Nitrate transboundary heavy pollution over East Asia in winter

High PM2.5 concentrations of around 100 μg/m were observed twice during an intensive observation campaign in January 2015 at Fukuoka (33.52°N, 130.47°E) in western Japan. These events were analyzed comprehensively within a regional chemical transport model and by synergetic ground-based observations with state-of-the-art measurement systems, 20 which can capture the behavior of secondary inorganic aerosols (SO4, NO3, and NH4). The first episode of high PM2.5 concentration was dominated by NO3 (type N), and the second episode by SO4 (type S). The concentration of NH4 (the counterion for SO4 and NO3) was high for both types. Sensitivity simulation in the chemical transport model showed that the dominant contribution was from transboundary air pollution for both types. To investigate the differences between these types further, the chemical transport model results were examined, and backward trajectory analysis was used to provide 25 additional information. During both types of episodes, high concentrations of NO3 were found above China, and an air mass that originated from northeast China reached Fukuoka. The travel time from the coastline of China to Fukuoka differed between types: 18 h for type N and 24 h for type S. The conversion ratio of SO2 to SO4 (Fs) was less than 0.1 for type N, but reached 0.3 for type S as the air mass approached Fukuoka. The higher Fs for type S was related to the higher relative humidity and concentration of HO2, which produces H2O2, the most effective oxidant for the aqueous-phase production of 30 SO4. Analyzing the gas ratio as an indicator of the sensitivity of NO3 to changes in SO4 and NH4 showed that the air mass over China was NH3-rich for type N, but almost NH3-neutral for type S. Thus, although the high concentration of NO3 above China gradually decreased during transport from China to Fukuoka, higher NO3 concentrations were maintained during transport owing to the lower SO4 for type N. In contrast, for type S, the production of SO4 led to decomposition of NH4NO3 and more SO4 was transported. Notably, the type-N transport pattern was limited to western Japan, and especially 35


Introduction
Particulate matter (PM) presents major environmental problems globally, especially in East Asia.A typical example is the episode of severe air pollution that occurred in January 2013 above China (e.g., Wang et al., 2014;Uno et al., 2014).During this episode, PM with aerodynamic diameters of less than 2.5 µm (PM 2.5 ) reached record-breaking concentrations of 772 µg m −3 on 12 January 2013 (Y.Pan al., 2016).Transboundary air pollution in downwind regions resulting from the severe air pollution in China is also an important environmental problem.For example, the possible long-range transport of PM 2.5 was based on the comparison of observations in metropolitan areas and remote islands in western Japan (Kaneyasu et al., 2014).They highlighted the dominant effect of the transboundary transport of sulfate (SO 2− 4 ) as a major PM 2.5 component in western Japan throughout most of the year.In spring, due to the prevailing westerly wind over East Asia, the transboundary air pollution of both aerosols and gases, e.g., carbon monoxide (CO) and ozone (O 3 ), has been thoroughly discussed (Itahashi et al., 2010(Itahashi et al., , 2013(Itahashi et al., , 2015;;Kanaya et al., 2016;Nagashima et al., 2010).In summer, the clean air mass from the oceans is moved over Japan by the southerly wind caused by the Pacific High; however, some studies have discussed the importance of transboundary air pollution from China over western Japan (Itahashi et al., 2012;Ikeda et al., 2014).Recently, 1-year source-receptor relationships for SO 2− 4 were evaluated, and China was identified as the main influence on downwind regions throughout the year with local sulfur dioxide (SO 2 ) emissions making an important contribution in summer (Itahashi et al., 2017).Compared with the analyses for spring and summer, transboundary air pollution events in winter are less well understood.
In this study, we also focused on nitrate (NO − 3 ), which is an important PM 2.5 component.The NO − 3 in PM 2.5 is produced via the reaction of gas-phase nitrate (nitric acid; HNO 3 ) and ammonia (NH 3 ), and this process is reversible.The reaction favors a shift toward the aerosol phase at low temperatures and high humidity (Seinfeld and Pandis, 2006).The simulated spatial distribution over East Asia showed the possible impact of transboundary NO − 3 pollution in winter over western Japan (Zhang et al., 2007;Ying, 2014).However, a quantitative evaluation over downwind regions was not presented in previous studies.This is partly because the model ability was not evaluated owing to the difficulty in measuring NO − 3 .Particulate NH 4 NO 3 may be volatilized after collection on the filter, either through an increase in the pressure drop across the particle-collecting medium or changes in the gas-aerosol equilibrium during sampling (Sickles II et al., 1999;Chang et al., 2000).This volatilization could occur even in winter because the temperature in the instrument shelter can be increased by heat from the pump.Therefore, the ground-based Acid Deposition Monitoring Network in East Asia uses the four-stage fil-ter pack method: NH 4 NO 3 is collected on the first filter and gas-phase HNO 3 and NH 3 are detected on the subsequent filters.The artifacts might not be significant; however, to avoid the possibility of volatilization, total nitrate (the sum of NO − 3 and HNO 3 ) has been used to evaluate the model ability in previous studies (e.g., Kajino et al., 2013).
To improve our understanding of the behavior of NO − 3 , accurate measurements and the evaluation of the model ability are needed.In this study, we used a state-of-the-art automated monitoring system for SO 2− 4 and NO − 3 , an aerosol chemical speciation analyzer (ACSA).This system measures SO 2− 4 and NO − 3 with high temporal resolution, and 1 h intervals were used in this study, minimizing the possibility of volatilization.In addition, the behavior of ammonium (NH + 4 ), which is the counterion for SO 2− 4 and NO − 3 , was captured by the well-validated NH x monitoring system.Therefore, the secondary inorganic aerosols (sulfate (SO 2− 4 ), nitrate (NO − 3 ), and ammonium (NH + 4 ); hereafter summarized as SNA) were fully observed by our synergetic monitoring system.The denuder filter pack (D-F pack) method with 6-8 h cycles was also used during the intensive observation period from 7-17 January 2015.This intensive observation was designed to capture the heavy PM 2.5 pollution episode in the wintertime and to support and validate the ACSA and NH x monitoring systems.Based on these measurement systems, gas-phase HNO 3 and NH 3 can be measured by the D-F pack method and the NH x monitor, respectively.The related gas-phase behavior analysis is valuable for improving our understanding of the formation of NO − 3 .The observations were conducted at the Chikushi Campus of Kyushu University, which is in the suburbs of Fukuoka (33.52 • N, 130.47 • E) in western Japan.The synergetic ground-based observation dataset was systematically interpreted by using the regional chemical transport model, and we also examined the impact of the domestic and transboundary air pollution in winter.Chemical transport model studies are one type of critical approach for analyzing the behavior of threedimensional air pollutants and estimating source impacts.A systematic comparison of model results with observations, including gas-phase precursors, will promote understanding and improve the model ability for Asia.This will also contribute to the Model Intercomparison Study for Asia (MICS-Asia), which focuses on providing a common understanding of the model performances and uncertainties, especially for models of long-range transport in Asia (Carmichael et al., 2002(Carmichael et al., , 2008;;Li et al., 2017).This paper is constructed as follows.Section 2 documents the observation dataset and the model simulation.Section 3 discusses the results with respect to the temporal variations at the observation sites and the model results combined with the backward trajectory analysis.Finally, a summary and conclusions are given in Sect. 4.

