Physicochemical Characteristics of PM 2.5 Based on Long-term Hourly Data at National Intensive Monitoring Sites in Korea

The objective of this study was to analyze the temporal and spatial characteristics of fine particulate matters by using huge hourly datasets of PM 2.5 , including chemical information monitored at the 6 national intensive monitoring sites (NIMSs) from 2013 to 2018 in Korea. Hourly PM 2.5 raw datasets were obtained from the National Institute of Environmental Research (NIER) in Korea. Monitoring sites included urban, rural/agricultural, industrial, and marine environments. Since the PM 2.5 concentration steadily decreased nationwide, each species concentration also decreased in general. One of key reasons for decreasing PM 2.5 might be explained by the implementation of domestic fine dust reduction policies and external influences such as PM 2.5 concentration reduction in China. It was observed that 45.0% of all datasets for 6 years were classified as good condition. The average sum of 14 elements over all sites in 2018 was calculated to be 501.5 ng/m 3 , and its mass ratio for PM 2.5 (21.9 μg/m 3 ) was 2.30%. The inorganic elements were generally higher in industrial/urban areas than in agricultural areas. In addition, the average TC (total carbon) over all 6 sites was 28.3% of PM 2.5 with the range of 23.6% to 31.4%. The TC in small urban areas was much higher than that in marine areas or even that in large, populated urban area/industrial areas. It seemed that the latter areas were better controlled than the former area in terms of combustion activities of fossil fuels. It is suggested that these results could be play an important role as important basic data to manage ambient air quality and establish effective emission reduction strategies in each region.


INTRODUCTION
Fine particulate matter (FPM) in the accumulation size mode remains and is transported in air much longer than that in other modes (McMurry et al., 2004;EPA, 1999).Thus, it can increase exposure time and directly affect health risks.Further, it is obvious that PM penetrates deep into the alveoli through the upper and lower respiratory tracts of the human body and causes pulmonary diseases such as emphysema or asthma, or cardiovascular diseases (Dockery and Stone, 2007).In addition, it reduces visual range and adversely affects crop growth (KOSAE, 2018).The WHO (World Health Organization) estimated that PM in the air environment contributed to 3.2 million premature deaths worldwide in 2010, mostly from cardiovascular diseases, and 223,000 deaths from lung cancer (WHO, 2013).Related to outdoor and indoor air pollution, WHO recently evaluated 7 million deaths worldwide in 2016 due to cardiovascular disease, stroke, respiratory disease and cancer (WHO, 2021a), and then strengthened air quality guidelines for PM 2.5 down to 5 μg/m 3 for the annual mean and 10 μg/m 3 for the 24-hr mean in Sep.2021 (WHO, 2021b).
The National Ambient Air Quality Standard (NAAQS) in Korea for PM started in 1983 as total suspended particulate (TSP) with a diameter of 500 μm or less.TSP was deleted from the NAAQS, and only PM 10 has been regulated since 2001 (Kim, 2013).In 2015, the NAAQS for PM 2.5 was set to an annual mean of 25 μg/m 3 and a 24-hr mean of 50 μg/m 3 .However, since the standard was too loose compared to WHO or other foreign standards, the MOE (Ministry of Environment) strengthened the PM 2.5 standard a few years later, and the annual and 24-hr standards have been changed to 15 μg/m 3 and to 35 μg/m 3 , respectively, since March 2018 (MOE, 2018).According to a report (MOE, 2020a), annual average level for PM 2.5 slightly improved to 26 μg/m 3 in 2016, 25 μg/m 3 in 2017, and 23 μg/m 3 in 2018, but the levels still far exceeded the latest standard of 15 μg/m 3 .Due to significant social problems related to PM 2.5 in the last decade in Korea, the "Special Act on PM Reduction and Management" was enacted on Feb 2019 (MOE, 2019a).By law, the MOE had to prepare a comprehensive PM management plan by November 2019.The goal of the plan was to reduce the average level of PM 2.5 nationwide from 26 μg/m 3 in 2016 to 16 μg/m 3 in 2024 during the period of 2020 to 2024 (MOE, 2019b).
PM is not a simple compound, but a mixture of particles with various sizes containing various chemical species.Usually, PM is composed of sulfate, nitrate, ammonium, and other ionic species, carbon species (such as organic, inorganic, and elemental carbons) and other inorganic elements.Properly providing supplementary meteorological parameters, the chemical species in PM (PM 10 and PM 2.5 ) provide key information on where the PM came from, what the PM sources were, how the PM formed, and even when the PM formed.For example, PMs composed of many elements such as Si, Al, Fe, K, Ca are mainly emitted from a soil-related source, sulfate is from combusting sulfur-containing fossil fuels, and V and Ni are from oil combustion sources (Hopke, 1985).
Recently, there have been many studies to characterize chemical species of PM 2.5 nationwide in Korea such as a study of chemical characteristics of PM 2.5 in Seoul in 2019 (Um et al., 2020), a study of the origins and distributions of atmospheric ammonia in Jeonju during 2019-2020 (Park et al., 2020), a study on the characteristics of PM 2.5 chemical compositions and high-concentration episodes from 2013 to 2016 in Jeju (Kim et al., 2020), a study on PM 2.5 and its chemical compositions from 2017 to 2018 in Jeonju (Jo et al., 2018), estimation of emission source contribution of OC and EC in the spring of 2016 in Seoul (Ham et al., 2017), and a study on the distribution of heavy metals during 2013 and 2014 in Gyeonggido (Kim et al., 2014).
The purpose of this study is to analyze the temporal and spatial characteristics of fine particulate matters by using huge hourly datasets of PM 2.5 , including chemical information monitored at the 6 national intensive monitoring sites (NIMSs) in Korea during the period of 2013 to 2018.These research results can play an important role as basic data for air quality management and establishment of effective emission reduction strategies in each region.

