3.1. General Trends
The average PM
2.5 concentration during the study period was 29.4 (±16.8) μg m
−3 (
Table 3): about two times higher than the current annual NAAQS of South Korea (note that the annual PM
2.5 NAAQS changed from 25 μg m
−3 to 15 μg m
−3 in 2018). In addition, approximately 24% of the samples exceeded the current daily standard of 35 μg m
−3. The seasonal averaged PM
2.5 values were the highest in winter, followed by spring, fall, and summer (
Figure 2,
Table 3; note that spring, summer, fall, and winter are indicated as March-May, June-August, September-November, and December-February, respectively). The same trend was observed by other researches in South Korea [
16,
26]. Of the two winters included in the study period, that between 2016 and 2017 showed higher average PM
2.5 concentrations (36 ± 20 μg m
−3 vs. 30 ± 16 μg m
−3 of the winter between 2015 and 2016) and the maximum absolute PM
2.5 concentrations (84 μg m
−3 vs. 61 μg m
−3 of the winter between 2015 and 2016).
The Mann-Kendall trend analysis method was used for the mass of PM
2.5 and its chemical constituents. Many constituents including carbonaceous compounds (OC, EC, and WSOC) and cationic components (Na
+, NH
4+, and K
+) showed a decreasing trend during the sampling period, whereas the remaining constituents and PM
2.5 showed no trend at a significance level of 0.05 (
Table 4). Carbonaceous compounds were measured from November 2016 to October 2017; therefore, the decreasing trend reflects the seasonal variation—higher concentration in late fall and winter and lower concentration in summer (
Figure 2). The sampling period of about 22 months was not long enough to determine the temporal trend over years, thus, there is a need for long-term monitoring in the future.
During the whole sampling period, there were significant negative correlations between the PM
2.5 concentrations and temperature (Pearson r = −0.289,
p-value = 0.001) and wind speed (Pearson r = −0.332,
p-value < 0.001), and a statistical significant multiple linear regression was identified based on the wind speed and temperature data (Equation (4)): PM
2.5 concentrations generally increased in winter when the wind speed was low.
In the above equation, Ta and WS indicate the atmospheric temperature (°C) and the wind speed (m s−1), respectively. Both of these variables and a constant in Equation (4) were statistically significant (p-value < 0.001).
In this study, seven ionic constituents were determined and measured: their total contribution to PM
2.5 mass was found to be 28.5%. The average concentrations of NO
3−, SO
42−, and NH
4+ at the KNU site were 2.0 (±2.7) μg m
−3, 2.2 (±1.8) μg m
−3, and 2.9 (±1.7) μg m
−3, respectively (
Table 3). The contribution of NH
4+ to PM
2.5 mass (12.7% ± 13.2%) was the highest among the seven ionic constituents, followed by SO
42− (9.2% ± 7.4%) and NO
3− (7.2% ± 7.2%). As anticipated, the proportion of SO
42− and NO
3− among the ionic compounds was enhanced in summer and in winter, respectively (
Figure 3). Meanwhile, the proportion of NH
4+ has generally decreased in winter and increased in spring. K
+, which is typically used as a biomass burning marker, then increased in late fall and winter and was not abundant enough to be detected between May and September, along with Mg
2+ and Ca
2+. There was a positive correlation between Mg
2+ and Ca
2+ (Pearson r = 0.58,
p-value< 0.001), indicating that they were affected by the same factors such as crustal dust. The highest average concentrations of Mg
2+ and Ca
2+ were obtained in spring (
Table 3), probably because the soils that are frozen during winter are thawed in spring and subsequently suspended from the surface to the atmosphere [
17].
The proportion of Na
+ did not show any significant trend; however, there were four extraordinarily high peaks observed on March 16, 22, and 28, 2016, and January 10, 2017, showing Na
+ concentrations above 3.0 μg m
−3 (
Figure 2). Ming et al. (2007) [
27] identified the possible source of Na
+ using the Na
+/Ca
2+ ratio because the two main sources include marine aerosol and crustal dust. In this study, a higher concentration of Na
+ than Ca
2+ (
Table 3), resulted in an average Na
+/Ca
2+ ratio of 10.2 ± 11.8. The Na
+/Ca
2+ ratio was high, showing 24.8, 99.2, and 40.8 on March 16, 22, and 28, 2016, respectively, whereas it was 8.1 on January 10, 2017. Three high Na
+ samples obtained in 2016 were likely to be affected by sea-salt aerosol because the back-trajectories of air-parcels showed a long residence time over the ocean (
Figure 4a). On the other hand, the sample obtained on January 19, 2017 also showed the highest concentrations of both Mg
2+ and Ca
2+, indicating the effect of crustal dust. The corresponding 72-h back-trajectories originated from Russia and passed through Mongolia, China, and North Korea before arriving at the sampling site (
Figure 4b).