Observation sites
The synergetic observations for capturing SNA behavior were conducted at the Chikushi Campus of Kyushu University located in the suburbs of Fukuoka (33.52 • N, 130.47 • E).Fukuoka is the largest center of commerce on the island of Kyushu.The population of Fukuoka is 1.5 million and that of the Fukuoka metropolitan area is 2.5 million.This is the fourth largest metropolitan area in Japan after Tokyo (34.8 million), Osaka (12.2 million), and Nagoya (5.5 million).This site is an urban site.In addition to the observations at Fukuoka, observations from the Goto Islands (32.68 • N, 128.83 • E) and Tsushima Island (32.20 • N, 129.28 • E) were also used.The Goto Islands are located in the East China Sea, 190 km southwest of Fukuoka, and have a population of 70 000.Tsushima Island is located in the Tsushima Strait, 140 km northwest of Fukuoka, and has a population of 34 000.These islands have negligible anthropogenic emission sources and are regarded as remote sites.In addition to these three sites over Kyushu, to investigate the regions affected by transboundary air pollution, observations from Tottori (35.54 • N, 134.21 • E) in western Japan were also used.Tottori has a population of 190 000, and this site is also regarded as a remote site in western Japan.The locations of these four observation sites over Japan are shown in Fig. 1.In addition to these observations over Japan, PM 2.5 observations in China from the US Embassy in Beijing and the US consulates in the provincial capitals of Shanghai and Shenyang were used.The locations of these three sites over China are shown in Fig. 2.

Aerosol chemical speciation analyzer (ACSA)
An ACSA-12 monitor (Kimoto Electric Co., Ltd., Osaka, Japan) for PM 10 and PM 2.5 , which were separated by a US Environmental Protection Agency inlet and a virtual impactor, were measured with high temporal resolution (Kimoto et al., 2013).PM was collected on a tape filter made of Teflon (PTFE).Hourly observations were conducted for SO 2− 4 and NO − 3 at Fukuoka.The mass concentrations of PM were determined by using the beta ray absorption method.The ACSA-12 measured NO − 3 using an ultraviolet spectrophotometric method and SO 2− 4 by turbidimetry after the addition of BaCl 2 to form BaSO 4 and polyvinyl pyrrolidone as a stabilizer.The analytical period was within 2 h of sampling; therefore, the volatilization of particulate NH 4 NO 3 after collection was regarded as small compared with the traditional filter pack observation method.An ACSA has been tested (Osada et al., 2016) and used to analyze the severe winter haze in Beijing (Zheng et al., 2015;Li et al., 2016) and to identify the aerosol chemical compositions at Fukuoka (X.Pan et al., 2016).

NH x monitor
The behavior of NH 3 and NH + 4 is also important because they are the counterions for SO 2− 4 and NO − 3 .The concentrations of gaseous NH 3 and aerosol NH + 4 were measured with a semicontinuous microflow analytical system (MF-NH 3 A; Kimoto Electric Co., Ltd.; Osada et al., 2011) at Fukuoka.The atmospheric NH x was dissolved in ultrapure water with a continuous air-water droplet sampler and quantified by fluorescence (excitation, 360 nm; emission, 420 nm) of the ophthalaldehyde-sulfite-NH 3 reaction product (Genfa et al., 1989).Two inlet lines were used to differentiate the total amounts of NH x and particulate NH + 4 after gaseous NH 3 was removed by a phosphoric acid-coated denuder from the sample air stream.The cutoff diameter of the inlet impactor was about 2 µm.

