Seasonal Characteristics of Volatile Organic Compounds in Seoul, Korea: Major Sources and Contribution to Secondary Organic Aerosol Formation

Volatile organic compounds (VOCs) are major pollutants that cause air pollution and are precursors that react in the air to produce secondary organic aerosol (SOA). This study attempted to elucidate the distribution characteristics of VOCs in the atmosphere of Seoul by measuring 34 types of VOCs in real time in the winter of 2020 and summer of 2021. The objectives of this research are as follows: (1) understand the characteristics of VOCs in Seoul and the difference between winter and summer compositions, (2) identify the main sources of VOCs in winter and summer, and (3) estimate the contribution of VOCs to the SOA formation potential in Seoul. Total VOC concentrations were found to be higher in summer (7.61 ± 4.22 ppb) than in winter (6.28 ± 4.11 ppb). To further specify the cause of the difference in major VOC components in winter and summer, a cause analysis was performed using the ratio between marker components, and an emission source analysis of VOCs was performed by applying the positive matrix factorization model (PMF). The source distribution of VOCs in Seoul was attributed to five factors: solvent usage, vehicle exhaust, industry/burning of fossil fuels, petrochemical industry, and road emission (winter)/gasoline-related (summer). The contribution of VOCs to SOA formation was estimated using the secondary organic aerosol formation potential. The results showed that toluene was the primary contributor to SOA formation in both winter and summer. In the summer, solvent usage containing high proportion of ethylbenzene and xylenes contributed more than twice as much to SOA formation compared to the winter.


INTRODUCTION
Volatile organic compounds (VOCs) generally refer to all organic compounds with a boiling point below 50°C-260°C at standard atmospheric pressure and a melting point below room temperature, including alkanes, alkenes, alkynes, aromatics, alcohols, aldehydes, ketones, and halogenated VOCs (Zhang et al., 2017). VOCs are ubiquitous in the atmosphere and are emitted from various natural and anthropogenic sources such as vehicle exhaust, oil refineries, paint/solvent use, construction materials, tobacco smoke, various trees or forests, and wildfires (Bari and Kindzierski, 2018;Na et al., 2005). Additionally, ambient VOCs chemically interact with sunlight and nitrogen oxides to form ground-level ozone and are a representative precursor of secondary organic aerosol (SOA). VOCs may adversely affect human health by causing breathing difficulties, headaches, and carcinogenesis (He et al., 2015;Keller and Burtscher, 2017;Srivastava et al., 2018;Wang et al., 2013;

Measurements
The measurements were conducted on the roof of the Seoul Metropolitan Air Environment Research Institute (37.61°N, 126.93°E), which is located in northwestern Seoul, as shown in Fig. 1. Residential, commercial, and green were distributed around the sampling site. Measurements were performed for 32 days (December 15, 2020-January 15, 2021) during winter and for 30 days (June 1, 2021-June 30, 2021) during summer.
For measurement, VOCs corresponding to C6-C12 in the atmosphere were observed at 30-min intervals using a real-time ozone precursor analyzer Netherlands). When the sample collected from GC955-611 was concentrated in an internal pre-concentration tube, it was separated from column SY-1 (30 m in length, 0.32 mm in internal diameter; Synspec, Netherlands) and analyzed using a photoionization detector. A total of 34 compounds (34 VOCs) were analyzed using GC955-611, including 16 aromatic VOCs, such as benzene and toluene; 16 alkanes, including n-hexane and n-octane; and two alkenes, cyclohexane and methylcyclohexane. During the winter and summer measurement periods, the 34 VOCs were calibrated using standard mixed gas (Rigas, Korea) before sampling. Most of the correlation coefficients for the standard gas containing 34 VOCs were greater than 0.99.

PMF Model
The positive matrix factorization (PMF) model, first introduced by Paatero and Tapper (1994), is a statistical analysis that excludes the negative value of the factor load by considering the uncertainty and standard deviation of the measured data value and computes it as a positive value to enable quantitative estimation. The ultimate goal of PMF analysis is to quantify the effects of specific sources on receptors, and it is widely used to identify multiple sources of pollutants such as aerosols and VOCs (Su et al., 2019;Zheng et al., 2019).
