Characterization of Organic Aerosol Particles Observed during Asian Dust Events in Spring 2010

Between 20 March and 4 April, 2010, 24-hr PM2.5 measurements were carried out at an urban site in Gwangju, Korea, to examine the variations in the composition of its chemical constituents among dust storm (DS) events, haze pollution, and typical sampling (TS) conditions. A very heavy Asian DS (DS1) and a minor DS (DS2) occurred at the site on 20 and 23 March, 2010, respectively. The concentrations of organic matter, NO3, SO4, and the predicted crustal material made a contribution to the observed PM2.5 of 12.9, 9.7, 12.9, and 36.9% for DS1, 23.8, 17.1, 9.9, and 29.7% for DS2, and 24.9, 20.2, 10.3, and 17.6% for the haze event, respectively. Over the study period, the water-soluble organic carbon (WSOC)/OC and hydrophilic WSOC (WSOCHPI)/WSOC ratios ranged from 0.35 to 0.54 and from 0.15 to 0.61, respectively, with the highest ratios occurring during DS1. The concentration of WSOCHPI during DS1 was about 2–3 times higher than that during the TS period. The strong correlation between WSOC/OC and WSOCHPI/WSOC (R = 0.84) clearly indicates that the increase in the WSOC/OC ratio can be attributed to the increased concentration of WSOCHPI, suggesting that the aerosol sample collected during DS1 was more aged or atmospherically processed than the samples collected during the TS. In addition to the association between the WSOC/OC and WSOCHPI/WSOC ratios, strong correlations between the oxalate and SO4 concentrations (R = 0.74), and between the WSOCHPI and SO4 concentrations (R = 0.69), suggest that the WSOCHPI observed at the site was produced by atmospheric transformation processes similar to those seen with SO4 and oxalate. In contrast, the hydrophobic WSOC (WSOCHPO) concentration dominated in the other sampling periods, except for DS1, and accounted for 71.1–84.7% of WSOC. Based on the results of previous studies, the higher WSOCHPO/WSOC ratio during the TS indicates that the primary combustion emissions were important sources of the WSOCHPO fraction at this site.


INTRODUCTION
The Asian continent is an important source of atmospheric aerosols and the increasing occurrence of Asian dust storms (DSs) has led to severe problems such as poor air quality, visibility degradation, respiratory illness, and damage to animals, plants, crops, historical buildings, monuments and industrial activities (Lee et al., 2006).The arid and semiarid areas in north or northwestern China are major sources of dust emissions in East Asia.Mineral dust particles affect the energy balance of the Earth, both directly by scattering and absorbing light, and indirectly by altering cloud microphysical properties, and hence climate (Duce, 1995;Schwartz, 1996;Satheesh and Moorthy, 2005).It has been postulated that DSs over parts of China and Mongolia conversion in the atmosphere (Seinfeld and Pandis, 1998).Primary OC includes plant waxes, resin residues and longchain hydrocarbons, while the secondary OC are typically multifunctional oxygenated compounds, which are watersoluble (Jaffrezo et al., 2005), including carboxylic acids, alcohols, carbonyls, amines, polyols, saccharides, and organic nitrates.A significant portion, typically 10-70%, of particulate organic aerosols (OAs) is water-soluble (Jaffrezo et al., 2005;Park and Cho, 2011).The water-soluble OC (WSOC) can affect the hygroscopicity of aerosols and could be important in determining their ability to serve as cloud condensation nuclei, which may be important to the Earth's radiative budget (Novakov and Penner, 1993;Saxena et al., 1995;Facchini et al., 1999).Since secondary OAs (SOAs) are a major source of WSOC, the study of WSOC is one method for investigating SOAs (Saxena and Hildemann, 1996).Observations of atmospheric OAs have been conducted during Asian dust events to determine the chemical characteristics of the atmospheric organic compounds in the northwestern Pacific region and their sources in aerosol particles (Gao et al., 2003;Guo et al., 2004b;Simoneit et al., 2004;Yang et al., 2004;Park et al., 2006).However, little investigation has been conducted into the chemical properties of isolated chemical groups of WSOC in Asian dust particles.
