Optical Properties of Size-Resolved Aerosol Chemistry and Visibility Variation Observed in the Urban Site of Seoul , Korea

Long-term visibility observations, including that of aerosol chemistry, is necessary to improve the visibility of the mega-city of Seoul. In this study, the contributions of size-resolved aerosols to light extinction were estimated during an extensive visibility monitoring period, from 2007 to 2009. Optical measurements of a light extinction coefficient, a light scattering coefficient, and a light absorption coefficient were made using a transmissometer, nephelometer, and aethalometer. Size-resolved aerosol measurements, including the collection of submicron (Dp < 1.0 μm), fine (Dp < 2.5 μm), and coarse (2.5 < Dp < 10 μm) particles were conducted for the determination of their mass extinction coefficients and contributions of chemical components to light extinction. A total of 386 measurement data sets were used to construct the predictive mass extinction coefficients for the size-resolved particles using regression analysis. The mass extinction coefficients of the sized-resolved aerosols of PM1.0, PM2.5, and PM10 were found to be 8.7 ± 0.8, 4.7 ± 0.2, and 2.7 ± 0.2 m/g, respectively. The aerosol light extinction varied with aerosol size distribution, the chemical composition of the aerosol, and ambient relative humidity. It was found that the ammonium sulfates were the largest contributor to light extinction and visibility impairment due to aged aerosols in the urban atmosphere of Seoul.


INTRODUCTION
Over the past decades, the tendency of air pollution condition of the major cities of Korea became changed with the fuel supply system.The air pollution level for the primary pollutants such as concentrations of sulfur dioxide (SO 2 ), the carbon monoxide (CO), and the lead (Pb) decreased continuously with strong and sustained implementation of environmental considerations into energy policy such as supply expansion of low sulfur oil fuel and LNG (liquefied natural gas), and attachment obligation of a three-way catalytic converter to vehicles etc., which was attainable for most of the cities to the national air quality standards.However the secondary pollutants such as concentrations of particulate matters below 2.5 µm (PM 2.5 ) and ozone (O 3 ) are mostly generated by the physical and chemical reaction of their precursors in the ambient air rather than directly emitted from combustion processes of fossil fuels (Steinberger and Balmor, 1973;Pun and Seigneur, 1999).In particular, the particulate matters can cause light attenuation in the atmosphere, then they can lead to haze (Waggoner and Weiss, 1980;Malm et al., 1994).
Visibility varies with aerosol size distribution, the chemical composition of aerosol, and ambient relative humidity (Chung et al., 1999;Hand et al., 2002;Yuan et al., 2006;Kim, 2007;Xu et al., 2012).Visibility impairing aerosol can be affected by local-scale air quality control policy as well as regional or global-scale transport of air pollutants (Hoffer et al., 1967;Haywood et al., 2003;Chow et al., 2004;Zhang et al., 2012).Improvement of visibility is, therefore, not an easy task for our governments.Long-term visibility observation including aerosol chemistry is essential to accomplish the task.In this study, the amount of light attenuation by size-resolved aerosols and their contribution to light extinction were estimated during the extensive visibility monitoring periods from 2007 to 2009.The effective and desirable management policies provide people with a substantial improvement of atmospheric environment.Visibility management is one of them and makes it easy for the public to understand impact of air quality improvement.The goal of this atmospheric visibility monitoring is to provide optical properties of sizeresolved aerosol and to estimate light extinction coefficient by their chemical composition.