1 Data Collection and Analytical Methods
The national intensive monitoring sites (NIMSs) have been operated by the Ministry of Environment (MOE) since 2013 and were developed (i) to describe the status of the ambient air environment nationwide including urban/agricultural/industrial/marine areas, (ii) to measure various ionic species and trace elements contained in long-range transported PM, and finally (iii) to identify the reasons for high PM episodes in Korea.There are a total of six sites such as the Joongbu Site (NIMS-JB), the Seoul Metropolitan Site (NIMS-SM), the Honam Site (NIMS-HN), the Baengnyeongdo Site (NIMS-BN), the Youngnam Site (NIMS-YN), and the Jeju Site (NIMS-JJ) (MOE, 2019c).Recently, the MOE has made the hourly raw data containing chemical information of PM 2.5 available.Fig. 1 shows the locations of these six NIMSs in Korea, and Table 1 provides information on location, site category, and monitoring purpose for each NIMS (NIER, 2019).In summary, NIMS-SM and NIMS-JB were established to represent air in large and medium urban areas, respectively, NIMS-HN is used to monitor rural/agricultural areas, NIMS-YN to monitor industrial air, and NIMS-BN and NIMS-JJ to monitor background marine air.
The MOE provided raw sample datasets monitored during 6 years from January 1, 2013 to December 31, 2018 for all 6 NIMSs.The hourly datasets for PM 10 and PM 2.5 as well as chemical information on PM 2.5 were analyzed in this study.PM 10 and PM 2.5 were separately monitored at each NIMS via size selection by an inertial impactor type inlet and a cyclone type inlet, respectively.After collecting the PM on each filter, mass concentration was determined by a beta-ray absorption method, which detects beta-rays before and then determines PM concentration by beta-ray attenuation after sampling with a high-sensitivity detector (NlER, 2019).Ionic species in PM 2.5 were analyzed by ion chromatography after PM collection.Carbon species were analyzed by a thermal/optical transmittance method and a non-dispersive infrared method.Finally, non-destructive x-ray fluorescence spectroscopy (XRF) was employed for trace elements.Further detailed analytical and sampling methods in this study can be found in the literature (NIER, 2019).