The worst DS event in history was recorded on 20 March, 2010 in Korea, especially the southern region, including Daegu, Busan, Ulsan, Gwangju and Jeju, with hourly maximum PM 10 concentrations of 2,400-3,100 μg/m 3 .This DS (DS1) originated from the Gobi desert of Mongolia on 18-19 March, passing through the highly industrialized regions in China (e.g., Beijing), and reached the Korean peninsula in the afternoon of 20 March.Zhao et al. (2011) investigated the chemical characteristics of elemental species and water-soluble ionic components in PM observed during 21-23 March, 2010, when a heavy dust episode occurred, in the coastal city of Xiamen in southeastern China.In their work, hourly highest PM 10 and PM 2.5 concentrations were 990 and 455 μg/m 3 in the afternoon of 21 March, 2010, respectively.In the present study, 24-hr PM 2.5 samples were collected between 20 March and 4 April 2010 at an urban site in Gwangju, Korea, and used to determine the concentrations of PM 2.5 mass, elemental species, OC and elemental carbon (EC), ionic species, oxalate, WSOC, and two isolated water-soluble OAs.During the measurement period, the very strong DS1, a minor DS (DS2), and haze pollution event occurred on 20, 23 and 24 March 2010, respectively.The study objectives are to evaluate the chemical compositions of PM 2.5 samples during the study period and investigate the variations in chemical composition of PM 2.5 observed during the three aerosol pollution events, especially in OAs.

Ambient Sampling
Between 20 March and 4 April 2010, starting at about 09:00 a.m., 24-hr PM 2.5 samples were taken at an urban site (3511'N, 12654'E) in Gwangju, Korea.The sampling site was located on rooftop of a three-story building in a University, approximately 150 m from a two-lane road carrying heavy traffic during rush-hour, and greatly influenced by dust and anthropogenic emissions from northern China (Park et al., 2005(Park et al., , 2007(Park et al., , 2012a)).During the measurement period, daily average temperatures ranged from 4.8 to 13.7C and relative humidity varied between 39.5% and 68.5%.The prevailing surface wind was typically northwesterly, with a daily mean wind speed of 1.5-3.3m/s, with the highest speed occurring on 20 March.
Three PM 2.5 low-volume samplers were deployed to collect ambient aerosol particles at a flow rate of 16.7 L/min.The particle samples collected on 47 mm quartz fiber filters in the first and second samplers were used to analyze the OC and EC contents, water-soluble ionic components, total WSOC, and isolated WSOC.A multi-channel parallel plate carbon denuder was placed upstream of the filter-pack to remove semi-volatile organic vapors.All the quartz fiber filters used were pre-cleaned by baking at 500°C for 10 hr to minimize carbon blanks.These fiber filters were then placed in Petri dishes with prebaked aluminum foil liners.The Petri dishes were wrapped with Teflon tape and aluminum foil and then stored in a freezer until required for field measurements.All aerosol particle samples were refrigerated after collection.Aerosol samples in the third sampler were collected on 47 mm Teflon filters (Teflo membrane, 2.0-μm pore size) and used to determine the PM 2.5 mass and elemental species.Field blanks for the quartz fiber and Teflon filters during the measurement period were prepared three times by loading the filter into the sampler without the vacuum pump turned on.Hourly concentrations of ambient air pollutants -PM 10 , sulfur dioxide (SO 2 ), ozone (O 3 ), carbon monoxide (CO), and nitrogen oxides (NO x ) -were also measured using an ambient air monitoring system from the Ministry of the Environment, at a location about 2.0 km from the aerosol sampling site.

Chemical Analyses OC, EC, Ionic Species, and Elemental Species
One quarter or half of the quartz filter from the first sampler was analyzed for OC and EC using the National Institute for Occupational Safety and Health (NIOSH) thermal-optical transmittance protocol at Sunset Laboratory Inc. (NC office, USA).Details of the OC and EC analyses are given in Park et al. (2012a).However, the OC concentrations determined in this study may be overestimated as carbonate carbon can be evolved during OC under NIOSH protocol, especially during DS event.In the thermal manganese dioxide oxidation (TMO) protocol (Fung, 1990), a filter disk is acidified with HCl to remove carbonate carbon before OC/EC analysis.The linear regression between OC from TMO and OC from TOT conducted for five Asian dust storm samples collected in Korea during the ACE-Asia campaign in 2001 had a slope of 0.97 and an R 2 of 0.82 (not shown here).In this study, the method detection limit (MDL) of a substance was calculated as the average blank value of the substance plus three times the standard deviation of the blanks.The MDL for the OC and EC measurements were 0.25 and 0.04 μg C/m 3 .The OC and EC measurements had precisions of 6.5 and 9.6%, respectively.
