Measurement report: Source characteristics of water-soluble organic carbon in PM2.5 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 PM2.5 (particulate matter, aerodynamic diameter < 2.5 μm) from one suburban and one rural site in Japan. The average δCWSOC was −25.2 ± 1.1‰ and −24.6 ± 2.4‰ at the suburban 10 site and rural site, respectively. At the suburban site, the δCWSOC was consistent with the δC of levoglucosan, a tracer of biomass burning, and a high correlation was found between WSOC concentration and non-sea-salt potassium concentration, another tracer of biomass burning, suggesting that the main source of WSOC was biomass from burning of rice straw. At the rural site, the δCWSOC was significantly heavier in winter (−22.6 ± 1.3‰) than in summer (−27.4 ± 0.7‰; p < 0.01). The heavy δCWSOC in winter was a result mainly of biomass burning and the aging of OC during long-term transport, whereas the 15 light δCWSOC in summer was a result mainly of the formation of secondary organic aerosols from biogenic volatile organic compounds. Thus, our δCWSOC approach was useful to elucidate the sources and atmospheric processes that contribute to seasonal variations of WSOC concentrations.


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(Duarte et al., , 2015Pietrogrande 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 (δ 13 C) 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 δ 13 C of its aerosols (e.g., Kawashima and Haneishi, 2012;Zhao et al., 2018). In contrast, because OC reacts in the atmosphere, its δ 13 C 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 δ 13 C of WSOC (δ 13 C WSOC ) in PM (Kirillova et al., 2010(Kirillova et al., , 2013Suto and Kawashima, 2018;Zhang et al., 2019). In addition, various approaches have been used; for example, the δ 13 C 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 δ 13 C WSOC values that use wet oxidation with liquid chromatography and IRMS (LC-IRMS) , 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 δ 13 C 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(Kirillova et al., , 2014bBosch et al., 2014, Dasari et al., 2019, Europe (Fisseha et al., 2006(Fisseha et al., , 2009Kirillova et al., 2010), and the United States (Wozniak et al., 2012a, b) have been reported (Table S1 in the Supplement). For example, the δ 13 C of total carbon (δ 13 C TC ) and δ 13 C WSOC of total suspended particles (TSPs) was observed from September 2009 to October 2010 in Hokkaido, Japan (Pavuluri and Kawamura, 2017). Both δ 13 C TC and δ 13 C WSOC were heavier in winter than in summer, demonstrating seasonal variation. The authors concluded that the reason why δ 13 C WSOC was heavy in win-ter was because of the greater release of 13 C 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 δ 13 C WSOC was −20.4 ± 0.5 ‰ in Sinhagad and −18.4 ± 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 δ 13 C 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 δ 13 C was an effect related to the aging of OC during long-range transport of aerosol. A recent study reported that the enrichment of δ 13 C 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 δ 13 C 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 δ 13 C TC and δ 13 C 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  for long-term observation of δ 13 C WSOC .
2 Materials and experimental methods

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; 36 • 4 N, 140 • 4 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 approxi-mately 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; 39 • 23 N, 140 • 4 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.

Stable carbon isotope ratio analysis
Determination of δ 13 C 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 CO 2 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 CO 2 from the EA was ionized, and the δ 13 C value was determined by means of IRMS; data acquisition was performed with Isodat software (ver. 3.0, Thermo Fisher Scientific).
Determination of δ 13 C WSOC was performed at Akita Prefectural University using the wet-oxidation-IRMS method Suto and Kawashima, 2018). A portion of each quartz fiber filter (14.13 cm 2 ) 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 δ 13 C WSOC analysis, because the difference between the δ 13 C 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, < ±0.3 ‰; accuracy, < ±0.3 ‰) for levoglucosan and oxalic acid were 1111 and 1133 ng C, respectively.
Stable carbon isotope ratios are expressed in δ notation in per mil (‰): where R( 13 C/ 12 C) sample and R( 13 C/ 12 C) std (= 0.0111802) are the 13 C/ 12 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.

Chemical analysis
For determination of OC and EC concentrations, a portion of each quartz fiber filter (0.53 cm 2 ) 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 cm 2 ) 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 − , NO − 2 , NO − 3 , SO 2− 4 ) and cations (Na + , NH + 4 , K + , Mg 2+ , Ca 2+ ) 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.
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 leastpolluted 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 OC/EC and WSOC/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 The OC/EC ratio is an indicator of the source of carbonaceous particles (Chow et al., 1996). The average OC/EC ratio was 3.8 ± 1.4 in Tsukuba and 5.1 ± 1.9 in Yurihonjo. The higher OC/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 OC/EC ratio is due to the formation of secondary organic aerosols and biomass burning (Chow et al., 1996).