2 Data Pretreatment Steps for Characterization
of PM and Chemical Species in PM 2.5 Statistical analysis was performed after proper pretreatments for sample datasets monitored at hourly intervals from 2013 to 2018 over 6 NIMSs.Initially when either a PM 10 or PM 2.5 mass datum was missing, it was removed from each sample dataset at each NIMS.A datum was also removed when the PM 2.5 mass was higher than the PM 10 mass, that is, when F/C ratio>1 (i.e., when the mass ratio of fine particles/coarse particles = PM 2.5 /   2. By using the remaining samples, PM 10 and PM 2.5 mass behaviors, F/C ratios, and relationships with meteorological parameters were investigated for each site. In Study-2 for chemical characterization, all raw datasets consisting of PM 2.5 mass, 8 ionic species (SO 4 2-, NO 3 -, NH 4 + , Cl -, Na + , K + , Mg 2+ , and Ca 2+ ), and 2 carbon components (OC, EC) were examined during the 2 nd pretreatment step.When any one of those 10 chemical species was missing or if the sum of them was greater than PM 2.5 mass, the hourly dataset for that time was deleted from the analysis.The number of raw samples was 305,208, but only 166,331 samples (54.5%) remained after the pretreatment.In Study-3, the mass ratio for each inorganic element was analyzed based on samples measured only in 2018.Among the initial raw data of 17 inorganic elements (i.e., Si, S, K, Ca, Ti, V, Cr, Mn, Fe, Ni, Cu, Zn, As, Se, Br, Ba, and Pb), three elements (S, K, and Ca) were excluded since those overlapped with ion species mentioned above.For the 3 rd data pretreatment step, all the corresponding raw samples were deleted during this pretreatment step if 1) any of the 24 species including 14 elements was missing in the same sam-ple, 2) PM 2.5 measurement was missing or its mass concentration was zero, or 3) if the F/C ratio exceeded 1.After the step, the number of initial samples was 52,560, but only 27,910 samples (53.1%) were left for the next analysis.

RESULTS AND DISCUSSION
3. 1 Physical Characteristics of PM (PM 10 and PM 2.5 ) for Study-1 For 6 years from 2013 to 2018, the average annual range of PM 10 concentration at each NIMS was 28.9 to 50.9 μg/m 3 , and PM 2.5 was 16.1 to 31.9 μg/m 3 .Table 3 summarizes 6-year average statistics for PM, F/C ratio, and meteorological parameters at each site.For the PM levels, it was observed that NIMS-JB> NIMS-SM> NIMS-HN>NIMS-BN>NIMS-YN>NIMS-JJ.The NIMS-JB located on a medium sized urban city in the middle of Korea showed the highest PM level.However, the NIMB-JJ located on the biggest southern island showed the lowest PM level.As shown in Table 3, PM 2.5 levels over the country exceeded the 2021 WHO guidelines as well as the 2018 Korean NAAQS.
In our study, the average range of the F/C ratio was 0.55 to 0.64 over 6 NIMSs, as shown in Table 3.We observed that the F/C ratio tended to be higher as the average concentrations of PM 10 and PM 2.5 increased together.This means that the increase rate of PM 2.5 is generally faster than that of PM 10 .Further, the ratio was higher in urban and agricultural areas than in marine and industrial areas.It is obvious that there are enormous PAL sources (point, area, line source) emitting primary PM 2.5 as well as man-made gaseous pollutants forming secondary PM 2.5 in densely populated urban areas.However, there are few distinct PAL sources in sparsely populated background marine areas.Though many point sources exist in industrial areas (like near NIMS-YN), it seems that the sources in that area are well controlled by strict emission regulations.On the other hand, even though there are few point and line sources in agricultural areas like near NIMS-HN, the area sources emitting gaseous pollutants like NH 3 might be a major contributor to the formation of secondary PM 2.5 , and there are many scattered area sources emitting various gaseous and particulate matter from illegal burning of agricultural wastes and solid refused fuel (SRF) in that area.According to the CAPSS (Clean Air Policy Support System) data of the Ministry of Environment as of 2018, NH 3 emissions from the agricultural sector were the highest in Honam region with 67,836 tons, followed by the Youngnam, Joongbu, and the metropolitan area at 59,930 ton, 59,877 ton, and 42,677 tons, respectively (NAEIRC, 2019).
