Measurement report: Source characteristics of water-soluble organic carbon in PM_2.5 at two sites in Japan, as assessed by long-term observation and stable carbon isotope ratio

The sources and seasonal trends of water-soluble organic carbon (WSOC) in carbonaceous aerosols are of significant interest. From July 2017 to July 2019, we collected samples of PM_2.5 (particulate matter, aerodynamic diameter<2.5 µm) from one suburban and one rural site in Japan. The average δ¹³C_WSOC was $-\mathrm{25.2}\pm \mathrm{1.1}$ ‰ and $-\mathrm{24.6}\pm \mathrm{2.4}$ ‰ at the suburban site and rural site, respectively. At the suburban site, the δ¹³C_WSOC was consistent with the δ¹³C of burned C₃ plants, and a high correlation was found between WSOC concentrations and non-sea-salt potassium concentrations; these results suggest that the main source of WSOC at this site was biomass burning of rice straw. At the rural site, the average δ¹³C_WSOC was significantly heavier from autumn to spring ($-\mathrm{23.9}\pm \mathrm{2.1}$ ‰) than in summer ($-\mathrm{27.4}\pm \mathrm{0.7}$ ‰) (p<0.01). The δ¹³C_WSOC from autumn to spring was consistent with that of biomass burning of rice straw, whereas that in summer was considered to reflect mainly the formation of secondary organic aerosols from biogenic volatile organic compounds (VOCs). The heaviest δ¹³C_WSOC ($-\mathrm{21.3}\pm \mathrm{1.9}$ ‰) was observed from February to April 2019, which may be explained by long-range transport of C₄ plant burning such as corn from overseas. Thus, the present study indicates that δ¹³C_WSOC is potentially useful for elucidating the sources and atmospheric processes that contribute to seasonal variations of WSOC concentration.

1 Introduction

Particulate matter (PM) has deleterious effects on human health and contributes to climate change (Pope et al., 1995; Lohmann and Feichter, 2005). A major component of PM_2.5 (particulate matter, aerodynamic diameter<2.5 µm) is carbonaceous aerosol, which comprises organic carbon (OC) and elemental carbon (EC) (Chow et al., 1993; Malm et al., 2004; Pöschl, 2005). The OC in carbonaceous aerosol can be further classified as water-insoluble organic carbon (WIOC) and water-soluble organic carbon (WSOC) (Sullivan and Weber, 2006). WIOC is produced mainly by the combustion of fossil fuels and contains compounds such as alkanes (Pöschl, 2005). WSOC is emitted primarily from combustion processes, industrial process, and natural sources; it can also be formed through secondary processes such as homogeneous gas-phase or heterogeneous aerosol-phase oxidation (Claeys et al., 2004; Koch et al., 2007; Schichtel et al., 2008). WSOC accounts for 20 %–80 % of the total OC in carbonaceous aerosol depending on the location and season (Decesari et al., 2001; Sullivan et al., 2004; Du et al., 2014; Duarte et al., 2015; Zhang et al., 2019). In addition, an average of 74 % of all WSOC is contained in fine particles (Yu et al., 2004). WSOC is hygroscopic; therefore, it enhances the capability of aerosols to act as cloud condensation nuclei, which affects climate change (Padró et al., 2010; Asa-Awuku et al., 2011). Therefore, source contributions of WSOC have been of significant interest for decades. A common approach for estimating the source contributions of WSOC is the use of a positive matrix factorization model. Using this approach, the annual contributions of biomass burning and secondary processes to WSOC in Beijing, China, were estimated to be 40 % and 54 %, respectively (Du et al., 2014). Similarly, in Helsinki, Finland, the contribution of secondary organic aerosols (SOAs) to WSOC is reported to be high in summer (78 %) but low in winter (28 %) (Saarikoski et al., 2008). WSOC is known to contain various oxygenated compounds, including dicarboxylic acids, ketocarboxylic acids, aliphatic aldehydes, alcohols, saccharides, saccharide anhydrides, aromatic acids, phenols, amines, amino acids, organic nitrates, and organic sulfates (Duarte et al., 2007, 2015; Pietrogrande et al., 2013; Timonen et al., 2013; Chalbot et al., 2014). However, the precise molecular composition of WSOC is poorly understood because of the large number of compounds involved and the difficulties involved in identifying the individual components.

The stable carbon isotope ratio (δ¹³C) of carbonaceous aerosols can provide useful information about a sample of PM (e.g., Widory et al., 2004; Fisseha et al., 2009; Cao et al., 2011; Gensch et al., 2014). For example, because EC is unreactive, it is possible to identify its source directly from the δ¹³C of its aerosols (e.g., Kawashima and Haneishi, 2012; Zhao et al., 2018). In contrast, because OC reacts in the atmosphere, its δ¹³C provides information not only about the source of the PM but also about any atmospheric processing it has undergone (e.g., Cao et al., 2011; Ni et al., 2018). In recent years, some groups have examined the δ¹³C of WSOC (δ¹³C_WSOC) in PM (Kirillova et al., 2010, 2013; Suto and Kawashima, 2018; Zhang et al., 2019). In addition, various approaches have been used; for example, the δ¹³C_WSOC of ambient aerosols has been examined by means of wet oxidation with GasBench and isotope-ratio mass spectrometry (IRMS) (Fisseha et al., 2006) and by means of combustion with an elemental analyzer and IRMS (EA–IRMS) (Kirillova et al., 2010). Recent advances have afforded highly sensitive analytical methods for determining δ¹³C_WSOC values that use wet oxidation with liquid chromatography and IRMS (LC–IRMS) (Suto and Kawashima, 2018), GasBench and IRMS (Zhang et al., 2019), or total organic carbon analyzer and IRMS (Han et al., 2020); however, combustion-based approaches remain the most widely used.