Denuder filter pack method (D-F pack)
During the intensive observation period from 7-17 January 2015, D-F pack measurements were conducted at Fukuoka to validate the ACSA and NH x monitor measurement systems.An annular denuder multistage filter sampling system was used for HNO 3 and size-segregated aerosol sampling.The sampling interval was 6-8 h.At the inlet, coarsemode aerosols were removed by nuclepore membrane filters (111114; Nomura Micro Science Co., Ltd., Atsugi, Japan; pore size 8 µm), and then gas-phase HNO 3 was collected with the annular denuder (2000-30x242-3CSS; URG Co., Chapel Hill, NC, USA) coated with NaCl (Perrino et al., 1990).Fine-mode aerosols were collected with a PTFE filter (J100A047A; ADVANTEC, Tokyo, Japan; pore size 1 µm), and a nylon filter (66509; Pall Co., Port Washington, NY, USA) captured volatized NO − 3 from the PTFE filter (Appel et al., 1981;Vecchi et al., 2009).Because of the use of an annular denuder system before PTFE filter collection, the adsorption of HNO 3 , which would otherwise lead to the overestimation of NO − 3 , was low.The sample airflow rate was 16.7 L min −1 (1 atm, 25 • C).Under these conditions, the aerodynamic diameter of 50 % cutoff for the nuclepore filter was about 1.9 µm (John et al., 1983).The samples were analyzed by ion chromatography (IC).Comparing size-segregated SO 2− 4 and NO − 3 data based on the D-F pack method with data from the ACSA showed systematic differences.A linear regression analysis of fine-mode and coarsemode SO 2− 4 and NO − 3 showed good correlation, but the slope was not uniform.Fine-mode aerosols were underestimated by the D-F pack relative to the ACSA measurements, and coarse-mode aerosols were overestimated by the D-F pack relative to the ACSA measurements.However, by summing the fine-and coarse-mode data, the slope between the D-F pack and ACSA results was close to unity.The difference in the cutoff diameter of the D-F pack method was considered to be the most important factor in explaining the differences between the results from the D-F pack and the ACSA.More

PM-712
Hourly PM 10 and PM 2.5 concentrations were measured by a PM monitor (PM-712; Kimoto Electric Co., Ltd.) at the Goto Islands, Tsushima Island, and Tottori.The PM-712 used the beta ray attenuation method to measure the mass concentrations of PM 10 and PM 2.5 .The ionic constituents of the species on the PTFE tape filters were also analyzed by IC to compare the aerosol behavior at Fukuoka and the other sites.At all sites, the sample spots collected on the tape filter were covered with polyester tape to avoid contamination and cross talk interference during storage.The sampling duration for the PM-712 tape filters was 1 h, except for Tottori where it was 3 h.For chemical analysis, four consecutive 1 h tape samples were combined into one sample for the Goto Islands and Tsushima Island.For Tottori, tape samples for 1 or 0.5 days were combined.Because of a temperature change during PM sample storage, some NO − 3 may have escaped via volatilization of HNO 3 from the sample.Therefore, NO − 3 data from the Goto Islands, Tsushima Island, and Tottori were not used.NH + 4 is also affected by volatilization when it forms NH 4 NO 3 , but it preferentially forms (NH 4 ) 2 SO 4 .The data for NH + 4 were used to indicate the concentration trapped by SO 2− 4 .Hereafter, we call these datasets from the PM tape samples "tape filters".

Beta attenuation monitors (BAMs)
Hourly PM 2.5 concentrations in China were measured by beta attenuation monitors (BAMs) (1020; Met One Instruments, Inc., Grants Pass, OR, USA) at the US Embassy in Beijing on 8 April 2008, at the US consulates in Shanghai on 21 December 2011, and at the US consulates in Shenyang on 22 April 2013 (Ministry of Environment, 2016).In the BAM technique, PM is collected on a quartz filter tape over a given time interval and the attenuation of beta rays through the sample is measured and correlated directly with the PM mass.The details and the statistical analysis results of the BAM observation at the US Embassy and consulates are found in the work of San Martini et al. (2015).

Multi-angle absorption photometer (MAAP)
Observations of black carbon (BC), a primary aerosol that directly reflects local emission contributions, from Fukuoka and the Goto Islands were also used to distinguish domestic and transboundary air pollution.BC is observed by using a multi-angle absorption photometer (MAAP) (MAAP5012; Thermo Fisher Scientific, Waltham, MA, USA) (Petzold et al., 2005).In this method, the absorbance of the particles deposited on the filter is distinguished from scattering by reflectance measurements at multiple angles and by transmittance.This is to minimize the effects of coexisting aerosol particles other than BC on filter-based absorption photometers.The comparison measurements of BC from the Goto Islands were previously performed by Kanaya et al. (2013), and they reported that the BC MAAP measurements were strongly correlated with measurements by other techniques but had a positive bias.From the results reported by Kanaya et al. (2013), the MAAP absorption cross section of 6.6 m 2 g −1 was systematically increased to 10.3 m 2 g −1 at 639 nm.There were no MAAP BC measurements from the Goto Islands before 11 January 2015 during the intensive observation period.