A correlation result among annual averages of PM 10 , PM 2.5 , and F/C ratios showed that PM 10 and PM 2.5 were highly correlated with r = 0.941 and p-value = 0.000.However, PM 2.5 and F/C ratio were moderately correlated with r = 0.737 and p-value = 0.000.Furthermore, PM 10 and F/C ratios were poorly correlated with r = 0.513 and p-value = 0.002.In addition, PM 10 and PM 2.5 were higher at NIMS-JB and NIMS-SM, both located in urban areas, where WS was quite low at 1.35 m/s and 2.11 m/s, respectively, than at NIMS-JJ (in a marine area), where PM 10 and PM 2.5 were the lowest with an average WS of 3.08 m/s.Thus, it seems to be that WS inversely impacted PM 10 and PM 2.5 by dispersed bulk motion.However, even though NIMS-BN located in far western marine areas had the highest wind speed of 4.1 m/s, elevated levels of PM 10 and PM 2.5 were observed compared to NIMS-YN in land industrial area and NIMS-JJ in a southern marine area.It might be directly affected by PM flowing from outside since NIMS-BN is 180 km away from mainland China and only 14 km away from North Korea.According to Im et al. (2021), high PM and PM 2.5 during the period of 2015 to 2020 were affected by incoming PM from outside because there are no distinct emitting sources inside the island.
As another correlation result among annual averages of PM 10 , PM 2.5 , and WS, when all average data over all sites were used, both PM 10 and PM 2.5 were weakly but negatively correlated with WS with r = -0.436,p-value = 0.009 and r = -0.518,p-value = 0.001, respectively, at 95% significance level, as shown in Fig. 2  However, since NIMS-BN had a big standard deviation and the fastest WS when calculating annual averages of PM, we reanalyzed the relationships between PM and WS after excluding NIMS-BN.The result showed that the correlation coefficient was negative but stronger than before, as shown in Fig. 2(c) and (d).PM 10 and PM 2.5 were clearly negatively correlated with WS by r = -0.742,p-value = 0.000 and r = -0.778and p-value = 0.000, respectively.The point was that less scattered plots were obtained when NIMS-BN was excluded in the analysis.
The air quality of NIMS-BN was likely influenced by wind direction rather than wind speed because the site is located on a small island is surrounded by China, South and North Korea and it has few internal emission sources, as mentioned above.Fig. 3 shows annual average concentrations of PM 10 and PM 2.5 at each NIMS, which show an overall decreasing trend.It must be noted that the annual average PM was calculated with sample datasets after performing the data pretreatment step, as shown in Table 2. Thus, 277,681 out of 305,208 datasets were utilized in this analysis.As shown in Fig. 3, the annual PM 10 and PM 2.5 at NIMS-JB gradually decreased after peaking in 2014 at 61.3 μg/m 3 and 38.4 μg/m 3 , respectively.They were 45.3To comprehensively understand the air quality status, we initially analyzed all 277,681 hourly PM 2.5 samples over 6 NIMSs and classified the samples into 5 conditions according to the forecast criteria by the guideline of the MOE (MOE, 2018): good condition (PM 2.5 ≤15 μg/m 3 ), moderate (15-35 μg/m 3 ), bad (35-75 μg/m 3 ), very bad (75-150 μg/m 3 ), and the worst warning class (PM 2.5 >150 μg/m 3 ).As a result, the air quality condition for 6 years is shown in Fig. 4. The results were as follows: 45.0% of the hourly PM 2.5 samples were sorted into good condition, 38.0% in normal, 15.2% in bad, 1.8% in very bad, and 0.1% in the worst warning conditions.The figure shows the trend of the conditions for each NIMS and plots (a) through (f ) contain annual trends for each NIMS.The sums of bad and worse conditions (that is, highly polluted conditions) were 34.6%, 30.0%, and 25.3% observed at NIMS-JB, NIMS-SM, and NIMS-HN, respectively, while 17.4% and 16.4% were observed at NIMS-BY and NIMS-YN, respectively, and lastly NIMS-JJ was only 8.8%.It is obvious that the good/moderate conditions are increasing in all sites, and the bad/very bad/worst conditions are continuously decreasing.Specifically, in the case of NIMS-SM in Seoul, the percentage of good/normal conditions in 2018 increased by 82.4% compared to 2013.The improved