Visibility Measurement and Aerosol Monitoring
The extensive visibility monitoring was performed to  , 2005).The transmitter was moved at the calibration site during the calibration period then it was installed 852 m away from the receiver site (59 m above ground level) across the downtown.Calibration was performed during a clear daytime to avoid the thermal effect by measuring raw reading values in lamp intensity and the path length, and by calculating the atmospheric transmittance.The calibration path length between the transmitter and receiver unit was 267 m.The calibration number produced from this procedure was used in this study.The nephelometer measured the light scattering coefficient of a continuously sampled known volume of ambient air.b scat was recorded at the wavelength of 550 nm without heating device and size-selective inlet.Thus b scat inferred light scattering by ambient particles.The nephelometer calibration was performed in order to correct the zero count and the span count value of the instrument using clean air and CFC-12 gas, respectively.Scattering range and accuracy of the NGN-2 were reported to be 10 to 7000 Mm -1 and 10% of true value for air near Rayleigh, respectively (OPTEC, 1993).Light extinction and scattering measurement were performed approximately every 1 minute interval during the intensive monitoring periods.b abs was inferred from an external black carbon (BC) concentration data with a PM 10 inlet from the aethalometer.The aerosol is collected on an area of a quartz fiber filter at 4 liters per minute.The aethalometer measured the optical attenuation of the aerosol deposited on the filter by detecting the intensity of light transmitted through the spot on the filter.In this study absorption efficiency of 10 m 2 /g was used to calculate the atmospheric light absorption coefficient from BC (Parungo et al., 1994).Uncertainty in the aethalometer measurement was reported to be ± 50 ng/m 3 (Magee Scientific, 1996).The calibration procedure 'Optical Test Strip Procedure' was used to perform systematically at routine intervals in the aethalometer (Hansen et al., 1982).
In order to estimate the light extinction coefficient using mass concentrations of size-resolved aerosols, aerosol samples were collected during the intensive monitoring periods.Size-resolved aerosol measurements including collection of submicron (D p < 1.0 µm; PM 1.0 ), fine (D p < 2.5 µm; PM 2.5 ), and coarse (2.5 < D p < 10 µm) particles were conducted for the determination of the gravimetric mass and the elemental, ionic, and carbonaceous concentrations using model URG-2000-30EHB two PM 1.0 cyclones (URG, Chapel Hill, USA) and a URG (University Research Glassware) model VAPS 2000J, respectively.Filter-based aerosol collections carried out from 8 A.M. to 6 P.M. with 2-hour sampling interval and from 6 P.M. to the next day 8 A.M. with 14-hour sampling interval to investigate the optical properties of aerosols and their impact on visibility reduction.The sampling methods and the analytic parameters for aerosol collection are summarized in Table 1.

Analytic Methods
A 47-mm-diameter, 0.4-µm pore-sized polycarbonate filter (Nuclepore) was used to measure the gravimetric mass and the elemental concentrations of Al, Si, P, S, K, Ca, Ti, Cr, Mn, Fe, Cu, Zn, Se, Cd, and Pb using the PIXE method (Cohen et al., 1996).The PIXE spectrums of elements for aerosol samples were obtained using a 1-mm-diameter, 4-to-5-nA beam, 3.0-MeV proton beam from a 3-MV Tandetron accelerator at the Electrostatic Accelerator Research Center, National Center for Inter-university Research Facility, Seoul National University.The proton beam was extracted into air through a 7-µm-thick Kapton window, and the samples were placed 3 cm downstream from the window.The Xrays were detected using a Si (Li) detector.The spectral peak area for each element was extracted using the analysis program WinQXAS (IAEA) (Kim et al., 2008).Because the target thickness of the aerosol for each Nuclepore filter sample varied, the peak area was converted to relative concentrations by comparing it to a NIST (National Institute of Standards and Technology) standard reference material SRM1648 (National Institute of Standards & Technology, 1991).It was assumed that the aerosols were uniformly deposited on the filter.The elemental concentrations were, then, obtained from an average of duplicate determination.The mass concentrations of ammonium nitrate aerosol, volatile nitric acid, and ammonia gas were determined using the filter-followed-denuder systems.The two denuders in series coated with sodium carbonate and citric acid, respectively.The samples were analyzed for the mass concentrations of the major light attenuating ions (Na + , NH 4 + , NO 3 -, SO 4 2-, Cl -, etc.) using a GAT (gamma analysen technik GmbH) model DKK-TOA IA-300 ion chromatography (GAT, Bremerhaven, Germany) after being extracted with 10 ml of distilled water.Each sampled filter was first put into a 20 mL vial, wetted in 1 mL of HPLC grade methanol, and then mechanically extracted for 30 minutes with 9 mL distilled deionized water.Two denuders were installed upstream of the polycarbonate filter to eliminate over-estimated effect by acidic gas (H 2 SO 4 and HNO 3 , etc.).242 mm and 150 mm annular denuders were prepared with zero gas generated by a drying train consisting of silica gels, potassium permanganates, activated carbons, and citric acids.Once sampled, denuders were extracted with  (Kim, 2007).Carbonaceous compounds of exposed quartz fiber filters were analyzed to quantify the mass concentrations of elemental and organic carbon (EC & OC) in the submicron regime with the PM 1.0 cyclone.And they were measured semi-continuously with a model-4 semi-continuous OC-EC field analyzer (Sunset Laboratory, Oregon, USA) in the fine and coarse regime by a thermal optical transmittance (TOT) method.The sample is heated in four increasing temperature steps to quantify all organic carbon on the filter.As the organic compounds are vaporized, they are immediately oxidized to carbon dioxide in an oxidizer oven.After the carbon dioxide is reduced to methane, the methane is detected by a flame ionization detector.Elemental carbon is oxidized to carbon dioxide when the sample oven temperature is stepped up to 850°C (Birch and Cary, 1996).In order to estimate light absorption by nitrogen dioxide (NO 2 ), its concentration was obtained from the hourly averaged data measured at Mapo and Joongu sites by the Korean Ministry of Environment.Air temperature and relative humidity (RH) were measured using a model H21 relative humidity data logger (Onset, Massachusetts, USA).In addition, the RH data measured by the Korea Meteorological Administration were used for meteorological analysis.A self-contained automatic camera for the visibility monitoring was installed and operated at the top of the Seoul Tower.Scenic images of diurnal visibility variation were taken every ten minutes using a model EOS-10D digital camera and a model TC-80N3 timer remote controller (Cannon, Tokyo, Japan) at the perceived visibility monitoring site.They were used to estimate the optical properties of the atmosphere under certain conditions.

