Title : Springtime carbon emission episodes at the Gosan background site revealed by total carbon , stable carbon isotopic composition , and thermal characteristics of carbonaceous particles

Reviewer #1 (Comments): This paper reports on the analysis of total suspended matter from filter samples taken in April/May 2007 and April 2008 at the Gosan site on Jeju Island. The samples were analysed with respect to total carbon, organic and elemental carbon, total nitrogen and citric and oxalic acid concentration. The thermal characteristics of organic and elemental carbon are reported for individual filter samples. Furthermore the carbon isotopic composition of total carbon and oxalic and citric acid were measured and used for the interpretation of the aerosol sources. Aerosol particles influence climate and have adverse health effects. Elucidating the source of particles and identifying potential strategies for reducing atmospheric burdens of particles is an up to date topic. A particular challenge in this respect is the organic fraction of aerosols which has a vast variety of sources and is known to change properties and composition upon atmospheric aging. Therefore the paper in principle merits publication in ACP. Unfortunately the paper has several weaknesses which should be corrected before publication.


Introduction
Carbonaceous aerosols that comprise elemental carbon (EC) and organic carbon (OC) have large impacts on human health (Baltensperger et al., 2008), visibility impairment (IMPROVE, 2006), and radiation budget in the atmosphere (Forster et al., 2007).Organic aerosols are primarily emitted from various sources and secondarily produced in the atmosphere by oxidation of volatile organic compounds followed by condensation on pre-existing particles and/or nucleation.Novakov and Penner (1993) determined the relative contributions of sulfate and organic aerosols to cloud condensation nuclei (CCN) concentrations at a marine site known to be influenced by anthropogenic emissions, and found that organic aerosols account for the major part of both the total aerosol number concentration and the CCN fraction.Thus, in regions that are affected by anthropogenic pollutants, organic aerosols may play at least as important a role as sulfate aerosols in determining the climate effect of clouds (Novakov and Penner, 1993).Model simulation results indicate that organic aerosols can enhance cloud droplet concentrations and are therefore an important component of the aerosol-cloud-climate feedback system (O'Dowd et al., 2004).
Jeju Island, Korea, is located at the boundary of the Yellow Sea and the East China Sea and is surrounded by mainland China, the Korean peninsula, and Kyushu Island, Japan.The Gosan site is located on the western edge of Jeju Island facing the Asian continent and is isolated from residential areas on the island (Kawamura et al., 2004).In order to understand physicochemical and radiative properties of anthropogenic aerosols under Asian continental outflow, several international experiments have been conducted at the Gosan site, such as ACE-Asia (Aerosol Characterization Experiment-Asia) (Huebert et al., 2003) and ABC-EAREX 2005(Atmospheric Brown Cloud-East Asia Regional Experiment 2005) (Nakajima et al., 2007).To better understand the link between chemical and physical properties of aerosols mainly transported from the Asian continent and regional climate change, sources and formation mechanism of secondary aerosols should be investigated.In addition, contributions of local effects on the Gosan site aerosols should be qualitatively and quantitatively evaluated.
Pollen is one of the important sources of bioaerosols (Solomon, 2002).These particles can cause serious allergic problems to human health (Solomon, 2002) and visibility impairment (Kim, 2007).Most pollen emission events in Korea occur during the blooming season of plants from March to May (Oh et al., 1998).Because airborne pollen can be transported very long distances (Porsbjerg et al., 2003;Rousseau et al., 2008), airborne pollen is not only a local problem but also a regional and intercontinental problem.However, studies regarding the impact of pollen on the aerosol chemical composition at the Gosan site are rare (Jung and Kawamura, 2011).
Stable carbon isotopic compositions (δ 13 C) of total carbon (TC) are very useful for investigating sources and the longrange atmospheric transport of organic aerosols (Cachier et al., 1986;Narukawa et al., 2008;Miyazaki et al., 2010).On the basis of the δ 13 C values and Na + /TC ratios, Narukawa et al. (2008) estimated the contribution of marine organic matter in TC in the high Arctic at Alert during spring.Miyazaki et al. (2010) estimated marine-derived carbon in TC over the western North Pacific using δ 13 C values.Also, Kawamura and Watanabe (2004) developed a novel method for compound specific carbon isotope compositions of dicarboxylic acids and related compounds using gas chromatography/isotope ratio mass spectrometry (GC/irMS).Since then, δ 13 C values of dicarboxylic acids and related compounds have successfully been used to assess the extent of photochemical processing of aerosols during their long-range atmospheric transport (Wang and Kawamura, 2006;Aggarwal and Kawamura, 2008).
In this study, elevated concentrations of total carbon were often found in the total suspended particulate (TSP) samples collected at the Gosan site during spring of 2007 and 2008.Using remote monitoring and comprehensive in-situ chemical and δ 13 C analyses, we categorize the carbon emission episodes to three types: long-range transport anthropogenic pollution (LTP) from Asian continent, Asian dust (AD) accompanying with the LTP, and local airborne pollen episodes.We discuss the measured δ 13 C values and the thermal evolution patterns obtained from heating of carbonaceous particles.Based on a carbon isotope mass balance equation, we quantify the TC fraction of local pollen.Using HCl fume treatment of the dust-enriched samples, we quantify carbonate carbon in the TC and discuss removal of carbonate via the reaction with acidic gases during the longrange atmospheric transport.