2 Chemical Characteristics of 8 Ionic Species
and EC/OC in PM 2.5 for Study-2 PM 2.5 is emitted directly from various sources as primary aerosols such as PAL sources, which are mostly emitted by combustion of fossil fuels (Kondratyev et al., 2006).PM 2.5 is also generated as a secondary aerosol in the forms of sulfate (SO 4 2-), nitrate (NO 3 -), ammonium (NH 4 + ) by chemical reactions with various precursor gases like sulfur oxides (SO x ), nitrogen oxides (NO x ), and ammonia (NH 3 ) in the troposphere (EEA, 2020).In the presence of moisture (H 2 O) in the atmosphere, the SO 4 2-and NO 3 -react with NH 3 to generate particulate ammonium sulfate ((NH 4 ) 2 SO 4 ) and ammonium nitrate (NH 4 NO 3 ).In addition, secondary PM 2.5 is also generated by a photochemical reaction between ozone (O 3 ) and various volatile organic carbons (VOC) in the form of secondary organic aerosols (SOAs) in the atmosphere (McMurry et al., 2004).
Among the chemical species measured hourly at 6 NIMSs, the 6-yr average concentrations (5-yr average in the case of NIMS-YN) for each species in PM 2.5 were calculated and then plotted in , Cl -, Na + , K + , Mg 2+ , and Ca 2+ ), and 2 carbon components (OC, EC).Again, each concentration for each species was a mean calculated based on the hourly samples during the period of 2013 to 2018 except NIMS-YN during 2014 to 2018.In addition, Fig. 6 is a location map together with chemical compositions in terms of average ratios of each species to PM 2.5 .The 'others' in the figure stands for the average percentage of all other unmeasured species except the 10 measured species in PM 2.5 .
The average mass ratio of SO 4 2-was the highest among all 10 species measured.The SO 4 2-ratios at NIMS-SM and NIMS-HN were higher than the other NIMSs.It is known that SO 2 as a precursor of SO 4 2-is emitted from power plants, industrial facilities, and various fugitive area sources using fossil fuels such as coal, oil, and bio-SRF (biomass-solid refuse fuel) containing sulfur.Even though the forming rate in winter is slower than in other seasons (Hodan and Barnard, 2004), a huge amount of SO 4 2-was emitted from local areas in addition to longrange transport from the outside regional area due to an increase in burning activity in winter (Park et al., 2019).The Korean MOE reported that SO 2 is mainly emitted from the power generation sector in the energy industry and other sectors in petroleum product and manufacturing industries (MOE, 2020a).The European Environment Agency (EEA) estimated that SO 2 is emitted from energy supply (47%), manufacturing and industrial sectors (33%) (EEA, 2020).
As shown in Fig. 5, the average concentrations of NO 3 in 2018 ranged from 1.14 to 7.56 μg/m 3 over NIMSs, in the order of NIMS-SM (7.56 μg/m 3 )>NIMS-JB (4.92 μg/m 3 )> NIMS-HN (4.14 μg/m 3 )> NIMS-YN (2.76 μg/m 3 )> NIMS-BN (2.36 μg/m 3 )> NIMS-JJ (1.14 μg/m 3 ).The average mass ratio of NO 3 -in PM 2.5 was 14.2%, showing the second highest ratio among 10 species.Particularly, NIMS-SM in Seoul was observed to have a 21.5% mass ratio, which was remarkably high compared to the other NIMSs.It is obvious that gaseous NO and NO 2 as a precursor of particulate NO 3 -are widely emitted from mobile sources in metropolitan areas.In the USA, SO 4 2-was higher in the eastern part of the country than in the west, whereas NO 3 -was a specific major pollutant in western urban areas (Hodan and Barnard, 2004).The EEA showed that NO x was emitted mainly from mobile sources (39%), energy supply, manufacturing and industrial sectors, and agriculture (15% each) (EEA, 2020).According to the CAPSS in Korea (MOE, 2020b), mobile sources including passenger cars were the largest emitters, followed by freight cars> ships>RVs>construction equipment.Point sources such as power plants and chemical manufacturing processes were the next highest emitter of NO x .
The OC/EC ratios were calculated for all samples over all sites, and the ratios were in the range of 2.86 for the large urban NIMS-SM to 4.14 at the industrial NIMS-YN with an average of 3.56 in Korea.Tan et al. (2009) reported that OC/EC ratios ranged from 2.8 to 6.2 with an average of 4.7 during normal and hazy days in Guangzhou, China.Part et al. (2001) measured ratios at the Sihwa industrial area in 1998-1999 in Korea.The ratios ranged from 4.4 to 12.0 and were higher than those for other urban and rural environments.They concluded that the ratio from primary emissions was influenced by temporal fluctuations of factory activities in the industrial complex and meteorological conditions.On the other hand, average OC/EC ratio was 6.6, a ratio similar to that of biomass burning emissions in Agra, India (Pachauri et al., 2013).
Again, as can be seen in Fig. 6, the average concentration for other minor species such as Cl -, Na + , K + , Mg 2+ , and Ca 2+ was in the range of 0.01 to 0.54 μg/m 3 , accounting for only 0.1 to 1.9% of the mass ratio in PM 2.5 .The order of their mass ratios was Cl ->Na + >K + >Ca 2+ > Mg 2+ .Their annual average trends for each NIMS is shown in Fig. 7. Overall, PM 2.5 concentrations (i.e., the dotted line in the figure), tend to decrease with small fluctuations.SO 4 2-at NIMS-YN and NIMS-JJ, NO 3 -at NIMS-SM and NIMS-HN, NH 4 + at NIMS-YN and NIMS-JJ, and OC at NIMS-SM increased, but all the other species showed stagnant or decreasing trends at all sites.Especially, NO 3 -at NIMS-SM in Seoul increased significantly in 2018.However, SO 4 2-in 2018 decreased at all 5 sites except at NIMS-YN, where it increased by 15.4% in 2018 as compared to 2013.NO 3 -in 2018 was increased by 8.4% to 33.4% at all 5 sites except NIMS-SM as compared to 2013.Also, OC in 2018 was increa-sed by 5.0% and 0.9% at NIMS-SM and NIMS-HN, respectively.With the ratio of NO 3 -to SO 4 2-(N/S ratio), the contribution of mobile and stationary pollutants in the air can be assessed qualitatively.In this study, the average N/S ratio was 0.35 to 1.29, which was high in the Seoul metropolitan and Joongbu area, and low in Baengnyeong-do and Jeju-do.The N/S ratio in the Seoul metropolitan area continued to increase except in 2015, and it showed a maximum value of 2.0 in 2018.This means that the Seoul metropolitan area and the Joongbu area have a large effect of mobile emission sources, and it can be confirmed that the fixed pollution source has a large effect on Baengnyeongdo and Jeju Island, which has a relatively small population.The impact of major pollutants by region requires a more in-depth analysis through an receptor model analysis in the future.
Fig. 8 shows stacked plots containing annual trends for 10 chemical compositions in PM 2.5 at each NIMS.After calculating mass ratios for each species to PM 2.5 , its annual trend from 2013 to 2018 (2014 to 2018 at NIMS-YN) was plotted for each site.The mass ratios for each species in PM 2.5 must be an essential entry in PM 2.5 source inventory and then it can be directly used to identify the PM 2.5 emission sources, especially in the field of receptor modeling like source apportionment studies.
When looking over the annual concentration trends of 5 major species at each NIMS shown in Fig. 8  + , and OC showed high reductions of 23.7%, 13.5%, and 14.0%, respectively.On the other hand, only NO 3 -increased by 14.6%.Trends in terms of mass ratio at each site are presented in Table S1, which provides all statistics for 5 major species in detail.Briefly, the mass ratio at NIMS-JB showed an increasing trend in 2018 compared to 2013, except for EC.That is, the mass ratio of EC to PM 2.5 decreased slightly from 0.048 in 2013 to 0.043 in 2018.At NIMS-SM in Seoul, the mass ratio of SO 4 2-to PM 2.5 decreased from 0.214 in 2013 to 0.122 in 2018.However, NO 3 increased slightly from 0.236 to 0.249, and OC increased from 0.098 to 0.131.At NIMS-HN in an agricultural area, NO 3 -increased from 0.131 to 0.204 and NH 4 + increased slightly from 0.121 to 0.136.On the other hand, at

