Seasonal Variation of PM 2 . 5 and Its Major Ionic Components in an Urban Monitoring Site

The ionic composition of PM2.5 samples was investigated by their datasets of cationic (Na+, NH4+, K+, Mg2+, and Ca2+) and anionic components (Cl−, NO3−, and SO42−) along with relevant environmental parameters collected from an urban monitoring site in Korea at hourly intervals in 2010. The mean (and SD) annual concentration of PM2.5 was computed as 25.3 μg m−3 with the wintertime maximum. In addition, sum concentrations (neq m−3) of five cationic species (291) were slightly lower than 3 anionic species (308). Most cations exhibited the highest seasonal values in spring, while anions showed more diversified seasonal patterns. According to PCA, five major source categories were apparent with the relative dominance of secondary inorganic aerosols (SIA). The results of our study suggest consistently that the distribution of ionic constituents in an urban area is affected by the combined effects of both natural and anthropogenic processes.


INTRODUCTION
As seen from elsewhere on the globe, many cities in Korea suffered from severe air pollution due to the combined effect of both local (e.g., vehicular traffic) and distant sources (e.g., transport from the industrial areas in China) (Han et al., 2008).In an effort to reduce environmental problems associated with air pollution, the Korean Ministry of Environment (KMOE) established special act plan for the improvement of Metropolitan Air Quality on 1 January 2005.The main objective of such plan was to suppress the annual PM 10 levels from 60 (in 2005) to 40 μg m -3 (in 2014) (Heo et al., 2009).However, in spite of various efforts, the PM 2.5 levels in Seoul are still higher than many other cities in developed countries (Kim et al., 2007).
In this study, eight major ionic components of PM 2.5 including five cations (Na + + , NH 4 + + , K + + , Mg 2 + + , and Ca 2 + + ) plus three anions (Cl -, NO 3 -, and SO 4 2-) were measured routinely at hourly intervals over one year period (1st January to 31st December in 2010) at Gwang Jin Goo district of Seoul, Korea.Relevant parameters were also monitored concurrently to include both airborne criteria pollutants and basic meteorological parameters.Note that most of the previous studies of PM compositions had to rely on data sets collected over daily intervals.In this respect, our hourly measurement data should be useful enough to describe the short-term variability of ionic constituents.The data acquisition at such short interval should be important, as the air pollution in Seoul is heavily dependent on highly dynamic source activities (e.g., traffic conditions).These data sets were analyzed to describe the fundamental factors controlling the distribution of major ionic constituents in a number of aspects in relation to the changes in the relevant environmental parameters.The main objective of the present study is to investigate the seasonal variability of PM 2.5 and its ionic constituents in air and to relate such results with its source/sink processes.The results of this study will thus help us gain valuable insights into the temporal variability of key chemical components of PM 2.5 in an urbanized environment.

MATERIALS AND METHODS
In this study, ambient air samples were collected on a hourly basis for one year period (1 st January to 31 st December, 2010) at air quality monitoring station located at Goo Eui dong in Gwang Jin district (127� 05.44′, 37� 32.40′) of Seoul, Korea.Gwang Jin area, located on the north side of the Han River of Seoul is home to couple of universities and many residential apartments along with a huge amusement park named as Children's grand park.Air quality of the area is known to be affected by traffic activities from adjacent roads without any major sources (Kim et al., in press).The analysis of PM 2.5 and its ionic constituents was made continuously by a model ADI 2080 online analyzer (MARGA, Applikon Analytical B.V Corp., Netherlands).This system consists of a sampling box and an analytical box with the particle collection efficiency of 99.7%.Ambient air is drawn into the sampling box at a flow rate of 1 m 3 h -1 through the inlet.The gaseous components are also trapped by the liquid film (0.0035% H 2 O 2 ) formed by one Wet Rotating Denuder (WRD).In the meantime, particles in residual airflow go through the supersaturated steam (0.0035% H 2 O 2 , 120-140� C).They are erupted out from one Steam Jet Aerosol collector (SJAC) and pooled into its collector.These liquefied samples flow through glass filters, and are stored in syringe pumps of the analytical box for the final detection by Ion Chromatography (IC).As such, MARGA utilizes a WRD to collect acid gases and ammonia by diffusion into an aqueous film, while particles passing through the WRD are collected in a SJAC.A supersaturated environment is created within the SJAC which grows particles by a process known as deliquescence; it then allows them to be collected subsequently by inertial separation.As cooling takes place, steam condenses and washes the collected particles into an aqueous sample stream.The soluble ionic species in aqueous solutions from the WRD and SJAC are subsequently analyzed by IC (ion chromatography).