2 Samples and methods TSP sampling was carried out at the Gosan supersite (33.17 • N, 126.10 • E) on Jeju Island, located approximately 100 km south of the Korean peninsula (Fig. 1), over successive periods that integrated 2-5 days from 23 March to 1 June 2007 and from 16 to 24 April 2008.TSP samples were collected on pre-combusted quartz fiber filters (20 × 25 cm) using a high volume air sampler (Kimoto AS-810) at a flow rate of 50 m 3 hr −1 on the rooftop of a trailer (∼3 m above the ground).Before and after sampling, the filter samples were stored in clean glass jars (150 ml) with a Teflon-lined screw cap at −20 • C prior to analysis.Field blank filters were collected every month.Hourly PM 10 mass data was obtained from the National Institute of Environment Research at the Gosan Observatory.Measurements of wind direction and wind speed at 10 m above ground level were obtained from the Korea Meteorological Administration at the Gosan Observatory.A total of 32 filter samples were analyzed in this study.

Airborne pollen and tangerine fruit samples
Three types of pollen samples were prepared and analyzed in this study.Two authentic standard pollen samples from Japanese cedar (Pollen cedar ) and Japanese cypress (Pollen cypress ) were obtained from the WAKO Chemical Co. (product No. 168-20911 for Japanese cedar and 165-20921 for Japanese cypress).Additionally, airborne pollen (Pollen Gosan ), which mainly originated from Japanese cedar trees planted around tangerine farms on Jeju Island, were separated from an aliquot of the TSP filter sample collected during a severe pollen episode (KOS751, 16-21 April 2008).An aliquot (15 cm 2 ) of the KOS751 filter sample was placed in a glass vial (50 ml) with a Teflon-lined screw cap and the pollen grains were separated by mild vibration using an automatic vibrator (Iuchi, HM-10) for 5 min.The separated pollen grains were then transferred to a pear shape flask for chemical analysis.In order to track the sources of the dicarboxylic acids and related compounds and airborne pollen, tangerine fruit produced from Jeju Island was also prepared and analyzed for dicarboxylic acids and related compounds (Jung and Kawamura, 2011) and for carbon isotopic compositions of total carbon, oxalic acid, and citric acid.In order to prevent possible contamination over the surface of tangerine fruit, the tangerine surface was mildly washed three times using ultra pure organic-free Milli-Q water.

Organic and elemental carbon analysis
Organic carbon (OC) and elemental carbon (EC) were analyzed with a Sunset carbon analyzer using the thermal-optical transmittance (TOT) protocol for pyrolysis correction (Birch and Cary, 1996;Miyazaki et al., 2009).A 2.14 cm 2 punch of the quartz filter was placed in a quartz boat inside the 1  C are defined as EC1, EC2, EC3, EC4, EC5, and EC6, respectively.The pyrolized organic carbon (PC), which was converted from OC in the inert mode of the analysis, was corrected by monitoring the transmittance of a pulsed He-Ne diode laser beam through the quartz fiber filter.External calibration was performed before the analysis using a known amount of sucrose.The detection limits of OC and EC, which are defined as three times of standard deviation of field blanks, were 0.26 and 0.01 µgC m −3 .However, these values are quite small compared to the instrument's minimum quantifiable level of 0.5 µgC m −3 given by the manufacturer.Therefore, 0.5 µgC m −3 was considered as the detection limit of both OC and EC.The analytical errors, which are defined as the ratio of the standard deviation to the average value obtained from the triplicate analyses of the filter sample, of OC and EC measurements were 5 % and 3 %, respectively.

Determination of water-soluble inorganic ions
To measure water-soluble inorganic ions, an aliquot (2.01 cm 2 ) of the quartz filter was extracted with 10 ml of the Milli-Q water under ultrasonication (30 min) and then passed through a disk filter (Millipore, Millex-GV, 0.45 µm).The concentrations of cations and anions in the water extracts were measured using an ion chromatography (Metrohm,761).The sodium (Na + ), calcium (Ca 2+ ) and ammonium (NH + 4 ) were determined us-ing an Metrosep C2 column (150 mm) with 4 mM tartaric acid/1 mM 2,6-pyridinedicarboxylic acid as an eluent (flow rate: 1.0 ml min −1 , sample loop volume: 200 µl, time eluted: 30 min).The nitrate (NO − 3 ) and sulfate (SO 2− 4 ) were determined using a Shodex IC SI-90 4E column with 1.8 mM Na 2 CO 3 /1.7 mM NaHCO 3 as an eluent (flow rate: 1.2 ml min −1 , sample loop volume: 200 µl, time eluted: 15 min).The analytical errors of the water-soluble inorganic ions were less than 1.4 % based on the triplicate analyses of filter sample.SO 2− 4 and Ca 2+ concentrations were corrected for sea-salt fraction using Na + as a sea-salt tracer in this study.It was found that ∼95 % of SO 2− 4 and ∼92 % of Ca 2+ were attributed to non-sea-salt SO 2− 4 (nss-SO 2− 4 ) and non-sea-salt Ca 2+ (nss-Ca 2+ ).All the concentrations of OC, EC, and water-soluble inorganic ions reported here were corrected against the field blanks.