3 Chemical
Characteristics of 8 Ionic Species, EC/OC, and 14 Elements in PM 2.5 for Step 3 Mass ratios of each chemical species in PM 2.5 is as important as the mass concentration of each species when characterizing PM 2.5 physicochemical properties, apportioning quantitative PM 2.5 sources, and seeking proper control measures.Among the chemical species measured at 6 NIMSs, samples for a total of 24 chemical species including 8 ionic species (SO 4 2-, NO 3 Cl -, Na + , K + , Mg 2+ , and Ca 2+ ), 2 carbon components (OC, EC), and 14 other elements (Si, Ti, V, Cr, Mn, Fe, Ni, Cu, Zn, As, Se, Br, Ba, and Pb) were analyzed after proper data pretreatment, as described in §2.2.Again, it is noted that the following study is focused on 2018 because the samples for all 14 inorganic elements have been provided at all 6 NIMSs since then.
Table S2 shows a summary of average concentration in unit of μg/m 3 and mass ratio for major chemical species and the other species groups measured at each NIMS in 2018.For the inorganic elements in Table S1, the sum of averaged 14 elements was calculated by 501.5 ng/m 3 and its mass ratio to PM 2.5 (21.9 μg/m 3 ) was 2.30%.The sum of mass ratios for each element was observed in the range of 1.74% to 2.82% over 6 NIMSs in the order of NIMS-YN (2.82%)> NIMS-JB (2.81%)> NIMS-JJ (2.60%)>NIMS-BN (1.92%)>NIMS-SM (1.90%)> NIMS-HN (1.74%).The elements were generally higher in industrial/urban areas than in agricultural areas.According to Matawie et al. (2015), the mass fraction was higher in mobile sources in Raipur, India.
According to a report from the Seoul Research Institute for Health and Environment (SRIHE) (Um et al., 2020), the NO 3 -in Seoul was 24% of the PM 2.5 automobile exhaust source.In addition, SO 4 2-accounted for 14%, NH 4 + 12%, OC 16%, and EC 12%.All the other inorganic elements accounted for only 2% of PM 2.5 .Thus, they reported that all the measured species including ions, EC/OC, and elements accounted for 85% of PM 2.5 mass.As PM 2.5 mass increased, the mass ratio of ion species showed a tendency to increase in general.
Carbon components are usually classified as OC and EC.Since OC detected only carbon by our analytical method, some elements such as O, H, and S bound to carbon were excluded during component analysis.When estimating organic matter (OM), a reasonable constant is often multiplied by OC to make up for unmeasured chemicals in PM 2.5 .Um et al. (2020) used 1.8 as an appropriate constant in Seoul city, and then they estimated carbon component such that OM = 1.8 × OC, and then TC (total carbon) = OM + EC = 1.8 × OC + EC.After using the constant in 2018, a result showed that TC in Seoul accounted for 32% of PM 2.5 compared to the initial 28% (i.e., 16% of OC plus 12% of EC).Fig. 9 shows average chemical compositions in PM 2.5 at each NIMS for 24 chemical species after converting OM into 1.8 × OC from data monitored hourly in 2018.
Back to our study in Korea, we multiplied OC by a constant of 1.8 as suggested by Um et al. (2020).The average TC at all 6 sites accounted for 28.3% of PM 2.5 with a range of 23.6% at NIMS-BN to 31.4% at NIMS-JB, as shown in Table S2.Thus, the TC in a mediumsized urban area was much higher than those in a marine area and even in large urban and industrial areas.It seemed that the latter areas were better controlled than the former area in terms of combustion activities of fossil fuels.In addition, the measured fractions of PM 2.5 for 24 species were in the range of 67.1% at NIMS-BN to 89.7% at NIMS-JJ with a mean of 82.7% for all 6 sites.Both sites showing minimum and maximum values were in the same marine environment even though both sites showed much lower PM 2.5 levels than inland sites.We wondered what kinds of unidentified components accounted for 32.9% of PM 2.5 mass at NIMS-BN, far exceeding the other 5 sites' average of 14.2%.According to Almeida et al. (2006), the unmeasured fraction in PM 2.5 might be partly due to the presence of water associated with PM and errors in the estimation of chemical species.In addition, Tsai and Kuo (2005) reported that water content in PM 2.5 was higher in coastal areas than in urban areas.Further, it was higher in spring than in winter and higher at night than in the daytime.Many other earlier studies focused on water content and hygroscopic growth of PM under ambient conditions (Ueda et al., 2000;Meng et al., 1995a).Generally, when RH was high, hygroscopic components like NH 4 NO 3 in PM took up water from the atmosphere so that particles might gain mass several times their dry weight, and then PM 2.5 might contain water as an unidentified component under very humid conditions (Kajino et al., 2006).Perrino et al. (2016) insisted that water content in PM was roughly proportional to the soil content from long-range transported desert dust (about 5%), but dust did not contribute to water content.They also argued that the water content under winter stability conditions was dependent on NH 4 NO 3 and constituted up to 22% of the total PM 10 .
For reference during the study year 2018, the NIMS-BN on site had an annual average F/C ratio of 0.51, a temperature of 6.1°C, precipitation of 831.1 mm/yr, a WS of 4.3 m/s, a WD of 227.0°, and a RH of 67.9%.However, the NIMS-JJ site had a F/C ratio of 0.47, a temperature of 14.5°C, precipitation of 1,345.8mm/yr, WS of 3.1 m/s, WD of 194.4°, and RH of 68.6%.Also, Asian yellow dust events occurred 7 times in 2018 at NIMS-BN, but none occurred at NIMS-JJ (KMA, 2021).The NIMS-BN environment was colder, showed much less precipitation, and much faster WS than the other 5 sites.Thus, probable reasons for the smallest unidentified fraction at NIMS-BN were: 1) more water content due to a humid marine environment as well as frequent Asian dust occurrences, 2) colder weather condition below zero to facilitate water binding and building up more hygroscopic NH 4 NO 3 in winter, and 3) unaccounted species from China and nearby North Korea with fastest WS compared to the other sites.