Light Attenuation by Size-Resolved Aerosols and Optical Closure
The average light extinction coefficient, light scattering coefficient, and light absorption coefficient were measured to be 356 ± 284, 268 ± 228, and 32 ± 22 Mm -1 during the intensive visibility monitoring periods (IVP) and the hourly averaged b ext , b scat , and b abs varied from 30 to 1558 Mm -1 , from 21 to 1267 Mm -1 , and from 4 to 118 Mm -1 , respectively.The estimated visual range based on Koschmieder' formula ranged from about 2.5 km to about 130 km and the average visual range was approximately 10.5 ± 9.7 km during the IVP.According to Kim et al. (2006), it was reported that the average b ext and visual range observed to be 501 ± 299 Mm -1 and 7.8 ± 6.7 km at Seoul from 2002 to 2004, respectively.The average mass concentrations of PM 2.5 , PM 10 (D p < 10 µm), and relative humidity (RH) were reported to be 35.3 ± 16.8, 77.6 ± 43.5 µg/m 3 , and 61.4 ± 19.1% at that moment, respectively.During the intensive visibility monitoring periods, visibility of Seoul was found to be improved in comparison to the results from 2002 to 2004.However the average mass concentration of fine particulate matters slightly increased to 38.0 ± 27.3 µg/m 3 and did not attain the USEPA PM 2.5 standard in this study.USEPA retained the annual PM 2.5 National Ambient Air Quality Standards at 15 µg/m 3 and lowered the 24-hour PM 2.5 standard to 35 µg/m 3 .According to Ministry of Environment of Korea, the annual and the 24-hour PM 2.5 National Ambient Air Quality Standards will be retained at 25 and 50 µg/m 3 after 1 January 2015, respectively.Furthermore, the average mass concentration of submicron aerosol (20.1 ± 13.9 µg/m 3 ) was relatively high.The average mass concentrations of PM 1.0 , PM 2.5 , and PM 10 were measured to be the highest values of 35.2 ± 22.6, 63.6 ± 38.1, and 116.7 ± 70.7 µg/m 3 for the 6th IVP, as shown in Fig. 2. Meanwhile, the average relative humidity decreased to 59.1 ± 18.5% comparing with that of 2006 (61.4 ± 19.1%) in this study.The water loss by fine particulate matters in the atmosphere may reduce the light scattering coefficient (Horvath, 1996;Tang, 1996).But there are diverse factors such as chemical composition of aerosol and meteorological condition to affect the light attenuation.Light extinction depends on the mass extinction efficiency of the chemical species as well as relative humidity.Visibility variation in Seoul was relevant with lower relative humidity and variation of chemical speciation.Visibility in Seoul was, consequently, still not good.
During the IVP, measured light attenuation coefficients represent optical closure between measurements of light extinction, light scattering, and light absorption coefficients as shown in Fig. 3.The three light attenuation coefficients of b ext , b scat , and b abs were continuously measured and they were found indicating good agreement among the measurements from optical instruments.The light absorption coefficient by NO 2 (b NO2 ) was obtained from 0.33 times the hourly averaged NO 2 concentration (ppb) and then it was included in the scatter plot.The slope of the regression and the coefficient of the determination (R 2 ) for the IVPs ranged from 0.69 to 0.93 and from 0.86 to 0.99, respectively.And the average values of the slope and R 2 for the entire IVP are 0.81 and 0.96, respectively, as summarized in Table 2.They provide a measure of how well aerosol optical properties are likely to be predicted for visibility modeling by this study.A scatter plot between b ext and the sum of b scat and b abs is shown in Fig. 3.The average relative error (|b ext -(b scat + b abs )|/b ext ) was calculated to be 19 ± 14% from the scatter plot.The relative errors for the 2nd, the 3rd, and the 5th IVP show relatively higher values of 25.4 ± 21.0%, 27.2 ± 14.6%, and 23.5 ± 10.8%, respectively.They can be produced from heterogeneous aerosol distribution between the open path measurement of b ext and point measurements of b scat and b abs .In this study we will focus on the estimation of the light extinction coefficient from the size-resolved aerosol mass and the RH.A revised equation for estimating the light extinction coefficient was suggested using various visibility impairing aerosol components a few years ago (Pitchford et al., 2007).The equation provided more detailed information about aerosol optical properties.It is necessary to approximately estimate light extinction for visual air quality using the size-resolved aerosol mass and the RH which can be measured continuously in the ambient air.The uncertainty of the light extinction estimation can be indicated from the optical closure analysis.The average light scattering coefficient of 268 ± 228 Mm -1 accounted  (b NO2 ) and coarse dust particles (Kim, 2007).However, the NGN-2 nephelometer used to measure b scat in this study contains b Ray .And it was known that existing integrating nephelometers were not ideal and typically integrate over scattering angles ranging from 7° to 170° (Abu-Rahmah et al., 2006).The underestimation of the sum of b scat and b abs might be caused by absence of light attenuation by b NO2 and coarse dust particles as well as the forward scattering truncation in this study.In addition, the discrepancies due to the differences in the area and the point measurement techniques between them cannot be ruled out.Considering b NO2 , the reconstructed errors for the IVPs range from 0.05 ± 0.02 to 0.22 ± 0.11 and the average value of them shows 0.12 ± 0.06 as summarized in Table 2.