3.2. Sourcing the Measured Species
Correlation coefficients of 7 ionic compounds were identified (
Table 5). NO
3− was correlated with all other ionic constituents, and the highest correlation coefficient was observed with K
+. SO
42− showed the strongest correlation with NH
4+ which also correlated with NO
3−, indicating the existence of (NH
4)
2SO
4 and NH
4NO
3 in the PM
2.5. Na
+ was rather weakly but statistically correlated with NO
3−, NH
4+, K
+, and Ca
2+, suggesting that it was possibly affected by biomass burning and crustal dust. K
+ was highly correlated with NO
3−, but it showed a relatively weak correlation with SO
42−, Na
+, NH
4+, and Mg
2+, which indicates that KNO
3 was one of the major constituents of the PM
2.5.
The SO
42− concentrations obtained in this study are significantly lower compared to those reported in a previous study performed at the same site between 2013 and 2014 (3.9 (±3.6) μg m
−3 in Cho et al., 2016 [
28]; check
Table 3 for our results). This was probably because, according to the National Air Pollutants Emission Service of the Republic, SO
X emissions have continuously decreased in South Korea since 2011. However, the NH
4+ concentrations obtained in this study (
Table 3) were considerably higher compared to those measured in 2013 and 2014 (2.0 (±1.9) μg m
−3 in Cho et al., 2016 [
28]). Here, we calculated an NH
4+ availability index:
J (Equation (5)) (
Figure S1 of the Supplementary Material) [
29]. The value of this index generally exceeded 100%, indicating an NH
4+ surplus and the basicity of SO
42− and NO
3−. In this city, Chuncheon, a large portion of land cover is represented by agricultural areas. From there, significant amounts of NH
3 are emitted due to fertilizer application and cattle farms.
In the above equation, [NH
4+], [SO
42−], and [NO
3−] represent the molar concentration of NH
4+, SO
42−, and NO
3−, respectively.
J was the lowest in winter (182% ± 144%), despite still indicating, on average, an NH
4+ surplus (
Figure S1). The correlation between [NH
4+] and 2[SO
42−]+[NO
3−] was very good during winter (when the
J values were relatively low), indicating that NH
3 was combined with H
2SO
4 and HNO
3 in the form of (NH
4)
2SO
4 and NH
4NO
3, respectively. However, the correlation between [NH
4+] and 2[SO
42−]+[NO
3−] was clearly lower in other seasons (
Figure 5). In cities with size and emission characteristics similar to Chuncheon, NH
3 concentrations were reportedly higher in spring and lower in fall and winter [
30]. In South Korea, NH
3 has been observed to be lower in summer than in spring due to the occurrence of intensive monsoon-related raining episodes [
30]. The poorer correlation observed between the two same variables in spring, summer, and fall (
Figure 5) suggest that NH
3 remained excessive even when combined with H
2SO
4 and HNO
3 and, hence, it likely existed also in other forms (e.g., NH
4Cl) [
31]. Ammonia can react with hydrochloric acid to form NH
4Cl [
32,
33], and according to Du et al. (2010) [
31], NH
4Cl is a common constituent of aerosols in non-sulfate NH
4+, although NH
4NO
3 is more favorably formed than NH
4Cl. Unfortunately, Cl was not measured in this study. Therefore, the possible existence of NH
4Cl should be further investigated in future research. Notably, the correlation between [NH
4+] and 2[SO
42−]+[NO
3−] in fall (
Figure 5) became much stronger when excluding only three points. The three samples that had exceptionally low
J in fall (
Figure 5) due to the low NO
3− and SO
42− and high NH
4+ concentrations showed no notable features. However, relatively high Na
+, Mg
2+, and Ca
2+ and low K
+ concentrations in those samples, compared to other samples obtained in fall indicate that the suspension of soil treated by nitrogen-based fertilizers was a possible source [
34].
