Sources of High Sulfate Aerosol Concentration Observed at Cape Hedo in Spring 2012

Received: September 18, 2018
Revised: December 26, 2018
Accepted: January 23, 2019

Download Citation: RIS|BibTeX|https://doi.org/10.4209/aaqr.2018.09.0350  

Cite this article:

Itahashi, S., Hatakeyama, S., Shimada, K. and Takami, A. (2019). Sources of High Sulfate Aerosol Concentration Observed at Cape Hedo in Spring 2012. Aerosol Air Qual. Res. 19: 587-600. https://doi.org/10.4209/aaqr.2018.09.0350

HIGHLIGHTS

• High concentrations of SO₄^2– were found during spring 2012 at Cape Hedo.
• SO₄^2– sources for these episodes were estimated by the tagged tracer method.
• Volcanoes had a large effect in March.
• Long-range transport from China was dominant in April.
• Injection height was important for better model prediction of volcano sources.

Keywords: Air quality model; Source apportionment; Tagged tracer method; East Asia; Cape Hedo Atmosphere and Aerosol Monitoring Station (CHAAMS).

INTRODUCTION

Asian air quality is currently the subject of intense interest due to its regional-to-global impacts (Carmichael et al., 2009). Understanding aerosol transformation processes during long-range transport (LRT) is essential for air quality and regional climate change. To achieve this, an intensive observation campaign has been conducted continuously at Cape Hedo Atmosphere and Aerosol Monitoring Station (CHAAMS), Okinawa, Japan. CHAAMS is at the northern edge of Okinawa Prefecture (26.87°N, 128.25°E, 60 m above sea level), far from the largest city in Okinawa, Naha City (http://www.nies.go.jp/asia/hedomisaki/home-e.html). There are no large industrial or residential areas near CHAAMS. Air masses are transported to CHAAMS from China, Korea, Japan, Southeast Asia, or the Pacific Ocean depending on the wind pattern. The location of CHAAMS is suitable for capturing the atmospheric pollutants that travel via LRT and has already revealed the behavior of aerosols affected by LRT (Takami et al., 2007; Hatakeyama et al., 2011; Shimada et al., 2015; Shimada et al., 2016; Itahashi et al., 2017a). At CHAAMS, sulfate aerosol (SO₄^2–), which is mainly produced by the oxidation of sulfur dioxide (SO₂), was the dominant aerosol component (e.g., Tatsuta et al., 2017), as it is across Japan (Chatani et al., 2018).

Because of the long lifetime of SO₄^2– in the atmosphere, LRT is important in downwind regions. To estimate the impact of LRT quantitatively, an air quality model can provide valuable insights (Itahashi et al., 2012a, 2018a). Based on the tagged tracer method, which assigns tracers to track sources, we have been investigating the LRT of SO₄^2– over East Asia (Itahashi et al., 2017b; Itahashi, 2018). There has been a lack of comprehensive information at specific receptor sites and detailed time-scales. At CHAAMS, continuous observation campaigns of about 1 week long have been conducted over 6 years from 2010 to 2015. During these campaigns, the highest SO₄^2– concentrations were found in spring 2012 (see “Methods” for detailed description). In this work, we evaluated the sources of SO₄^2– during the high concentration episodes based on an air quality model with the tagged tracer method. In our previous study, which revealed the source apportionment at CHAAMS in the autumn (Itahashi et al., 2017a), seven sources of anthropogenic emissions from China, Taiwan, the Korean Peninsula, Japan, and ships, as well as natural emissions, were set. To build on our previous study, more sources were set for a detailed analysis. In China, one of the important sources of SO₄^2–, source regions were divided into provinces, with a total of 31 regions. For volcanoes as natural sources, the source apportionment to each volcano was investigated. Throughout this study, non-sea salt SO₄^2– (referred to hereafter as SO₄^2–) was calculated by the conservative assumption of Na⁺ as a sea salt tracer (Itahashi et al., 2017a) and analyzed to exclude the effect of sea salt.