Total carbon, total nitrogen, and carbon isotope analysis
Total carbon (TC) and total nitrogen (TN) were measured by an elemental analyzer (EA) (Carlo Erba, NA 1500) whereas stable carbon isotope (δ 13 C) analyses were conducted using the same EA interfaced to an isotope ratio mass spectrometer (irMS) (Finnigan MAT Delta Plus) (Kawamura et al., 2004).An aliquot of filter sample (2.01 cm 2 ) was cut using a circular cutter with a diameter of 1.6 cm and placed in a tin cup, introduced into the EA and then were oxidized in a combustion column packed with chromium (III) oxide at 1020 • C. Nitrogen oxides coming from the combustion column were reduced to molecular nitrogen (N 2 ) in a reduction column packed with metallic copper at 650 • C. The derived N 2 and carbon dioxide (CO 2 ) were isolated using a gas chromatograph (GC) installed inline with the EA and then measured with a thermal conductivity detector.An aliquot of CO 2 gases was then introduced to the irMS through an interface (ThermoQuest, ConFlo II).The carbon isotopic composition (δ 13 C) relative to the Pee Dee Belemnite (PDB) standard was calculated using the standard isotopic conversion equation as follows.
External calibration was conducted using known amounts of acetanilide in order to calculate mass concentrations of TC and TN and δ 13 C of TC (δ 13 C TC ). 5 standards ranging from 0.2 mg to 0.6 mg of acetanilide were prepared and analyzed by the EA-irMS.Acetanilide was purchased from Thermo Electron with a δ 13 C TC of −27.26 ‰.The mass concentrations and δ 13 C TC values reported here were corrected against the field blanks using isotope mass balance equations (Turekian et al., 2003).The blank levels of TC and TN mass concentrations were 2.0 % and 1.3 % of the measured concentrations, respectively.The analytical errors for TC and www.atmos-chem-phys.net/11/10911/2011/TN mass concentrations based on the triplicate analyses of the filter sample were 2.3 % and 5.2 %, respectively.The standard deviation of δ 13 C TC based on the triplicate analyses of the filter sample were 0.03 ‰ and its analytical error was 0.1 %.In order to remove carbonate carbon from the dustenriched filter samples, aliquots of the samples were treated with HCl fumes by the method described by Kawamura et al. (2004) and Kundu et al. (2010).Each filter cut from the dust-enriched samples was placed in a 50 ml glass vial and exposed to the HCl fumes overnight in a glass desiccator (10 l).The HCl treated filter samples were then analyzed for TC mass and δ 13 C TC as described above.
The δ 13 C values of the water-soluble fraction of the pollen samples were also analyzed in this study as follows.Aliquots of the pollen samples were extracted with the Milli-Q water under ultrasonication (5 min × 3 times) and then passed through a disk filter (Millipore, Millex-GV, 0.45 µm) to remove water-insoluble suspended particles and filter debris.After concentrating the filtered water extracts using a rotary evaporator, the samples were applied with a micro glass syringe to the prebaked quartz filters and then dried overnight using silica gel in a glass desiccator.Finally, the prepared samples were analyzed for δ 13 C TC using the same technique used for the bulk filter sample analysis.The blank level of TC mass was less than 2.0 % of the measured mass.The δ 13 C TC values reported here were corrected against the blank using isotope mass balance equations (Turekian et al., 2003).

Stable carbon isotopic compositions of the major dicarboxylic acids and related compounds
The δ 13 C values of the major dicarboxylic acids and citric acid were measured using the method developed by Kawamura and Watanabe (2004).Diacids and citric acid in the TSP samples were reacted with 14 % BF 3 in 1-butanol at 100 • C for 60 min to form butyl esters (Kawamura, 1993;Jung and Kawamura, 2011).After an appropriate amount of internal standard (n-alkane C 13 ) was spiked to an aliquot of the derivatives, 13 C/ 12 C ratios of the esters were measured using a GC (Hewlett-Packard, HP6890) interfaced to the irMS.Peak identification was performed by comparing the GC-irMS retention times with those of authentic standards.Identification of the esters was also confirmed by mass spectra of the sample using a GC-mass spectrometry (Thermo Trace MS) system (Jung and Kawamura, 2011).The δ 13 C values of the esters relative to the PDB standard were calculated from 13 C/ 12 C ratio of the C 13 standard and its δ 13 C.The internal standard was purchased from the WAKO Chemical Co. and its δ 13 C is −27.24 ‰ (Kawamura and Watanabe, 2004).The δ 13 C values of free organic acids were then calculated by an isotopic mass balance equation using the measured δ 13 C of the derivatives and the derivatizing agent (1-butanol) (Kawamura and Watanabe, 2004).Each sample was analyzed in duplicate, and the average δ 13 C values of the quantified compounds are reported.Prior to actual sample analysis, we confirmed that the δ 13 C values of the working standards (a mixture of normal C 16 -C 30 alkanes) were equivalent to the theoretical values within an analytical error of 0.2 ‰.Around 0.55-2.83ng of the working standards were injected to the GC-irMS.The working standards were purchased from the biogeochemical laboratories at Indiana University and their δ 13 C ranged from −33.24 ‰ to −28.49 ‰ (http://php.indiana.edu/∼ aschimme/n-Alkanes.html).In this paper we report δ 13 C values for oxalic and citric acids.
3 Air mass backward trajectories and aerosol optical thickness (AOT) Air mass backward trajectories and satellite aerosol optical thickness (AOT) can be utilized to characterize potential source regions and transport pathways of air masses.Air mass backward trajectories that ended at the measurement site were computed for 500 m height above ground level using the HYSPLIT (HYbrid Single-Particle Lagrangian Trajectory) backward trajectory analysis (Draxler and Rolph, 2011;Rolph, 2011).All calculated backward trajectories extended 96 h backward with a 1-h interval.Up to 20 % errors of the traveled distance are typical for those trajectories computed from analyzed wind fields (Stohl, 1998).Thus, calculated air mass pathways indicate the general airflow pattern rather than the exact pathway of an air mass.Because a longrange transported haze layer was frequently observed between ∼0.2 km and 3 km elevation over the Korean peninsula (Noh et al., 2009;Yoon et al., 2008), even though aerosol sampling was conducted at ∼3 m height above the ground level, backward trajectories that ended at 500 m height were used in this study by assuming complete vertical mixing below a boundary layer height.Because the Gosan site is located at the western edge of Jeju Island and there are no high mountains within several kilometers of the site, local geographical conditions rarely affect the backward trajectory calculation.