Chemical transport model
The chemical transport model simulation was performed by using the Community Multiscale Air Quality (CMAQ) modeling system version 4.7.1 (Byun and Schere, 2006) with nesting over East Asia.The meteorological fields of CMAQ were prepared with the Weather Research and Forecasting Model version 3.3.1 (Skamarock et al., 2008) with analysis nudging applied to the National Centers for Environmental Prediction final operational global analysis data.The model domain covers the whole of East Asia with an 81 km horizontal grid resolution and a 95 × 75 grid centered at 30 • N and 115 • E on a Lambert conformal projection.The nested domain covers eastern China and the whole of Japan with a 27 km horizontal grid resolution and a 145 × 145 grid.The vertical grid for sigma-pressure coordinates extends to 50 hPa with 37 layers with nonuniform spacing.The lateral boundary condition was as given by the global chemical transport model of Geos-Chem (Uno et al., 2014).The simulation period was from 1-17 January 2015, and the first 6 days were discarded as model spin-up time.The dry deposition velocity of HNO 3 over land was increased by a factor of 5 based on the model intercomparison results (Shimadera et al., 2014;Morino et al., 2015).
Emissions were set as follows.Anthropogenic emissions and natural sources of NO x from soil were obtained from the latest Regional Emission inventory in ASia (REAS) version 2.1 (Kurokawa et al., 2013), which covers 2000 to 2008.Therefore, the emissions for January 2008 were used in this study.This is because satellite observations of the NO 2 column showed a decreasing trend in NO x emissions from China of −6 % yr −1 after 2011, and the levels for 2015 are similar to those for 2009 (Irie et al., 2016).In contrast, the NO 2 column over Japan decreased until 2013, and then began to increase from 2013 owing to the change in power plant use after the Fukushima Daiichi nuclear disaster (Morino et al., 2011).The level of the NO 2 column over Japan in 2015  (Xia et al., 2016).The assumption about the 2008 level of SO 2 emissions may overestimate the actual status.In Japan, SO 2 emissions are increasing for the same reasons as for NO x ; however, there are no reliable references for the current status of SO 2 emissions.Considering these factors in the variation of NO x and SO 2 emissions over China and Japan, it was assumed that the 2008 emissions would be within the range of uncertainty of the bottom-up emission inventories.Because REAS does not consider the monthly variation in NH 3 , we used the monthly variation estimated by Huang et al. (2012).The annual NH 3 emission amount was also adjusted to match the estimate of Huang et al. (2012).Other anthropogenic emissions, such as BC, organic carbon (OC), and volatile organic compounds (VOC), were also assumed to be at the 2008 level due to the lack of information needed to update the status.The anthropogenic emission amounts of SO 2 , NO x , and NH 3 in January 2015 are listed in Table 1. the biogenic emissions were prepared from the Model of Emissions of Gases and Aerosols from Nature (MEGAN; Guenther et al., 2012).As a biogenic source, dimethylsulfide (DMS) was not treated in the modeling system.The biomass burning emissions were used from the climatological database of REanalysis of the TROpospheric chemical composition (RETRO; Schultz et al., 2008).The volcanic activity data were taken from the Ace-Asia and TRACE-P Modeling and Emission Support System (ACESS; Streets et al., 2003) and were modified by the volcanic activity observation data from the Japan Meteorological Agency (JMA) for available volcanoes (JMA, 2016).The 14 main active volcanoes in Japan, along with and Mount Mayon and Mount Bulusan on the island of Luzon in the Philippines, were considered.The model simulation using the above dataset is referred to as the base case simulation.The modeling system domain with overlaid anthropogenic NO x emissions is shown in Fig. 3.
To investigate whether the effect of domestic or transboundary air pollution is dominant, we also conducted a sensitivity simulation in which the anthropogenic emissions in Japan are switched off.Here, the anthropogenic emissions were taken from the REAS inventory; because of this, emis- sions from agriculture were included.Shipping emissions were not treated in the sensitivity simulation.In terms of O 3 , which is involved in complex nonlinear chemistry, larger nonlinearities in the atmospheric concentration response to emission variation for China, but not Japan, were clarified due to the higher amount of emissions from China than from Japan (Itahashi et al., 2015).Therefore, the sensitivity simulation was designed to remove the anthropogenic emissions in Japan instead of those in China.Based on the differences between the base case simulation and this sensitivity simulation, the domestic contribution from Japan was estimated.

Meteorological conditions
The meteorological conditions during the intensive observation campaign from 7-17 January 2015 are shown in Fig. 4 with the observations and model results.The meteorological observation stations of the JMA in the corresponding nested model grid of Fukuoka were used.Temperatures (Fig. 4a) were around 5 • C at night and 10 • C during the day in January 2015.On 9 January, the temperature was nearly 0 • C at Fukuoka.For the wind field (Fig. 4b and c), because of the dominance of the northwesterly wind system from the Asian continent in winter, the wind direction was generally 270-360 • (west to north) and the wind speed was around 5 m s −1 , with the exception of 9 and 12-15 January.On 9 January when the coldest temperature during the intensive observation campaign was observed, the wind speed was less than 1 m s −1 and the wind direction was from the south.From 13-15 January, the wind speed was also low at 2-3 m s −1 , and the wind direction was easterly, caused by a warm front passing over the south of Kyushu on 14 January.After the warm front had passed, the relative humidity was close to 100 % on 15-16 January (Fig. 4d) with maximum rainfall of 10 mm h −1 on 15 January (Fig. 4e).The model tended to underestimate the precipitation amount, as we have reported in a previous study (Itahashi et al., 2014).Comparing the observations with the model results shows that our modeling system generally captures the observed meteorological variations during this episode.