METHODS

Observations

During the intensive observation campaign at CHAAMS, 24-h sampling was performed from 10:00 local time (LT) to 10:00 the next day. A total of 15 campaigns were conducted from 2010 to 2015. PM_2.5 was collected on Teflon filters (203 × 254 mm, WP-500-50; Sumitomo Electric, Osaka, Japan) using a high-volume air sampler (HV-1000F; Sibata, Tokyo, Japan) for 24 h. The PM_2.5 sampler was equipped with a PM_2.5 impactor developed by Kaneyasu (2010). The samplers were operated at 740 L min^–1 for 24 h. After sampling, filters were stored in a freezer at –20°C. After the campaign, the samples were transported to the lab and the chemical components were analyzed. To analyze the ionic composition, filter samples were ultrasonically extracted with distilled water (specific resistivity > 18 mol L^–1 cm^–1) for 20 min and distilled water (10 mL) in a polypropylene tube. Cl^–, NO₃^–, SO₄^2–, Na⁺, NH₄⁺, K⁺, Mg²⁺, and Ca²⁺ were analyzed by ion chromatography (LC10AD; Shimadzu, Kyoto, Japan). Anions were separated by an anion analysis column (IC-SA2; Shimadzu, Kyoto, Japan) and detected by a conductivity detector (CDD-10ASP; Shimadzu, Kyoto, Japan) with a suppressor. Cations were separated by a cation analysis column (IC-C4; Shimadzu, Kyoto, Japan) and detected by a conductivity detector (CDD-6A; Shimadzu, Kyoto, Japan) without a suppressor. The eluent was 1.8 mol L^–1 Na₂CO₃/1.7 mmol L^–1 NaHCO₃ for anions, and 18 mmol L^–1 methanesulfonic acid for cations. The concentration of each ionic component (detection limit for each ion of 0.01 µg m^–3) and element was calculated by subtracting the blank value (Yumoto et al., 2015; Taniguchi et al., 2017; Tatsuta et al., 2017; Shimada et al., 2017). The SO₄^2– concentrations obtained from the 15 field campaigns are summarized in Fig. 1. The average concentrations for each campaign were from 0.34 to 6.97 µg m^–3, and the average concentration of all observations was 3.13 µg m^–3. The maximum daily mean concentrations for each campaign are also shown. The highest concentration was observed during the campaign in spring 2012. These campaigns were conducted from March 10 to 17 and from April 22 to 28. The maximum concentrations were 16.91 µg m^–3 on March 15 and 15.87 µg m^–3 on April 23.

Fig. 1. SO₄^2– concentration observed by EANET and during intensive field campaigns at CHAAMS from 2010 to 2015.

At Cape Hedo, network observations by the Acid Deposition Monitoring Network in East Asia (EANET) have been conducted by the four-stage filter pack method with a sampling period of 2 weeks. Measured SO₄^2– on a Teflon filter in the first stage was analyzed by ion chromatography (EANET, 2013). The monthly, minimum, and maximum values for 2-week observation periods are shown in Fig. 1. The value for one observation campaign conducted in January 2010 did not correspond to the monthly mean obtained by EANET, and this can be explained by the short duration of the campaign, with a total of three observations. Except for this one case, the average observation campaign conducted in an approximately 1-week period and the monthly mean observation by EANET were generally consistent. During spring 2012, there was generally a correspondence between average concentrations; however, the daily maximum value was approximately 3 times greater than the average value. Because these concentrations were much higher than those of other campaigns and other monthly averaged EANET data, we sought to explain these episodes based on the air quality model.