The EPA PMF 5.0 user guide was followed for data preprocessing, 24 and 27 of the 34 measured compounds in winter and summer, respectively, were used, excluding ten species in winter and seven species in summer owing to their low signal-to-noise ratio. The method detection limit for the instrument was not measured; lower detection limit (LDL) values were used instead. For the input data, the missing value was replaced with the median value of the component, and a concentration corresponding to one-fourth of LDL was entered for a value smaller than LDL. If the uncertainty was ≤ LDL, LDL was multiplied by 5/6. If a concentration value exceeded LDL, the uncertainty was calculated using Eq. (1) (Norris and Duvall, 2014): The error fraction can be estimated from the precision of replicate measurements, but when the results for the precision are not available, it can be used the relative standard deviation or 5 to 20% of photochemical activity or VOC concentration (Abeleira et al., 2017;Xie et al., 2021). In this study, 10%, which is the most frequently used value for the error fraction was applied to all species (Huang et al., 2022;Liu et al., 2016;Song et al., 2019;Xie et al., 2021;Yang et al., 2022). The PMF models are described in greater detail by Paatero and Tapper (1994) and Norris and Duvall (2014). Information on the values used in the process of operating the PMF in this study is presented in Table 1.

Contribution of VOCs to SOA
VOCs are among the most important precursors of SOA, which constitute a large proportion of PM2.5 (Zhang et al., 2017). Since the early 1990s, several studies have been conducted to characterize the yield of SOA from VOCs using chamber and chemical models (Ng et al., 2007;Odum et al., 1996;Pandis et al., 1992). In this study, two methods of SOAFP (secondary organic aerosol formation potential, SOAI) and SOAP (secondary organic aerosol potential, SOAII) were compared to investigate SOA formation by VOCs. The first method calculated SOAI using the SOA yield of each compound reported in Shin et al. (2013). In this study, SOAI was obtained by multiplying the VOC concentration value with the SOA yield, using Eq. (2): where VOCi represents the measured concentration (ppb) of VOC individual compounds i, SOAyield,i (µg m -3 ppm -1 ) is the SOA yield produced by i components of VOCs (Table 2). The second method involved calculating the SOAII to obtain toluene-based SOA, which was then compared with the trends for the generated secondary aerosols. The concept of SOAII was developed by Derwent et al. (2010) to reflect the tendency of each organic compound to form SOA on the same mass emission basis as toluene. SOAII differs from the existing SOAI as it is based on the toluene SOA yield (100). Toluene was selected as the base compound for SOAII because its emission properties are better known than those of other compounds and it is an important precursor for SOA formation (Hui et al., 2019;Zhao et al., 2023). The method of calculating SOAyield-to-toluene,i is shown in Eq.
here, the SOAyield-to-toluene,i values used the values reported by Derwent et al. (2010). The SOAII for each compound was calculated by multiplying SOAyield-to-toluene,i with the concentration of each VOC species observed in this study as follows Eq. (4): where the VOCi represents the concentrations (ppb) of VOC each compound i, and the unit of 16.2 a The SOAyield,i values (unit: µg m -3 ppm -1 ) basically followed the those of Shin et al. (2013). b The SOAyield-to-toluene,i values followed that reported by Derwent et al. (2010) and the unit of SOAyield-to-toluene,i was not available. c 0, denotes that there is no contribution to the SOA. d -, denotes that no value has been reported. e The SOAyield,i or SOAyield-to-toluene,i was calculated to the m-xylene value.
SOAyield-to-toluene,i was not available ( Table 2). The SOAI and SOAII values were calculated based on individual compounds for which yield values were available.
Additionally, the secondary organic carbon (SOC) method based on the ratio of organic carbon (OC) and elemental carbon (EC) included in PM2.5 was also compared. The PM2.5, OC, and EC data used in this study were measured simultaneously during the measurement period and location and were provided by the National Institute of Environmental Research, Korea. SOC can be obtained using several methods. In this study, SOC was estimated from the measured OC and EC concentrations (µg m -3 ) according to the method of Yoo et al. (2020) using Eq. (5) by employing the EC trace method: where (OC/EC)pri was calculated through regression analysis of the OC and EC concentrations with an OC/EC ratio lower than 5%.