PM 2.5
The temporal variation in PM 2.5 over Japan at Fukuoka, Tsushima Island, the Goto Islands, and Tottori are presented in Fig. 1.The PM 2.5 observation data are taken from the ACSA at Fukuoka and from the PM-712 at other sites.The temporal resolution is 1 h for all observations.During the analyzed period of 7-17 January 2015, episodic PM 2.5 peaks reached around 100 µg m −3 at Fukuoka twice.The first peak, observed at 12:00 LT on 11 January (shown in blue in Fig. 1), reached a maximum concentration of 86.4 µg m −3 at Fukuoka and 105.1 µg m −3 at the Goto Islands.During this first peak, the concentration at Tsushima Island was 63.9 µg m −3 , which was lower than at the other remote island sites in the Goto Islands, and there was no distinctive peak at Tottori.The second peak, observed at 00:00 LT on 17 January (shown in red in Fig. 1), reached a maximum concentration at Fukuoka of 106.2 µg m −3 .During this second peak, the remote sites of the Goto Islands and Tsushima Island also recorded high PM 2.5 concentrations of 104.8 and 89.1 µg m −3 , respectively, and the PM 2.5 concentration reached 37.6 µg m −3 at Tottori.In Fig. 1, we show the model results as black lines.Generally, the model captured the observed temporal PM 2.5 behavior, although it underestimated the first peaks at Fukuoka and the Goto Islands and the second peak at the Goto Islands.The timing of the high PM 2.5 concentration was reproduced well by the modeling system.A statistical analysis of the model reproducibility demonstrated that all paired datasets for PM 2.5 showed good correlations between the observations and the model at the four sites in Japan, with a correlation coefficient (R) of 0.86.The mean fractional bias (MFB) and mean fractional error (MFE) were −42.6 and 67.4 %, respectively, and these results satisfied the model performance criteria (MFB ≤ ± 60 % and MFE ≤ + 75 %) proposed by Boylan and Russell (2006).Figure 1 also shows the model results of a sensitivity simulation performed by switching off the Japanese anthropogenic emissions.The sensitivity simulation results are shown as dotted black lines, and the difference between the base case and the sensitivity simulation is shown in gray, which indicates the domestic contribution of Japan.Except for Fukuoka, there were few domestic contributions for PM 2.5 ; therefore, transboundary air pollution was dominant in January 2015 at remote sites in western Japan.At Fukuoka, although domestic contributions for PM 2.5 were found in some cases (from 8-10 and on 14 January), the concentration of PM 2.5 was lower compared with the two peaks.During the two episodes when PM 2.5 concentration reached around 100 µg m −3 over Japan, the model simulation suggested that the effect of transboundary air pollution was dominant, even at Fukuoka.
The temporal variations in PM 2.5 over China at Beijing, Shanghai, and Shenyang are shown in Fig. 2. The PM 2.5 observation data are taken from the BAMs, and the temporal resolution is 1 h.The model results of a sensitivity simulation suggested that the impact on three Chinese sites from Japanese anthropogenic emissions was negligible during this period.At Beijing, there were high concentrations of PM 2.5 that correspond to the high concentration of PM 2.5 found over Japan.One high concentration was approximately 300 µg m −3 on 10-11 January and another was around 600 µg m −3 on 16 January.These peak times were almost 1 day before the high concentration was observed over Japan.At Shanghai, there were two clear peaks with a PM 2.5 concentration of 200 µg m −3 on 11 and 17 January.The time corresponded well to the peak time over Japan.At Shenyang, where the local emissions from domestic sources were dominant in winter, the temporal variation was complex compared with Beijing and Shanghai.PM 2.5 showed sharp peaks several times with concentrations of around 300 µg m −3 , whereas the model only showed gentle peaks.An analysis of the model reproducibility showed that all for PM 2.5 paired datasets for the observations and the model at three sites over China, R was 0.73 and MFB and MFE were −9.8 and 46.8 %, respectively; these numbers satisfy the model goal criteria (MFB ≤ ± 30 % and MFE ≤ ± 50 %) proposed by Boylan and Russell (2006).The evaluation of the model performance over China supports the discussion on downwind regions.