The δ¹³C_WSOC of particles of various sizes collected at various times of the year in East Asia (Miyazaki et al., 2012; Kirillova et al., 2014a; Pavuluri and Kawamura, 2017; Yan et al., 2017; Suto and Kawashima, 2018; Zhang et al., 2019; Han et al., 2020), South Asia (Kirillova et al., 2013, 2014b; Bosch et al., 2014, Dasari et al., 2019), Europe (Fisseha et al., 2006, 2009; Kirillova et al., 2010), and the United States (Wozniak et al., 2012a, b) have been reported (Table S1 in the Supplement). For example, the δ¹³C of total carbon (δ¹³C_TC) and δ¹³C_WSOC of total suspended particles (TSPs) was observed from September 2009 to October 2010 in Hokkaido, Japan (Pavuluri and Kawamura, 2017). Both δ¹³C_TC and δ¹³C_WSOC were heavier in winter than in summer, demonstrating seasonal variation. The authors concluded that the reason why δ¹³C_WSOC was heavy in winter was because of the greater release of ¹³C by fossil fuel combustion and biomass burning. Similarly, Kirillova et al. (2013) collected TSP samples from January 2008 to April 2009 in Sinhagad, India, and Hanimaadhoo Island, Maldives. The average δ¹³C_WSOC was $-\mathrm{20.4}\pm \mathrm{0.5}$ ‰ in Sinhagad and $-\mathrm{18.4}\pm \mathrm{0.5}$ ‰ in Hanimaadhoo Island, which are heavier than values reported in other studies. In addition, aerosols reaching Hanimaadhoo Island after long-range over-ocean transport were enriched by 3 ‰–4 ‰ in δ¹³C_WSOC relative to the aerosols collected in Sinhagad. Based on these findings, Kirillova et al. (2013) reported for the first time that this enrichment of δ¹³C was an effect related to the aging of OC during long-range transport of aerosol. A recent study reported that the enrichment of δ¹³C_WSOC between source site (Delhi, India) and receptor site (Hanimaadhoo Island, Maldives) is caused by an aging effect during long-range transport (Dasari et al., 2019).

The combustion method, which is widely used at present, requires more pretreatment time, because samples of PM are extracted, dehydrated with a freeze drier, dried, and then measured by EA–IRMS. The wet-oxidation–IRMS method described above does not require a drying stage during sample preparation; therefore, the total analysis time is markedly reduced compared with the combustion method. In addition, this newer approach is highly sensitive, so only small amounts of sample are needed compared to the combustion method. However, despite these improved approaches and the significant interest in the seasonal trends and source apportionment of WSOC, no studies have examined the change of δ¹³C_WSOC in PM_2.5 over a long period of time to understand seasonal variability. As mentioned above, the small particle size PM_2.5 contains large amount of WSOC, so further investigations are needed. Here, we investigated the seasonal trends of WSOC at one suburban site and one rural site in Japan. Samples of PM_2.5 were collected from July 2017 to July 2019 at both sites, and δ¹³C_TC and δ¹³C_WSOC values, as well as carbon component and water-soluble ion concentrations, were determined. We then characterized the source of WSOC and any atmospheric processes it had undergone using isotope-based approaches. We believe that this is the first report of the use of the wet-oxidation–IRMS method (Suto and Kawashima, 2018) for long-term observation of δ¹³C_WSOC.

2 Materials and experimental methods

2.1 Sampling sites and sample collection

Samples of PM_2.5 were collected at one suburban site and one rural site in Japan (Fig. S1 in the Supplement). The suburban site (Tsukuba; $\mathrm{36}{}^{\circ }{\mathrm{4}}^{\prime }$ N, $\mathrm{140}{}^{\circ }{\mathrm{4}}^{\prime }$ E) was on the rooftop of a 25 m high building at the Japan Automobile Research Institute in Tsukuba City, Ibaraki Prefecture, Japan. Tsukuba is a suburban city located in the inland Kanto plain approximately 60 km northeast of the Tokyo metropolitan area. This site is surrounded by residential areas and forests, and there is a road in front of the building. PM_2.5 samples were collected approximately every 10 d from 19 July 2017 to 12 July 2019. The rural site (Yurihonjo; $\mathrm{39}{}^{\circ }{\mathrm{23}}^{\prime }$ N, $\mathrm{140}{}^{\circ }{\mathrm{4}}^{\prime }$ E) was on the campus of Akita Prefectural University in Yurihonjo City, Akita Prefecture, Japan. Yurihonjo is located 370 km northwest of Tsukuba and about 5 km away from the coast. The sampling site had no local pollutant sources such as large factories. Every year from December to February, the site is covered with several centimeters of snow (Japan Meteorological Agency, 2020). PM_2.5 samples were collected approximately every 14 d from 11 August 2017 to 5 July 2019.

At both sites, the PM_2.5 samples were collected by using high-volume samplers (HV-1000F, Sibata Scientific Technology, Saitama, Japan) equipped with a PM_2.5 impactor (HV-1000-PM_2.5, Sibata Scientific Technology) at a flow rate of approximately 1000 L min⁻¹. The samples were collected on quartz fiber filters (20.3 cm×25.4 cm, 2500QAT-UP, Pallflex, Putnam, USA) that had been prebaked at 550 ^∘C for 4 h before use. After sampling, the filters were kept in a freezer at −30 ^∘C. A total of 107 PM_2.5 samples (62 samples from Tsukuba and 45 samples from Yurihonjo) were collected. PM_2.5 mass concentration was analyzed gravimetrically by using an electronic balance before and after sampling.