Air Quality Model Simulation

The model simulation was conducted with the regional chemical transport model of the Comprehensive Air Quality Model with Extensions (CAMx) version 6.40 (ENVIRON International Corporation, 2016). CAMx was configured with a 36-km horizontal resolution and 220 × 140 grid points, centered at 35°N and 115°E on a Lambert conformal projection, covering East Asia. The vertical resolution was 37 non-uniformly spaced layers from the surface to 50 hPa. This modeling domain was the same as in our previous study (Itahashi et al., 2017a). Gas-phase chemistry was updated to use SAPRC-07 (Carter, 2010), and the aerosol chemistry used the CF scheme, which treats the size distribution as two static fine and coarse modes. CAMx was driven by the meteorological fields calculated by the Weather Research and Forecasting Model version 3.6.1 (Skamarock et al., 2008). The initial and lateral boundary conditions were prepared from the global chemical transport in the Model for Ozone and Related chemical Tracers version 4 (MOZART-4) (Emmons et al., 2010). For the simulations of March and April 2012, a 1-week spin-up time was used.

The emissions inventory for 2012 was prepared as follows. Anthropogenic emissions were based on Hemispheric Transport of Air Pollution (HTAP) version 2.2 for the target year of 2010 (Janssens-Maenhout et al., 2015). China is a particularly important source of anthropogenic emissions globally, although its emissions have gradually decreased after 2005 (Itahashi et al., 2012b). The satellite-based SO₂ column observed by the Ozone Monitoring Instrument (OMI) was used to evaluate the emissions variation (Itahashi et al., 2018b). First, for each Chinese province in 2010, the amount of anthropogenic SO₂ emissions was compared with the SO₂ column (Fig. 2(a)), and the results showed moderate correspondence with a correlation coefficient of 0.59. Second, the variation in the SO₂ column from 2010 to 2012 was calculated by averaging across Chinese provinces. Over eastern-central China, the SO₂ column increased, whereas over other provinces it decreased (Fig. 2(b)). Finally, the variation ratio over each province was used to estimate anthropogenic SO₂ emissions in 2012. The total amount from China was 76.2 Gg day^–1, which was slightly lower (–2.8%) than 78.4 Gg day^–1 in 2010. This decreasing ratio was fitted within two estimates (–0.7% and –7.9%) based on the scenario analysis of the bottom-up emission inventory (Xia et al., 2016). For the Republic of Korea, the variation from 2010 to 2012 of +4.0% officially reported by the National Institute of Environmental Research (NIER, 2018) was used, and the estimated SO₂ emissions were 1.2 Gg day^–1. For Japan, SO₂ emissions may have increased owing to higher power plant use after the Fukushima Daiichi nuclear disaster (e.g., Morino et al., 2011), but emissions were kept at the 2010 level of 1.9 Gg day^–1 due to the lack of detailed information. For other countries in the modeling domain, SO₂ emissions were also not adjusted. Ship emissions did not change dramatically from 2010 to 2015 unless the clean scenario was adopted (Eyring et al., 2005); hence, we did not change the ship emissions. The anthropogenic SO₂ emissions were calculated for this simulation (Fig. 2(c)). We adjusted the amount of SO₂ emissions and did not change the spatial distribution pattern and emission source categories. Biogenic emissions were prepared from the Model of Emissions of Gases and Aerosols from Nature (MEGAN) (Guenther et al., 2012). Biomass burning emissions were taken from the Global Fire Emissions Database (GFED) version 4.1 (van der Werf et al., 2017). SO₂ emissions from volcanoes were updated from our previous study (Itahashi et al., 2017a) based on the OMI-constrained volcanic SO₂ emission database (Carn et al., 2017). In total, 21 active volcanoes were considered in the modeling domain, including 2 volcanoes in the Philippines, 1 volcano in the Northern Mariana Islands, and 3 volcanoes in eastern Russia. For the remaining 15 volcanoes located in Japan, SO₂ emissions were replaced if Japan Meteorological Agency (JMA) observations were available (JMA, 2018). For the JMA observations, the SO₂ emissions were measured repeatedly at each observation, and minimum, mean, and maximum values were provided. The plume height for each volcano was also provided. Monthly average mean values were used because the measurements were not taken periodically. 7 of the Japanese volcanoes were located on Kyushu on the western edge of the Japanese archipelago and close to CHAAMS (Fig. 3). The details of volcanic SO₂ emissions from these 7 volcanoes are as follows. In 2012, the largest emissions were from Sakurajima, located in southern Kyushu, and were 2.81 and 1.91 Gg day^–1 in March and April, respectively. These emissions alone surpassed the amount of anthropogenic emissions from Japan. The second and third largest emissions of around 0.7 Gg day^–1 were from Suwanosejima and Asosan. The next largest were 0.45 Gg day^–1 from Satsuma-Iojima. The emissions of the other 3 volcanoes, Kirishimayama, Kuchinoerabujima, and Kujusan, were relatively small. Though there were no observations at Satsuma-Iojima and Kuchinoerabujima in the study period, two observations were conducted in this year and the observed average values were used (JMA, 2018). There were no reports from JMA of the emissions from Suwanosejima and Kujusan; thus, the amount of emissions from Suwanosejima was taken from the latest summary report averaged between 2000 and 2006, and considering the current low activity of Kujusan, the lowest value after 2004 was taken (Mori et al., 2013). Kirishimayama had a massive eruption in 2011, but the activity settled down in spring 2012 (Mori and Kato, 2013).