Characteristics of VOCs during Winter and Summer
The concentrations of VOCs measured in Seoul during the winter of 2020 and summer of 2021 and the winter and summer ratios are listed in Table 3. The temperature at the sampling site using the automatic weather system (AWS) data of the Korea Meteorological Administration (KMA) was −5.80 ± 7.16°C in winter and 21.3 ± 3.94°C in summer, showing a large difference, and the 2.29 ± 1.16 375 ± 277 1.21 a S/W ratio of measured concentration by each compound, summer/winter ratio. b -, denotes that the value is zero or has no yield. c N.D., denotes that the value is not detected. seasonal difference was statistically significant (p < 0.05). Insolation in Seoul, where automated synoptic observing system (ASOS) data were measured, was 0.82 ± 0.67 MJ m -2 (8:00 AM-6:00 PM) in winter and 1.25 ± 1.05 MJ m -2 (6:00 AM-8:00 PM) in summer (KMA, https://data.kma.go.kr). In the case of the 34 VOCs, the arithmetic average and standard deviations were found to be 6.28 ± 4.11 and 7.61 ± 4.22 ppb during winter and summer, respectively, and the concentration in summer was not only 1.2 times higher than that in winter but also statistically significant (p < 0.05). The correlation (r) between the 34 VOCs and temperature and insolation was low at 0.21 and 0.10, respectively, but it was statistically significant (p < 0.05).
Toluene had the highest concentration among the individual compounds in both winter and summer. The ratio of BTEX (benzene, toluene, ethylbenzene, and xylenes) to 34 VOCs was 45.4% (2.85 ppb) and 47.3% (3.60 ppb) during winter and summer, respectively, accounting for the majority of the 34 VOCs. Comparing the results of this study with those of studies conducted in other urban cities revealed that the BTEX concentrations in Seoul was lower than that in Chengdu (5.29 ppb, measured in 2016) in China, but higher than that in Beijing (1.87 ppb, measured in 2018) and New York (3.52 µg m -3 ≒ 0.82 ppb under standard condition, measured during 2015-2019) (Li et al., 2022;Song et al., 2018;Paul and Bari, 2022). As suggested by Li et al. (2022), global VOC concentrations have gradually decreased in recent times, and the reduction of VOC emissions in New York has also shown a decreasing trend in recent years, as reported by Paul and Bari (2022). Compared with the VOCs measured in this study (2020)(2021), the concentration of BTEX (11.4-13.6 ppb) measured in Seoul during 1997-1999 was approximately five times higher (Na et al., 2005). In a study by Shin et al. (2013), which was performed in Seoul from 2004 to 2008, the concentration of VOCs (40.9 ppb) was approximately 10 times higher than that in this study; however, the proportions of aromatic VOCs to 34 VOCs were similar. The overall concentration has decreased owing to Korea's air quality policy since 2005. In 2003, the Ministry of Environment (MOE) of Korea regulated a law to improve the air environment in the SMA and manage air pollution sources. As reported by Kim and Yeo (2013), primary pollutants such as SO2 and CO have clearly decreased since the 2000s compared to the 1980s primarily owing to various policies implemented in the early 2000s. Additionally, annual average concentrations of benzene have shown a tendency to rapidly decrease since 2009.
The compound with the highest summer-to-winter (S/W) ratio was 2,3,4-trimethylpentane, followed by ethylbenzene and cyclohexane. 2,3,4-trimethylpentane has been reported to be highly correlated with gasoline vehicle emissions (Chang et al., 2006;Song et al., 2020). The concentration of 2,3,4-trimethylpentane during winter was low, and the S/W ratio was relatively high. Ethylbenzene and cyclohexane are known ozone precursors (Grosjean and Grosjean, 1997;Jia and Xu, 2013), and their potential sources are paint solvent usage and industrial production (Song et al., 2018). Conversely, compounds with high concentrations during winter were benzene and m-diethylbenzene. Benzene is classified as a specified hazardous air pollutant in the Clean Air Conservation Act of the MOE of Korea and an ozone precursor (Ng et al., 2007). Although m-diethylbenzene was reported to be toxic by Hartwig (2019), this was insufficient to classify it as a carcinogen. Similar to benzene, m-diethylbenzene showed a relatively low S/W ratio owing to its low concentration in summer. Fig. 2 shows the diurnal patterns of the hourly average for 34 VOCs and BTEX during the summer and winter. The 34 VOCs in winter increased between 8:00 AM-11:00 AM during rush hour, and the highest concentration (7.14 ± 3.91 ppb) was measured around noon. This is likely affected by increased human activity and traffic in the morning. Subsequently, the 34 VOCs decreased and showed a minimum concentration of 5.58 ± 3.44 ppb around 7:00 PM. In summer, the 34 VOCs showed an increasing pattern similar to that in winter in the morning; however, they increased steadily after noon till around 5:00 PM and had the highest concentration of 9.60 ± 5.13 ppb. This is presumably due to the stronger insolation and temperature in summer than in winter. By comparing the diurnal patterns of the individual BTEX components, the summer and winter trends of each BTEX component were confirmed to be similar; however, the diurnal patterns between the components were found to be different. Although the concentration of benzene in winter was significantly high compared to summer (p < 0.05), the diurnal patterns between summer and winter were similar unlike other components. The emission sources of benzene are known to be vehicles and heating (Elbir et al. 2007); thus, the concentration increased because of heating in winter.