SNA
The temporal variations in SNA are shown in Figs. 5 and 6.In Fig. 5, SO 2− 4 and NH + 4 are shown for four sites in Japan.At Fukuoka, the ACSA and D-F pack observations are shown for SO 2− 4 .The NH x monitor and D-F pack results are shown for NH + 4 .The temporal resolutions of the ACSA and the NH x monitor were 1 h, and those of the D-F packs were 6-8 h depending on the samples.For the Goto Islands, Tsushima Island, and Tottori, the PM-712 tape filter data were used.The temporal resolution was 4 h at the Goto Islands and Tsushima Island and 1 or 0.5 days at Tottori.In 3 analysis at the Fukuoka site was used.At Fukuoka, the SNA concentration contributed 52 and 46 % of the PM 2.5 concentration in the first and second episodes, respectively.For SO 2− 4 (Fig. 5; left), the concentration during the second episode was larger than during the first episode at Fukuoka, the Goto Islands, and Tsushima Island.At Tottori, there was no peak for the first episode for SO 2− 4 .In contrast to SO 2− 4 , a higher NO − 3 concentration was observed during the first episode instead of the second episode (Fig. 6).NH + 4 showed high concentrations during both episodes because it is the counterion for SO 2− 4 and NO − 3 (Fig. 5; right).Based on the analysis of the PM 2.5 (Fig. 1) and SNA (Figs. 5 and  6) observations, the PM 2.5 concentrations were similar during the episodes on 11 and 17 January; however, the main component of SNA was NO − 3 during the first episode and SO 2− 4 during the second episode.At Fukuoka, the relative portions of SO 2− 4 , NO − 3 , and NH + 4 within the PM 2.5 were respectively 18, 20, and 14 % during the first episode, and 27, 6, and 14 % during the second episode.Therefore, the first episode (shown in blue in Figs. 1, 5, 6, 7 and 8) is referred to as "type N", and the second episode (shown in red in Figs. 1, 5, 6, 7 and 8) is referred to as "type S" hereafter.
The model results for the base case and sensitivity simulations are overlaid with the same temporal resolution as the observations in Figs. 5 and 6.The model tended to underestimate the SO 2− 4 concentration (Fig. 5; left); however, the model reproduced the features of types N and S, and the sensitivity simulation indicated the dominance of transboundary air pollution for SO 2− 4 during the intensive observation campaign in January 2015, even at Fukuoka.For NO − 3 (Fig. 6), the model reproduced the features of the type N and S peaks well, although the model overestimated the dip in NO − 3 concentration found from the evening of 10 January to before the type N episode.The D-F pack observations generally underestimated NO − 3 compared with the ACSA observations because of the difference in the cutoff diameter between these measurement systems.Except for the type N and S episodes, domestic contributions were seen for NO − 3 from 8-10 and on 14 January.However, the sensitivity simulation confirmed that transboundary NO − 3 air pollution was dominant for types N and S. Because NH + 4 is the counterion for both SO 2− 4 and NO − 3 , small domestic contributions for NH + 4 were observed at Fukuoka (Fig. 5; right).This result corresponded to the domestic contribution for NO − 3 .For the other three remote sites, there were no domestic contributions for NH + 4 .The behavior of SNA, gas-phase HNO 3 , and NH 3 was analyzed comprehensively based on the NH x monitor and D-F pack observations (Fig. 6) to support our understanding of NO − 3 behavior.There are few synergetic analyses including gas-phase behavior over the downwind region of the Asian continent.Peaks for gas-phase HNO 3 were found for types N and S, whereas the concentration of gas-phase NH 3 was nearly zero (less than 1 µg m −3 for 24 h average) for types N and S on 8, 10, 12, and 15 January (the green arrows in Fig. 6).Type S, in particular, showed an almost zero concentration of NH 3 .The concentration of total am- monia showed distinct peaks for types N and S; therefore, the nearly zero concentration of NH 3 suggested the full conversion of NH 3 to produce NH + 4 as a counterion for SO 2− 4 and NO − 3 .The sensitivity simulation, in which Japanese anthropogenic emissions were switched off, clarified the different features of related gas-phase species.The base case simulation and sensitivity simulation were similar for HNO 3 , suggesting that it originated from transboundary air pollution.A slight increase in HNO 3 in the sensitivity simulation was found from 8-10 and on 12 January (the red arrows in Fig. 6).These were the complex cases connected to overseas and domestic emissions.If there are no Japanese NH 3 emissions, the transported HNO 3 cannot produce NO − 3 in Japan, and so it remains as gas-phase HNO 3 .The synergetic analysis for gas-phase HNO 3 and NH 3 indicated that abundant HNO 3 was transported from abroad and reacted with do-mestic NH 3 , producing NO − 3 from 8-10 January.Compared with these cases, domestic HNO 3 and NH 3 produced NO − 3 on 14 January (orange arrows in Fig. 6).The concentrations were lower than for types N and S, which were dominated by transboundary air pollution.

Coarse-mode aerosols
Coarse-mode aerosols were also partly analyzed in this study.Because of the effect of transboundary air pollution on HNO 3 (Fig. 6), we focused on coarse-mode NO − 3 .Coarse-mode NO − 3 is produced by reactions of HNO 3 with mineral dust or sea salt particles.In general, mineral dust mainly has an effect in spring over East Asia (Itahashi et al., 2010), whereas sea salt particles play an important role throughout the year.Recently, we reported the importance of coarse-mode NO − as an atmospheric input in East Asian ocean regions (Itahashi et al., 2016).Figure 7 shows the modeled and observed coarse-mode NO − 3 , Na + , and Cl − .The ACSA and D-F pack observations are shown for coarse-mode NO − 3 , and the D-F pack observations are shown for coarse-mode Na + and Cl − .During the intensive observation period in January, coarsemode NO − 3 also showed high concentrations for types N and S of around 10 µg m −3 (on 9-10 January, around 5 µg m −3 ).Based on the model results and because the domestic contribution for HNO 3 was found on 14 January (Fig. 6), the domestic contribution for coarse-mode NO − 3 was found only on 14 January, but the concentration was below 1 µg m −3 .Na + and Cl − from sea salt particles also had peaks for types N and S. Sea salt particles are mechanically produced by high winds; therefore, these peaks generally corresponded to high wind speeds (Fig. 4b).High winds were observed on 15 January; Na + and Cl − peaks occurred, but the coarse-mode NO − 3 concentration was close to zero.This was because there was no HNO 3 to react with NaCl from 12-15 January or the wet deposition of coarse-mode NO − 3 with the precipitation from noon on 14 January to the evening of 15 January.For coarse-mode NO − 3 , transboundary air pollution was the dominant factor.This means that a large amount of HNO 3 was transported from abroad (Fig. 6), reacted with sea salt particles over the ocean, and reached Fukuoka in the air mass.