OC concentrations were relatively higher in Chuncheon than in other cities of South Korea, although their ECs were comparable [
35,
36], resulting in high OC/EC ratios (average = 9.6 ± 4.5). OC corresponded to approximately 40% of PM
2.5 mass, ranging from 36.5% in winter to 42.4% in spring. However, OC concentrations were the highest in winter and the lowest in fall (
Table 3). Moreover, the OC/EC ratios were about two times higher in summer (16.1 ± 5.0 in average of ratio) compared to the other seasons (8.0 ± 2.6), suggesting a major presence of secondary OC (SOC) in summer [
37]. The correlation coefficient between OC and EC was also determined to be lower in summer (Pearson r = 0.68) than in the other seasons (Pearson r = 0.82) (
Figure S2 of the Supplementary Material), likely indicating that a significant portion of the SOC was produced in summer. The high OC/EC ratios observed in all seasons implied the presence of other major sources besides mobile sources. The OC/EC ratios of vehicle exhaust emissions typically range from 1.0 to 4.2, while those of solid fuel combustions (including biomass and coal burning) are in a much higher range [
37,
38,
39,
40,
41]. Using the EC tracer method (SOC = OC
tot − [OC/EC]
pri), first suggested by Turpin and Huntzicker (1995) [
42], we calculated the average POC and SOC concentrations as 5.3 (±2.4) μg m
−3 and 3.9 (±2.5) μg m
−3, respectively. The [OC/EC]
pri indicates the OC/EC ratio directly emitted from combustion sources, which has often been estimated from the minimum OC/EC ratio [
42]. In this study, the minimum OC/EC ratio for the month was used as the [OC/EC]
pri for each month. The POC concentrations were comparable to the SOC concentrations in summer and fall, while they have significantly exceeded SOC values in spring and winter (
Figure S3 of the Supplementary Material). The average WSOC concentration was 4.4 (±2.4) μg m
−3, and the contribution of WSOC to OC ranged from 36.2% in fall to 57.5% in spring (no WSOC data were available for the summer season).
3.3. Effect of Humidity
Previous studies have shown that RH affects the gas-particle portioning of semi-volatile species, including NH
3, HNO
3, HCl, and some organic acids [
43,
44,
45]. A large number of lakes and reservoirs are located within Chuncheon, where the sampling site is located; therefore, fog episodes tend to be frequent and RH tends to be high. As indicated by Equation (1), PM
2.5 was negatively correlated with temperature and wind speed, but it was not correlated with RH during the whole sampling period. However, when considering only the data obtained during winter, the PM
2.5 concentrations and the PM
2.5/PM
10 ratio showed both a significant positive correlation with RH (Pearson r = 0.69 between PM
2.5 concentrations and RH, Pearson r = 0.79 between PM
2.5/PM
10 ratio and RH, both
p-values < 0.001) (
Figure S4 of the Supplementary Material). Among the PM
2.5 components, NO
3− (Pearson r = 0.60), SO
42− (Pearson r = 0.55), NH
4+ (Pearson r = 0.60), K
+ (Pearson r = 0.55), and OC (Pearson r = 0.68) were determined to be clearly correlated with RH at a significance level of 0.01, indicating that high RH promoted the formation of secondary inorganic and organic PM
2.5 during winter. In addition, NO
3− was highly correlated with K
+ (Pearson r = 0.74,
p-value < 0.001) and NH
4+ (Pearson r = 0.88,
p-value < 0.001), while it did not show any correlation with either Ca
2+ or Mg
2+, suggesting the possible formation of non-volatile KNO
3 salt under high RH in winter. These results imply that high RH likely provides suitable conditions for HNO
3, SO
2, and NH
3 to be condensed on humid particles in winter [
45,
46,
47,
48]. The prominent correlation between gaseous NO
2 and NO
3− (Pearson r = 0.70) and the high NO
3−/NO
2 ratio observed in winter (NO
3−/NO
2 = 0.051, 0.019, 0.025, and 0.074 ppm ppm
−1 in spring, summer, fall, and winter, respectively) also suggest the effective conversion of gaseous NO
2 to NO
3− during winter.
Although the formation of secondary sulfate and secondary organic aerosols in the aqueous phase is significant in summer [
49,
50], in this study, there was no statistical correlation between RH and PM
2.5 and its chemical constituents in summer. This could be because the number of samples in summer (
Table 3) were not enough to show clearly the effect of RH, or it may be difficult to determine the influence of RH on PM
2.5 and/or its constituents because the RH was relatively consistent in summer (the coefficient of variation (standard deviation/average) was 0.13 and 0.21 in summer and winter, respectively). Therefore, there is a need for further research on this.