Source Apportionment by PMF Model
Source apportionment was performed using PMF, and five factors were identified according to the contributions of the representative markers (Fig. 3). VOC sources in winter and summer were classified into five factors. Four factors (solvent usage, petrochemical industry, industry/burning of fossil fuel, and vehicle exhaust) were common; however, road emission and gasoline-related sources were divided into different factors in winter and summer.
Factor 1 was found to be highly loaded with ethylbenzene and xylenes in both winter and summer, contributing to 1.27 ppb (20.7%) and 2.44 ppb (28.4%) of the total, respectively. These species are commonly emitted from paints, building coatings, and asphalt (Dehghani et al., 2017;Liu et al., 2008;Won et al., 2021b;Zheng et al., 2019), and factor 1 represents "Solvent usage". The daily variation of solvent usage in winter and summer was similar to the insolation pattern, increased from the morning, showed the highest concentration around noon (1.63 ± 1.41 and 2.84 ± 1.93 ppb in winter and summer, respectively), and showed a tendency to decrease (Fig. 4). Xylenes have a shorter lifetime (11.8 hour, 19.4 hour, and 20.3 hour in m-xylene, p-xylene, and o-xylene, respectively) compared to ethylbenzene (1.6 days) which is more stable in the atmosphere (Ho et al., 2004). Therefore, the m, p-xylene-to-ethylbenzene (X/E) ratio was used as an indicator to determine whether the air mass was affected by local sources or photochemical aging. If the X/E ratio is less than 3, it can be regarded as aging due to long-range transport or photochemical activity (Hui et al., 2020;Ibragimova et al., 2021) and the ranges of X/E ratios were between 2.5 and 2.9 in urban areas (Hui et al., 2020). The average X/E ratio measured in this study was 2.53 and 0.99 in winter and summer, respectively, indicating that the air mass was generally aging, and that the photochemical reaction. It has been reported that ethylbenzene and xylenes are easily removed by OH radicals, and their concentrations tend to decrease when photochemical activity is strong (Hui et al., 2020). In Fig. 2, ethylbenzene and xylenes were decreased around noon in summer and this was might due to strong photochemical activity.
Factor 2 had a large proportion of toluene, 3-methylpentane, and n-hexane in both winter (1.64 ppb, 26.5%) and summer (2.16 ppb, 25.1%). These compounds are associated with vehicle emissions, gasoline, and diesel fuel evaporation (Chan et al., 2006;Hui et al., 2020;Song et al., 2018). As for the vehicle exhaust source, the increase in rush hour in winter can be clearly observed in Fig. 4. Therefore, this factor was classified as "Vehicle exhaust". In summer, n-heptane, 2,2,4trimethypentane, and methylcyclohexane were highly loaded, and these components have also been reported to be affected by vehicle emissions, fuel additives, and diesel evaporation (Liu et al., 2008;McCarthy et al., 2013;Zweldlinger et al., 1990). In the case of Factor 3, benzene, n-nonane, n-decane, and C8-C9 aromatic VOCs such as ethyltoluenes and trimethylbenzenes had the highest contribution in both winter (1.93 ppb, 31.3%) and summer (1.92 ppb, 22.3%). C8-C9 aromatics, n-nonane, and n-decane are related to vehicle and industrial production processes (Hui et al., 2020;Song et al., 2018), and benzene is an indicator of fossil fuel combustion emissions. In northern China, the use of fossil fuels for heating has been shown to increase benzene concentrations during winter . The meteorological data of the KMA revealed that the winter contribution rate in Seoul may have increased owing to the influence of heating fuel used in northeast China, as the northwest wind mainly blows in winter. Industry/burning of fossil fuel sources gradually increased from around 9:00 AM and peaked in the afternoon (Fig. 4). There was no significant change from the evening to the morning, and the range of variations was larger in summer than in winter. The daily variations of Factor 3 in summer were estimated to indicate emissions during industry operating hours, excluding long-range transport by heating effects in winter. Therefore, Factor 3 was identified as "Industry/burning of fossil fuels". The ratio of cyclohexane was highest in factor 4, which is known to be a tracer of petrochemical complexes (Hsu et al., 2018;Hui et al., 2020;Liu et al., 2008). In Fig. 3, the VOCs concentrations in Factor 4 were similar in both winter and summer, as 0.99 ppb and 0.71 ppb, respectively. The petrochemical industry sources should be steady emissions without diurnal variation in both winter and summer, indicating that there was no local impact. Therefore, Factor 4 was classified as the "Petrochemical industry". The measurement site is not only influenced by a large-scale coal-fired power plant located approximately 50 km away in the southwest (Won et al., 2021a) but it also appears that the winter contribution may have increased because of the impact of air mass transport from northeast China in winter, as noted previously.