BC
To support the discussion of the domestic and transboundary contributions to SNA, the behavior of BC at Fukuoka and the Goto Islands is shown in Fig. 8.The sensitivity simulation would suffer from a nonlinear chemical response if complex atmospheric chemistry were involved; hence, we focused on BC, which is a primary aerosol.The temporal variation of BC also showed distinctive peaks for types N and S at Fukuoka and the Goto Islands.The model results reproduced these peaks well, and the sensitivity simulation also suggested the dominance of transboundary air pollution for both peaks N and S. The temporal variation at the Goto Islands showed only two peaks of types N and S, although many short-term peaks were seen at Fukuoka.The sensitivity simulation confirmed that domestic air pollution contributed to these shortterm peaks at Fukuoka; however, the model could not fully capture the peaks observed on 7, 13, and 14 January.To improve the performance of the model to capture these shortterm peaks, a higher resolution model simulation and a revision of the emission inventory are needed.An analysis of the primary aerosol confirmed that transboundary air pollution was dominant for types N and S in January 2015.
Consequently, the well-validated model simulation indicated that two high PM 2.5 episodes with concentrations of around 100 µg m −3 occurring over western Japan in January were dominated by NO − 3 for the first peak (type N) and by SO 2− 4 for the second peak (type S); the NH + 4 concentration was high for both types.The model sensitivity simulation clarified that these high SNA concentrations in the type N and S episodes were dominated by transboundary air pollution.In addition to the transport of SNA, abundant gas-phase HNO 3 and coarse-mode NO − 3 reacted with sea salt particles over the ocean and were also transported to western Japan.NH 3 , which mainly came from domestic emissions, showed concentrations of around zero during type N and S events, suggesting that NH 3 was depleted to neutralize SO 2− 4 and NO − 3 .

Trajectory analysis
Analyzing the synergetic observations at Fukuoka and the other three remote sites in Japan with the regional chemical transport model demonstrated that the two PM 2.5 episodic peaks were dominated by transboundary heavy pollution, even at Fukuoka.The two peaks had different SNA compositions.The first episode on 11 January showed a high NO − 3 (type N) concentration and the second episode on 17 January was dominated by SO 2− 4 (type S).The differences in these episodes were investigated further by a model simulation combined with a backward trajectory analysis.The spatial distributions of SO 2− 4 and NO − 3 during type N and S patterns are shown in Figs. 9 and 10, respectively.
In type N (Fig. 9; right), the model results showed that a low SO 2− 4 concentration of less than 5 µg m −3 and a

high NO −
3 concentration of more than 10 µg m −3 covered Fukuoka.The spatial distribution patterns indicated an outflow of SO 2− 4 and NO − 3 from continental Asia to western Japan.The dominance of transboundary air pollution suggested by these spatial distributions was consistent with the model sensitivity simulation results (Figs. 5 and 6).Highconcentration regions of SO 2− 4 and NO − 3 stretched from the eastern coastline of China to the East China Sea and western Japan.The spatial distribution implied the direct transport from continental Asia to the downwind regions.In addition, the high-concentration region stretched from eastern China to western Japan, which is consistent with the corresponding PM 2.5 peak on 11 January at Shanghai and over Japan.To investigate the air mass origin for type N, the HYSPLIT backward trajectory (Stein et al., 2015) was analyzed over 72 h starting from Fukuoka (T N in Fig. 9; left).The backward trajectory during type N transport suggested that the air mass originated from Shaanxi Province and passed over Shanxi Province, southern Hebei Province, and Shandong Province, and then reached Fukuoka.The traveling time from the coast of China to Fukuoka was about 18 h.The distance from the coastline of China to Fukuoka is approximately 1000 km, so the air mass speed for type N was 55.6 km h −1 .Figure 9 (left) shows the spatial distribution when the air mass was located over China.A high concentration of NO − 3 of more than 60 µg m −3 occurred over the east coast of China before the air mass arrived in Fukuoka, whereas the SO 2− 4 concentration was as high as 10 µg m −3 above the East China Sea.(types N and S).The outflow concentration of SO 2− 4 was lower on 11 January for type N and larger on 17 January for type S; a high concentration of more than 15 µg m −3 reached 130 • N (Fukuoka) and a concentration of around 10 µg m −3 reached 134 • N (Tottori) for type S (Fig. 14a).The outflow of NO − 3 was observed over the Goto Islands, Tsushima Island, and Fukuoka on 11 January for type N, whereas the high concentration of over 10 µg m −3 was limited to the East China Sea region on 17 January for type S. A concentration of NO − 3 of more than 5 µg m −3 did not reach 134 • N (Tottori) for type N (Fig. 14b).The outflow analysis suggested that SO 2− 4 can be transported over longer distances, whereas transboundary NO − 3 air pollution is limited to western Japan, especially over Kyushu.

Conclusion
Using state-of-the-art observation systems to capture SNA behavior and a chemical transport model, two episodes of high PM 2.5 concentrations of around 100 µg m −3 were analyzed that occurred in winter over western Japan.The first episode (11 January) was dominated by NO − 3 (type N) and the second episode (17 January) by SO 2− 4 (type S).The chemical transport model captured the behavior of SNA and the related gas-phase species of HNO 3 and NH 3 as well as coarse-mode NO − 3 observed over Japan.The model also reproduced PM 2.5 variation over China.To evaluate the domestic contributions, a sensitivity analysis was performed.In this analysis, the anthropogenic emissions from Japan were switched off in the chemical transport model.The results showed some domestic contributions for NO − 3 , although the

Figure 1 .
Figure 1.The temporal variation in PM 2.5 over Japan at Fukuoka, the Goto Islands, Tsushima Island, and Tottori from 7-17 January 2015.The blue and red shading shows the episodes focused on in this study.The red lines indicate the observations.The black lines indicate the base case simulation, and the dotted black lines indicate the sensitivity simulation in which the anthropogenic emissions from Japan were switched off; the differences between these results, shown in gray, represent local contributions.