CONCLUSIONS
The chemical compositions and their mass ratios in PM 2.5 are particularly important in fundamental studies on identifying and determining PM 2.5 emission sources.For this study, the PM was monitored hourly for 6 years at 6 different environmental areas (urban/agricultural/ industrial/marine areas) in Korea.PM 2.5 levels in Korea far exceeded the 2021 WHO guidelines and slightly exceeded the 2018 Korean NAAQS.When classifying PM 2.5 samples into 5 classes of air quality conditions, we observed that 45.0% of all datasets were assorted in the good class.It was found that PM 10 and PM 2.5 were negatively correlated with WS; however, PM was influenced by wind direction rather than wind speed in certain areas.Since annual PM 2.5 concentration steadily decreased nationwide, each species was also decreased.One of the reasons for decreasing PM 10 and PM 2.5 might be caused by changes in citizens' awareness of PM due to enforcement of various laws and policies in Korea.In addition, it was found that the PM 2.5 concentration in China decreased by more than 50% during the same period, which is considered to be an important external factor in the reduction of PM 2.5 in Korea.The chemical mass ratio of PM 2.5 significantly changed based on area.For example, at NIMS-SM in the metropolitan area, the mass ratio of SO 4 2-decreased from 0.214 in 2013 to 0.122 in 2018.However, NO 3 -increased slightly from 0.236 to 0.249, and OC increased from 0.098 to 0.131.At NIMS-HN in an agricultural area from 2013 to 2018, NO 3 -increased from 0.131 to 0.204, NH 4 + and OC increased slightly from 0.127 to 0.136.On the other hand, at NIMS-BN in the western marine area, SO 4 2-decreased from 0.185 to 0.153, and NO 3 -greatly increased from 0.081 to 0.137.The elements were higher in industrial/urban areas than in agricultural areas.In addition, the average TC for 6 sites was 28.3% of PM 2.5 .The TC in a mediumsized urban area was much higher than those in marine areas or even in large populated urban areas and industrial areas.It seemed to be that the latter areas were better controlled than the former area in terms of combustion activities of fossil fuels.
The chemical species in PM 2.5 such as various ions, carbons, and elements provide fundamental information to help qualitatively identify emission sources, quantitively determining the sources, suitably managing PM action plans, and reasonably establishing government policies.For the efficient control and management of air pollutants and to establish a reasonable air environment policy, it is necessary to present and implement fine dust management measures and countermeasures that match the circumstances of each local government.It is considered that the concentration trend analysis of PM 2.5 and chemical components in six regions in Korea can serve as an important basic data for each region's air pollutant emission reduction and management plan.Other ions are Cl -, Na + , K + , Mg 2+ , and Ca 2+ ; b) TC (total carbon) = OM (organic matter) + EC, where OM = 1.8 × OC; c) Element stands for Si, Ti, V, Cr, Mn, Fe, Ni, Cu, Zn, As, Se, Br, Ba, and Pb; d) The rest = Unidentified species = PM 2.5 -(SO 4 2-+ NO 3 -+ NH 4 + + Other ions + TC + Elements)

