3.1. General Trends
In this study, 363 samples of PM
2.5 were collected throughout the whole sampling period. The average PM
2.5 concentration was 27.8 ± 18.8 μg m
−3, which exceeds the annual NAAQS of Korea of 25 μg m
−3 (Note that the annual PM
2.5 NAAQS has been strengthened to 15 μg m
−3 in 2018). Seasonal PM
2.5 concentrations showed the highest value in the winter (40.5 ± 22.1 μg m
−3) and the lowest value in the summer (16.9 ± 11.1 μg m
−3) (
Table 3, ANOVA test,
p-value < 0.001), which is frequently observed in other studies in Korea [
14]. The percentage of the samples exceeding the daily NAAQS of Korea of 50 μg m
−3 (Note that the daily NAAQS has been strengthened to 35 μg m
−3 in 2018) was 10.2% (
N = 37). Most of the high concentration events (>50 μg m
−3) occurred in the spring (
N = 10) and winter (
N = 25) while the average PM
2.5 concentration was 68.0 ± 16.1 μg m
−3. Among 37 high concentration events, 7 fog events and 25 mist events occurred, which indicates that a high content of water vapor and low wind speed characterized environmental conditions that were conducive to PM
2.5 accumulation. Previous studies showed that the humidity significantly affected the secondary organic aerosols especially those formed by aromatic volatile hydrocarbons [
22,
23] as well as the secondary inorganic aerosols such as (NH
4)
2SO
4 and NH
4NO
3 [
24]. There are many large artificial reservoirs in this city frequently causing fogs, which may be an important factor leading to the high PM
2.5 concentration in this city despite the low PM
2.5 emissions compared with other cities. Yearly average PM
2.5 concentration was the highest in 2013. This also coincided with the highest relative humidity and the lowest average wind speed (
Figure 2).
Some previous research has suggested that PM
2.5 concentration is affected by meteorological factors including temperature, wind speed, and relative humidity (RH) [
25,
26]. There were statistical correlations between PM
2.5 and temperature, wind speed, and RH at a significance level of 0.05 (Pearson R = −0.431,
p-value < 0.001 for temperature, R = −0.364,
p-value < 0.001 for wind speed, R = 0.118,
p-value = 0.029). When the data were limited to winter only, the effect of RH on PM
2.5 was more evident, which showed a Pearson R of 0.478 (
p-value < 0.001). There was no correlation between PM
2.5 and temperature. We found a statistical significant multiple linear relationship between PM
2.5 with RH and wind speed in the winter.
where RH and WS indicate the atmospheric relative humidity (%) and wind speed (m s
−1), respectively. The multiple linear equation fits the data well (R
2 = 0.261, R = 0.511,
p-value < 0.001) and both variables of RH and WS and constant were statistically significant (
p-value < 0.001). When each of the RH and WS was used as a single independent variable, the PM
2.5 regression equation was still significant with a Pearson correlation coefficient of 0.478 (R
2 = 0.228,
p-value < 0.001) for RH and0.439 (R
2 = 0.193,
p-value < 0.001), which was somewhat lower than that of the multiple regression. This result indicates that the RH can play a significant role on the high PM2.5 concentration often observed during winter at this site. Significant correlation between PM2.5 and RH was not observed during any other season.
The average OC and EC concentrations were 7.8 ± 4.6 μg m
−3 and 1.2 ± 0.9 μg m
−3, which contribute 27.9% and 4.4% of the PM
2.5 mass, respectively (
Table 3). Both the OC and EC concentrations were the highest in the winter (OC = 10.7 ± 5.4 μg m
−3, EC = 1.8 ± 1.3 μg m
−3) and the lowest in the summer (OC = 5.6 ± 3.0 μg m
−3, EC = 0.7 ± 0.3 μg m
−3) (
Table 3, ANOVA test,
p-value < 0.001). However, the OC contribution to PM
2.5 mass was the highest in the summer (33.5%) while the EC contribution was the highest in the fall (5.5%) (
Table 3). Over the whole sampling period, the average OC/EC ratio was 7.7, which was considerably higher than those measured in other major cities in Korea (
Table 4). In addition, the fraction of OC in total PM
2.5 was also markedly higher at this site at approximately 28% while those in other sites ranged from 10% to 22%.
These results indicate that secondary OC was important and/or primary OC emitted by sources besides vehicular emission was predominant. Previous studies have stated OC/EC ratios of 1.0 to 4.2 for diesel-vehicle and gasoline-vehicle exhaust [
37,
38], from 2.5 to 10.5 for residential coal smoke [
39], from 16.8 to 40 for wood combustion [
40], from 32.9 to 81.6 for kitchen emissions [
41], and approximately 7.7 for biomass burning [
37,
40]. The seasonally averaged OC/EC ratios were 7.6, 10.0, 5.9, and 6.8 for spring, summer, fall, and winter in this study while the high OC/EC ratio in the summer was thought to be due to active photochemical reactions.