δ 13 C TC and δ 13 C WSOC
To our knowledge, this is the first report of a 2-year-long observation of δ 13 C TC and δ 13 C WSOC in PM 2.5 at two sites simultaneously. The δ 13 C 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 δ 13 C TC was −25.7 ± 0.7 ‰ (−26.9 ‰ to −24.0 ‰) in Tsukuba and −24.7 ± 1.6 ‰ (−27.3 ‰ to −20.4 ‰) in Yurihonjo (Table 1 and Fig. 2). Previous studies have reported the average δ 13 C TC of TSP in Sapporo, Japan (−24.8 ± 0.68 ‰) (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.
In the present study, the average δ 13 C WSOC was −25.2 ± 1.1 ‰ (−26.7 ‰ to −21.8 ‰) in Tsukuba and −24.6±2.4 ‰ (−28.4 ‰ to −19.8 ‰) in Yurihonjo (Table 1 and Fig. 2). The δ 13 C WSOC of PM 2.5 , which was the particle size examined in the present study, was −25.4 ± 1.0 ‰ in Delhi, India (Dasari et al., 2019), and −24.2 ± 0.6 ‰ in Bhola, Bangladesh (Dasari et al., 2019), which are very close to our δ 13 C WSOC values. The δ 13 C WSOC of TSP was −24.2 ± 3.4 Seasonal variations in δ 13 C TC and δ 13 C WSOC in PM 2.5 δ 13 C TC and δ 13 C WSOC at Tsukuba showed no other clear seasonal variation, but they became slightly heavy from February to April 2019 (Fig. 2a). In contrast, the δ 13 C TC and δ 13 C WSOC at Yurihonjo were heaver from autumn to spring than in summer (Fig. 2b), and they showed a significant seasonal variation (δ 13 C TC ; p < 0.01, δ 13 C WSOC ; p < 0.01) compared to those in Tsukuba. In addition, δ 13 C WSOC became heavier from February to April 2019 as in Tsukuba. At both study sites, δ 13 C WSOC was usually heavier than δ 13 C TC , but in summer δ 13 C WSOC was comparable to or lighter than δ 13 C TC (Tsukuba; p < 0.01, Yurihonjo; p < 0.01). The seasonal trends of δ 13 C TC and δ 13 C WSOC observed in the present study were compared with those reported from previous long-term observations. No seasonal variation for δ 13 C 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 δ 13 C TC and δ 13 C 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 δ 13 C WSOC is usually heavier than δ 13 C TC and that this tendency is reversed in summer. Together, these findings imply that δ 13 C 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 δ 13 C TC and δ 13 C WSOC were 2.9 ‰ and 4.9 ‰ in Tsukuba and 7.0 ‰ and 8.6 ‰ in Yurihonjo, respectively. The variation of δ 13 C WSOC was larger than that of δ 13 C TC at both sites, with both variations larger in Yurihonjo. In previous studies, the variation of δ 13 C TC was reported as 2.5 ‰ in Sapporo (Pavuluri and Kawamura, 2017), and that of δ 13 C WSOC was reported as 5.5 ‰ in Sapporo (Pavuluri and Kawamura, 2017) and 6.5 ‰ in Seoul (Han et al., 2020). The variation of δ 13 C EC of PM 2.5 was only 1.6 ‰ in Japan (Kawashima and Haneishi, 2012) and 3.7 ‰ in China Zhao et al., 2018). In the present study and these previous studies, the variation of δ 13 C WSOC was larger than that of δ 13 C EC , regardless of sampling site. The reason for this is likely that δ 13 C WSOC is affected not only by the source characteristics but also by atmospheric processing. The reasons underlying the seasonal trend observed for δ 13 C WSOC are further discussed in Sect. 3.5.1 and 3.5.2.
3.5 Determination of seasonal trends and sources of WSOC using δ 13 C WSOC

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 -δ 13 C 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 . The nss-K + parameter is a tracer of biomass burning that excludes K + from seawater (nss-K + = [K + ]−0.0335×[Na + ]) (Lai et al., 2007). A weak correlation (r = 0.18) was found between the annual-average WSOC concentration and annual-average δ 13 C 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).
The average δ 13 C WSOC was −25.2 ± 1.1 ‰ in Tsukuba (Table 1). Because C 3 and C 4 plants have different metabolic pathways, their δ 13 C values range from −34 ‰ to −24 ‰  (Smith and Epstein, 1971). When C 3 plants are burned in the laboratory, there is no difference between the δ 13 C of the produced particles and that of the original C 3 plants (Turekian et al., 1998;Das et al., 2010). However, the particles produced by burning C 4 plants are 3.5 ‰ lighter than the original C 4 plants (Turekian et al., 1998). Therefore, the δ 13 C of C 4 plants was estimated to be −22.5 ‰ to −9.5 ‰. The δ 13 C of C 3 and C 4 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 δ 13 C WSOC at Tsukuba indicates that the burning of C 3 plant biomass is the dominant source of WSOC at this site. Indeed, rice, a C 3 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 δ 13 C of levoglucosan emitted from the burning of C 3 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 −24.26 ± 0.09 ‰ (Sang et al., 2012). The average δ 13 C WSOC in Tsukuba was very close to this previously reported δ 13 C 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 δ 13 C values of levoglucosan. Further research is needed to determine the δ 13 C of the components of WSOC other than levoglucosan.