AOT values retrieved by the new version V5.2 of the NASA MODIS (Moderate Resolution Imaging Spectroradiometer) algorithm, called Collection 005 (C005) (Levy et al., 2007a, b), were used in this study.AOT data that are part of the MODIS Terra/Aqua Level-2 gridded atmospheric data product are available on the MODIS web site (http: //modis.gsfc.nasa.gov/).Remer et al. (2005) reported the associated errors of MODIS AOT with ±(0.05 + 0.15 • AOT) and ±(0.03 + 0.05 • AOT) over land and ocean, respectively.
Aerosol Ångström exponent (α) calculated from the AOT values at 440 and 870 nm measured by a sunphotometer were obtained from the Gosan AERONET site (http://aeronet.gsfc.nasa.gov).The α represents the wavelength (λ) dependence of AOT (=−dlogAOT/dlogλ).A small α indicates the presence of particles with a large size, and vice versa.In order to estimate α values during cloudy days of the sampling periods at the Gosan site, α values were also obtained from the Gwangju AERONET site ( 126• 50 E, 35 • 134 N), which is located ∼200 km north of the Gosan site.During the episodic periods, excellent agreement of the α values was observed between the two sites with the regression slope of 1.04 (R 2 = 0.95).Thus, we employed surrogate α values obtained from the Gwangju AERONET site for the cloudy days of the sampling periods (KOS606, 608, and 751) at the Gosan site (Table 1).Cloud-screened and quality-assured Level 2.0 sunphotometer data determined by the AERONET algorithm (Dubovik and King, 2000) were used in this study.

Categorization of the carbon emission episodes
The frequency distribution of TC mass concentrations at 2 µgC m −3 increments is shown in Fig. 2 with a peak value in the rage of 6-8 µgC m −3 .Gaussian fit of the frequency distribution showed a peak center at 7.2 µgC m −3 with the width of 3.1 µgC m −3 , representing background TC mass distributions during the entire sampling periods.The Gaussian fit was clearly separated from the total TC distribution with a threshold value of 10 µgC m −3 (Fig. 2).Thus, this study defined the carbon episode as an average mass concentration of TC >10 µgC m −3 .Three carbon episodes such as long-range transported pollutants (LTP), Asian dust accompanying with LTP (AD + LTP), and local pollen episodes were observed as marked in Fig. 3 and summarized in Tables 1-2.
particles and anthropogenic pollutants (Tables 1-2).Air mass backward trajectories during the AD + LTP episodes clearly showed that dust particles mainly originated from the Nei Mongol desert in China and were transported across the Yellow Sea to the measurement site (Fig. 4b).
Pollen episodes, which were mainly caused by pollen from Japanese cedar trees planted around tangerine farms, were observed at the Gosan site during mid-to late April of 2007 and 2008 (Jung and Kawamura, 2011).Jung and Kawamura (2011) reported the enhanced concentrations of citric acid, which may be directly emitted from tangerine fruit during the pollen episodes, likely adsorbed on the pollen from Japanese cedar trees and then transported to the Gosan site.Mass concentrations of citric acid during the pollen episodes (range: 20-320 µg m −3 ) were several dozen times higher than the LTP (range: 0.64-9.2µg m −3 ) and AD + LTP episodes (range: 3.4-5.8µg m −3 ) and non-episodes (range: 0.17-18 µg m −3 ).Identification of pollen episodes was based on daily human observation of pollen blowing and the microscopic image of pollens collected in the TSP samples (Jung and Kawamura, 2011).A total of 8 pollen-enriched TSP samples were collected during 11-23 April 2007 and 16-24 April 2008 (Table 1).Air mass backward trajectories that ended at the measurement site were computed for 500 m height above ground level.Yellow trajectories represent the KOS606 and KOS614 sampling periods while red ones represent the KOS621 and KOS623 sampling periods.See Table 1 for specifics of each sampling period.
Temporal variations of mass concentrations of citric acid and stable carbon isotopic composition during the pollen episodes showed gradual transition from the non-episodes to the pollen episodes (Fig. 3).Average mass concentrations of NO − 3 and SO 2− 4 during the pollen episodes were ∼2-3 times lower than those during the LTP and AD + LTP episodes (Table 2), implying that a relatively low impact of anthropogenic emissions from the Asian continent during the pollen episodes.Since it was cloudy during most of the pollen episodes (Jung and Kawamura, 2011), only a few MODIS images are available.MODIS AOT and α values during the selected day (19 April 2008) of the pollen episodes showed relatively low aerosol loadings over Korean peninsula (Fig. 5i, j).Air mass backward trajectories during the pollen episodes mainly originated from the northern part of China and the western North Pacific Ocean (Fig. 4c).Dominant wind directions during the pollen episodes were northerly and southeasterly.Relatively low wind speed was observed for southeasterly winds with a median of 5.1 m s −1 (range: 0.3-13.4m s −1 ) compared to northerly winds with a median of 7.1 m s −1 (range: 0.7-20.5 m s −1 ).