Fig. 2. (a) Relationship between Chinese province SO₂ emission amount and province-averaged SO₂ column. (b) Variation ratio of SO₂ column observed by OMI from 2010 to 2012. (c) Estimated SO₂ emissions in spring 2012 used in this study.

Fig. 3. Expanded map over Kyushu, Japan, showing the location of the volcanoes from which the SO₂ emissions considered in this study came.

To estimate the sources of SO₄^2–, we used the Particulate Source Apportionment Technology (PSAT) method in CAMx (Wagstrom et al., 2008). This method assigns tracers to track the sources; hence, we refer to this method as the “tagged tracer method.” This method was modified to use SAPRC-07 chemistry. The usefulness of the tagged tracer method for SO₄^2– over East Asia with respect to the validity of the linear assumption has been reported previously (Itahashi et al., 2017a, b; Itahashi, 2018). In this study, we prepared seven source groups: five sources for anthropogenic emissions from mainland China, Taiwan, the Korean Peninsula, Japan, and other regions in the modeling domain, including the lateral boundary condition; one source for ship emissions; and one source for volcano emissions as a natural source. In this work, the source region of Japan did not include Okinawa; therefore, the source apportionment for Japan represented LRT from mainland Japan. For the detailed study in March and April characterized by different sources, we set more source groups (see “Results and Discussion”).

RESULTS AND DISCUSSION

Model Evaluation and Estimation of SO₄^2– Sources at CHAAMS

For the meteorological field, the model results were compared with observations at CHAAMS (Fig. 4). Although the wind speed was slightly overestimated, its temporal variation was captured (Fig. 4(a)). The model overestimation was partly because CHAAMS was resolved as an ocean grid at the current model resolution, as in our previous study (Itahashi et al., 2017a). The wind direction was captured well by the model. In March, the wind direction was northerly, and then slowly changed from easterly to southerly from March 13 to 18. In April, the wind direction changed from northerly to easterly, then southerly from April 23 to 26, and repeatedly changed from April 27 to 29 (Fig. 4(b)). The temperature was 15°C in March and increased to around 20°C in April (Fig. 4(c)). In March, there were no precipitation events greater than 1 mm, but in April, several precipitation events occurred. On April 22, precipitation of 0.28 mm was observed at 05:00 LT and the model overestimated this event as 1.9 mm. In contrast, from late April 24 to early on the 25, there was high precipitation of 10.28 mm, but the model substantially underestimated the event as 1.3 mm (Fig. 4(d)). Capturing local precipitation events was difficult for the meteorological model. The relative humidity was just below 80% in March, although it reached 100% during the precipitation event on April 25, and then decreased to 60% for the rest of April (Fig. 4(e)).

Fig. 4. Temporal variation of the meteorological fields of (a) wind speed, (b) wind direction, (c) temperature, and (d) relative humidity for observations (gray) and the model (black).