Mass Extinction Coefficients for Size-Resolved Aerosols
The aerosol light scattering coefficient is influenced by the relative humidity in the ambient air.Wet particles attenuate more light than their dry equivalents.But they scatter more light in the appropriate particles size, which has a direct influence on visibility impairment (Sloane, 1984;Xue et al., 2011).Fig. 4 shows the scatter plot between the mass concentrations of PM 1.0 , PM 2.5 , and PM 10 and the light extinction coefficient divided by f(RH), which is a function of the growth factor.In this analysis, light extinction by the growth of hygroscopic particles was calculated using the RH-dependent function proposed by IMPROVE (1993).The light extinction coefficient of hygroscopic aerosol composes a RH-dependent scaling factor (f(RH)) which represents the relationship between RH and the scattering efficiency (Kim, 2007).The first 2-parameter fit for growth factors was discussed by Hänel (1976) as Eq.(1).
Light scattering depends on values of a and γ in the equation.And they can vary with aerosol type.In this study, they were 0.7 and 0.7 obtained for clean and mostly dry air in the national parks in U.S.A., respectively.A certain degree of error can be estimated since the light attenuation coefficients were measured at further polluted atmospheric condition in this study.The light attenuation coefficients affected by precipitation including rainfalls and snowfalls were excluded in this study.The investigation was based on extensive data base for each of the IVP.A total of 386 data sets except invalid ones were used to construct the predictive mass extinction coefficients for the size-resolved particles.Each sampling frequency for the IVP is summarized as shown in Fig. 2. The general regression model used in this study is defined by Eq. ( 2).
where the β i are the regression coefficients of independent variables, which mean the mass extinction coefficients of PM 1.0 , PM 2.5 , and PM 10 assuming the all particles are hygroscopic.They are defined as the ratio of the extinction coefficient to the aerosol mass concentration.And they are the significant factors for estimating the visibility (Kim, 2011).As shown in the all eight scatter plots in Fig. 4, the slope of the regression line of PM 2.5 fitted through 0 ranges from 4.4 to 5.0.And the slopes of PM 1.0 and PM 10 range from 8.2 to 10.5 and from 2.4 to 3.0, respectively.In general, the mass extinction coefficient of PM 2.5 was further investigated more than those of PM 1.0 or PM 10 .The useful estimation for visibility could be obtained from the fine mass rather than the coarse mass.Ozkaynak et al. (1985) found that the best fit between the fine mass and the extinction coefficient occurred from 3.3 to 6.1 m 2 /g and 4.8 m 2 /g on average.The mass extinction coefficient of PM 2.5 in Santiago was given the representative value of 5 m 2 /g (Trier and Horvath, 1993).And it was reported that the linear regressions yielded the fine mass extinction coefficients of 4.7 ± 0.4 m 2 /g in U.S.A. (Malm et al., 1996) and of 4.93 ± 0.69 m 2 /g in Austria (Trier et al., 1997).However, the mass extinction coefficients of PM 2.5 were reported to be more broad range of 4.7-8.6 m 2 /g in Seoul, Korea.The average mass extinction coefficient of PM 2.5 (β 2.5 ) was found to be 4.7 ± 0.2 m 2 /g during the entire IVP and they correlated well with R 2 = 0.78.This was similar to the results from the various urban observation reported in the literature, but it was relatively lower than the other previous results observed in Korea.The linear regression of the submicron particles to light extinction produced the submicron mass extinction coefficient (β 1.0 ) of 8.7 ± 0.8 m 2 /g and a good correspondence (R 2 = 0.80) was found in this study.The average β 10 was calculated to be 2.7 ± 0.2 m 2 /g.Similar values of β 10 were reported to be 2.6 ± 0.8 m 2 /g for a rural environment in Hyytiälä, Finland (Virkkula et al., 2011) and 2.5 ± 1.1 m 2 /g for an urban environment in Beijing, China (Jung et al., 2009).Exactly the mass extinction coefficient of wet particles was considered as a conversion factor to estimate the light extinction coefficient by their mass concentration in this study.Therefore these results have considerable uncertainties in   the smaller aerosol size regime due to overestimation of light extinction comparing with mass extinction efficiency under dry conditions.There are different methods to estimate the mass extinction efficiencies (Hand and Malm, 2007).The mass extinction coefficients of β 1.0 , β 2.5 , and β 10 obtained from this study may lead to differences from the mass extinction efficiencies reported in the literature.Nevertheless, the size-resolved mass extinction coefficients can provide the better information about the estimation of the aerosol light extinction in Korea.Various mass extinction coefficients for suspended particulate matters are summarized in Table 3.