1 General Tendency of PM 2.5 Concentration
In this research, the ionic components of PM 2.5 and relevant parameters were monitored at Gwang Jin district of Seoul at hourly intervals from 1 January to 31 December, 2010.To facilitate the data analysis, daily mean values for all PM 2.5 constituents were computed: such values were acquired, only when more than two thirds of data are available for the day (e.g., ¤18 hourly data).Hence, all the statistical treatment of data was basically made using these daily values, unless otherwise specified.The total number of daily values for different constituents was generally less than 340, as instruments were stopped occasionally either due to periodical maintenance or malfunctioning.Table 1 presents the summary of the cationic (Na + + , NH 4 , K + + , Mg 2 + + , and Ca 2 + + ) and anionic (Cl -, NO 3 -, and SO 4 2-) components of PM 2.5 measured in yearly and seasonal basis during the whole study period.The results are also computed for their derivatives (e.g., summation and ratio).
The annual concentration of PM 2.5 in our study was 25.3±16.0μg m -3 which exceeded the US EPA annual standard of 15.0 μg m -3 (Table 2).The annual mean (±SD) concentrations of 5 cations (Na + + , NH 4 + + , K + + , Mg 2 + + , and Ca 2 + + ) were 7.80±6.90,212±176, 11.1± 13.6, 14.3±9.11,and 39.5±28.7 neq m -3 , respectively.The concentrations for 3 anionic counterparts (Cl -, NO 3 -, and SO 4 2-) were 13.5±14.1,198±99.9,and 110±79.1 neq m -3 , respectively.To describe the fundamental aspects of their behavior and to validate the quality of our measurements, we checked for the extent of ionic balances using the molar concentration ratio between total cations and anions (» cations/» anions).The annual mean (and SD) of such ratios was less than unity (0.88±0.43) (or cation deficiency) which may signify that the study site is slightly acidic (Fig. 1).All of the ionic components in PM 2.5 exhibited mo-  B. Test of statistical significance for seasonal dominance.Although the results of PM 2.5 analysis are not consistent enough to cover various locations throughout Korea, gradual reductions in PM 2.5 levels across deca-dal periods are fairly apparent between different studies.The concentration of PM 2.5 was measured as 44.2 μg m -3 in Chongju, Korea in 1995-1996 (Hieu and Lee, 2010).The PM 2.5 values were measured as 27.6 μg m -3 in Ulsan in 2008 (Hieu and Lee, 2010).It's concentration in Daejon, another major city of Korea was 37.7±17.2μg m -3 in 2008 (Lim et al., 2011).In relatively clear areas like Deokjeok and Gosan in Korea concentrations of PM 2.5 were somewhat comparable with our data such as 25.3±22.et al., 2009).It is important to note that over the years, PM 2.5 has decreased consistently across Korea.Thus, it is not necessarily uncommon to find less PM 2.5 values in our experiment than the old studies.