The modeled SO₂ and SO₄^2– concentrations were compared with observations from CHAAMS (Fig. 5). The statistical analysis of the model results and observations based on the period mean, correlation coefficient (R) with significance level determined by Student’s t-test, mean fractional bias (MFB), and mean fractional error (MFE) are listed in Table 1. In Fig. 5, the symbol for SO₄^2– is colored according to the observed and modeled conversion ratio (Fs) from SO₂ to SO₄^2–, which is calculated from the molar concentrations by:

This ratio represents the ratio of SO₄^2– to the sum of SO₂ and SO₄^2–. Because the local SO₂ emissions at CHAAMS are negligible, Fs was used as the index of the LRT impact to identify whether the air mass was fresh (lower Fs) or aged (higher Fs). In our previous estimation of sources for the autumn case study at CHAAMS, Fs values below 70% were strongly related to volcanic contributions due to their fast transport at higher altitudes, and Fs near 100% indicated a well-aged air mass from LRT (Itahashi et al., 2017a).

Fig. 5. Temporal variation of (a) SO₂ and (b) SO₄^2– concentrations. The whiskers indicate the standard deviation of the daily mean value. Observations and model results are shown in white and black, respectively, in (a). Observed and modeled Fs values are shown in red in (b). (c) Temporal variation in modeled source apportionments.

For SO₂ (Fig. 5(a)), based on the environmental standard criteria in Japan, days when there were less than 20 h of hourly observations were regarded as deficit days and are not shown (i.e., March 13, 16, and 17; April 24 and 25). On March 11 and 15, the SO₂ concentration surpassed 1 ppbv, whereas in April, SO₂ concentrations were low at less than 0.5 ppbv. These observed features were captured well by the model. The statistical analysis in Table 1 shows that the observed and modeled mean concentrations were similar and were significantly correlated (R = 0.90, p < 0.001). During the observation period in 2012, there were unusually high concentrations of SO₄^2– in March and April (Fig. 5(b)), as shown in Fig. 1. On March 15, the daily mean SO₄^2– concentration was 16.91 µg m^–3. However, the model did not reproduce this high concentration. In April, the highest daily mean SO₄^2– concentration of 15.87 µg m^–3 was observed on April 23 and the second highest of 8.08 µg m^–3 was observed on April 22, and these high concentrations were simulated well by the model. From the statistical analysis (Table 1), MFB and MFE for SO₄^2– were within the performance goal criteria proposed by Boylan and Russell (2006) for models. During the observation campaign in spring 2012, Fs values were well reproduced (Table 1), suggesting that the air mass aging at CHAAMS was captured by the model in addition to the concentrations of SO₂ and SO₄^2–.

The sources of SO₄^2– estimated by the tagged tracer method are shown as the daily mean in Fig. 5(c). High concentrations on March 14 and 15 were characterized by lower Fs values (Fig. 5(b)) and the wind direction was northerly to easterly (Fig. 4(b)); hence, volcanoes were identified as an important source. By contrast, during the high-concentration episode in April, featuring a well-aged air mass of higher Fs (Fig. 5(b)) and change in wind direction from north to south, the anthropogenic emissions from China were identified as the dominant source. In the following sections, we analyze these unusual high-concentration episodes that occurred in March and April further by focusing on volcanic emissions and anthropogenic emissions from China, respectively.