A half or third quarters of the quartz filter was analyzed for eight ionic species (Cl -, NO 3 -, SO 4 2-, Na + , NH 4 + , K + , Mg 2+ , and Ca 2+ ) and oxalic acid using a Metrohm 861 ion chromatography.Each filter was first put into a 40 mL vial and extracted with 20 mL of ultrapure distilled de-ionized water (DDW) via ultrasonication for 60 min.All the extracts were filtered using a 0.45 μm membrane filter.The MDL values for NO 3 -, SO 4 2-, NH 4 + , and oxalic acid were measured to be 0.07, 0.14, 0.06, and 0.01 μg/m 3 , respectively.A detailed description of the ion chromatography system used in this study can be found elsewhere (Park et al., 2012a).
The Teflon filters from the third sampler were weighed before and after the sample collection with a microbalance (Sartorius MC5) of 1-μg sensitivity and analyzed for twenty-three elemental components.To analyze the trace elements, each Teflon filter was placed in a 60-mL lowpressure Teflon digestion vessel.After 5-mL aqua solution of HNO 3 and HCl with a 3:1 ratio was added to the vessel, it was capped to prevent the aqua solution from being volatilized.The vessel was heated to 150-180°C and held for 4 hr to facilitate complete dissolution.After cooling to room temperature, the solution was volatilized by a heater.Finally, the concentrated solution was diluted by adding 20 mL of 1% HNO 3 to the vessel.The elements Na, Mg, Al, Si, S, K, Ca, Ti, Mn, Fe, Cu, Zn, and Ba were determined by Inductively Coupled Plasma-Atomic Emission Spectrometry (ICP-AES, Shimadzu, ICPS-1000III).The rest of the elements (Sc, V, Cr, Ni, As, Zr, Cd, Sb, Se, and Pb) were analyzed by Inductively Coupled Plasma-Mass Spectrometry (ICP-MS, Thermo Elemental, XSERIES).When elemental species were analyzed by the ICP-MS, internal standard materials of 115 In and 205 Tl were dispensed to the samples and standard solution to correct for the instrument drift and matrix effect.The filter blanks were also analyzed, and the average filter blank value was used as a background subtraction for each sample filter.The MDL of Na, Mg, Al, Si, S, K, Ca, Sc, Ti, V, Cr, Mn, Fe, Ni, Cu, Zn, As, Zr, Cd, Sb, Ba, Se, and Pb for 24-hr measurements were 7. 7, 11.4, 10.4, 34.8, 26.9, 6.1, 31.9, 0.1, 3.1, 0.1, 3.0, 0.1, 12.1, 0.5, 2.1, 6.7, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, and 0.4 ng/m 3 , respectively.Precision, defined as the relative standard deviation, showed values of < 20% for all elements.The accuracy of the extraction and analysis used in this study was also tested by NIST standard reference material (SRM) 1648a.The average recoveries of SRMs were compared with their reported values.Approximately 100 ± 10% was recovered for all elements, except Si, of which about 65% ± 10% was recovered.These results indicate that our analytical techniques for the determination of the elemental composition of atmospheric aerosols were reliable.

Group Separation of WSOC Fractions Using XAD Resin Column
Quartz filter samples from the second sampler were extracted in 40 mL of DDW using ultrasonication for 60 min.The water extracts filtered using a 0.45 μm syringe membrane filter were analyzed to determine the amount of WSOC using a total organic carbon analyzer (TOC, Sievers 5310C, USA).Details of the TOC analysis are described in Park et al. (2012a).The WSOC detection limit was measured to be 0.22 μg C/m 3 .Also the WSOC measurement showed a precision of < ± 5%.The remaining extracts were used to group isolate the original WSOC into hydrophilic (WSOC HPI ) and hydrophobic (WSOC HPO ) fractions using a macro-porous nonionic resin (XAD7HP) column.
In some studies, an XAD-8 resin has been used to separate WSOC HPO fractions from ambient air samples (Duarte and Duarte, 2005;Sullivan and Weber, 2006a).Because the XAD-8 resin is no longer commercially available, for the group separation of the WSOC in this study, an XAD7HP (Rohm & Haas France S.A.S) and an aqueous chromatography column (6 mm ID × 10 cm long, Spectrum Laboratories, Inc., Houston, TX, USA) were used with TOC detection.The XAD7HP is a polymeric adsorbent available as white insoluble beads.It is a nonionic, aliphatic acrylic crosslinked polymer which derives its adsorptive properties from its macroreticular structure, high surface area and the aliphatic nature of its surface.