2.2 Stable carbon isotope ratio analysis

Determination of δ¹³C_TC was performed at the Japan Automobile Research Institute using EA–IRMS (EA IsoLink, Thermo Fisher Scientific, Bremen, Germany; Delta V Advantage, Thermo Fisher Scientific, respectively). Portions of quartz filter (5–10 mg) were packed into a tin cup. The samples were combusted instantaneously with oxygen in the EA, and the carbon was converted to CO₂ via an oxidation/reduction tube of the EA. The oxidation/reduction tube and the packed column were maintained at 1020 and 60 ^∘C, respectively. The flow rate of ultra-high-purity helium during the analysis was 180 mL min⁻¹. The CO₂ from the EA was ionized, and the δ¹³C value was determined by means of IRMS; data acquisition was performed with Isodat software (ver. 3.0, Thermo Fisher Scientific).

Determination of δ¹³C_WSOC was performed at Akita Prefectural University using the wet-oxidation–IRMS method (Kawashima et al., 2018; Suto and Kawashima, 2018). A portion of each quartz fiber filter (14.13 cm²) was extracted in 5 mL of Milli-Q water under ultrasonic agitation for 30 min. The extract was filtered through a syringe filter (Chromatodisc type A 0.45 µm, GL Sciences, Japan) to remove insoluble material. The PM_2.5 samples were not decarbonated before δ¹³C_WSOC analysis, because the difference between the δ¹³C_WSOC with and without hydrochloric acid pretreatment was within 0.2 ‰. A high-performance liquid chromatography (HPLC) system (Shimadzu Co.) was coupled to the IRMS instrument (Isoprime, Elementar UK, Manchester, UK) via a LiquiFace interface (Elementar UK). The HPLC system consisted of a column pump (LC-10ADvp), oxidation pump (LC-10ADvp), post-column pump (LC-10ADvp), autosampler (SIL-10ADvp), degasser (DGU-14A), and UV detector (SPD-10ADvp). The injection volume was 100 µL. The HPLC flow rate (without column), the sodium peroxodisulfate flow rate, and the post-column flow rate were 0.5, 0.4, and 0.3 mL min⁻¹, respectively. Sodium peroxodisulfate (0.5 M) and phosphoric acid (0.2 M) were mixed and then degassed in an ultrasonic bath for 1 h. One run took about 6 min. The trap current was set at 300 µA. The limits of detection (precision, $<\pm \mathrm{0.3}$ ‰; accuracy, $<\pm \mathrm{0.3}$ ‰) for levoglucosan and oxalic acid were 1111 and 1133 ng C, respectively.

The IRMS instrument and the data acquisition system were controlled by IonVantage NT software (ver. 1.5.4.0., Isoprime). The HPLC system was controlled by LCsolution software (ver. 1.25, Shimadzu Co.).

Stable carbon isotope ratios are expressed in δ notation in per mil (‰):

$$\begin{array}{}\text{(1)}& {\mathit{\delta }}^{\mathrm{13}}\mathrm{C}\left[\mathrm{‰}\right]=\left({\displaystyle \frac{R{\left({}^{\mathrm{13}}\mathrm{C}/{}^{\mathrm{12}}\mathrm{C}\right)}_{\text{sample}}}{R{\left({}^{\mathrm{13}}\mathrm{C}/{}^{\mathrm{12}}\mathrm{C}\right)}_{\text{std}}}}-\mathrm{1}\right),\end{array}$$

where $R({}^{\mathrm{13}}\mathrm{C}{/}^{\mathrm{12}}\mathrm{C}{)}_{\text{sample}}$ and $R({}^{\mathrm{13}}\mathrm{C}{/}^{\mathrm{12}}\mathrm{C}{)}_{\text{std}}$ (=0.0111802) are the ${}^{\mathrm{13}}\mathrm{C}{/}^{\mathrm{12}}\mathrm{C}$ ratios for the sample and the standard (Vienna Pee Dee Belemnite), respectively. For all samples, the EA–IRMS and wet-oxidation–IRMS data were measured in triplicate.

A two-point linear calibration was carried out for δ¹³C (Coplen et al., 2006). For EA–IRMS, δ¹³C_TC values and three internal laboratory standards were calculated by using the following international isotopic standards: IAEA-CH-3 (cellulose, ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}=-\mathrm{24.724}$ ‰), IAEA-600 (caffeine, ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}=-\mathrm{27.771}$ ‰), and USGS24 (graphite, ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}=-\mathrm{16.049}$ ‰). These standards were obtained from the International Atomic Energy Agency (Vienna, Austria). As a check of instrumental stability, an isotope working standard (l-alanine, SI Science Co., Tokyo, Japan; ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}=-\mathrm{19.9}$ ‰) was analyzed after every nine samples. For wet-oxidation–IRMS, δ¹³C values were calculated by means of a two-point linear calibration method from international isotope standards of sucrose (IAEA-CH-6, ${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}=-\mathrm{10.449}$ ‰) and three internal laboratory standards for D-(+)-arabitol (${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}=-\mathrm{23.6}$ ‰), levoglucosan (${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}=-\mathrm{25.8}$ ‰), and oxalic acid (${\mathit{\delta }}^{\mathrm{13}}\mathrm{C}=-\mathrm{28.7}$ ‰) obtained from EA–IRMS. Ultrapure water was prepared with a Milli-Q system (18.2 MΩ cm; Millipore, Bedford, MA, USA). To check instrumental stability, the laboratory standard of levoglucosan was analyzed after every nine samples. The average 1SD for δ¹³C_TC and δ¹³C_WSOC was 0.12 ‰ (<0.46 ‰) and 0.09 ‰ (<0.50 ‰), respectively, for all samples examined in the present study.