SO₄^2– Sources for the High-concentration Episode in March

Three unusually high concentrations of 11.68, 10.83, and 16.91 µg m^–3 were observed on March 11, 14, and 15, respectively; however, these concentrations were all underestimated by the model. On March 11, compared with the other two days of high concentrations, the standard deviation of the model used to derive the daily mean was large (Fig. 5(b)), and the hourly modeled SO₄^2– concentration ranged from 1.5 to 16.2 µg m^–3. In addition, the model slightly underestimated Fs. Therefore, a more stagnant air mass may explain this peak. Compared with March 11, the underestimation on March 14 and 15 was problematic because the model consistently underestimated the daily means, as suggested by the small standard deviation values. We analyzed the model results on these two days further to explain the reason of underestimation. The modeled spatial distributions on these days were mapped (Fig. 6). On both days, a high concentration of SO₄^2– greater than 10 µg m^–3 limited to over China and a high concentration from southern Kyushu to CHAAMS were found with the prevailing northerly to easterly wind (see also Fig. 4(b)). The source apportionments by the tagged tracer method indicated the important source as volcanoes (Fig. 5). On March 14 and 15, the contribution of volcanoes was the largest at 70.2% and 40.8%, surpassing that of anthropogenic emissions from China of 12.0% and 38.4%, respectively. Kyushu has many active volcanoes, and our previous studies demonstrated that volcanoes are an important source of SO₄^2– in Japan (Itahashi et al., 2017a, b; Itahashi, 2018). Therefore, we examined the source apportionment of individual volcanoes in Japan further. Considering the SO₂ emissions (Fig. 3), the source apportionments of four main active volcanoes of Sakurajima, Suwanosejima, Asosan, and Satsuma-Iojima on Kyushu are shown (Fig. 7). Sakurajima, which emitted the largest amount of SO₂ that surpassed the anthropogenic emissions from Japan (Fig. 3), was identified as the dominant volcano source in March (Fig. 7(a)). Its apportionments were estimated as 2.49 and 3.05 µg m^–3 on March 14 and 15, which were 34.7% and 32.6% of the modeled SO₄^2–concentration, respectively. The second largest SO₂ emitter, Suwanosejima, located nearest to CHAAMS (Fig. 7(b)) had a contribution of 1.28 µg m^–3 on March 14, whereas its contribution was negligible (0.05 µg m^–3) on March 15 due to the strong easterly wind around CHAAMS (Figs. 4(b) and 6). The spatial distributions of the source apportionments of Asosan and Satsuma-Iojima showed similar patterns to that of Sakurajima. Due to the smaller SO₂ emissions from these volcanoes, on March 14 and 15, the apportionments of Asosan were 0.21 and 0.37 µg m^–3, and those of Satsuma-Iojima were 0.84 and 0.38 µg m^–3, respectively. These four main volcanoes on Kyushu accounted for over 97% of the source apportionment of the total volcanic sources, and the spatial distribution of these volcanoes corresponded well to the high-concentration flow from Kyushu to CHAAMS (Fig. 6).

Fig. 6. Spatial distribution of modeled SO₄^2– concentrations on March 14 and 15, 2012.

Fig. 7. Spatial distribution of modeled source apportionments for volcanic sources of (a) Sakurajima, (b) Suwanosejima, (c) Asosan, and (d) Satsuma-Iojima on (top) March 14, 2012, and (bottom) March 15, 2012. The values of the source apportionments at CHAAMS are shown in the bottom-right corner of each panel. Note the different color scale.

Further sensitivity simulations were conducted for Sakurajima, which was the main source for the high-concentration episode in March. Considering the uncertainties of SO₂ emissions from volcanoes, sensitivity simulations were performed for the emission amount and injection height (Table 2). In these sensitivity simulations, the SO₂ emission amounts from Sakurajima were increased to the estimated maximum value of 4.00 Gg day^–1. This was 1.43 times the mean value. The JMA reported minimum, mean, and maximum SO₂ measurements for each observation, indicating the uncertainty of SO₂ measurements. A large eruption on March 12 was reported; therefore, we used the maximum values and this setting increased the SO₄^2– concentration (Case A). The other parameter considered in the sensitivity simulations was the injection height of volcanic SO₂ emissions, because the plume height of a volcano can vary widely. In the standard simulation, the average injection height was about 1000 m, and this value was reduced to 750 m (Case B) and increased to 1500 m (Case C) according to the estimated minimum and maximum injection heights.

