Variation of PM2.5 Chemical Compositions and their Contributions to Light Extinction in Seoul

The objective of this study was to determine comprehensive chemical components in PM2.5 from March 2011 till February 2012 in Seoul, South Korea, and their contributions to light extinction. Major chemical components in the aerosol were: ammonium sulfate, 30.3%; ammonium nitrate, 25.2%; organic matter, 21.3%; crustal mass, 16.9%; element carbon, 6.1%; and trace metals, 0.2%. PM2.5 mass concentrations and light extinction were mostly correlated in their diurnal and monthly variations, which indicates that the aerosol mass is the key variable in light extinction in Seoul. However, the aerosol size and composition (of PM2.5) also played significant roles in light extinction. We applied the IMPROVE algorithm to quantify the contributions of observed chemical components to light extinction. It was found that the IMPROVE formula tended to underestimate light extinction by up to 30% in urban conditions where large sources of organic matter (OM) and element carbon (EC) existed unless some revision was made before the light extinction calculations. The IMPROVE algorithm was further optimized for the observed light extinction for OM and EC. The revised light extinction efficiencies of OM and EC in Seoul increased by about 1.5–3 times of those in the original IMPROVE algorithm. The optimized IMPROVE scheme in this study reproduced the observed light extinction more accurately in Seoul. Overall, 41% of the contribution from OM and EC to light extinction in Seoul was close to 50% of the contribution from nitrate and sulfate, although the mass of the former contribution was only half of the latter.


INTRODUCTION
Aerosol has many known adverse effects on human health, global cooling and visibility (Ghim et al., 2005;Goldsberg et al., 2006;Huang et al., 2009;Hyslop 2009;Lee et al., 2013).Among them, visibility reduction is a major visual problem in many cities.Visibility impairment has been used as a visual indicator of air quality in urban areas (Watson, 2002).High PM episodes, such as sporadic Asian Dust events and frequent photochemical smog, are widely known to have the most significant impacts on visibility in Korea.
Visibility degradation is primarily attributed to the scattering and absorption of visible light, not only by suspended particles but also by gaseous pollutants in the atmosphere (Appel et al., 1985;Sun et al., 2006;Lee et al., 2015).Many studies have reported that particles smaller than fine particulate matter (PM) with an aerodynamic diameter of less than 2.5 um (PM 2.5 ) contribute the most to visibility degradation and extinction through scattering and absorption, respectively (Jung et al., 2009;Cao et al., 2012;Lin et al., 2012;Kim, 2015).PM in the atmosphere can be largely classified into three major components: watersoluble ions, carbon and heavy metals (Chow and Watson, 2001;Artinano et al., 2003;Chen et al., 2003;Querol et al., 2008).PM 2.5 , including major water-soluble ions, organic matter and black carbon, contributes to light absorption and scattering (Watson, 2002;Ramanathan and Carmichael, 2008).Recently, several studies have determined causes and characteristics of light extinction with chemical components in urban and background aerosols (Jolleys et al., 2015;Pant et al., 2015;Petit et al., 2015;Tao et al., 2015;Visser et al., 2015;Wang et al., 2015;Zheng et al., 2015).Among them, the Interagency Monitoring of Protected Visual Environment (IMPROVE) algorithm is a suitable tool for assessing light extinction by attributing major chemical components (Pitchford et al., 2007;Tao et al., 2015;Chen et al., 2016).
The aim of this study was to determine PM 2.5 mass concentrations and their chemical and optical components and to understand their contributions to aerosol light extinction and visibility in Seoul.

Sampling Sites
The Korea Ministry of Environment (KME) has maintained a comprehensive and intensive urban monitoring site at the Bulkwang district in Seoul since 2008, which is currently operated by the National Institute of Environment Research (NIER).It is located on the rooftop of a building (approximately 67 m above the ground, 37°61'N, 126°93'E) in the northwest part of Seoul, Korea.The western half of the surrounding site can be categorized as a typical urban area with mainly residential buildings and roads.However, the eastern sector is mostly encompassed by the foothills of Mt.Bukhansan (Fig. 1).