Fig. 1 .
Fig. 1.Location of six national intensive monitoring sites in Korea.

Fig. 4 .
Fig. 4. Annual trends of PM 2.5 air quality conditions.(a) (b) (c) (d) (e) (f) Fig. 5.It must be noted that the average concentration was calculated based on 166,331 samples over 6 NIMSs after performing data pretreatment steps.The figure shows histograms of stacked average concentrations for 10 chemical species including 8 ionic species (
-To monitor urban air pollutants -To identify the impact of local emission and long-range transport on local air quality -To monitor background air quality on the Southern coast of Korea -To identify the impact of long-range air pollutants transported from outside PM 10 ratio is over 1).After performing the 1 st pretreatment step for Study-1, 277,681 (91.0%) out of 305,208 raw sample datasets were left as shown in Table

Table 2 .
Numbers of sample datasets deleted and used at each NIMS after data pretreatment.

Table 3 .
A summary of statistics for 6-year averages of PM, F/C ratio, and meteorological parameters (Study-1).

Table S1 .
Overall summary of annual average concentrations (μg/m 3 ) and mass ratios for 5 major chemical species (Study-2).

Table S1 .
Overall summary of annual average concentrations (μg/m 3 ) and mass ratios for 5 major chemical species (Study-2).

Table S2 .
A summary of average concentrations (μg/m 3 ) and mass ratios for major species, Total Carbon, 5 other ions and 14 elements (Study-3). a)