The results of these sensitivity simulations as changes from the standard simulation (Fig. 7(a)) are shown in Fig. 8. The spatial distributions on March 14 and 15 and the curtain plot averaged from 126° to 133° longitude (region in the pink rectangle) to cover the volcanic plume within 3 km are shown. As expected, the maximum SO₂ emissions from Sakurajima increased the SO₄^2– concentration at CHAAMS in Case A. These increases were 0.58 and 1.02 µg m^–3 on March 14 and 15, respectively. The curtain plot clarified that these increases were effectively found within the planetary boundary layer (PBL) around 1000 m over the ocean. Therefore, the lowest injection height in Case B increased the contribution from Sakurajima to the SO₄^2– concentration at CHAAMS further, by 1.07 and 0.95 µg m^–3 on March 14 and 15, respectively. In contrast, the highest injection height, approximately 1500 m, in Case C was above the PBL; hence, the effects were over a broader downwind region, and the impact on the SO₄^2– concentration at CHAAMS was reduced. The statistical analysis confirming the model performance in March 2012 is given in Table 3. In the standard simulation, the model performed within the performance goal criteria proposed by Boylan and Russell (2006), and the performance was improved further in Cases A and B, whereas it deteriorated by a non-significant amount in Case C. The sensitivity simulations demonstrated the importance of the injection height as well as the emission amount in accurately capturing the impact of volcanoes.

Fig. 8. Change of source apportionment from Sakurajima as a (top) spatial distribution and (bottom) curtain plot on March 14 and 15, 2012, through (a) Case A, (b) Case B, and (c) Case C sensitivity simulations. The changed source apportionment values at CHAAMS are shown in the bottom-right corner of each spatial distribution panel. For the curtain plot, the source apportionment values are averaged over 126° to 133° longitude (pink rectangle) and are shown from 0 to 3 km. The thick line indicates the modeled PBL height.

SO₄^2– Sources for the High-concentration Episode in April

Source apportionments for China were found throughout April 2012, and the high concentration observed on April 12 was captured well by the model. A similar dominant effect from China was previously reported based on aerial observations over the East China Sea during spring (Hatakeyama et al., 1997, 2011). We have already reported seasonal and annual mean source apportionments over Japan (Itahashi et al., 2017b). To improve our knowledge of source apportionment at specific sites, we have reported the source apportionment at CHAAMS during the intensive field campaign in autumn 2015 (Itahashi et al., 2017a). To achieve detailed source apportionment for China, mainland China was separated into 31 regions (22 provinces, 4 municipalities, and 5 autonomous regions). Due to the horizontal grid resolution of 36 km, the special administrative regions of Hong Kong and Macau were included in Guangdong Province. We focused on provincial-scale source apportionment, and then we estimated and compared the averaged and daily specific source apportionment.

The modeled spatial distributions of SO₄^2– on April 22 and 23, when the highest concentration was obtained, are shown in Fig. 9. The isolated high-concentration plume over the East China Sea to Kyushu and CHAAMS was found on April 22, and the polluted air mass from northeastern China to the Korean Peninsula, and western Japan was observed on April 23. The source apportionments of mainland China are shown in Fig. 10. The high-concentration plume in Fig. 9 corresponded well with the source apportionment from China. The apportionments for China were 7.26 and 8.84 µg m^–3 on April 22 and 23 and these were 76.8% and 64.8% of the modeled SO₄^2– concentration, respectively. Throughout April 2012, mainland China was the largest SO₄^2– source with a contribution of 3.82 µg m^–3 accounting for 62.1% of the modeled concentration.

Fig. 9. Spatial distribution of modeled SO₄^2– concentration on April 22 and 23, 2012.

Fig. 10. Spatial distribution of modeled source apportionment for anthropogenic emissions from (a) mainland China, (b) Shandong Province, and (c) Jiangsu Province on (top) April 22, 2012, and (bottom) April 23, 2012. The source apportionment values at CHAAMS are shown in the bottom-right corner of each panel. The color scale in (a) is different from those in (b, c). 