Average mass concentrations of TC during the nonepisodes (avg.7.1 ± 1.7 µgC m −3 ) were ∼2 times lower than those during the carbon episodes (Table 2).Average mass concentrations of nss-Ca 2+ during the non-pollen episodes (avg.1.4 ± 0.76 µg m −3 ) were ∼5 times lower than those during the AD + LTP episodes (avg.7.5 ± 0.2 µg m −3 ).Mass concentrations of citric acid during the non-episodes (range: 0.17-18 µg m −3 ) were several dozen times lower than www.atmos-chem-phys.net/11/10911/2011/ the pollen episodes (range: 20-320 µg m −3 ) and almost no pollen grains were observed from the microscopic image of the TSP samples.These results indicated that airborne pollens and dust particles rarely had an impact on the TSP samples during the non-episodes.Air mass backward trajectories during the non-episodes mainly originated either from the northern part of China or the western North Pacific Ocean (Fig. 4d), indicating that anthropogenic pollutants emitted from the eastern part of China had rarely impacted on the TSP samples during the non-episodes.

TC and TN mass concentrations during the carbon episodes
Similar TC mass concentrations were obtained during the three carbon episodes, with averages ranging 15 ± 6.0 µgC m −3 to 16 ± 6.7 µgC m −3 (Table 2).Similar amounts of OC and EC as well as TC during the LTP and AD + LTP episodes as shown in Table 2 indicated that the AD + LTP episodes in spring are frequently accompanied with anthropogenic pollutants emitted from the industrial regions of East China and Northeast China.However, relatively high mass concentrations of TN (avg.11 ± 8.2 µgN m −3 ) were obtained during the LTP episodes, followed by the AD + LTP episodes (avg.7.7 ± 4.0 µgN m −3 ) and the pollen episodes (avg.4.8 ± 1.5 µgN m −3 ).Much higher TC/TN mass concentration ratios (avg.3.5 ± 1.2, range: 1.8 to 5.3) were observed during the pollen episodes than in those during the LTP episodes (avg.1.6 ± 0.58, range: 0.96 to 2.3) mainly due to the enhanced organic carbon mass of the airborne pollen.The highest OC/EC ratios (avg.5.8 ± 2.6, range: 3.8 to 12) were also obtained during the pollen episodes, confirming the enhanced organic carbon mass of the airborne pollen.
Excellent correlation (R 2 = 0.95) was obtained between mass concentrations of TN and TC during the LTP periods (Fig. 6a), implying similar sources of TN and TC.A strong correlation between mass concentrations of TN and TC was also obtained during the LTP plus non-episodic periods (R 2 = 0.82) (Fig. 6b), suggesting that aerosols during the non-episodic periods in spring might be influenced by the LTP aerosols from the Asian continent.However, poor correlations of TN versus TC mass concentrations were obtained during the pollen episodes (Fig. 6a).Even though similar levels of TN were obtained during the pollen episodes, TC values were highly variable, ranging 7.5 to 28 µgC m −3 mainly due to different strengths of pollen transport and different meteorological conditions (Doskey and Ugoagwu, 1989;Puc and Wolski, 2002;Palacios et al., 2007).TC and TN mass concentrations obtained during the LTP plus nonepisodic periods are compared to those from previous studies in the Asian continent (Fig. 6b).TC and TN concentrations were obtained from the Hua mountain site in China in winter (size cut: PM 10 , sampling flow rate: 100 l min −1 , integration time: 10 h) (Li et al., 2011) and three cities in China; TN and TC values on the Asian continent were obtained from urban areas of China: Shanghai in winter and spring (size cut: PM 2.5 , sampling flow rate: 0.4 l min −1 , integration time: weekly) (Ye et al., 2003), Nanjing in winter (PM 2.5 , 1110 l min −1 , 12 h) (Yang et al., 2005), Baoji in spring (PM 10 , 100 l min −1 , 8 h) (Wang et al., 2010) as well as the Hua mountain site in winter (PM 10 , PM 10 , 100 l min −1 , 10 h) (Li et al., 2011).The analytical errors for TC and TN mass concentrations at the Gosan site were 2.3 % and 5.2 %, respectively.
Using the regression slope of TN versus TC mass concentrations during the LTP plus non-episodic periods that may represent normal atmospheric condition at the Gosan site in spring, the contribution of the airborne pollen carbon in TC can be roughly calculated as follows; TC pollen (µgC m −3 ) = TC(µgC m −3 ) −(0.83 • TN (µgN m −3 ) + 3.79) (1) The contribution of the airborne pollen carbon to TC was roughly estimated to be 20 % to 71 % with an average of 46 ± 19 %.
TC/TN mass concentration ratio (3.9) in the strong AD + LTP episode sample (KOS603) was much higher than that (1.2) in the weaker AD + LTP episode sample (KOS627) (Fig. 6a).The enhanced mass concentration of TC in the KOS603 sample may be in part attributed to the presence of carbonate carbon in dust particles, which wasn't removed completely via the reaction with nitrogen dioxide (NO 2 ), nitric acid (HNO 3 ), and sulfur dioxide (SO 2 ) (Zhang et al., 1994;Mamane and Gottlieb, 1989;Underwood et al., 2001) during the long-range atmospheric transport.The enhanced carbonate carbon in TC during the severe AD + LTP episode will be discussed in detail in Sect.4.3.