Details for pre-treatment of the resin and the group separation of WSOC fractions are given in Park et al. (2012a).Briefly, the cleaning cycle of the resin-packed column consisted of alternating 0.1 M NaOH and 0.1 M HCl for 20 min each, all at flow rates of 2.0 mL/min, before the water extracts were continuously pumped onto the resin column.An intermediate cleaning step between alkali and acid treatments was performed with DDW.This NaOH-HCl cycle was repeated three times.After all the washing procedures were finished, the DDW passing through the column was analyzed using the TOC analyzer to check the quality of the washing.The TOC levels before and after passing through the resin column were measured to have concentrations of < 15 and < 10 ppb C, respectively.The average blank values were used as the background for subtraction from each filter sample.The aqueous sample solution prior to loading onto the cleaned resin column was adjusted to pH 2 using HCl, and then introduced to the column at a rate of 2.0 mL/min.The flow rate over the column was changed from the cleaning phase to sample loading.The sample flow rate through the column was maintained at 1.3 mL/min for 20 min.The organic compounds that penetrate the XAD7HP column were referred to as the hydrophilic fraction, while the compounds retained on the column at pH 2 were referred to as the hydrophobic components.The column was eluted to pH 13 to recover the amount of the adsorbed hydrophobic fraction.According to the previous recovery experiments (Park et al., 2012a), the hydrophobic compounds were not completely recovered from the column at pH 13 eluent, with recovery efficiencies ranging from 75% to 85% depending on the organic compounds.Similar results were also obtained by other researchers (Sullivan and Weber, 2006a;Miyazaki et al., 2009).Thus, in this study, the difference between the total WSOC and WSOC HPI was defined as the WSOC HPO fraction.Park et al. (2012b) evaluated the penetration efficiencies of the XAD7HP resin column using a variety of water-soluble organic compounds.The results indicated that the WSOC HPI fraction included aliphatic dicarboxylic acids and carbonyls (< 4 carbons), amines and saccharides, while WSOC HPO included aliphatic dicarboxylic acids (> 4-5 carbons), phenols, aromatic acids and cyclic acid, as well as Suwannee River fulvic acid.As discussed above, some artifacts may have remained in the fractionation process.The analytical uncertainties of the total WSOC and WSOC HPI measurements made with the field blanks were estimated at ± 5% and ± 9%, respectively.

Air Mass Pathways of the Three Aerosol Events and Time-Series Plot of PM 10 Concentration
The Korean Meteorological Agency (KMA) reported that a heavy DS event (DS1), a minor DS event (DS2), and a haze pollution episode occurred on 20, 23, and 24 March, 2010, respectively, during the study period.A detailed description on the dust storm and haze pollution events classified in this study is given below.To examine the origin of the air mass for the classified aerosol particle pollution events, 3-day backward air trajectories were calculated using the HYSPLIT model for three altitudes (500, 1000, and 1500 m) above ground level (AGL) (Draxler and Rolph, 2012).Fig. 1 shows the air parcel transport pathways for the indicated three event days (March 20,23,and 24).The PM 10 forecast charts calculated during the Asian DS periods  Therefore, it can be concluded that the two DS events had different PM 2.5 chemical composition characteristics.The air mass calculated for 24 March originated from the northeastern region of China, or Yellow Sea (500 m AGL), and passed over North Korea, the East Sea of Korea, and the eastern inland region of Korea prior to reaching the sampling site.On March 24 th , a haze pollution episode was observed at the site (http://www.kma.go.kr/we ather/observation).This air mass may have contained a mixture of local haze and long-range transported pollution.Fig. 3 shows the temporal variation in hourly PM 10 mass concentration during the measurement period.A very heavy Asian DS originated from the Gobi desert of Mongolia on 18-19 March (DS1, see Fig. 1), reached theGwangju sampling