3.4. Effect of Biomass Burning
The outskirts of Chuncheon (where the sampling site is located) consist mainly of forests and agricultural areas. There, the open burning of agricultural residue and household wastes frequently occurs. WSOC served as a proxy for secondary OC in many previous studies [
14,
51,
52], but it is also known to be directly emitted, along with K
+, during biomass burning [
15,
51,
53]. In this study, WSOC was significantly correlated with EC (Pearson r = 0.58) and K
+ (Pearson r = 0.52). This correlation became much stronger during fall and winter (Pearson r = 0.79 with EC and Pearson r = 0.83 with K
+, both
p-values < 0.001), when the burning of agricultural residue actively occurred. These results suggest that biomass burning was an important source of carbonaceous PM
2.5 in Chuncheon, especially during cold months. Particles derived from the burning of fresh biomass mostly consisted of K
+, Cl
−, and OC, but can be readily converted to KNO
3 and K
2SO
4 via heterogeneous reactions during their transport [
54,
55]. SO
42− (Pearson r = 0.63) and NO
3− (Pearson r = 0.74) were determined to be highly correlated with K
+ in winter, suggesting the importance of aerosol produced from the burning of aged biomass. In addition, the deliquescence relative humidity (DRH) of KCl is 84.3%, which was lower than that of KNO
3 (97.4%) and K
2SO
4 (92.5%) [
56,
57]. Therefore, at RH > ∼84%, chlorine gases can be released to the atmosphere, while KNO
3 and K
2SO
4 still exist in their solid forms, resulting in a strong correlation of K
+ with both NO
3− and SO
42−. These results suggest that biomass burning and high RH might have a synergistic effect, enhancing PM
2.5 concentrations in the city.
3.5. High Concentration Episode
In this study, a high concentration episode (HCE) was defined for daily PM
2.5 concentrations higher than the 24-h NAARS (35 μg m
−3). In total, 33 samples were identified as corresponding to a HCE. Of these 13, 12, 6, and 2 samples were obtained in winter, spring, fall, and summer, respectively. The average PM
2.5 concentration was 47.9 (±11.1) μg m
−3 during the HCE, ranging from 36.3 to 83.8 μg m
−3. Among them, the samples with the highest PM
2.5 concentrations were those obtained on February 4, 2017 (83.8 μg m
−3) and January 19, 2017 (81.5 μg m
−3) (
Figure 2). Despite the similar PM
2.5 concentrations, the components of these samples showed different concentrations. On February 4, 2017, the OC, EC, K
+, and NH
4+ concentrations all showed an increase (
Figure 2); moreover, the HCE sample collected on this day was characterized by the highest OC/EC ratio (OC/EC = 9.7), possibly indicating biomass burning as a source of PM
2.5. On the other hand, NO
3− and SO
42− reached their highest concentrations on January 19, 2017, indicating that the gas-particle partitioning of HNO
3 and heterogeneous reactions in the nitrate and sulfate formation were facilitated by low atmospheric temperatures (−2.4°C: the lowest temperature observed among HCE) and relatively high RH (77%: the average RH was 63% during winter) [
58]. NH
4+ was also high, but its increase was not noticeable compared to those of NO
3− and SO
42−, resulting in an NH
4+ deficit (NH
4+ availability index,
J = 69.7%). On January 19, the SO
42− concentration (9.3 μg m
−3) was approximately two times higher than on February 4, 2017 (4.4 μg m
−3), despite the similar PM
2.5 concentrations of the corresponding samples. The increase of SO
42− concentrations was possibly supported by the back-trajectories that originated from northeastern China and Mongolia on January 19, while the back-trajectories were much shorter and stayed for a longer time over the ocean on February 4 (
Figure 6). Previous studies have also highlighted an increase of SO
42− concentrations in parallel with long-range transport from China [
16,
59].
All of the PM
2.5 components significantly increased during HCEs (by a factor of 1.3 to 2.4, compared to non-HCEs) (
Figure 7). NO
3− was determined to be the most important component for the HCE samples, and its concentration increased by approximately 2.4 times; moreover, this component showed the strongest correlation with PM
2.5 mass (Pearson r = 0.77) among all components in the HCE samples, suggesting that the source increasing NO
3− should have been crucial for reaching the high PM
2.5 concentration observed in Chuncheon. With respect to mass, the highest enhancement was observed for OC: it increased by 4.6 μg m
−3 (
Figure 7). K
+ was also significantly higher in the HCE, rather than in the non-HCE samples (by a factor of 1.7), although its contribution to PM
2.5 mass was imperceptible.