The last factor was named "Road emission" (0.34 ppb, 5.54%) and "Gasoline-related" (1.38 ppb, 16.0%) because the winter and summer factors displayed notable differences. In winter, n-octane exhibits a high loading, which is known to be released from asphalt or paint (Hui et al., 2020;Liu et al., 2008;Zheng et al., 2018). At a distance of 150 m from the sampling site, there is a round-trip, six-lane road that is paved with asphalt. In Fig. 4(a), daily variation of the road emission source is not observed, and because the contribution is low, it appears to be a steady emission. This is similar to a study that showed that asphalt sources do not have a large daily change compared to other sources and have the smallest contribution (Zheng et al., 2018). In summer, the contribution of 3-methylhexane and 2,3,4-trimethypentane was high loading; moreover, they have been reported to be emitted from vehicle emissions or industrial sources (Chang et al., 2006;Song et al., 2018;Song et al., 2020;Zweldlinger et al., 1990). According to previous studies, 3-methylhexane and 2,3,4-trimethylpentane showed good correlation with methyl tert-butyl ether, a motor vehicle indicator, and were classified as potential VOC tracers that could distinguish gasoline vehicles (Chang et al., 2006;Song et al., 2020). Unlike the vehicle exhaust of Factor 2, there is no noticeable daily variation (Fig. 4(b)); thus, it appears to be steadily emitted, which is presumed to be related to the presence of three gas stations within a radius of 1 km. For this factor, a detailed discussion through data analysis, such as using the conditional probability function, is apparently necessary in the future.

SOA Formation from Each VOCs
According to Li et al. (2020), VOCs and PM2.5 show good seasonal correlation, and identification of the precursor VOCs of SOA is critical in controlling PM2.5. In particular, anthropogenically and biologically derived VOCs can produce SOA via photochemical reactions. Although there was no seasonal variation in the concentration of EC (0.88 µg m -3 in winter, 0.74 µg m -3 in summer), the concentration of OC was higher in the summer (3.66 µg m -3 ) than it was in the winter (3.34 µg m -3 ), and its daily variation followed a pattern similar to that of PM2.5 in summer (Fig. 5(a)). In other words, the concentration of SOC in OC was higher in summer (2.46 µg m -3 ) than in winter (1.33 µg m -3 ) and correlated well with PM2.5, implying that secondary formation was active in summer. The correlations (r) of PM2.5 and SOC were high, at 0.72 and 0.73 in both winter and summer, respectively (p < 0.05). SOAI and SOAII were higher in summer (2.29 µg m -3 in SOAI and 375 ppb in SOAII) than in the winter (2.20 µg m -3 in SOAI and 309 ppb in SOAII), similar to the SOC. In Fig. 5(b), the results confirmed that the trends of the three methods (SOAI, SOAII, and SOC) were similar, and the validity of the SOA yield method was confirmed. The two SOA yield methods showed trends similar to SOC and PM2.5, and the correlation (r) with SOC was high for SOAI and SOAII, at 0.47 and 0.53, respectively (p < 0.05). Toluene is the primary contributor to both SOAI and SOAII (Table 3 and Fig. 6); however, the SOAI method indicates that the absolute concentration of individual compounds when applying SOA yield makes a significant contribution to SOA formation. Conversely, because SOAII sets toluene as 100 and applies a relative value to SOA formation, aromatic VOCs were predominant. The correlation between SOAI and SOAII is strong, at 0.97 and 0.98 in winter and summer, respectively (p < 0.05), which suggests that both methods are reliable for representing SOA formation. The compound that formed the most SOA was toluene, and the top 10 compounds (including toluene, benzene, xylenes, ethylbenzene, n-decane, and 1,2,4-trimbenzene) accounted for more than 95% of the SOAI and SOAII (Table 3). In particular, BTEX was found to be involved in the formation of most SOA, accounting for 63% of SOAI and 92% of SOAII. According to the study by Li et al. (2020), it was found that the top 10 contributing species, which included BTEX, were all aromatic VOCs, which accounted for more than 95% of the total SOA. Additionally, Zhang et al. (2017) reported that BTEX contributes to SOA formation by 80% or more. When the diurnal pattern for the top 10 compounds was examined, as shown in Fig. 6. Although, the methods used to estimate SOA differed, the trends of concentrations were similar. The highest SOA production was observed at noon in winter; however, the diurnal variation was not significant. Conversely, during summer, the diurnal pattern rapidly increased, and SOA production was active roughly 9:00 AM-5:00 PM. In Fig. 6(a), it can be observed that for SOAI, toluene and benzene were major contributors to SOA formation during the winter, while toluene and n-decane played significant roles in SOA formation during the summer. In Fig. 6(b), toluene was found to contribute significantly to SOA formation in both winter and summer. Additionally, in summer, ethylbenzene and xylenes appeared to be mostly involved in SOA formation.
