variations of Fe and Mn solubilities for the CTMs. To evaluate this development of CTMs and understand SO42– formation processes, the coordinated observation of Δ17O(SO42–) and TMI solubilities along trajectories of long-range-transported air mass from upwind to downstream in East Asia will be one significant approach. These accurately modeled SO42– formation processes over East Asia are necessary to build SO2 emission regulation strategies along with carbon neutrality[60] because the unknown response to SO2 emission reduction has already been reported in Western countries.[27,28]

The changes in SO42– oxidation processes also alter the size distribution of SO42– and hence direct and indirect radiative forcing,[61,62] closely related to climate aspects. The reduction of SO42– in future atmospheric conditions (i.e., higher CO2 concentration and lower SO2 concentration) will increase atmospheric warming compared with the current atmospheric conditions through the slow climate response.[63,64] The reduction of SO42– will also reduce atmospheric acidity and alter the magnitude, distribution, and deposition mode of nutrients supplied to the ocean in the coming decades.[65] As we verified the important role of Fe as the catalyst on the SO42– formation, the declined acidity in the future will relate to weakening the role of TMI-related SO42– formation in the downstream of dust sources over East Asia. Given that a significant emission reduction of CO2 combined with a well-designed emission pathway of SO2 is required,[66] our findings on the role of dust and TMI solubilities and the way to improve the modeling of SO42– formation will contribute to better emission regulations required for air quality and climate change.