2 Seasonal Variation of PM 2.5 Constituents
As the distribution of PM 2.5 is known to exhibit strong seasonal variations, so we may expect such pattern from its diverse constituents (Fig. 2).In order to have a thorough understanding of the temporal pattern between different ionic constituents, the seasonal patterns of PM 2.5 and its constituents were examined.The PM 2.5 concentration at our study site was the highest in winter with the value of 29±14 μg m -3 (range 3.28 -81) (Table 2).Such wintertime enhancement in PM 2.5 levels is comparable to those seen in many previous findings in Korea (Kim et al., 2004).It is interesting to note that all of the criteria pollutants (namely PM 10 , PM 2.5 , SO 2 , NO 2 , NO x , and CO), except O 3 consistently show wintertime maxima.The observed pattern in our study might also reflect the combined effects of source mechanisms (the use of fossil fuel for house heating and cold start of car engines) and meteorological conditions (e.g., reduced mixing heights in winter and enhanced wet depositions during summer) (Kim et al., 2007).In addition, enhanced PM levels (mainly crustal components) are commonly observed in the spring months (March and April) due to the common occurrences of Asian Dust (AD) event (Kim et al., 2003).However, in our study period of 2010, such signals were not strong enough.
According to our analysis, most ionic species tend to exhibit strong seasonality across four seasons, namely spring (March to May), summer (June to August), fall (September to November), and winter (December to February).Although some cations (Na + + , Mg 2 + + , and Ca 2 + + ) showed their peak occurrences in spring, others (NH 4 + + and K + + ) did in fall.However, this seasonal dominance is not statistically significant, if comparison is made between the highest seasonal mean and the one next to it (by the z statistics test at p⁄0.05).The observed fall peak of NH 4 + + , although insignificant statistically, is different from the finding of Lee et al. (1999) who reported winter time maximum in Seoul.
The concentrations of anions (Cl -, NO 3 -, and SO 4 2-) exhibited more diversified patterns with their maximum in winter (26.6), spring (22.0), and summer (133 neq m -3 ), respectively.The seasonal patterns of the former two are statistically significant, while it is not the case for SO 4 2-.The highest nitrate concentration is observed in spring closely followed by winter in our study.This is somewhat unusual, as it tends to peak in winter by enhanced consumption of fossil fuel and increased NO x levels.It is known that nitrate in the atmo-sphere can result mainly from the atmospheric conversion of nitrogen gases (NO and NO 2 ) and ammonia (NH 3 ) (Seinfield and Pandis, 1998).Moreover, the partitioning of inorganic NO 3 -(from HNO 3 and ammonium nitrate) is generally favored in cold weather (Khan et al., 2010).As such, its minimum values are typically observed in summer (Lee et al., 1999).
As a secondary pollutant, SO 4 2-was found to be the highest in summer.This pattern of SO 4 2-appears to be a common trend in many previous studies conducted elsewhere (Khan et al., 2010).This relative sulfate enhancement in summer can be attributed at least partially to such effect as increased photochemical activity (Husain and Dutkiewicz, 1990).In addition, sea salt emission may also make a certain contribution to the total SO 4 2-concentration, if one considers the distance between the study site and seashore (35 km).However, its winter time enhancement associated with high coal combustion was also reported (Cyrys et al., 1995).Sulphate is suspected to come from long-range transport as well as regional sources (Yao et al., 2002).Our wintertime maximum of Cl -concentration appears to match with the seasonal pattern seen in China (Yao et al., 2002).It is well known that high concentration of Cl - in the winter months is known to be associated with coal burning.Despite the observed complexities among individual ionic components, the sum concentrations of both cationic and anionic components peaked consistently in spring (292 and 345 neq m -3 , respectively).It is also interesting to find that the sum ionic ratios (» cations/» anions) were the highest in fall (1.03±0.65) but the lowest in winter (0.78±0.16).This observation may signify the fact that overall deviation from the unity in its ratio should have proceeded most significantly during winter with increased acidity.