site at 14:00 on 20 March with 1-hr PM 10 concentration of 123.0 μg/m 3 , and had completely disappeared by 01:00-02:00 on 21 March.On March 20 th , the PM 10 concentration began to increase at 14:00 LT to 123.0 μg/m 3 , reached a maximum concentration of 1,864 μg/m 3 at 22:00 LT, and then decreased to 36.0 μg/m 3 at 01:00-02:00 LT on 21 March.The 1-hr PM 10 concentration was maintained within the range of 30-85 μg/m 3 on 21 March.On March 20 th , 24-hr average PM 10 and PM 2.5 concentrations were 350.9 and 91.9 μg/m 3 , respectively.In the DS2, the PM 10 concentration started to increase at 14:00 LT (119 μg/m 3 ), reached a maximum level of 292 μg/m 3 at 17:00-18:00 LT, and then decreased.KMA reported the time period between 14:00 and 22:00 on March 23 as the "dust storm" event.After then, no noticeable change in PM 10 concentration was observed due to the occurrence of local haze.The pollution episode observed at 23:00 on March 23 was associated with the haze that lingered over Gwangju until 05:00 on March 25, and which elevated the PM 10 and PM 2.5 levels.The hourly PM 10 concentration between 23:00 on 23 March and 09:00 on 24 March ranged from 93.0 to 135.0 μg/m 3 .During this DS event, 24-hr average PM 10 and PM 2.5 concentrations were 140.3 and 51.7 μg/m 3 , respectively.As a result, DS2 (09:00 on March 23-09:00 on March 24) was characterized by aerosols derived from both Asian dust and local haze pollution.After 09:00 on March 24, hourly PM 10 concentration ranged from 90 μg/m 3 (05:00 on March 25) to 197 μg/m 3 (18:00 on March 24).The aerosol sample collected between 09:00 on March 24 and 09:00 on March 25 was attributed to locally produced pollution.

Chemical Properties of PM 2.5 for the Three Aerosol Pollution Episodes
Table 1 lists the concentrations of the chemical constituents in PM 2.5 for all measurement days.Reconstructed PM 2.5 mass concentrations based on the concentrations of the measured chemical species were estimated to verify the reliability of the measurement results.The reconstructed PM 2.5 mass (PM 2.5Rec ) was computed by summing the concentrations of organic matter (OM), EC, Cl -, NO 3 -, SO 4 2-, NH 4 + , crustal material, and others.OM was estimated by applying a factor of 1.6 (i.e., urban aerosols) to the OC (Turpin and Lim, 2001).The concentration of crustal material was also predicted by using the following formula: Crustal material (μg/m 3 ) = 2.20[Al] + 2.49[Si] + 1.63[Ca] + 2.42[Fe] + 1.94[Ti] (Malm et al., 1996).The reconstructed PM 2.5 concentration was strongly correlated with the observed one (PM 2.5obs ) according to the following relationship: PM 2.5Rec (μg/m 3 ) = 0.92 PM 2.5obs (μg/m 3 ), R 2 = 0.96, which shows the good agreement between the two methods.The reconstructed PM 2.5 mass concentrations ranged from 75.1 to 108.9% of the measured concentrations.As shown in Table 1, the 24-hr PM 2.5 mass concentration for DS1, DS2, and the haze pollution that occurred on March 20, 23, and 24 was 91.9, 51.7, and 61.6 μg/m 3 , respectively.For DS1, the concentrations of OC, EC, NO 3 -, SO 4 2-, NH 4 + , and the predicted crustal material were 8.5, 1.9, 10.3, 13.7, 5.4, and 39.1 μg/m 3 , accounting for 12.9 (based on OM), 1.8, 9.7, 12.9, 5.1, and 36.9% of the measured PM 2.5 concentration, respectively.The equivalent concentrations for DS2 were 7.7, 2.9, 8.9, 5.1, 3.9, and 15.4 μg/m 3 , contributing 23.8 (based on the OM), 5.6, 17.1, 9.9, 7.6, and 29.7% to the PM 2.5 mass, respectively.During the haze event, the OM concentration contributed 24.9% to the PM 2.5 mass, followed by NO 3 -(20.2%),crustal material (17.6%),SO 4 2-(10.3%),NH 4 + (9.2%), and EC (3.4%).The crustal material contribution to PM 2.5 during DS1 and DS2 was much lower than those (50-70% of the PM 2.5 mass) measured at different regions during the DS periods of ACE-Asia (Xu et al., 2004;Park et al., 2007).An interesting feature of the haze episode was its highly elevated contributions of OM, NO 3 -, and crustal material.The elevated crustal material concentration (10.9 μg/m 3 ) observed for this haze episode was likely attributed to the remaining crustal materials from DS2 on the previous day.The composition of elemental species in PM 2.5 