2.3 Chemical analysis

For determination of OC and EC concentrations, a portion of each quartz fiber filter (0.53 cm²) was examined using a thermal-optical carbon analyzer (model 2001, Desert Research Institute), and the samples were processed according to the IMPROVE (Interagency Monitoring of Protected Visual Environments) thermal desorption and optical reflectance method with a 550 ^∘C temperature, split for OC and EC (Chow et al., 2001). The limits of detection for OC and EC were determined as 3 times the standard deviation of a blank filter, and they were 0.02 and 0.02 µg m⁻³, respectively. These limits of detection were sufficiently low (Yamagami et al., 2019). For determination of WSOC concentrations, a portion of each quartz fiber filter (1.58 cm²) was extracted with 8 mL of ultrapure water for 30 min at room temperature. The water extracts were passed through a polyvinylidene difluoride filter (pore size 0.20 µm, GE Healthcare, USA) to remove insoluble materials, and then the filtrate was analyzed using a total organic carbon analyzer (TOC-L, Shimadzu, Kyoto, Japan). The limit of detection was determined as 3 times the standard deviation of a blank filter, and it was 0.03 µg m⁻³, which was sufficiently low (Du et al., 2014). Quantification of the major water-soluble anions (Cl⁻, ${\mathrm{NO}}_{\mathrm{2}}^{-}$, ${\mathrm{NO}}_{\mathrm{3}}^{-}$, ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$) and cations (Na⁺, ${\mathrm{NH}}_{\mathrm{4}}^{+}$, K⁺, Mg²⁺, Ca²⁺) was achieved by ion chromatography (Integrion RFIC; Thermo Fisher Scientific Inc., Sunnyvale, CA, USA). Details of the water-soluble ion analysis method are described in Supplement Sect. S1.

3 Results and discussion

3.1 Mass concentrations of PM_2.5 at the study sites

The average mass concentrations of PM_2.5 during the observation period were 19.7±8.2 µg m⁻³ (range, 7.1–46.6 µg m⁻³) in Tsukuba and 11.2±4.7 µg m⁻³ (5.7–23.4 µg m⁻³) in Yurihonjo (Table 1). The average mass concentration of PM_2.5 in Tsukuba was higher than the air quality standard for the annual average of Japan (15 µg m⁻³) by the Ministry of the Environment and those at other residential sites across Japan (annual average in 2018: 11.2 µg m⁻³) (Ministry of the Environment, 2019). In Yurihonjo, the average mass concentration of PM_2.5 was lower than the air quality standard for the annual average of Japan, and it was comparable with those at other residential sites across Japan.

Table 1 Seasonal average concentrations of PM_2.5, EC, OC, and WSOC; $\mathrm{OC}/\mathrm{EC}$ and $\mathrm{WSOC}/\mathrm{OC}$ ratios; and δ¹³C_TC and δ¹³C_WSOC values for PM_2.5, in Tsukuba and Yurihonjo, Japan.

A previous study reviewed the annual PM_2.5 concentrations in 45 global megacities in 2013 (Cheng et al., 2016). The five most-polluted megacities were Delhi, India; Cairo, Egypt; and Xi'an, Tianjin, and Chengdu, China (PM_2.5 annual average concentration, 89–143 µg m⁻³). The five least-polluted megacities were Toronto, Canada; Miami, Philadelphia, and New York, United States; and Madrid, Spain (PM_2.5 annual average concentration, 7–10 µg m⁻³). The mass concentration of PM_2.5 at both sites in the present study was much closer to that determined for the least-polluted megacities than that determined for the most-polluted megacities. The mass concentrations of PM_2.5 in Tsukuba and Yurihonjo were significantly higher in winter and spring than in summer and autumn (p<0.01). The mass concentrations of PM_2.5 were consistent with the seasonal variation for nearby sites of the Atmospheric Environmental Regional Observation System (AEROS) provided by the Ministry of the Environment (Ministry of the Environment, 2021).

3.2 Concentrations of EC, OC, and WSOC, as well as OC $/$ EC and WSOC $/$ OC ratios

The concentrations of EC, OC, and WSOC, as well as the $\mathrm{OC}/\mathrm{EC}$ and $\mathrm{WSOC}/\mathrm{OC}$ ratios, at the study sites are summarized in Table 1. The concentrations of the carbon components (EC, WIOC, and WSOC) by season are shown in Fig. 1. The sum of EC and organic matter (1.6×OC concentration) (Turpin and Lim, 2001) accounted for an average of 32 % of the PM_2.5 mass concentration in Tsukuba and 25 % in Yurihonjo. Thus, the contribution was slightly higher at Tsukuba than at Yurihonjo. The average EC concentration during the observation period was 0.9±0.40 µg m⁻³ (0.4–2.40 µg m⁻³) in Tsukuba and 0.3±0.10 µg m⁻³ (0.2–0.60 µg m⁻³) in Yurihonjo. These values are comparable to those reported for Nagoya (1.10 µg m⁻³) (Yamagami et al., 2019) and Niigata (0.50 µg m⁻³) (Li et al., 2018), Japan, and lower than that reported for Xi'an, China (7.60 µg m⁻³) (Zhao et al., 2018). The EC concentration contributed an average of 5 % to the PM_2.5 mass concentration in Tsukuba and 3 % in Yurihonjo. Currently, EC concentrations in Japan are decreasing as a result of Japanese government regulations on emissions from diesel vehicles (Yamagami et al., 2019). The average OC concentration during the observation period was 3.2±1.4 µg m⁻³ (1.0–6.60 µg m⁻³) in Tsukuba and 1.5±0.80 µg m⁻³ (0.6–4.20 µg m⁻³) in Yurihonjo. The OC concentration contributed an average of 28 % to the PM_2.5 mass concentration in Tsukuba and 22 % in Yurihonjo. The higher percentage contribution to the PM_2.5 mass concentration from OC than EC was in agreement when compared to other studies (contribution of OC and EC concentration in PM_2.5 concentration: 20 % and 6 % in Korea) (Park and Cho, 2011).