Chemical Analysis
This study utilized a year of hourly chemical compositions of PM 10 , PM 2.5 , aerosol optical properties, major gaseous species and visibility at the Bulkwang supersite from March 2011 till February 2012.Mass concentrations of PM 10 and PM 2.5 were measured by two independent β-ray absorption instruments, FH62C-14 (Thermo Instruments Inc.) and BAM1020 (MetOne Instruments Inc.), respectively.Chemical compositions of PM 2.5 were determined using an Ambient Ion Monitor (URG 9000D, URG Co.) for ionic species, a semi-continuous carbon aerosol analyzer (SOCEC, Sunset Laboratory Inc.) for carbonaceous components, and an online X-ray fluorescence (Xact 620, Cooper Co.) for trace elements.
The Ambient Ion Monitor was used to analyze watersoluble inorganic ions in aerosol and gaseous species.Ambient air with aerosol and gaseous species flowed into the diffusion denuder via a PM 2.5 inlet at a flow rate of 3 L min -1 .To collect aerosol selectively, gaseous species such as NH 3 , HCl and HNO 3 were separated by a diffusion denuder coated with 30% hydrogen peroxide and analyzed their ions by ion chromatography (ICS-2000, Dionex Co.).The separated aerosol samples were later collected in the supersaturated humidity chamber and sampling syringes and subjected to an additional ion chromatography set (Park et al., 2012).We utilized ionic mass balance in each sample to ensure the required data quality under quality control procedures recommended by EANET (Acid Deposition Monitoring Network in East Asia) (Wang et al., 2007).
Carbonaceous aerosol was analyzed using an SOCEC analyzer on an hourly basis with modified NIOSH method 5040.Samples were collected on 47 mm quartz filters (TISSUQUARTZ, Gelman Sciences) at a flow rate of 8 L for 45 minutes and placed in an oxidation oven after removing gaseous organic carbon by organic denuders.The oven was first purged with helium and heated to 840°C using a stepped temperature ramp scheme.In the presence of helium, organic compounds and pyrolysis products were desorbed from the filter, transferred and oxidized to CO 2 in an MnO 2 oven.Oxidized CO 2 gas from the organic carbon was then measured with a nondispersive infrared (NDIR) detector system.In the second temperature ramp, with an oxidizing gas stream in the oxidizing oven, any elemental carbon was oxidized off the filter.CO 2 from the elemental carbon was then detected in the same manner as the organic carbon, as described previously (Shin et al., 2016).
Elements, namely, Si, K, Ca, Ti, V, Cr, Mn, Fe, Co, Ni, Zn, As, Se, Br, Cd, Ba and Pb, were analyzed using online XRF on an hourly basis.Air driven subsequently with 2.5 µm size cut cyclone was collected on Teflon filter tape for 45 minutes at the flow rate of 16.7 L min -1 .After sampling, high-energy X-ray was emitted to the filtered samples, and the excited energy on a specific wavelength band was then identified and quantified based on the EPA Method IO 3.3.Meteorological data were obtained from the Bulkwang weather station near the monitoring site, which is operated by KOREA Environment Corporation (K-ECO).

Optical Analysis
Aerosol light scattering and absorption were measured by using a three-wavelength integrating nephelometer (Model 3563, TSI Inc.) and a seven-wavelength aethalometer (AE31, Magee Sci.), respectively.The nephelometer was used to measure the light scattering coefficient at wavelengths of 450, 550 and 700 nm with a one-minute temporal resolution of 20 L min -1 flowrate.The light absorption was measured at seven-wavelengths (370, 470, 520, 590, 660, Fig. 1.Location of a monitoring site for this study in Seoul, Korea.Park et al., Aerosol and Air Quality Research, 18: 2220-2229, 20182222 880 and 950 nm) with quartz filters sampled every hour with a flow rate of 5 L min -1 .

Aerosol Mass Concentrations
The annual mean concentrations of PM 10 and PM 2.5 at the Bulkwang site in Seoul were 44.5 and 25.7 µg m -3 , respectively.The seasonal PM 10 and PM 2.5 were 49.6 and  32.8 µg m -3 in winter, 63.7 and 28.9 µg m -3 in spring, 31.6  and 19.2 µg m -3 in summer and 33.6 and 21.9 µg m -3 in autumn, respectively (Table 1).Higher concentrations of PM 10 and PM 2.5 in winter and spring were attributed to the seasonal increase in primary and secondary aerosol emissions in both local and transported sources (Hagler et al., 2006;Park et al., 2012;Lee et al., 2013;Kim et al., 2014).The ratio of PM 2.5 /PM 10 also showed a distinct seasonal variation: spring (0.51) < summer (0.56) < autumn (0.65) < winter (0.66).The increase of the fine size contribution in winter was often observed in Asia due to an increase in precursor emissions and very dry seasonal weather (Yang et al., 2005).Large-size springtime aerosol, often impacted by Asian Dust, was the major cause of the lowest PM 2.5 /PM 10 in spring.Although PM 2.5 /PM 10 was minimal in the spring and summer, its ratio elevated rapidly to 0.63 in the late spring and early summer, prior to the summer monsoon.