In fact, aerosol is mixed externally or internally with hygroscopic and non-hygroscopic chemical species (Adachi et al., 2011).There are sulfates, nitrates, some of organics and sea salt aerosols in hygroscopic species.In this study, The RH-dependent scaling factor was used to estimate the scattering efficiencies of the major hygroscopic species of sulfates and nitrates except organics and sea salt aerosol.Therefore, the mass fractions (f i ) are defined as the ratios of the sum of the mass concentrations of sulfates and nitrates out of the size-resolved aerosol mass.The mass fractions were used to calculate the reconstructed mass extinction coefficients of PM 1.0 , PM 2.5 , and PM 10 using Eq.(3) as follow.
where RM i is the reconstructed particle mass concentrations for PM 1.0 , PM 2.5 , and PM 10 and PM i is the mass concentrations of them.Fig. 5 shows the scatter plot between the RM 1.0 , RM 2.5 , and RM 10 and the light extinction coefficient divided by the f(RH) during the entire IVP.The light extinction coefficients had a good correlation with the RM 1.0 (R 2 = 0.82), RM 2.5 (R 2 = 0.82), and RM 10 (R 2 = 0.77).Considering the mass fraction of the hygroscopic aerosols to light scattering, the slopes of the regression lines of RM 1.0 , RM 2.5 , and RM 10 fitted through 0 range from 10.3 to 15.0, from 5.5 to 8.2, and from 3.4 to 4.7, respectively, as shown in Fig. 5.In Table 4, the sum of sulfates and nitrates contributes approximately 27% to the average mass concentration of PM 1.0 .And they account for 29% and 22% of the average PM 2.5 and PM 10 mass during the entire IVP, respectively.The reconstructed β 1.0 , β 2.5 , and β 10 vary with the mass fractions of f 2.5 , and f 10 as shown in Fig. 5 and summarized in Table 4.They were relatively higher during the 3rd IVP and relatively lower during the 8th IVP.The reconstructed mass extinction coefficients of PM 1.0 , PM 2.5 , and PM 10 considering the mass fraction of the hygroscopic aerosols are higher than their mass extinction coefficients.Consequently, the linear regressions for RM 1.0 , RM 2.5 , and RM 10 yielded the average reconstructed β 1.0 (12.8 ± 1.7 m 2 /g), β 2.5 (6.7 ± 0.9 m 2 /g), and β 10 (4.0 ± 0.5 m 2 /g), respectively, when the average mass fractions of f 1.0 , f 2.5 , and f 10 were 0.27 ± 0.16, 0.29 ± 0.13, and 0.22 ± 0.09, respectively.The reconstructed β 1.0 , β 2.5 , and β 10 were found to vary substantially with aerosol hygroscopicity as expected.However, the f(RH) depends on chemical composition, and the dry and wet particle size distributions.As illustrated by Pitchford et al. (2007), f(RH) for ammonium sulfate and nitrate varies significantly as a function of the particle size distribution.Light extinction by the growth of hygroscopic particles was recalculated using the ratio of wet to dry scattering proposed by Pitchford et al. (2007) because the f(RH) in IMPROVE ( 1993) is specific to the PM 2.5 size fraction and its chemical composition at IMPROVE sites in the U.S. Pitchford et al. (2007) suggested the ratios of wet to dry scattering for the large particles (f L (RH)) and the small particles (f S (RH)) in the fine regime.Therefore, the measured PM 1.0 was assumed to be the small particles in this study and light extinction by the growth of hygroscopic particles was calculated using the f S (RH).The small particles were calculated by applying Mie theory to lognormal size distributions with specified geometric mean diameters and standard deviations (Pitchford et al., 2007).The mass extinction coefficient of PM 1.0 (β′ 1.0 ) considering the f S (RH) for total particles (b ext /f S (RH) versus PM 1.0 ) and the reconstructed β′ 1.0 considering the f S (RH) for sulfates and nitrates particles (b ext /f S (RH) versus RM 1.0 ) need to be estimated with a certain uncertainty.β′ 1.0 varied from 6.5 to 2.5 ± 1.1 f 3.4 ± 1.2 f < 40 f Urban (Seoul, Korea) 3.5(2.4-4.0)g 7.2(6.2-8.6)g 61.4 ± 19.1 g 2.7 ± 0.4 h 6.4 ± 1.6 h 57.4 ± 13.6 h 2.8 j 4.7 j 8.8 j 60.1 ± 14.8 j Urban (Incheon, Korea) 3.7(2.9-4.3)g 6.8(5.0-7.5)g 64.5 g Urban (Gwangju, Korea) 2.6 ± 0.4 i 5.6 ± 1.1 i 57.8 ± 10.2 g a Trier et al., 1997, b Trier et al., 1993, c Ozkaynak et al., 1985, d Malm et al., 1996, e Virkkula et al., 2011, f Jung et al.,  2009, g Kim et al., 2006, h Kim, 2007, i Kim et al., 2001, j Kim, 2011.8.5 m 2 /g and the average β′ 1.0 was calculated to be approximately 7.3 ± 0.6 m 2 /g as summarized in Table 5.
The average correspondence between PM 1.0 mass and light extinction (R 2 ) was 0.81 ± 0.08.These mass extinction coefficients of PM 1.0 were relatively lower than those considering the f(RH) because the value of the f S (RH) was higher than that of the f(RH).The average reconstructed β′ 1.0 was also lower value of 11.8 ± 1.5 m 2 /g (R 2 = 0.81) than reconstructed β 1.0 .It was suggested that estimation of light extinction using the mass extinction coefficient of PM 1.0 might be roughly performed in two ways associated with aerosol hygroscopicity by application of the particlesize-dependant ratios of wet to dry scattering f S (RH), which was the function of the growth factor.