Figure 1 Concentrations of EC, WIOC, and WSOC of PM_2.5 from July 2017 to July 2019 in (a) Tsukuba, Ibaraki, and (b) Yurihonjo, Akita, Japan.

The $\mathrm{OC}/\mathrm{EC}$ ratio is an indicator of the source of carbonaceous particles (Chow et al., 1996). The average $\mathrm{OC}/\mathrm{EC}$ ratio was 3.8±1.4 in Tsukuba and 5.1±1.9 in Yurihonjo. The higher $\mathrm{OC}/\mathrm{EC}$ ratio at the rural site (Yurihonjo) than at the suburban site (Tsukuba) was comparable with the results of other studies (Ho et al., 2006; Zhang et al., 2008). This was likely because primary emissions, such as EC, are low at rural sites, meaning that the OC is larger in comparison. The high $\mathrm{OC}/\mathrm{EC}$ ratio is due to the formation of secondary organic aerosols and biomass burning (Chow et al., 1996).

The average WSOC concentration during the observation period was 1.2±0.40 µg m⁻³ (0.4–2.40 µg m⁻³) in Tsukuba and 0.8±0.50 µg m⁻³ (0.3–2.60 µg m⁻³) in Yurihonjo. These values were similar to those reported for Sapporo (1.00 µg m⁻³) (Pavuluri and Kawamura, 2017) and Maebashi (2.30 µg m⁻³) (Kumagai et al., 2009), Japan, but lower than those reported for Beijing, China (7.20 µg m⁻³) (Du et al., 2014), and Gwangju, South Korea (3.70 µg m⁻³) (Park and Cho, 2011). The WSOC concentration at Tsukuba was significantly higher in autumn and winter than in spring and summer (p<0.01), whereas that in Yurihonjo was significantly higher in spring than in the other seasons (p<0.05). The average $\mathrm{WSOC}/\mathrm{OC}$ ratio was 0.4±0.1 in Tsukuba and 0.5±0.1 in Yurihonjo. This is consistent with previous studies that showed that the average $\mathrm{WSOC}/\mathrm{OC}$ ratio was higher at rural sites than at urban sites (Kumagai et al., 2009; Ram and Sarin, 2010). This is also the same as the trend we found for $\mathrm{OC}/\mathrm{EC}$ ratio in the present study.

3.3 δ¹³C_TC and δ¹³C_WSOC

To our knowledge, this is the first report of a 2-year-long observation of δ¹³C_TC and δ¹³C_WSOC in PM_2.5 at two sites simultaneously. The δ¹³C_WSOC values reported from previous studies conducted at various sampling sites and examining various particle sizes are summarized in Table S1. In the present study, the average δ¹³C_TC was $-\mathrm{25.7}\pm \mathrm{0.7}$ ‰ (−26.9 ‰ to −24.0 ‰) in Tsukuba and $-\mathrm{24.7}\pm \mathrm{1.6}$ ‰ (−27.3 ‰ to −20.4 ‰) in Yurihonjo (Table 1 and Fig. 2). Previous studies have reported the average δ¹³C_TC of TSP in Sapporo, Japan ($-\mathrm{24.8}\pm \mathrm{0.68}\mathrm{‰}$) (Pavuluri and Kawamura, 2017), and of PM_2.5 in Sanjiang Plain, China (−24.2 ‰) (Cao et al., 2016), and these values are comparable to our present values.

Figure 2 δ¹³C_TC and δ¹³C_WSOC of PM_2.5 from July 2017 to July 2019 in (a) Tsukuba, Ibaraki, and (b) Yurihonjo, Akita, Japan.

In the present study, the average δ¹³C_WSOC was $-\mathrm{25.2}\pm \mathrm{1.1}$ ‰ (−26.7 ‰ to −21.8 ‰) in Tsukuba and $-\mathrm{24.6}\pm \mathrm{2.4}$ ‰ (−28.4 ‰ to −19.8 ‰) in Yurihonjo (Table 1 and Fig. 2). The δ¹³C_WSOC of PM_2.5, which was the particle size examined in the present study, was $-\mathrm{25.4}\pm \mathrm{1.0}\mathrm{‰}$ in Delhi, India (Dasari et al., 2019), and $-\mathrm{24.2}\pm \mathrm{0.6}\mathrm{‰}$ in Bhola, Bangladesh (Dasari et al., 2019), which are very close to our δ¹³C_WSOC values. The δ¹³C_WSOC of TSP was $-\mathrm{24.2}\pm \mathrm{1.59}\mathrm{‰}$ in Sapporo, Japan (Pavuluri and Kawamura, 2017), $-\mathrm{24.0}\pm \mathrm{1.5}\mathrm{‰}$ in Seoul, South Korea (Han et al., 2020), $-\mathrm{25.2}\pm \mathrm{0.2}\mathrm{‰}$ in Millbrook, USA (Wozniak et al., 2012a), and similar values were obtained for particles of different sizes. In these previous studies, most of the average δ¹³C_WSOC values were in the range of −25 ‰ to −24 ‰ regardless of particle size, although there were some heavy values such as those for Hanimaadhoo Island, Maldives ($-\mathrm{18.4}\pm \mathrm{0.5}\mathrm{‰}$), and Sinhagad, India ($-\mathrm{20.4}\pm \mathrm{0.5}\mathrm{‰}$) (Kirillova et al., 2013).