In this study, the formation of SOA was examined by applying the SOA yield to the concentration of each species obtained through factor profile by PMF modeling (Fig. 7). During winter, the industry/burning of fossil fuels accounted for the highest proportion of SOAI (42.9%, 0.89 µg m -3 ). However, there was no significant difference in vehicle exhaust (28.6%, 84.4 ppb), industry/burning of fossil fuels (27.0%, 79.8 ppb), and solvent usage (26.9%, 79.5 ppb) for SOAII in winter. It has been shown that the contribution of SOA in winter differs due to the higher SOA yield from benzene, n-nonane, and n-decane contained in the industrial emissions of SOAI compared to SOAII. Conversely, during summer, the SOAI (31.9%, 0.67 µg m -3 ) and SOAII (55.7%, 194 ppb) by the solvent usage source was the most significant. The benzene contributions from the industry/burning of fossil fuel source decreased rapidly during summer, while the ethylbenzene, and xylenes from solvent usage increased. Among the sources, vehicle exhaust sources showed a constant SOAI (20.3%, 0.42 µg m -3 in winter and 21.9%, 0.46 µg m -3 in summer) and SOAII (28.6%, 84.4 ppb in winter and 27.6%, 96.1 ppb in summer) in winter and summer, implying that a significant amount of SOA is formed from toluene emitted to solvent usage rather than vehicle exhaust. In addition, this implies that VOCs management can also be used to manage SOA and PM2.5 through the derivation of emission source priorities and components to be controlled by SOA formation for each source.

CONCLUSIONS
In this study, 34 VOC compounds were investigated using a high-resolution instrument in the winter of 2020 and summer of 2021 to identify the seasonal characteristics of VOCs and the emission sources of VOCs in Seoul and confirm the contribution of VOCs to SOA.
During the monitoring period, the average mixing ratio of VOCs in Seoul was 6.28 ± 4.11 ppb in winter and 7.61 ± 4.22 ppb in summer. Toluene occupied the highest proportion, followed by n-hexane, n-nonane, m, p-xylene, and 3-methylpentane, ethylbenzene. The emission sources in Seoul were divided into five factors following the implementation of the PMF model: solvent usage, vehicle exhaust, industry/burning of fossil fuels, petrochemical industry, and road emission (winter)/gasoline-related (summer). Compared to summer, the contribution of industry/burning of fossil fuel increased in winter. The proportion of solvent usage increased in summer. There was no seasonal difference in vehicle exhaust, but in the daily variation in winter, there was a clear increase and decrease during rush hours. The three methods (SOAI, SOAII, and SOC) of calculating SOA generated by VOCs showed similar results, and greater SOA formation occurred in summer than in winter. The compound that contributed the most to SOA formation was toluene, and the top 10 compounds, accounted for more than 95% of the SOA formed. By each emission source, industry/burning of fossil fuel and solvent usage had a high possibility of SOA formation in winter and summer, respectively. In the diurnal pattern, the production of SOA was significantly influenced by benzene in winter and n-decane, ethylbenzene, and xylenes in the summer. In Seoul, a considerable amount of SOA is understood to be produced by aromatic VOCs. Therefore, managing SOA through aromatic VOCs by a major sources reduction policy is expected to have some impact on reducing PM2.5.