Contribution of Size-Resolved Aerosols to Light Extinction
The visibility impairing aerosol species contribute to light extinction with their mass extinction efficiencies.They are classified into seven major types: ammonium sulfates (NHSO), ammonium nitrates (NHNO), organic mass by carbon (OMC), elemental carbon (EC), fine soil (FS), mineral dust (MD), and sea salt (SS).Pitchford et al. (2007) suggested hygroscopic growth factors for the large particles (f L (RH)) and the small particles (fS(RH)) in the fine regime for NHSO and NHNO and that for SS (f SS (RH)).The mass extinction efficiencies for NHSO L , NHNO L , and OMC L in the large particle size mode were estimated to be 4.8, 5.1, and 6.1 m 2 /g and those for NHSO S , NHNO S , and OMC S in the small particle size mode were 2.2, 2.4, and 2.8 m 2 /g, respectively, based upon the calculation by applying Mie theory to lognormal size distributions with specified geometric mean diameters and standard deviations.In this study, the measured PM 1.0 and the particles between PM 1.0 and PM 2.5 were assumed to be the small particles and the large particles in the fine regime, respectively.And the mass extinction efficiencies of elemental carbon (EC), fine soil (FS), SS, and coarse mass (CM) including b NO2 and b Ray were used to reconstruct the light extinction coefficient based on Eq. (4) (Pitchford et al., 2007).This assumption is not exactly consistent with the small and the large modes given by Pitchford et al. (2007) because they have geometric mean diameters of 0.2 and 0.5 µm, respectively.The mass scattering efficiency is a function not only of the scattering cross section, which increases as a function of size, but also of the number of particles in a particular size range.This study, therefore, was carried out by considering the uncertainty of the assumption.In addition, the modified IMPROVE equation was applied to avoids the discrepancy with Pitchford et al. (2007) with dry efficiencies of 3, 3, and 4 for NHSO, NHNO, and OMC to PM 2.5 including SS in the fine regime as shown in Eq. ( 5).The multiplier of 1.8 was used to convert OC to OMC for Eq. ( 4) and the multiplier of 1.4 was used for Eq. ( 5).Pitchford et al. (2007) suggested the multiplier of 1.8 to convert OC to OMC.Turpin and Lim (2001) recommended the use of OM/OC factors of 1.6 ± 0.2 and 2.1 ± 0.2 for urban and nonurban aerosol, respectively.The multiplier of 1.8 was used according to the revised IMPROVE equation.Therefore, the overestimation of light scattering by OMC can be generated by using the multiplier of 1.8 applying the urban area of Seoul.4) and 46% based on Eq. ( 5) to the light extinction.On average, the carbonaceous particles including OMC and EC accounted for both about 19% to light extinction.However, the contribution of OMC to light extinction was higher based on Eq. ( 4) than based on Eq. ( 5).The contribution of NHNO based on Eq. ( 4) was also higher values of 10.2% than that based on Eq. ( 5) as shown in Fig. 7.The contributions of ammonium sulfates, ammonium nitrate, and organic mass aerosols to light extinction were estimated relatively higher by applying Eq. (3) than by applying Eq. ( 4).In addition, the contributions of NHSO L , NHNO L , and OMC L in the large particle size mode to light extinction were relatively higher than those of NHSO S , NHNO S , and OMC S in the small particle size mode.The small particles size mode represents firstly emitted particles, whereas the large particle size mode represents aged aerosols (Pitchford et al., 2007).It was found that aged aerosols had an effect on visibility reduction in the urban atmosphere of Seoul.However uncertainty of the results cannot be ruled out because the assumption proposed in this study can overestimate light extinction in the large particle size mode due to inconsistency with the light scattering peak.