3.4 Seasonal variations in δ¹³C_TC and δ¹³C_WSOC in PM_2.5

δ¹³C_TC and δ¹³C_WSOC at Tsukuba showed no other clear seasonal variation, but they became slightly heavy from February to April 2019 (Fig. 2a). In contrast, the δ¹³C_TC and δ¹³C_WSOC at Yurihonjo were heaver from autumn to spring than in summer (Fig. 2b), and they showed a significant seasonal variation (δ¹³C_TC; p<0.01, δ¹³C_WSOC; p<0.01) compared to those in Tsukuba. In addition, δ¹³C_WSOC became heavier from February to April 2019 as in Tsukuba. At both study sites, δ¹³C_WSOC was usually heavier than δ¹³C_TC, but in summer δ¹³C_WSOC was comparable to or lighter than δ¹³C_TC (Tsukuba; p<0.01, Yurihonjo; p<0.01).

The seasonal trends of δ¹³C_TC and δ¹³C_WSOC observed in the present study were compared with those reported from previous long-term observations. No seasonal variation for δ¹³C_WSOC in the suburban site, Tsukuba, is comparable with that in TSP in Seoul, South Korea, from March 2015 to January 2016 (Han et al., 2020). Similarly, a clear trend for heavier ratios in winter than in summer for δ¹³C_TC and δ¹³C_WSOC in the rural site, Yurihonjo, is comparable with that in TSP reported for Sapporo, Japan, from September 2009 to October 2010 (Pavuluri and Kawamura, 2017). In both Yurihonjo and Sapporo, it was observed that δ¹³C_WSOC is usually heavier than δ¹³C_TC and that this tendency is reversed in summer. Together, these findings imply that δ¹³C_WSOC shows a weak seasonal trend in suburban or urban sites such as Tsukuba and Seoul but a clear seasonal trend in rural sites such as Yurihonjo and Sapporo.

The variations (difference between maximum and minimum value) of δ¹³C_TC and δ¹³C_WSOC were 2.9 ‰ and 4.9 ‰ in Tsukuba and 7.0 ‰ and 8.6 ‰ in Yurihonjo, respectively. The variation of δ¹³C_WSOC was larger than that of δ¹³C_TC at both sites, with both variations larger in Yurihonjo. In previous studies, the variation of δ¹³C_TC was reported as 2.5 ‰ in Sapporo (Pavuluri and Kawamura, 2017), and that of δ¹³C_WSOC was reported as 5.5 ‰ in Sapporo (Pavuluri and Kawamura, 2017) and 6.5 ‰ in Seoul (Han et al., 2020). The variation of δ¹³C_EC of PM_2.5 was only 1.6 ‰ in Japan (Kawashima and Haneishi, 2012) and 3.7 ‰ in China (Ni et al., 2018; Zhao et al., 2018). In the present study and these previous studies, the variation of δ¹³C_WSOC was larger than that of δ¹³C_EC, regardless of sampling site. The reason for this is likely that δ¹³C_WSOC is affected not only by the source characteristics but also by atmospheric processing. The reasons underlying the seasonal trend observed for δ¹³C_WSOC are further discussed in Sect. 3.5.1 and 3.5.2.

3.5 Determination of seasonal trends and sources of WSOC using δ¹³C_WSOC

3.5.1 Seasonal trends and sources of WSOC in Tsukuba

The average WSOC concentration in Tsukuba was significantly higher in autumn and winter than in spring and summer (p<0.01), and EC concentrations showed a similar significant seasonal trend (p<0.01) (Table 1). Table 2 shows the correlation coefficients between WSOC concentrations and three other parameters – δ¹³C_WSOC, EC concentration, and non-sea-salt potassium concentration (nss-K⁺) – for each season and for the whole year. The EC concentration is a tracer of combustion (Bond et al., 2007). The nss-K⁺ parameter is a tracer of biomass burning that excludes K⁺ from seawater (nss-${\mathrm{K}}^{+}=\left[{\mathrm{K}}^{+}\right]-\mathrm{0.0335}\times \left[{\mathrm{Na}}^{+}\right]$) (Lai et al., 2007). A weak correlation (r=0.18) was found between the annual-average WSOC concentration and annual-average δ¹³C_WSOC. The strong correlation that was found between the annual-average WSOC concentration and annual-average EC concentration (r=0.71) suggests that the WSOC at this suburban site is from combustion sources (e.g., fossil fuel and/or biomass burning). The strong correlations that were observed between WSOC concentrations and nss-K⁺ for every season (autumn, r=0.96; winter, r=0.83; spring, r=0.85; summer r=0.77; all p<0.01) further suggests that the WSOC at this site is a result of biomass burning. The dominant annual source for WSOC was consistent with that reported in Seoul by Han et al. (2020).

Table 2 Correlation (r) between WSOC concentration and the stated parameters.