3 Factors Controlling the Distribution of PM 2.5 and Its Constituents
There are many factors that can exert controls on the distribution of PM 2.5 and its ionic constituents.To learn more about the relationship between those components, their correlation patterns were examined by using their daily mean data on both yearly and seasonal intervals.Results shown in Table 3 indicate the frequent occurrences of strong correlations between many cationic and anionic species.Among the cations, NH 4 + + shows strong correlation with most cations (except Mg 2 + + ) and all anions; this trend is seen consistently throughout the seasons.Among the anions, SO 4 2-is strongly correlated with most cations in all seasons, while maximum frequency of strong correlation is observed in spring (Fig. 2).It is seen that Na + + also shows a strong correlation with Cl -(r= =0.31, p= =5.44E-09), suggesting the possible role of sea salt (NaCl) as the PM 2.5 source.If  ) also suggests a preferential formation of (NH 4 ) 2 SO 4 or NH 4 HSO 4 over more volatile NH 4 NO 3 .Direct emissions (e.g., NO 2 , SO 2 , CO, and PM 10 ) from motor vehicles are often explained in relation to traffic flow in Korea (Kim and Guldmann, 2011).Because diesel vehicles are still one of the major sources of SO 2 emission in urban areas of Korea (Korea Ministry of Environment, 2011), higher concentration of SO 4 2-can be maintained throughout the day.The concentration data of NH 4 + + show strong correlations with many other parameters (e.g., » cation (0.95), » anion (0.74), » all ions (0.89), and PM 10 (0.74)).It is thus reasonable to infer that NH 4 + + should be influenced by emissions from many anthropogenic sources in urban areas.In atmosphere, Ca 2 + + is generally believed to originate from soil, crustal material (e.g., calcite and dolomite), or construction activities (e.g, cement manufacturing).In our study, a strong correlation has been observed between Ca 2 + + and Mg 2 + + (r= =0.43, p= =1.22E-16).The relationship between Ca 2 + + and SO 4 2-(r= =0.24, p= =8.48E-06) also suggests the formation of CaSO 4 .Interestingly, a strong correlation (r = =0.52,p= =5.26E-25) between two anions namely NO 3 and SO 4 2-can be highly significant, if they exist as internal mixtures or if their concentrations are governed by common meteorological conditions (Zhuang et al., 1999).
The results of our correlation analysis were also evaluated after being divided into five different classes based on the magnitude of probability (P) such as: (1) no class (the weakest correlation range); P›10 -3 ; (2) class I (moderately weak correlation); 10 -5 ‹P⁄10 -3 ; (3) class II (10 -10 ‹P⁄10 -5 ); (4) class III (10 -10 ‹P⁄ 10 -5 ); and (5) class IV (the strongest correlation range): P⁄10 -15 .The results of this comparative analysis are summarized to allow direct comparison among different seasons (Table 3b).If we go through the correlation results derived from all yearly data, it is seen that among 66 matching pairs for each data group, the frequency of the strongest correlation pairs (IV) is the highest among all five classes for comparison.It is also seen that among 66 cases, majority of the values are found either in Class II ( 14), no class (20), and class IV (25).Among the seasonally grouped data sets, the strongest correlation pairs (IV) are found most frequently in spring with 17 out of 66 cases.In contrast, the least significant correlation case (no class) was dominant in winter with 37 out of 66 cases.

4 Source Identification of Constituents of
PM 2.5 using PCA For the effective management of air quality, it is important to characterize the sources of target pollutants.The fine particles are a mixture of primary and secondary (mainly anthropogenic origin) aerosols through vapor nucleation/condensation mechanisms (Kleeman and Cass, 1998).To learn more about the processes controlling the ionic composition of PM 2.5 , it is meaningful to assess relative contribution between different sources (Manoli et al., 2002).
Factor analysis (FA) is a useful tool to assess the relative roles between different emission sources without the aid of local source profile library.As part of FA, we conducted principal component analysis (PCA) (using SPSS v16) with varimax rotation using the daily average data of each constituent.As a type of multivariate analysis, PCA has been widely employed in air quality studies.The results yielded five source components, namely PC1 through PC5 which accounted for 30.6, 14.0, 13.3, 10.3, and 7.89% of the total valiance (Table 4).To simplify the interaction among the ions, only the positive r values are considered.PC1 is identified as secondary inorganic aerosol (SIA) component with high 'r' values of NH  (0.88) along with their summation terms.To have a better understanding of sources, PCA was performed on a seasonally divided data groups as well.The results showed the highest variance in summer (84.3%), followed by spring (82.6%), winter (79.5%), and fall (76.7 %).In each season, their main source was identified as secondary inorganic aerosol with the highest variance in spring (38.2%), followed by fall (36.2%),summer (34.6%), and winter (29.1%).In a recent study conducted at Corpus Christi, an industrial and coastal city in South Texas, US, secondary sulphate was also identified as the dominant source category (30.4%) of the apportioned mass which is comparable with our value of 30.6% (Karnae and John, 2011).A predominance of such sources as soil dust, traffic, oil/coal combustion, and road dust was seen in outdoor microenvironment in Daejon, one of the major cities in Korea (Lim et al., 2011).
The second main source (PC2) appears to be affected by the secondary nitrate with high 'r' value of NO 2 (0.73), NO x (0.80), and CO (0.66).In all seasons (except summer), secondary nitrate became the second main source.However, in summer, second source was identified as road dust with high r value of Na + + (0.71), K + + (0.64), Mg 2 + + (0.86), and Ca 2 + + (0.89).Third source (PC3) was unidentified with high r value of Cl -(0.67) and SO 2 (0.53).This source seems to be little ambiguous to a degree, although both SO 2 and Cl -can be generated from traffic emission.Hence, traffic activities can act as the source of this component.The fourth main source (PC4) was identified as the road dust with high r values of Na + + (0.57), K + + (0.53), Mg 2 + + (0.78), and Ca 2+ + (0.73).Road dust normally originates from multiple sources such as automobile exhaust, lubricating oil residues, tire and brake lining wear, street surface weathering, leaf detris, garden soil, etc (Rogge et al., 1993).In Mexico City, vehicular emission was also estimated as the most important source of PM 2.5 in an urban site (42%), followed by re-suspended dust (26%), secondary inorganic aerosols (11%), and industrial emission and food cooking (10% each).Result of factor analysis in Jeju Island, Korea indicated the existence of at least three major sources including sea-salt aerosol, secondary aerosol, and organic aerosol component in 1997 (winter) and 1998 (spring) (Kim et al., 2002).The observed difference between Jeju and our study may be explained by the fact that the target site in Jeju is a relatively clean area with low anthropogenic activities.