Stable carbon isotopic compositions of TC during the carbon episodes
The δ 13 C values of TC (δ 13 C TC ) during the carbon episodes are plotted as a function of TC mass concentrations in Fig. 7a and are summarized in Table 2.Because TC mass weighted average δ 13 C TC values during the different episodes agreed within 3 ‰ to the arithmetic means of δ 13 C TC , we used the arithmetic means when comparing different episodes.The δ 13 C TC values during the LTP episodes ranged from −23.5 ‰ to −23.0 ‰ with an average of −23.3 ± 0.3 ‰.Similar δ 13 C TC values were observed regardless of TC mass concentrations during the LTP episodes (Fig. 7a).However, the δ 13 C TC values during the pollen episodes were more negative, ranging from −26.2 to −23.5 ‰ with an average of −25.2 ± 0.9 ‰.The δ 13 C TC values for the AD + LTP episodes are relatively high, ranging −23.3 to −20.4 ‰ with an average of −21.8 ± 2.0 ‰.These results suggested that δ 13 C TC can be utilized as an indicator of the possible sources of the carbon episodes at the Gosan site, with relatively low values during the pollen episodes compared to those during the LTP and AD + LTP episodes.The δ 13 C TC during the strong AD + LTP episode (KOS603) was higher than during the weaker AD + LTP episode (KOS627) as shown in Fig. 7a.In order to quantify the effect of carbonate in dust particle on the TC mass and δ 13 C TC measurements, the HCl fume treated filter samples during the AD + LTP episodes were also analyzed for TC mass and δ 13 C TC .It was clearly observed that the TC mass concentration and δ 13 C TC in the KOS603 sample decreased from 18.8 µgC m −3 to 14.7 µgC m −3 and from −20.4 ‰ to −22.1‰ after the HCl fume treatment.However, those in the KOS627 sample were invariant before and after the HCl treatment (Fig. 7b).The removed carbon (RMD-C) that was calculated from the difference between TC mass concentrations before and after the HCl fume treatment was found to be 4.1 µgC m −3 for the KOS603 sample.
In order to characterize the carbonate carbon in the RMD-C, the δ 13 C of the RMD-C, (δ 13 C RMD−C ) was calculated using the isotopic mass balance equation as follows (Kawamura et al., 2004); where REM-C represents the remaining carbon after the HCl fume treatment.Isotope equilibrium exchange reactions within the inorganic carbon system "atmospheric CO 2 -dissolved bicarbonate -solid carbonate" lead to an enrichment of 13 C in carbonates (Hoefs, 1997).Thus, carbonate carbon is isotopically heavy, with δ 13 C values of around 0 ‰ (Hoefs, 1997;Kawamura et al., 2004).The δ 13 C RMD−C in the KOS603 sample was calculated as −14.1 ‰.This value was higher than that of the REM-C while much lower than those (−1.3 to −0.3 ‰) of the standard Asian dust samples collected from the Gunsu Province, China (Kawamura et al., 2004), indicating that not only carbonate carbon but also volatile organic acids adsorbed on aerosol particles were removed by the HCl fume treatment.Kawamura et al. (2004) suggested that low molecular weight organic acids such as formic, acetic, and oxalic acids are possible candidates for the removed volatile and semi-volatile organic acids.By assuming that the δ 13 C of the removed organic acids has the same δ 13 C of the REM-C, carbonate carbon in the KOS603 sample was roughly estimated as 1.5 µgC m −3 (8 % in TC) using the isotopic mass balance Eq. ( 2) and the δ 13 C of the removed carbonate of −0.3 ‰ by Kawamura et al. (2004).Calcium carbonate in dust particle reacts with nitrogen dioxide (NO 2 ), nitric acid (HNO 3 ), and sulfur dioxide (SO 2 ) to produce calcium nitrate and calcium sulfate.(Zhang et al., 1994;Mamane and Gottlieb, 1989;Underwood et al., 2001).The presence of carbonate in the KOS603 sample indicated insufficient amounts of gas phase NO 2 , HNO 3 , and SO 2 to remove all carbonate during the long-range atmospheric transport.The much higher TC/TN ratio during the strong AD + LTP episode (KOS603) was also attributed to the remaining carbonate carbon.However, the negligible amount of carbonate in the KOS627 sample indicated that most of its carbonate was removed via the reaction with NO 2 , HNO 3 , and SO 2 during its long-range atmospheric transport.The relatively high TN mass in the KOS627 sample also supported the efficient removal of carbonate via reaction with HNO 3 .

δ 13 C values of TC in airborne pollen and tangerine fruit
A much lower δ 13 C TC value was obtained in the Pollen Gosan (−28.0 ‰) than in the authentic standard pollens; −25.4 ‰ for the Pollen cedar and −23.3 ‰ for the Pollen cypress (Fig. 8 and Table 3).The δ 13 C TC in the Pollen Gosan was slightly lower than the average value (−26.8 ‰) of 174 different species of pollen samples from C 3 plants across the US (Jahren, 2004).These differences may be attributed to the geographical difference of the pollen samples.Jahren (2004) reported that the δ 13 C TC values in the pollens from C 3 plants varied from −30.2 ‰ to −24.5 ‰ depending on their geographical locations and the types of C 3 plants.The Pollen Gosan was produced on Jeju Island, Korea while the authentic standard pollens were produced on Japan (personal communication to Wako Chemical Co.).Thus, the differences of δ 13 C TC values between the Pollen Gosan and the authentic standard pollens may be explained by the different geographical conditions.Interestingly, the δ 13 C TC in the tangerine peel (−28.1 ‰) was very similar to that in the Pollen Gosan (Fig. 8 and Table 3).A slightly higher δ 13 C TC value (−27.0 ‰) was obtained in the tangerine juice than in the tangerine peel.The δ 13 C values of the watersoluble fraction of the pollens were found to be −24.7 ‰, −26.1 ‰, and −25.0 ‰ for the Pollen Gosan , Pollen cedar , and Pollen cypress , respectively (Fig. 8 and Table 3).The δ 13 C TC value for the water-soluble fraction of the Pollen Gosan was higher than that of the bulk Pollen Gosan .However, the δ 13 C values of the water-soluble fraction in the Pollen cedar and Pollen cypress were slightly lower than the bulk pollen (Table 3).These results imply that the water-soluble fraction of the Pollen cedar and Pollen cypress might originate from different sources and be adsorbed to the pollens.