is also shown in Table 1, along with those reported during previous DS events (Kim et al., 2003;Xu et al., 2004;Park et al., 2007).Higher concentrations were clearly observed for the crustal elements, i.e., Al, Si, K, Ca, Ti, Mn, Fe, Zr, and Ba, during the two DS episodes on March 20 and 23.Al and Si concentrations were 4.8 and 6.1 μg/m 3 on March 20, accounting for 4.6 and 5.8% of the measured PM 2.5 mass, and 1.9 and 2.7 μg/m 3 on March 23, accounting for 3.6 and 5.1% of the PM 2.5 mass, respectively.The Al fraction in PM 2.5 measured during the two DS days was lower than the Al abundance (~8%) from the average continental crust (Mason and Moore, 1982;Taylor and McLennan, 1995), indicating that these dust samples may have been diluted or dry-deposited prior to arriving at the site from the desert sources.Our results were slightly higher, or similar to those reported by previous studies (Xu et al., 2004;Park et al., 2007), in which the Al fraction at the same Gwangju sampling site and near a desert source in northwestern China during ACE-Asia DS events contributed 3.6% and 3.1% to the PM 2.5 mass concentration, respectively.The enriched concentration of Zr observed during the two DS events suggest that this element could be used as a tracer of soil dust.The concentration of Al, the representative soil material, was strongly correlated with Mg (r = 0.95), Si (r = 0.99), K (r = 0.98), Ca (r = 0.99), Sc (r = 0.96), Ti (r = 0.98), Mn (r = 0.98), Fe (r = 0.99), Zr (r = 0.92), and Ba (r = 0.98), suggesting that all of these elements came from crustal sources.The concentrations of anthropogenically produced trace elements such as S, V, Ni, Zn, As, Cd, and Pb were also more enhanced during DS1 than during typical sampling periods.A previous study has indicated that DSs carry not only considerable amounts of crustal dust, but also large amounts of pollution aerosol when the air mass passes over long distances through the industrialized polluted regions (Sun et al., 2005).The concentrations of pollution elements S, V, Ni, Zn, As, Cd, and Pb during DS1 increased by 1.8-5.0times as Al rose from 0.4 μg/m 3 in the typical average local-pollution aerosol to 4.8 μg/m 3 in the DS aerosol.In comparison, during DS2, which included significant impacts of local pollution, these ratios ranged between 0.8 and 1.9 as Al ranged from 0.4 μg/m 3 to 1.9 μg/m 3 (Table 1).The increased concentrations of pollution elements during these two DS events reflect the impact of anthropogenic emissions from polluted regions of China during the transport of air mass, as shown by the results of air mass back-trajectory analysis.Typically, DSs led to low SO 4 2-and NO 3 -concentrations, but high concentrations of crustal species.In this study, SO 4 2-was highly correlated with NO 3 -with an R 2 of 0.55 for all dataset (without two DS events, R 2 = 0.86).This correlation may have resulted from their existence as internal mixtures or their concentrations being governed by common meteorological conditions.A noticeable feature of PM 2.5 observed during DS1 was that the SO 4 2-contribution to PM 2.5 was elevated, which implied the influx of dust aerosols along with sulfate aerosol particles produced through chemical transformation of SO 2 emitted from highly polluted regions in northeastern China during transport of the air mass.The concentrations of the SO 4 2-and NO 3 -species were mostly higher during the DS events than afterwards, suggesting that they were influenced by the long-range transported polluted aerosol sources.Similar results were also found in previous studies (Streets et al., 2003;Heald et al., 2006;Park et al., 2007).While the contributions of the OM and EC concentrations to PM 2.5 during DS1 were relatively low, the OC/EC ratio of 4.5 was quite high over the entire measurement period.For DS2, the contribution of NO 3 particles, together with the carbonaceous species particles, was almost twice that of SO 4 2-particles.This indicates that in addition to the dust particles transported from southeastern China, which actually originated in the deserts of Mongolia and northeastern China, local pollution affected the chemical composition of the PM 2.5 measured on 23 March.