CONCLUSIONS
In order to describe the distribution characteristics of PM 2.5 and its ionic constituents, we analyzed their hourly data sets initially collected at hourly intervals for one year period in 2010.The seasonal distribution of PM 2.5 pattern in our study was generally characterized by its maximum in winter.The concentration of cationic components normally peaked in spring, while their anionic counterparts showed more diversified patterns.The seasonal variation of most ions was rather insignificant statistically, in terms of z test between the highest mean value and the one next to it.
When our data were examined by correlation analysis, the results indicated that most of the inorganic ion pairs maintained significant correlation.The results of PCA showed the existence of 5 major sources with the dominance of secondary inorganic aerosols (SIA).Thus, the results of our study, if analyzed in relation with seasonal and diurnal groups demonstrate consistently that the hourly data measured during the study period can be closely tied with the signatures of both natural and anthropogenic processes.

Fig. 1 .
Fig. 1.Scattered plot of summation of cationic and anionic components of PM 2.5 at Gwang Jin, Seoul, Korea for the year 2010 excluding two outlying data on upper left and another two outlying data on lower right.

Fig. 2 .
Fig. 2. Normalized monthly distribution of ionic components of PM 2.5 at Gwang Jin district in Seoul, Korea during 2010.

*
Correlation is significant at 0.05 level of significance.** Correlation is significant at 0.01 level of significance.B. Evaluation of correlation frequence on a seasonal scale.parameters are examined carefully, several interesting trends emerge.The observed correlation of NH 4 + + with SO 4 2-(r= =0.85, p= = 1.22E-94) and NO 3 -(r= =0.50, p= =7.33E-23) may imply similarities in their emission sources.A strong correlation of SO 4 2-with NH 4 + + (relative to NO 3 -

Table 1 .
Statistical a Mean±SD (median); and b Range (number of observations).

Table 2 .
Statistical summary of ionic components of PM 2.5 , PM 10 , and environmental parameters measured concurrently at Gwang Jin district in Seoul, Korea during 2010 A. Seasonal mean values of PM and relevant parameters.check the statistical significance of seasonal peak values by comparing between the highest mean value and the one next to it.deratevariation with their relative standard error (RSE) less than 10% (i.e., maximum of 6.67% in case of K + + ).The concentrations of criteria pollutants (SO 2 , NO 2 , NO x , CO, and O 3 ) averaged as 3.89±1.68,31.9±10.
cZ statistics is used to

Table 3 .
Correlation analysis using daily mean values of chemical components of PM2.5 at Gwang Jin district in the year 2010.

Table 4 .
Results a Rotation converged in 7 iterations.