CONCLUSION
The extensive visibility monitoring was conducted to determine the optical properties of size-resolved aerosols affecting visibility in the urban area of Seoul, Korea from 2007 to 2009.From the optical closure analysis, the measured and reconstructed light extinction coefficients compared well with R 2 = 0.96 and a relative error of 8 ± 5% considering Rayleigh scattering and light absorption by NO 2 .The average light extinction coefficient and visual range were measured to be 356 ± 284 Mm -1 and 10.5 ± 9.7 km during the intensive visibility monitoring periods.The average PM 2.5 mass of 38.0 ± 27.3 µg/m 3 and the average PM 1.0 mass of 20.1 ± 13.9 µg/m 3 did not attain the EPA annual PM 2.5 standard.The ammonium sulfates were the largest contributor to light extinction.It was found that aged aerosols of NHSO L , NHNO L , and OMC L in the large particle size mode had an effect on visibility reduction in the urban atmosphere of Seoul.Based upon the regression analysis the mass extinction coefficients of sized-resolved aerosols of PM 1.0 , PM 2.5 , and PM 10 were produced to be 8.7 ± 0.8, 4.7 ± 0.2, and 2.7 ± 0.2 m 2 /g, respectively.Visibility can be estimated with them because the light extinction coefficient is related to aerosol mass.They were reconstructed to be 12.8 ± 1.7, 6.7 ± 0.9, and 4.0 ± 0.5 m 2 /g when the mass fraction of hygroscopic aerosols of sulfates and nitrates was approximately 27% in the PM 1.0 regime, 29% in the PM 2.5 regime, and 22% in the PM 10 regime, respectively.In particular, the mass extinction coefficient of PM 1.0 was calculated to be approximately 7.3 ± 0.6 m 2 /g considering the f S (RH) for total particles and 11.8 ± 1.5 m 2 /g considering the f S (RH) for sulfates and nitrates particles.These results imply that the mass concentrations of size-resolved aerosols are good surrogates for the light extinction coefficient.Simultaneously they may lead to uncertainty of visibility estimation.Therefore, further extensive study including routine measurements of sizeresolved aerosols will be necessary to reduce the variability of the visibility estimation.Estimation of the light extinction coefficient can provide proper information for visibility improvement in the urban atmosphere of Seoul.
made aerosol optical and chemical data at the visibility monitoring station of Yonsei University (N37°33′52′′, E126°56′14′′) and visual images at the perceived visibility monitoring site of the Seoul Tower (N37°33′05′′, E126°59′16′′) in Seoul, Korea from 19 May 2007 to 22 November 2009.The eight intensive monitoring periods were from 19 to 30 May 2007, from 6 to 16 July 2007, from 29 September to 8 October 2007, from 15 to 24 December 2007, from 4 to 9 July 2008, from 17 to 24 October 2008, from 11 to 19 May 2009, and from 16 to 22 November 2009.The seasonal intensive monitoring periods were randomly selected.The visibility monitoring station is about 4.6 km northeast away from the perceived visibility monitoring site as shown in Fig. 1.The visibility monitoring station consisted of four categories of instrumentations such as optical measurements, aerosol samplers, optical & chemical analyzers, and a meteorological measurement.And an automatic digital camera system was installed at the perceived visibility monitoring site.The optical measurements were conducted with model LPV-3 long-path transmissometer (OPTEC, Lowell, USA), model NGN-2 nephelometer (OPTEC, Lowell, USA), and model AE-16U aethalometer (Magee Scientific, Berkeley, USA), which produced a light extinction coefficient (b ext ), a light scattering coefficient for wet particles (b scat ), and a light absorption coefficient (b abs ).b ext was measured at the wavelength of 550 ± 50 nm with the open path in ambient air.Calibration for measurement of b ext was carried out using the differential path method (OPTEC