Chemical Evolutions of Water-Soluble Organic Aerosols (OAs)
The concentration of WSOC over the study period was in the range of 1.9-5.0μg C/m 3 .The WSOC concentration during DS1, DS2, the haze pollution episode and other sampling periods was 4.6, 3.4, 4.4, and 3.2 μg C/m 3 , respectively.The contribution of WSOC to OC varied between 34.9 and 54.3% with an average of 43.8%, being highest during DS1 as compared with other times.For DS2 and the haze pollution event, the WSOC contribution to OC was 44.2 and 45.5%, respectively.A similar behavior was observed for the oxalate, with the highest concentration (0.32 μg/m 3 ) occurring during DS1 and the second highest (0.28 μg/m 3 ) during the haze episode.Oxalate is typically the most abundant dicarboxylic acid in atmospheric aerosols and another important class of WSOC in the atmosphere (Narukawa et al., 2003).Secondary products are considered to be an important source of oxalate (Kawamura and Yasui, 2005;Wang et al., 2009).In addition to the secondary production, oxalate is emitted from vehicle exhausts (Kawamura and Kaplan, 1987), biomass burning (Narukawa et al., 1999) and biogenic activity (Kawamura et al., 1996).
As shown in Table 1, the concentrations of WSOC HPI and WSOC HPO were in the range 0.5-2.8 and 0.9-3.8μg C/m 3 , respectively, over the whole measurement period.The concentration of WSOC HPI contributed 15.3-60.6% of the total WSOC, as shown in Fig. 4. The WSOC HPO concentration was typically observed to be greater than the WSOC HPI , except for during DS1.For DS1, the WSOC HPI and WSOC HPO concentrations were 2.8 and 1.8 μg C/m 3 , respectively.The WSOC HPI on March 20 and 21 accounted for 60.6% and 51.2% of the total WSOC, respectively, about 2-3 times more than that of the WSOC HPI for the other sampling periods.For DS2 and the haze pollution event, the concentration of WSOC HPI contributed 31.3 and 37.1% to the total WSOC, respectively.In contrast, the WSOC HPO concentration dominated in the typical sampling periods, i.e., when the three aerosol pollution periods were excluded, accounting for 71.1-84.7% of WSOC with an average of 78.3%.The maximum fraction of WSOC HPI on March 20 suggests that the WSOC HPI in the PM 2. gained a significant contribution from atmospheric processing during the long-range transport of volatile organic species emitted from industrialized sources in northeastern China, as shown in the trajectory of air mass (Fig. 1).The relatively high WSOC HPI fraction on March 21 after DS1 had disappeared may have resulted from the remnant aged aerosol particles in the air and the clean atmospheric conditions.Previous urban measurements from Atlanta and St. Louis suggested that SOA formation processes produce higher WSOC HPI fractions, which include compounds with low molecular weights, such as aliphatic carboxylic acids and carbonyls (< 4 carbons), saccharides and amines, rather than a hydrophobic fraction (Sullivan and Weber, 2006a).Furthermore, some investigators indicated that more WSOC HPO , which is related to highly refractory compounds and tends to be less hygroscopic, including aliphatic carboxylic acids and carbonyls (> 3-4 carbons), was emitted directly from primary combustion sources (Yu et al., 2004;Yang et al., 2005;Sullivan and Weber, 2006a;Park et al., 2012a).Considering these previously reported results, the higher WSOC HPI /WSOC ratio on March 20, 21, 23, and 24 was attributed to further atmospheric transformation processing during the transport of air masses, while the higher WSOC HPO /WSOC ratio obtained from other sampling days suggests that the primary combustion emissions were important sources of the WSOC HPO fraction at the site.
Fig. 5 shows the relationships among total WSOC and two WSOC fractions for the complete dataset, indicating that WSOC was more highly correlated with the WSOC HPO fraction (R 2 = 0.46) than with the WSOC HPI fraction (R 2 = 0.39).However, when the two DS events were excluded from the analysis, the correlation coefficient between WSOC and WSOC HPO was significantly increased to an R 2 of 0.79.This implies that the WSOC HPO fraction was an important factor contributing to the OC in PM 2.5 at the site.As shown in Fig. 6, the EC concentration during the entire sampling period was moderately correlated with the total WSOC (R 2 = 0.42) and WSOC HPO (R 2 = 0.58).Regression analysis, performed after data points observed during the two DS events were removed, gave a tight correlation between EC and WSOC HPO (and WSOC) (WSOC HPO = 0.95EC + 0.47, R 2 = 0.70), clearly implying that the primary combustion emissions at the site were important sources of the WSOC HPO fraction.