In order to identify the characteristics of each HCE, we divided them into four groups: NO
3−–driven, OC-driven, crustal element-driven, and summer HCE. Notably, all the 10 components (shown in
Table 3) were analyzed only in 14 out of the 33 HCE samples. Six samples were classified as NO
3−–driven HCE. These showed significantly enhanced K
+ and NO
3− concentrations (
Figure 8), as well as the highest six K
+ concentrations among all the HCE samples. All HCE samples showed a significant increase in K
+ concentration (
Figure 7; nevertheless, the values of this parameter were particularly high in the six NO
3−–driven HCE samples (0.45 ± 0.08 μg m
−3,
Figure 8, suggesting the active formation of inorganic aerosol (e.g., NH
4NO
3 and KNO
3) as a result of biomass burning. Of these six samples, the three collected in spring (on March 16, 22, and 28, 2016) have showed PM
2.5 concentrations somewhat different from those of the four collected in winter: a distinct increase in Na
+ and NH
4+ occurred under low RH conditions (
Figure 8; notably, no carbonaceous compounds were analyzed in these samples). High Na
+ and NH
4+ concentrations might have been possibly derived from the trajectories of air-parcels that had a long residence time over the ocean (
Figure 4) and intense emissions of NH
3 in spring [
30], respectively.
Two HCE samples collected on 17 December 2016, and 4 January 2017, had very high OC fractions, contributing to 47.3% and 36.3% of the total PM
2.5, respectively, and were classified as OC-driven (
Figure 8). The nine PM
2.5 constituents analyzed in this study contributed to 83.7% of the total PM
2.5 mass of the HCE samples collected on 17 December 2016, indicating that metallic elements directly emitted from primary sources were insignificant, while secondary formation contributed considerably to the high PM
2.5 concentrations.
On 29 March 2017, crustal and/or ocean elements (e.g., Na
+, Mg
2+, and Ca
2+) clearly increased, showing much higher concentrations (∑3 elements = 2.9 μg m
−3) and contributed to a higher fraction of the total PM
2.5 mass (∑3 elements = 5.4%) compared to the other HCE samples (0.8 μg m
−3 and 1.7%) (
Figure 8). Two HCE samples collected in summer (on 15 and 30 June 2017) showed an increase in the SO
42− and OC concentrations (
Figure 8). The OC/EC ratios were also very high (13.1 and 9.6 on 15 and 30 June, respectively), and most of the endpoints of the 72-h backward trajectories were stagnant over the Yellow Sea (
Figure 6). These results suggest that a combination effect of active photochemical oxidation reactions and stagnant air can lead to high PM
2.5 concentration episodes in Chuncheon even during summer.
Back-trajectories were also calculated for the bottom 10% samples (when PM
2.5 < 11 μg m
−3), and they were relatively long and originated in the north, east, and south (
Figure S5 of the Supplementary Material) whereas the back-trajectories for HCEs were short and originated in the west.
3.6. Comparison with Other Studies
The characteristics of the chemical constituents of PM
2.5 varied in different regions. When the PM
2.5 concentration was high, the air masses generally transported from eastern China and Seoul (the capital of South Korea) before arriving at Chuncheon [
28,
60,
61]. Because Baengryeong island, located in the northernmost part of Korea, is the nearest region to China, it is the national background monitoring site in South Korea. Therefore, the PM
2.5 characteristics in eastern China, Baengryeong island, Seoul, and Chuncheon were compared. The concentrations of PM
2.5 and its ionic components were much higher than those in other sites (
Table 6). One of the notable results is that both the OC concentration and the OC/EC ratio (9.6 as in average of ratio) were significantly high in this study when compared to that in Beijing, China (3.7), Baengryeong island (3.5), and Seoul (2.4) (
Table 6). Considering the low OC/EC ratio generally in the metropolitans, such as Beijing and Seoul due to the large contribution of mobile sources [
37,
38,
39,
40,
41], the significantly higher OC/EC ratio in Chuncheon than in Baengryeong island indicates the importance of biomass burning and/or large fractions of aged aerosol. Because PM
2.5 generally increased with the long-range transport from China to Baengryeong island and Seoul, SO
42− greatly increased in the haze episode [
62]. However, in this study, NO
3− was the most important constituent along with K
+ and OC especially under the high RH condition, suggesting that the formation of secondary inorganic and organic components and biomass burning were the important processes for the high PM
2.5 episode. In a recent study performed in Beijing, China
[63], high nitrate to sulfate ratio (2.2) was observed, which increased up to 2.7 during haze events (
Table 6). In this study, an average of nitrate to sulfate ratio was 0.9 was obtained, which increased to 1.3 for HCEs, indicating that the increase in NO
3− was greater than that of SO
42− and other components and that NO
3− was concentrated as the PM
2.5 increased.