Chemical Composition of PM 2.5
For PM 2.5 , ionic species accounted for the most (57%) dominant component in aerosol, with 22% of SO 4 2-(5.58µg m -3 ), 19% of NO 3 -(4.83µg m -3 ) and 16% of NH 4 + (4.10 µg m -3 ).The ion component was the highest in spring.In another study conducted in Seoul, major ionic concentrations were also the highest in spring (Lee et al., 2009).Ionic species in aerosol is mainly of secondary origin, and its conversion rate from its precursor is highest in summer.However, frequent rainfall and the preferred partitioning of particulate nitrate to the gaseous phase in high-temperature conditions prevented secondary aerosol accumulation in summer.
Annual mean concentrations of OC and EC in PM 2.5 were 3.72 µg m -3 and 1.55 µg m -3 , respectively.As OC is mainly produced by photochemical reactions (Kim et al., 2007), OC showed a seasonal trend similar to that of ionic species, with the highest value in spring before the summer monsoon.In the case of EC, steady emissions from the industry and transportation sectors, and its non-hygroscopic property, which was less affected by precipitation, maintained relatively constant levels of concentrations throughout the year.
The ratio between OC and EC provides a basic criterion for determining the degree of secondary aerosol contribution when the OC/EC ratio is larger than 2 (Turpin and Huntzicker, 1995).The annual average ratio of OC/EC was 2.82, and the highest ratio was found in spring (3.86), followed by summer (3.16).The seasonal variation in OC/EC was well matched with those of water soluble ionic species, such as sulfate.However, the correlation between OC and EC was lowest in spring and summer.This might be due to the influence from more diverse local primary and transported sources in spring and summer, including biomass burning and other emissions of biogenic origin (Zhang et al., 2014).
The concentrations of the metal species in PM 2.5 were in the following order: Si > K > Fe > Zn > Ca > Pb > Br > Mn > Ti > Ba > As > V > Cd > Ni > Cr > Se > Co.While crustal and natural components in PM 2.5 showed a spring maximum (A in Fig. 2(a)), trace metals were highest in the winter (B in Fig. 2(b)) due to anthropogenic origins, mainly coal combustion in local and transported sources (Han et al., 2004).
The major components of aerosols can be determined from chemical speciation measurements.The following equations were used to reconstruct major mass components in aerosol (El-Zanan et al., 2005;Frank, 2006;Lowenthal and Kumar, 2006).Fig. 3 shows monthly variations in reconstructed components of PM 2.5 .Overall, AMSUL was the single largest component in aerosol (30.3%) and was followed by AMNIT (25.2%),OM (21.3%),CM a , (16.9%), EC (6.1%) and TM (0.2%).As we discussed in the previous section, sulfate aerosol was dominant in summer, but nitrate was dominant in winter.The calculated OM/OC in Seoul was in a rather low range, being 1.46, compared to other studies indicating values from 1.3 to 2.5 (El-Zanan et al., 2005;Cheung et al., 2011;Yang et al., 2012), which reflected the dominant primary organic aerosol formation in typical urban conditions.