As discussed above, a significant proportion of the WSOC HPI fraction of OC could be produced through processes such as SOA formation.The WSOC fraction could provide insights into the transformation processes of OA particles.A tight correlation between WSOC and OC was found with an R 2 of 0.81 (not shown here), which clearly revealed their similar chemical characteristic.As described above, the WSOC/OC ratio over the study period ranged from 0.35 on March 26 to 0.54 on March 20 during DS1.The corresponding WSOC HPI /WSOC ratio on those days was 0.15 and 0.61, respectively.In addition, the relatively high WSOC/OC and WSOC HPI /WSOC ratios on March 21, after DS1 had passed, of 0.54 and 0.51, respectively, implied that the aerosol sample collected on March 21 may have included the aged or chemically processed components of OC that remained after DS1 had passed.On March 23 and 24 during DS2 and the haze pollution event, the WSOC/OC ratio was 0.44 and 0.46 and the WSOC HPI /WSOC ratio was 0.31 and 0.37, respectively.As shown in Fig. 7, the strong correlation between the WSOC HPI /WSOC and WSOC/OC (R 2 = 0.84) clearly indicated that an increase in the WSOC/ OC ratio at the site resulted in the increased concentration of the WSOC HPI and was accompanied by the addition of the SOA mass.
To further investigate the formation routes of WSOC HPI , the relationships among WSOC HPI , oxalate, and SO 4 2-(and SO 4 2-/SO x (= SO 2 + SO 4 2- )) were analyzed for all the data, and the results are presented in Figs. 8 and 9.The concentration of oxalate, the major source of which results  from atmospheric processing, was strongly correlated with that of SO 4 2-(R 2 = 0.74), suggesting that the two chemical species were produced from similar atmospheric processes such as in-cloud or aerosol droplet processes.The in-cloud or heterogeneous formation of oxalate has been reported to have a good correlation between oxalate and SO 4 2- (Kerminen et al., 2000;Huang et al., 2006;Guo et al., 2010).The WSOC HPI and SO 4 2-concentrations were highly correlated with an R 2 of 0.69, as shown in Fig. 8. Similar to the oxalate formation process, this suggests that a significant fraction of the WSOC HPI observed at the site was produced by atmospheric transformation processes similar to SO 4 2-and oxalate, i.e., in-cloud processing and aerosol droplet process, as discussed above.The SO 4 2-/SO x ratio has been used as an indicator of the formation routes of SO 4 2- (Kaneyasu et al., 1995).In this study, SO 4 2-/SO x ranged from 0.19 to 0.54.The ratio during the DS1, DS2, and the haze pollution event was 0.54, 0.38, and 0.40, respectively, which were higher than the values during the other sampling periods.This suggests that the chemical components of the observed PM 2.5 during these three events had undergone further chemical processing, compared to those during the typical sampling periods.WSOC HPI was also highly correlated with SO 4 2-/SO x , with an R 2 of 0.63.This suggests that the Park and Cho, Aerosol and Air Quality Research, 13: 1019-1033, 2013  observed WSOC HPI could be produced via similar oxidation processes as those of SO 4 2-, i.e., by in-cloud aqueousphase oxidation, photochemical oxidation of SO 2 by OH radicals, or an aerosol droplet process.

SUMMARY AND CONCLUSION
A series of 24-hr PM 2.5 samples was taken at an urban site in Gwangju, Korea in the spring of 2010, to investigate the characteristics of OAs observed during the Asian DS events.During the study period, one strong dust storm (DS1), one minor dust storm (DS2), and a haze pollution event occurred on 20, 23, and 24 March, respectively.Interestingly, DS2 came from the southern region of China where it had been affected between 21 and 24 March by the heavy DS originating from desert regions of Mongolia and China (Zhao et al., 2011).
The most significant contributor during DS1 and DS2 was crustal material, contributing about 37 and 30% of the measured PM 2.5 mass, respectively.However, the air mass transport pathway was an important factor influencing the chemical composition of dust particles.The highest concentration of SO 4 2-and a relatively high NO 3 concentration were observed during DS1, when the air mass had originated over the Gobi desert of Mongolia and traveled over highly industrialized regions of northeastern China and the Yellow sea, while the highest NO 3 concentration (12.5 μg/m 3 ) was observed during the local

Fig. 1 .
Fig. 1.Pathways of air mass arriving at the sampling site on 20, 23, and 24 March.