Fig. 1 .
Fig. 1.The satellite image of the study area in Seoul, Korea.

Fig. 3 .
Fig. 3. Scatter plot between the light extinction coefficient (b ext ) and the sum of the light scattering coefficient for wet particles (b scat ), the light absorption coefficient (b abs ) by EC, and the light absorption coefficient by NO 2 (b NO2 ).
used to calculate the atmospheric light absorption coefficient from black carbon concentration measured by the aethalometer.

)
Relationships between the measured b ext and the reconstructed light extinction coefficients (b ext,R ) based on Eqs.(4) and (5) during the entire IVP are shown in Fig. 6.The average values of R 2 are 0.73 by applying Eq. (4) and 0.80 by applying Eq. (5).The average relative errors (|b extb ext,R |/b ext ) based on Eqs.(4) and (5) were calculated to be 27 ± 23 and 31 ± 27% from the scatter plots, respectively.The correspondence are acceptable although they are affected by data points where the b ext,R underestimate or overestimate the measured b ext .However, the uncertainty resulted from the reconstructed parameters for light extinction cannot be ruled out because observed light extinction by the ambient aerosol is approximated by assuming the aerosol being an external mixture.It is expected that the further visibility study including the size-resolved aerosol monitoring provides the better information on the light extinction equation.The average light extinction budgets based on Eqs.(4) and (5) for visibility impairing aerosols during the entire IVP are

Table 1 .
Sampling methods and filter treatments for aerosol monitoring at the visibility monitoring station.distilled deionized water by carefully shaking and rolling them for IC analysis.Blank tests were performed for all filter measurement

Table 2 .
Statistics of light attenuation coefficients and an optical closure.Relative error (|b

Table 3 .
Various mass extinction coefficients for suspended particulate matters.

Table 4 .
Mass fraction (f i ) and reconstructed mass (RM i ) derived from the sum of mass concentrations of sulfates and nitrates out of the size-resolved aerosol mass.