In the detailed source apportionment results in Fig. 11, each source apportionment for 31 source regions in China was averaged over April, and over April 22 and 23. The emission amounts estimated in this study are also shown in the order of the amount. Shandong and Jiangsu Provinces, which are the first and seventh largest anthropogenic SO₂ emission sources in China, were revealed as the important sources in April. The averaged source apportionment from Shandong Province was 1.22 µg m^–3, and that from Jiangsu Province was 1.12 µg m^–3. These apportionments accounted for relative percentages of 32.0% and 29.4%, respectively, of the total apportionments of anthropogenic emissions from China. Other important sources regions were Zhejiang, Hebei, and Anhui Provinces; these apportionments were less than 0.2 µg m^–3, and accounted for relative percentages of around 5%. The averages revealed that the emissions from Shandong and Jiangsu Provinces were dominant. However, we examined the day-to-day variation, especially during the high-concentration episode, in greater detail. The spatial distributions for Shandong and Jiangsu Provinces are shown in Figs. 10(b) and 10(c). On April 22, the apportionment from Shandong Province was 1.86 µg m^–3, accounting for a relative percentage of 25.6% of the total apportionments of anthropogenic emissions from China, and that from Jiangsu Province was 3.75 µg m^–3, accounting for 51.6%. Both apportionments were found around CHAAMS as an isolated air mass. On April 23, the apportionment from Shandong Province was 3.80 µg m^–3 accounting for 43.0%, and that from Jiangsu Province was 2.46 µg m^–3 accounting for 27.8%. These apportionments were found from the continent to the East China Sea and CHAAMS. The important source regions were still Shandong and Jiangsu Provinces; however, the first and second rankings were switched during the high-concentration episode on April 22 and 23. In spring, these two regions were key contributors to the SO₄^2– concentration at CHAAMS, and other seasons with different prevailing wind directions should be investigated further.

Fig. 11. Source apportionment averaged during the campaign periods of April (leftmost axis) and April 22 and April 23 (left axis). The 31 regions of China are in the order of estimated SO₂ emission amount (right axis).

CONCLUSIONS

The continuous observation campaign from 2010 to 2015 at Cape Hedo, in the northernmost part of Okinawa, Japan (CHAAMS), reported high-concentration episodes in spring 2012. In March and April of that year, the day-to-day variation in SO₄^2– at CHAAMS was evaluated using an air quality model with the tagged tracer method. The air quality model captured the observed meteorological conditions and the variation in SO₂, SO₄^2–, and Fs. The results demonstrated that the high-concentration episodes in March 2012 were caused by volcanoes in Japan, whereas those in April were caused by anthropogenic emissions from China.

In March, Sakurajima, a volcano in southern Kyushu, was the main SO₂ source, and its emissions surpassed the anthropogenic emissions from Japan in amount. Although the model performed within the performance goal criteria for March 2012, it underestimated the high-concentration peaks found on March 14 and 15. Therefore, sensitivity simulations of the emission amounts and injection heights for Sakurajima were performed. The uncertainty of SO₂ emission amounts from volcanoes was one of the main reasons for the underestimation, and considering the maximum emission amount improved the model performance. In addition, the results revealed the importance of the injection height of volcanic SO₂ plumes, and considering the minimum injection height, which is related to the PBL height, also improved the model performance for this episode.

In April, the SO₄^2– was mainly attributable to anthropogenic emissions from China, and the model captured the high-concentration episode on April 22 and 23 well. The well-aged air mass suggested by higher Fs was observed. During the observation period in April, Shandong and Jiangsu Provinces were detected as the main source regions, and their first and second rankings were switched during the April 22–23 episode. These day-to-day variations identified via source apportionment will improve our understanding of the high-concentration episodes at CHAAMS.

ACKNOWLEDGMENTS

This research was supported by the Global Environment Research Fund (5-1751 and 2-1703) of the Ministry of the Environment of Japan. The authors acknowledge using MOZART-4 data for the global chemical transport model (http://www.acom.ucar.edu/wrf-chem/mozart.html) for the lateral boundary condition. The authors also thank JMA and other institutions for performing SO₂ observations of Japanese volcanoes.