PC was predominant at this site and contributed 17%–50% (31.1% on average) of total OC. There was a distinct seasonal variation for PC, which was typically higher in the winter and the spring and lower in the summer (
Figure 3). A previous study found that the PC concentration was enhanced during the harvest season and smoke events [
42], which indicates a possible relation with biomass burning. The second-most contributor to OC was OC1 and its fraction of total OC showed a contrasting seasonal trend, which exhibited a high contribution in the summertime (
Figure 3). Therefore, the organic carbons with low molecular weight were predominantly formed secondarily via photochemical reactions. A few previous studies [
43,
44,
45] suggested that primary OC1 and OC2 were mainly emitted from the combustion of solid fuels such as biomass and coal while OC3 and OC4 were predominantly emitted from the combustion of gasoline and diesel or from road dust. In this study, the concentration of OC1 + OC2 surpassed OC3 + OC4 (
Figure 3), which possibly indicates that biomass and/or coal burning was more important than vehicular emissions in terms of the primary OC concentration.
3.2. Primary Organic Carbon (POC) and Secondary Organic Carbon (SOC)
Secondary organic carbon (SOC) was estimated by using the following equations as Turpin and Huntzicker (1995) [
46] suggested.
where
OCpri,
OCsec, and
OCtot represent the concentrations of primary OC, secondary OC, and total OC, respectively. In Equation (2),
a is the OC concentration emitted from non-combustion sources, which has often been negligible in previous studies [
46].
(OC/EC)pri indicates the OC/EC ratio directly emitted from combustion sources, which has often been estimated from the minimum OC/EC ratio [
2]. In this study, Deming regression was used to estimate the
a and the
(OC/
EC)pri of the samples with the lowest 10% of the OC/EC ratio values for each season, which was used in the studies of Chu (2005) and Saylor et al. (2006) [
47,
48]. The Deming regression equations derived for each season are shown in
Table 5 (and
Figure S1), which indicated that the (OC/EC)
pri ranged from 3.2 to 4.7 and that the POC from non-combustion sources was estimated to be negligible due to being <0.2 μg m
−3 for all seasons.
Throughout the sampling period, the average POC and SOC concentrations were 4.7 ± 3.3 μg m
−3 and 3.4 ± 2.7 μg m
−3, respectively (
Table 3). SOC contributed approximately 42% of total OC throughout the sampling period, which increased to 51% in the summer time. This suggests that SOC was being actively formed via photochemical reactions. Although the SOC fraction in the summertime was considerably higher than those during other seasons, it was relatively lower than those observed in other cities and ranged from 53% to 63% [
28,
31,
49]. This indicated that the contribution from primary combustion sources was also important throughout the season.
Among all of the OC fractions, the best correlation with POC was found with PC (Pearson R = 0.78) while the highest correlation coefficient for SOC was observed with OC1 (Pearson R = 0.67), which indicated that a predominant fraction of PC was directly emitted from combustion sources and that a significant portion of the semi-volatile organic carbon such as OC1 was formed secondarily by reactions in the ambient air.
3.4. Polycyclic Aromatic Hydrocarbons
∑PAHs concentrations were significantly higher in the winter than in other seasons because PAHs are mainly emitted from incomplete combustion and are also semi-volatile compounds. Since PAHs are semi-volatile, their concentrations were high in the winter and low in the summer. Low molecular weight (LMW) PAHs with two or three rings are likely to exist as gases due to their high vapor pressure and result in low concentrations in particulate matter (
Table 6). On the other hand, high molecular weight (HMW) PAHs with five or six rings tend to exist in the particulate phase due to their low vapor pressure and they are predominant in the atmosphere. ∑PAHs were positively correlated with EC (
r = 0.74,
p-value < 0.001) and PM
2.5 (
r = 0.66,
p-value < 0.001) while the correlation with OC was not statistically significant (
r = 0.33,
p-value = 0.051). According to Di Wu et al. (2014) [
52], carcinogenic PAHs (C-PAHs) include BaA, BbF, BkF, BaP, DaA, and IcP. Although the summed concentration of carcinogenic PAHs was significantly higher in the winter (8.7 ± 5.4 ng m
−3) than in the spring (2.7 ± 2.5 ng m
−3) or summer (1.0 ± 0.6 ng m
−3), its fraction in ∑PAHs was 46.1%, 57.0%, and 66.4% in the winter, the spring, and the summer, respectively. This indicates that the toxicity per PAHs mass was the highest in the summer. The main PAH species emitted from combustion sources include FLT, PYR, CHR, BbF, BkF, BaA, BaP, IcP, and BgP [
53] and their summed concentration was 15.6 ± 7.7 ng m
−3 in the winter, which contributed 82.7% of ∑PAHs and 0.7 ± 0.5 ng m
−3 in the summer. This adds 49% of ∑PAHs. The top 10% of ∑PAHs samples showed 27.6 ± 6.8 μg m
−3 on average and mostly occurred in January. Compared with other samples, most PAH species showed significantly higher concentrations for the top 10% of ∑PAHs samples. The highest increase was observed for FLT followed by IcP, but FLU, DaA, and BgP did not increase for the top 10% of ∑PAHs samples (
Figure 5). IcP and FLT are mainly emitted from the combustion of solid fuels such as coal and biomass [
54,
55] rather than petrogenic and pyrogenic sources, which shows that coal and/or biomass combustion greatly enhanced the concentrations of ∑PAHs.