The average δ¹³C_WSOC was $-\mathrm{25.2}\pm \mathrm{1.1}$ ‰ in Tsukuba (Table 1). Because C₃ and C₄ plants have different metabolic pathways, their δ¹³C values range from −34 ‰ to −24 ‰ for C₃ plants and from −19 ‰ to −6 ‰ for C₄ plants (Smith and Epstein, 1971). When C₃ plants are burned in the laboratory, there is no difference between the δ¹³C of the produced particles and that of the original C₃ plants (Turekian et al., 1998; Das et al., 2010). However, the particles produced by burning C₄ plants are 3.5 ‰ lighter than the original C₄ plants (Turekian et al., 1998). Therefore, the δ¹³C of C₄ plants was estimated to be −22.5 ‰ to −9.5 ‰. The δ¹³C of C₃ and C₄ plant burning has been estimated to be −34.7 ‰ to −25.1 ‰ and −19.3 ‰ to −16.1 ‰, respectively (Kawashima and Haneishi, 2012; Garbaras et al., 2015; Guo et al., 2016). Thus, the average δ¹³C_WSOC at Tsukuba indicates that the burning of C₃ plant biomass is the dominant source of WSOC at this site. Indeed, rice, a C₃ plant, is Japan's largest crop followed by barley and wheat (Ministry of Agriculture Forestry and Fisheries, 2020). In Ibaraki Prefecture, where Tsukuba City is located, the crop acreage and harvest of rice were 68 400 ha and 358 400 t, respectively, in 2018 and were the largest in the Kanto region (Ministry of Agriculture Forestry and Fisheries, 2020). In addition, according to a field investigation, biomass burning in Tsukuba is predominantly the burning of rice straw and rice hulls from September to October Tomiyama et al. (2017). Using radiocarbon analysis, which can distinguish between biogenic and anthropogenic sources, a higher proportion of OC in PM_2.5 collected in Tokyo, Japan, in 2014 was reported to be biogenic from autumn to winter than in summer (Hoshi and Saito, 2020).

The main chemical component generated by the breakdown of cellulose by biomass burning is levoglucosan, which can be used as a tracer of biomass burning (Simoneit et al., 1999). The δ¹³C of levoglucosan emitted from the burning of C₃ plants such as peanut, mulberry, China fir, Chinese red pine, Chinese guger tree, and chestnut are reported to range from −26.05 ‰ to −22.60 ‰, with that from rice straw reported to be $-\mathrm{24.26}\pm \mathrm{0.09}$ ‰ (Sang et al., 2012). The average δ¹³C_WSOC in Tsukuba was very close to this previously reported δ¹³C of levoglucosan from the burning of rice straw. However, levoglucosan accounts for only about 3.8 % of the WSOC in urban areas of Japan (Kumagai et al., 2010). Therefore, it is difficult to accurately identify the sources of WSOC using only the δ¹³C values of levoglucosan. Further research is needed to determine the δ¹³C of the components of WSOC other than levoglucosan.

3.5.2 Seasonal trends and sources of WSOC in Yurihonjo

In Yurihonjo, the correlation between WSOC concentrations and EC concentrations was highest in winter (r=0.87, p<0.01), followed by autumn (r=0.83, p<0.01) and spring (r=0.64, p<0.05), and lowest in summer (r=0.24) (Table 2). This suggests that WSOC at this rural site was mainly from combustion sources in autumn and spring. In addition, the correlation between WSOC concentrations and nss-K⁺ concentrations was very high in autumn (r=0.93), winter (r=0.99), and spring (r=0.80; all p<0.01) but not in summer (r=0.40). These strong correlations from autumn to spring suggest that during that time the WSOC came mainly from combustion sources such as biomass burning. The average δ¹³C_WSOC at Yurihonjo for autumn and spring, $-\mathrm{23.9}\pm \mathrm{2.1}$ ‰, suggests that biomass burning of C₃ biomass such as rice straw and rice hulls may be the dominant source of WSOC, as was found in Tsukuba.

In Akita Prefecture, where Yurihonjo is located, the crop acreage of rice was 87 700 ha in 2018, and the rice harvest was 491 100 t (Ministry of Agriculture Forestry and Fisheries, 2020). From February to April 2019, the δ¹³C_WSOC was the heaviest ($-\mathrm{21.3}\pm \mathrm{1.9}$ ‰ ), and WSOC concentrations were markedly increased compared with the previous months (average, 1.5±0.7 µg m⁻³) (Figs. 1b and 2b). A moderate correlation between WSOC concentrations and δ¹³C_WSOC values was observed for this time period (r=0.54, p=0.27). This δ¹³C_WSOC value indicates a heavy δ¹³C source such as C₄ plants (e.g., corn and grass), but no evidence of burning of C₄ plants during this period was observed around the sampling site at Yurihonjo. Northeast China is the largest producer of corn in China (MWCACP, 2021), and biomass burning is used for heating in winter (Chen et al., 2017). Satellite imagery revealed a number of fire spots in that part of China from February to April 2019 (NASA, 2021) (Fig. S2 in the Supplement). Backward air-mass trajectories showed that air masses at Yurihonjo during this period originated mainly from areas in northeast China (Fig. S3 in the Supplement). Consistent with this finding, Uranishi et al. (2020) reported from an analysis using the Community Multiscale Air Quality model that particles from biomass burning in northeast China were transported to Akita Prefecture in February and March 2019. The correlation between Na⁺ and Cl⁻ concentrations was highest from winter to spring 2019 in Yurihonjo (r=0.98, p<0.01), suggesting the influence of sea salt. Recently, aerosol photochemical aging during long-range transport has been shown to selectively enrich the ¹³C content in organic aerosols, leading to heavier δ¹³C values (Kirillova et al., 2013; Bosch et al., 2014; Dasari et al., 2019; Zhang et al., 2019). In a field study, the isotope fractionation values for δ¹³C_WSOC were estimated to be enriched by 3 ‰–4 ‰ because of aging during transport (Kirillova et al., 2013). The combination of isotopic ratio and concentration measurements (Figs. 1 and 2) together with the evidence of prevailing biomass burning activities (Fig. S2) and the results of the backward trajectory analysis (Fig. S3) suggest that the heavier δ¹³C_WSOC from February to April 2019 at Yurihonjo was the result of C₄ plant combustion rather than aging during long-range transport.