In order to estimate the contribution of the airborne pollen carbon to aerosol TC during the pollen episodes, the carbon masses derived from the airborne pollen were determined using the isotopic mass balance equation in Eq. ( 2).Since most of aerosols at the Gosan site in spring were influenced by the long-range transport of anthropogenic aerosols Table 3. Carbon isotope compositions of TC and oxalic and citric acids in selected samples during the pollen episodes, pollens, and tangerine fruit. 1 Pollen Gosan represent airborne pollens separated from the KOS751 sample.
2 WS represents water-soluble fraction of sample.
3 Pollen cedar and Pollen cypress represent authentic standard pollens from Japanese cedar and Japanese cypress, respectively. 4The analytical error for δ 13 C of TC measurement was 2.3 %. 5 Average ± Standard deviation.Oxalic acid: (COOH) 2 , Citric acid: from the Asian continent as discussed in Sect.4.2, we assumed that carbonaceous particles during the pollen episodes were mainly from the airborne pollen and the long-range transported organic pollutants.Thus, the δ 13 C TC in the Pollen Gosan (−28.0 ‰) and average δ 13 C TC during the LTP episodes (−23.3 ‰) were used as two end members to calculate the airborne pollen carbon mass (TC pollen ).The TC pollen concentrations were determined to be 0.4 to 16.7 µgC m −3 (4 to 62 % in TC) with a median of 5.1 µgC m −3 (42 %) during the pollen episodes (Fig. 9).The median value of the TC pollen fraction in TC was quite similar to that obtained using the TN and TC regression approach in Eq. ( 1).Thus, it was found that ∼42 % of TC in the TSP samples at the Gosan site could be attributed to the airborne pollens during the pollen episodes.The rest of TC can be explained by long-range transported anthropogenic OC and EC.Local pollen can enhance the mass of the organic aerosols and may overestimate the relevant radiative forcing at the Gosan site.These results can provide useful information for accurately qualifying and quantifying the impact of the long-range transported pollutants from the Asian continent in spring.

Enhanced mass concentrations of citric acid and their sources during the pollen episodes
The cultivating area of tangerines in Jeju Island is ∼209 km 2 , which accounts for ∼11 % of the total area of the island.In order to dissipate the strong winds from the Pacific Ocean, all tangerine farms are surrounded by Japanese cedar trees.
Pollen in the air on Jeju Island in April was mainly from the Japanese cedar trees (Agricultural Research Institute in Jeju special self-governing province, personal communication, 2011).Because tangerines are widely cultivated in the coastal areas of Jeju Island, the distance of the nearest tangerine farms to the sampling site is between several hundred meters to several tens of kilometers, depending on the wind direction.Jung and Kawamura (2011) measured the elevated mass concentrations of atmospheric citric acid (range: 20-320 ng m −3 ) in the TSP samples during the pollen episodes.They postulated that citric acid that may be directly emitted from tangerine fruits was likely adsorbed on pollens emitted from Japanese cedar trees planted around tangerine farms and then transported to the Gosan site.In order to track the source and transport mechanism of citric acid, δ 13 C TC values during the pollen episodes were plotted as a function of the fraction of citric acid carbon (citric acid-C) in TC mass (Fig. 10).It was clearly evident that the δ 13 C TC decreased as citric acid-C/TC mass ratios increased.Because the airborne pollen showed much lower δ 13 C TC values than the LTP particles (Tables 2-3), the decrease of δ 13 C TC with an increase of citric acid-C/TC ratio demonstrates an increased contribution of airborne pollen to aerosol TC.These results indicated the positive correlation between citric acid and airborne pollen concentration, suggesting that citric acid emitted from tangerine fruit might be adsorbed on airborne pollen and then transported to the Gosan site.Divergence of the δ 13 C TC values at a certain level of the citric acid-C/TC ratios as shown in The δ 13 C values of oxalic and citric acids in the selected samples during the pollen episodes, authentic standard pollens, and tangerine fruit are shown in Table 3.The δ 13 C of citric acid (−27.4‰) in the tangerine peel was similar to the δ 13 C TC in the tangerine peel (−28.1 ‰) and the Pollen Gosan (−28.0 ‰).However, the δ 13 C values of citric acid in the Pollen Gosan (−26.3 ‰) and KOS751 samples (−25.8 ‰) were slightly higher than that in the tangerine peel.The pollen-enriched TSP samples were collected ing spring of 2007 and 2008 whereas tangerine fruit was produced during early winter of 2010.Thus, these differences can be partially explained by seasonal and annual variations of δ 13 C of citric acid in the tangerine peel.The δ 13 C values of n-alkanes for individual lipids from the leaves of Quercus castaneifolia showed ∼2.5 and 5.2 ‰ differences for C 29 and C 31 n-alkanes, respectively, in autumn leaves compared with leaves sampled at the start of the growing season (Lockheart et al., 1997).Thus, the seasonal difference further supports the assumption that citric acid in the Pollen Gosan and KOS751 samples originated from the tangerine peel (tangerine fruit) and then were transported to the Gosan site after adsorbing to the pollen from Japanese cedar trees planted around tangerine farms.