Particle Light Extinction
Annual means of the hourly light scattering (b sp ) and absorption coefficient (b ap ) by aerosol were 132.2 Mm -1 and 56.2 Mm -1 , respectively.Fig. 4 depicts the diurnal and monthly variations in particle light extinction b p (= b sp + b ap ), which showed strong correlations during the morning traffic hours and winter season when aerosol concentrations peaked.Interestingly, the monthly aerosol light extinction showed another sharp increase in the early summer while PM 2.5 decreased.We didn't analyze the aerosol size fraction in this study.However, the fraction of PM 2.5 in PM 10 increased sharply in the early summer (Fig. 4).This early summer increase in aerosol light extinction was likely attributable to an increase of the small size fraction in aerosol, though their compositions didn't change that much (Fig. 3).Using the IMPROVE algorithm, known as a reference scheme for calculating the aerosol light extinction with the chemical species below (Pitchford et al., 2007), we assessed the relationship between particle light extinction (b p ) and observed chemistries.
The light scattering of sulfate (assumed to be neutralized as AMSUL), nitrate (assumed to be neutralized as AMNIT), and organic mass were divided into two size modes according to IMPROVE's size fraction scheme.The apportionment of the total sulfate concentration into smalland large-size fractions was estimated using the following equations: The same equations were used to calculate the nitrate and organic size fractions.If the total concentration of a compound was less than 20 µg m -3 , the fraction in the large mode was estimated by dividing the square of the total concentration of the component by 20 µg m -3 (Chen et al., 2016).The relative humidity function f(RH) was also implemented to adjust the hygroscopic fraction of sulfate, nitrate and sea salt in different humidity conditions.We calculated b p with hourly aerosol chemical speciation and compared it with the observed hourly light extinction  using the IMPROVE scheme (Fig. 5(a)).The measured b p was correlated reasonably well with the calculated one.However, b p calculated by the IMPROVE algorithm was likely to underestimate the aerosol light extinction by up to 30%, with an obtained slope of 0.69.As stated before, an average of 1.46 for the OM/OC ratio in Seoul indicated the relatively fresh sources of OM (Turpin and Lim, 2001;Rees et al., 2004).This might indicate that the IMPROVE scheme, which was originally modeled for the regional haze with relatively aged air masses, might not be adequate for simulating light extinction in Seoul, where relatively large sources of OM and EC are present and light extinction is more likely influenced by them.In order to reduce the systematic bias in calculating light extinct, we implemented a linear regression for the observed b p along with a dependent variable to optimize the coefficients of OM and OC variables in the IMPROVE formula.The newly optimized coefficients obtained from the linear regression model are listed in Table 2.There were significant increases for the OM coefficients in the revised equation, especially for small size OM and EC.Fig. 5(b) shows the relationship between the observed values and those calculated with revised coefficients in the IMPROVE scheme.Although the correlations (or precision) did not change much, the slope (0.88) in the revised version was clearly improved from 0.69 in the original one.This implies that accuracy in determining aerosol light extinction could be improved by 28% when using this optimization procedure.Monthly variations in relative light extinction values by aerosol species with the revised IMPROVE scheme are shown in Fig. 6.AMSUL contributions to light extinctions were highest in summer, while the maximum contributions from AMNIT, OM and EC occurred in winter, in accordance with their monthly mass concentration variations in Fig. 3.The annual mean contributions of aerosol chemical components to light extinction are listed for the original coefficients in the IMPROVE scheme and for the revised ones in Table 3.Generally, our results using the original IMPROVE scheme agreed more appropriately with other studies, particularly those of Wang (2003).However, it should be mentioned that such a comparison might turn out differently if other studies utilized similar optimizations for chemical component coefficients, especially for OM and EC in the IMPROVE algorithm.Thus, care should be taken in the application of the IMPROVE algorithm to estimate light extinction in high OM and EC urban conditions, and an adjustment of their coefficients may be required to reduce the bias in light extinction calculations.
While the observed aerosol light extinction was closely related to the PM 2.5 mass concentration, the composition and size of the latter also played significant roles.In order to assess the aerosol light extinction by observed chemical species, the IMPROVE algorithm was used, and the results were compared with the observed aerosol light extinction.The IMPROVE algorithm with observed aerosol chemical concentrations reproduced the observed aerosol light extinction reasonably well.However, the IMPROVE algorithm underestimated light extinction by up to 30%.The optimization of the OM and EC coefficients in the original IMPROVE formula was made by a linear regression method.We found significant increases in OM and EC coefficients, especially for a small amount of OM.The values for light extinction determined by measured components in this study and by the original IMPROVE algorithm fell within the range of those found in urban studies, especially in Korea and China.However, we conclude that the coefficients in the original IMPROVE formula can be further revised to estimate the light extinction more accurately in urban conditions, where large sources of OM and EC exist, using a simple optimization process.Kim, 2009, d Tao et al., 2014, e Wang et al., 2015, f Deng et al., 2016.

AmmoniumFig. 2 .
Fig. 2. Monthly variations for metallic species.Shaded part indicates Asian dust (A) for (a) crustal species and anthropogenic influenced periods (B) for (b) other species.

Fig. 4 .
Fig. 4. Diurnal variations of observed (a) b p and PM 2.5 , and monthly variations of observed (b) b p , PM 2.5 and fraction of PM 2.5 in PM 10 .

Fig. 5 .
Fig. 5. Relationship between (a) measured b p and calculated b p with IMPROVE equation, and (b) as same with revised IMPROVE equation for OM and EC in this study.

Fig. 6 .
Fig. 6.Monthly contributions of chemical components to aerosol light extinction with revised IMPROVE coefficients.

Table 1 .
Seasonal and annual chemical concentrations in PM 2.5 , light scattering (b sp ) and absorption (b ap ) with means and standard deviations.

Table 2 .
Original coefficients of OM and EC in IMPROVE formula and their optimized values in Seoul.

Table 3 .
Statistical summary of OC and EC concentrations in PM 2.5 and their contributions to aerosol light extinction in urban sites of Korea and China.