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 δ 13 C WSOC at Yurihonjo for autumn and spring, −23.9±2.1 ‰, suggests that biomass burning of C 3 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 δ 13 C WSOC was the heaviest (−21.3 ± 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 δ 13 C WSOC values was observed for this time period (r = 0.54, p = 0.27). This δ 13 C WSOC value indicates a heavy δ 13 C source such as C 4 plants (e.g., corn and grass), but no evidence of burning of C 4 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 airmass 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 13 C content in organic aerosols, leading to heavier δ 13 C 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 δ 13 C 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 δ 13 C WSOC from February to April 2019 at Yurihonjo was the result of C 4 plant combustion rather than aging during long-range transport.
The δ 13 C 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 δ 13 C values for alkanes in tunnel, gas station, underground garage, and refinery air samples are reported to range from −28.6 ± 1.8 ‰ to −27.3 ± 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 δ 13 C 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 δ 13 C 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 δ 13 C for SOA in the atmosphere. In a laboratorybased experiment, the δ 13 C 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(Irei et al., , 2011. Assuming that this isotopic fractionation of toluene applies also to all other potential components, the δ 13 C 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 δ 13 C WSOC in Yurihonjo during summer (−27.4 ‰). This estimated δ 13 C 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 δ 13 C 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 δ 13 C of 2-methyltetrol is close to the δ 13 C 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 δ 13 C WSOC values (average −25.6 ± 0.7 ‰) were observed from June to September; the authors concluded from positive matrix factorization modeling data that biogenic SOAs (isoprene SOA and αand/or β-pinene) were the dominant source of WSOC in the summer, which is consistent with the findings of the present study.

Conclusion
The WSOC concentration, δ 13 C TC , and δ 13 C 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 δ 13 C WSOC was −25.2 ± 1.1 ‰ (−26.7 ‰ to −21.8 ‰) at the suburban site and −24.6 ± 2.4 ‰ (−28.4 ‰ to −19.8 ‰) at the rural site. The δ 13 C TC and δ 13 C WSOC at the suburban site showed no clear seasonal variations, but they were slightly heavier from February to April 2019. In contrast, the δ 13 C TC and δ 13 C WSOC at the rural site were heaver from autumn to spring than in summer, and they showed a significant seasonal variation (δ 13 C TC , p < 0.01; δ 13 C WSOC , p < 0.01). Using δ 13 C 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 δ 13 C 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 δ 13 C WSOC (−21.3 ‰ ± 1.9 ‰) was from February to April 2019 and may reflect long-range transport of particles resulting from the overseas burning of C 4 plants such as corn. Thus, we were able to use a δ 13 C WSOC -based approach to understand the sources and atmospheric processes that contribute to the WSOC concentrations at the two study sites.
Data availability. Data are available from the corresponding author on request (nsuto@jari.or.jp).
Author contributions. NS and HK were involved in research planning and experimental design. NS performed the sampling and measurements of δ 13 C TC , carbon components, and water-soluble ions. HK performed the sampling and measurements of δ 13 C WSOC . All authors clarified the experimental data and contributed to the writing of the paper.
Competing interests. The authors declare that they have no conflict of interest.
Disclaimer. Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Acknowledgements. This work was partially supported by the Japan Society for the Promotion of Science KAKENHI (grant nos. 19K20463, 18H03393, 17K12829, and 16KK0015). We acknowledge the use of data and imagery from NASA's Fire Information for Resource Management System (FIRMS) (https://earthdata.nasa. gov/firms, last access: 26 June 2021), part of NASA's Earth Observing System Data and Information System (EOSDIS). We thank emeritus professor Shigeki Masunaga of Yokohama National University for providing the high-volume samplers used in this research; we also thank Sae Ono, Momoka Suto, and Otoha Yoshida for collecting the aerosol samples and for helping with wetoxidation-IRMS analysis at Akita Prefectural University. Furthermore, we thank Akiyoshi Ito, Hiroyuki Hagino, Kazue Kagami, and Akemi Nakayama at the Japan Automobile Research Institute for their advice and help with chemical analysis. And, Yumi Sone from Thermo Fisher Scientific Inc., Japan, was very helpful with our EA-IRMS analysis. Finally, we thank ELSS, Inc. for editing the English of this article.
Financial support. This research has been supported by the Japan Society for the Promotion of Science KAKENHI (grant nos. 19K20463, 18H03393, 17K12829, and 16KK0015).
Review statement. This paper was edited by Rupert Holzinger and reviewed by Andrius Garbaras and one anonymous referee.