Figure 2 .
Figure 2. The temporal variation in PM 2.5 over China at Beijing, Shanghai, and Shenyang from 7-17 January 2015.The red lines indicate the observations by the BAMs at the US Embassy in Beijing and at the US consulates in Shenyang and Shanghai.The black lines indicate the base case simulation.

Figure 3 .
Figure 3.The modeling domain for the horizontal resolutions of (a) 81 km and (b) 27 km with anthropogenic NO x emissions.

Figure 4 .
Figure 4.The temporal variation in (a) temperature, (b) wind speed, (c) wind direction, (d) relative humidity, and (e) precipitation at Fukuoka from 7-17 January 2015.The gray and black coloring indicates the observations and the model results, respectively.
Fig. 6, NO − 3 , HNO 3 , NH 3 , and total ammonia (sum of NH + 4 and NH 3 ) are shown for Fukuoka.The ACSA and D-F pack observations for NO − 3 are shown, the D-F pack observations are shown for HNO 3 , and the NH x monitor observations are shown for NH 3 and total ammonia.Because of a temperature change during the PM-712 sample storage, the NO − 3 concentrations could have been affected by volatilization; hence, only the NO −

Figure 5 .
Figure 5.The temporal variation in SO 2− 4 and NH + 4 over Japan at Fukuoka, the Goto Islands, Tsushima Island, and Tottori from 7-17 January 2015.The blue and red shading shows the type N and S patterns focused on in this study.The red lines indicate SO 2− 4 observations by ACSA, and the green lines indicate NH + 4 observations by the NH x monitor at Fukuoka.The open circles are D-F pack observations at Fukuoka.The open squares are tape filter measurements at the Goto Islands, Tsushima Island, and Tottori.The black lines indicate the base case simulation, and the dotted black lines indicate the sensitivity simulation in which the anthropogenic emissions from Japan were switched off; the differences between these results, shown in gray, represent local contributions.

Figure 6 .
Figure 6.The temporal variation in NO − 3 , HNO 3 , NH 3 , and total ammonia at Fukuoka from 7-17 January 2015.The blue and red shading shows the type N and S patterns focused on in this study.The red lines indicate NO − 3 observations by the ACSA, the green lines indicate NH 3 and total ammonia observations by the NH x monitor, and the open circles indicate D-F pack observations.For NH 3 , periods of nearly zero concentration (24 h average of less than 1 µg m −3 ) are indicated by arrows.The black lines indicate the base case simulation, and the dotted black lines indicate the sensitivity simulation in which the anthropogenic emissions from Japan were switched off; the differences between these results, shown in gray, represent local contributions.

Figure 7 .
Figure 7.The temporal variation in coarse-mode NO − 3 , Na + , and Cl − at Fukuoka from 7-17 January 2015.The blue and red shading shows the type N and S patterns focused on in this study.The red lines indicate coarse-mode NO − 3 observations by the ACSA, and the open circles indicate D-F pack observations.The black lines indicate the base case simulation, and the dotted black lines indicate the sensitivity simulation in which the anthropogenic emissions from Japan were switched off; the differences between these results, shown in gray, represent local contributions.

Figure 8 .
Figure 8.The temporal variation in BC at Fukuoka and the Goto Islands from 7-17 January 2015.The blue and red shading shows the type N and S patterns focused on in this study.The magenta line indicates BC observations by MAAP.The black lines indicate the base case simulation, and the dotted black lines indicate the sensitivity simulation in which the anthropogenic emissions from Japan were switched off; the differences between these results, shown in gray, represent local contributions.

Figure 12 .
Figure 12.The path analysis of the model results along trajectory T S .(a) Temperature and relative humidity; (b) SO 2− 4 (with expansion at the bottom) and SO 2 ; (c) HO 2 concentration; (d) NO − 3 , HNO 3 , NO, NO 2 , other NO y (NO 3 , HNO 2 , N 2 O 5 , and PANs), and coarse-mode NO − 3 ; (e) NH + 4 and NH 3 ; (f) adjGR, F s , and F n .In (b), (d), and (e), BC and CO concentrations normalized to the maximum value are also shown.The time axis indicates the backward time from Fukuoka.The brown and blue bars at the bottom are schematics of the trajectory location over land and ocean.
humidity of around 70 % -Abundant NH 3 supply above China -No NH 3 in the gas phase -NH 3 -rich air mass maintaining neutralization of NO − 3 -NH 3 -neutral conditions during transport, SO 2− 4 neutralized Note: the parentheses indicate the multiplying factors compared with the status over China.The status is averaged over 6 h before the air mass leaves China, during the transport time from China to Fukuoka above the ocean, and over 3 h before the air mass reaches Fukuoka.Other NOy consists of NO, NO 2 , NO 3 , HNO 2 , N 2 O 5 , and PANs.

Table 1 .
The anthropogenic emission amounts used in this study.Units are Gg month −1 for January 2015.

Table 2 .
Summary of the path analysis for types N and S.