The δ¹³C_WSOC in summer was very light (−27.4 ‰) compared with the average value for the observation period. A weak correlation between WSOC concentrations and EC concentrations in summer (r=0.24; Table 2) suggests that the WSOC is derived from a non-combustion source. In general, the formation of WSOC involves atmospheric reactions such as the formation of SOAs, which are formed by oxidation of anthropogenic and biogenic volatile organic compounds (VOCs; Heo et al., 2013). Aliphatic hydrocarbons (e.g., alkanes and alkenes) and aromatics (e.g., benzene, toluene, ethylbenzene, and xylene) emitted from solvent evaporation and vehicle emissions are important anthropogenic VOCs and precursors of SOAs (Chen et al., 2010). The δ¹³C values for alkanes in tunnel, gas station, underground garage, and refinery air samples are reported to range from $-\mathrm{28.6}\pm \mathrm{1.8}$ ‰ to $-\mathrm{27.3}\pm \mathrm{2.1}$ ‰ (Rudolph et al., 2002). Toluene and xylene are the aliphatic hydrocarbons with the highest annual emissions in Japan (Japan Ministry of Economy Trade and Industry, 2020). The δ¹³C of toluene and xylene for gas station and vehicle emissions are reported to range from −27.7 ‰ to −23.8 ‰ (Rudolph et al., 2002; Kawashima and Murakami, 2014). Because VOCs in the atmosphere are oxidized by photochemical oxidants, the δ¹³C values of the residual VOCs become heavier via isotopic fractionation (Rudolph et al., 2000; Anderson et al., 2004); in other words, secondary production tends to result in a lighter δ¹³C for SOA in the atmosphere. In a laboratory-based experiment, the δ¹³C of SOA particles formed by photooxidation of toluene was 3 ‰ to 6 ‰ lighter than that of the precursor toluene, depending on the degree of oxidation (Irei et al., 2006, 2011). Assuming that this isotopic fractionation of toluene applies also to all other potential components, the δ¹³C of the emission source of VOCs at Yurihonjo would be approximately −24.4 ‰ to −21.4 ‰, as calculated by subtracting 3 ‰ to 6 ‰ from the average δ¹³C_WSOC in Yurihonjo during summer (−27.4 ‰). This estimated δ¹³C value of VOCs is heavier than those previously reported for anthropogenic VOCs. Therefore, anthropogenic VOCs were not considered to be the dominant source of WSOC at Yurihonjo.

At the global scale, biogenic VOC emissions are more than an order of magnitude greater than those of anthropogenic VOCs (Farina et al., 2010). Biogenic VOCs include isoprene, monoterpenes, and sesquiterpenes released from vegetation, with isoprene producing the most SOA (Atkinson and Arey, 1998). The oxidation product of isoprene is 2-methyltetrol, which is widely used as an organic tracer to evaluate the production of SOA from isoprene (Claeys et al., 2004). The average δ¹³C of 2-methyltetrol in aerosols in four forests in Sichuan Province, China, was −27.36 ‰ (−28.23 ‰ to −26.46 ‰) (Li et al., 2019). This average δ¹³C of 2-methyltetrol is close to the δ¹³C_WSOC detected in summer in Yurihonjo, suggesting the components produced by secondary reaction of biogenic VOCs make a large contribution to the WSOC in Yurihonjo during the summer. From a field study conducted in a forest in northern Japan, Miyazaki et al. (2012) reported that the lightest δ¹³C_WSOC values (average $-\mathrm{25.6}\pm \mathrm{0.7}$ ‰) were observed from June to September; the authors concluded from positive matrix factorization modeling data that biogenic SOAs (isoprene SOA and α- and/or β-pinene) were the dominant source of WSOC in the summer, which is consistent with the findings of the present study.

4 Conclusion

The WSOC concentration, δ¹³C_TC, and δ¹³C_WSOC of PM_2.5 were observed at one suburban and one rural site in Japan over a 2-year period. The average WSOC concentration during the observation period was 1.2±0.4 µg m⁻³ (0.4–2.4 µg m⁻³) at the suburban site and 0.8±0.5 µg m⁻³ (0.3–2.6 µg m⁻³) at the rural site. The δ¹³C_WSOC was $-\mathrm{25.2}\pm \mathrm{1.1}$ ‰ (−26.7 ‰ to −21.8 ‰) at the suburban site and $-\mathrm{24.6}\pm \mathrm{2.4}$ ‰ (−28.4 ‰ to −19.8 ‰) at the rural site. The δ¹³C_TC and δ¹³C_WSOC at the suburban site showed no clear seasonal variations, but they were slightly heavier from February to April 2019. In contrast, the δ¹³C_TC and δ¹³C_WSOC at the rural site were heaver from autumn to spring than in summer, and they showed a significant seasonal variation (δ¹³C_TC, p<0.01; δ¹³C_WSOC, p<0.01). Using δ¹³C_WSOC, carbon components, and water-soluble ions, the main source of WSOC at the suburban site was concluded to be local biomass burning of rice straw. At the rural site, the δ¹³C_WSOC from autumn to spring was concluded to reflect mainly the biomass burning of rice straw, whereas that in summer was considered to reflect mainly the formation of secondary organic aerosols from biogenic VOCs. The heaviest δ¹³C_WSOC ($-\mathrm{21.3}\mathrm{‰}\pm \mathrm{1.9}$ ‰) was from February to April 2019 and may reflect long-range transport of particles resulting from the overseas burning of C₄ plants such as corn. Thus, we were able to use a δ¹³C_WSOC-based approach to understand the sources and atmospheric processes that contribute to the WSOC concentrations at the two study sites.