The elevated δ 13 C values of oxalic acid were obtained for the authentic standard pollen samples; −5.0 ‰ for the Pollen cedar and 1.0 ‰ for the Pollen cypress .These high δ 13 C values of oxalic acid can be explained by the adsorption of aged oxalic acid on pollen before becoming air borne.The δ 13 C value of oxalic acid in the Pollen Gosan was higher than those in the tangerine peel and the KOS751 sample.Jung and Kawamura (2011) reported similar amounts of oxalic and citric acids in the tangerine peel, suggesting that not only aged oxalic acid but also directly emitted oxalic acid from the tangerine peel may be adsorbed on the Pollen Gosan and transported together.

Thermal evolution pattern of OC during the carbon episodes
Thermograms of OC and EC analysis for the carbon episode samples are shown in Fig. 11 and the results are summarized in Table 4. Unique evolution patterns of OC1 and OC2 were obtained depending on the types of the carbon emission episodes.However, OC3 and OC4 showed similar evolution patterns among the carbon episodes as shown in Table 4; OC3 fractions in total OC = avg.14 % to 20 % and OC4 fractions = avg.10 % to 19 %.A similar temperature dependent EC evolution was obtained during the carbon episodes.
Around 93 % of EC evolved at oven temperatures <700 • C; 39 % in EC1, 36 % in EC2, and 19 % in EC3 temperature steps.Even though sharp increase of the OC4 peak was not observed in Fig. 11 for all samples, the OC4 fractions were similar to the OC3 fractions as seen in Table 4.This discrepancy was attributed to the broad evolution of OC in the OC4 temperature step as shown in Fig. 11.Thermal evolution patterns of OC during the LTP episodes were subdivided to two groups: relatively higher OC evolution in the OC1 temperature step (24 ± 2 % in total OC) than the OC2 temperature step for the KOS606, 614, and 619 samples and similar OC evolution in the OC1 (16 ± 2 %) and OC2 temperature steps (21 ± 2%) for KOS621 and KOS623 samples.These subdivision coincided with the relatively high α values for KOS606, 614, and 619 samples (avg.1.28-1.40)relative to those for the KOS621 and KOS623 samples (avg.0.65-0.86)(Table 1).Air mass trajectories during the LTP episodes showed that air masses from East China impacted more on the KOS606, 614, and KOS619 samples (yellow and white trajectories in Fig. 4a).
OC2 (21 %) was much higher than OC1 (8 %) during the strong AD + LTP episode (KOS603) whereas similar amounts of OC1 (19 %) and OC2 (18 %) were obtained during the weaker AD + LTP episode (KOS627).Thermal evolution patterns of OC during the pollen episodes were clearly characterized by higher OC evolution in the OC2 temperature step (29 ± 6%) (Fig. 11b and Table 4).Mass concentrations of OC1 and OC2 in the TSP samples collected during the pollen episodes correlated well with R 2 of 0.77 (Fig. 12).Miyazaki et al. (2007) reported that medium molecular weight (>200 g mol −1 ) water-insoluble organic species such as nonacosane, docosanol, and hexadecanoic acid evolved mostly in the OC2 temperature step, with small fractions evolving in the OC3 temperature step.Sucrose has been identified as the major water-soluble organic compound in 15 pollen species tested by Hoekstra et al. (1992).91 % of the OC in sucrose compounds was evolved at the OC2 temperature step (Miyazaki et al., 2007).Thus, it was suggested that major fractions of the pollens collected at the Gosan site may have medium molecular level organic compounds.Thermal evolution patterns of OC during the LTP episodes showed   relatively constant OC2 mass concentrations but strongly variable OC1 mass concentrations (Fig. 12).Interestingly, PC fractions during the LTP episodes negatively correlated with OC2 fractions while positively correlated with OC1 fractions (Table 4).The thermal evolution pattern of OC during the strong AD + LTP episode (KOS603) was quite different from that during weaker AD + LTP episodes (KOS627).The OC2/OC1 mass ratio (2.7) during the strong AD + LTP episode showed much a higher value than that during the weaker AD + LTP episodes (1.0).Additionally, more PC was formed in the sample for the strong AD + LTP episode (43 %) than that for the weaker AD + LTP episodes (30 %).Miyazaki et al. (2007) reported that lower molecular weight (<∼180 g mol −1 ) water-soluble organic compounds evolved mostly at the OC1 temperature step while a higher oven tem- perature was needed for the higher molecular weight fraction of the organic aerosol, resulting in increased OC2 and OC3 fractions.Lim et al. (2010) suggested that organic acids are dominantly formed in cloud processing, whereas large multifunctional humic-like substances are dominantly formed in wet aerosols via radical-radical reactions.Thus, different evolution patterns of OC obtained for the LTP and AD + LTP episodes can be explained by different formation mechanisms of secondary organic aerosols and the effect of aging of organic aerosols during long-range atmospheric transport.Different sources of organic aerosols from the Asian continent may also contribute to the different thermal evolution patterns of OC.

Thermally resolved OC component versus stable carbon isotopic composition
The relations between the OC fractions and δ 13 C TC during the LTP and AD + LTP episodes are also examined.
In contrast to the pollen episodes, almost no correlations were obtained between the OC fractions and δ 13 C TC except for the OC1 fraction that gave a moderate correlation of R 2 = 0.59 (data are not shown).The moderate correlation between the OC1 fraction and δ 13 C TC was attributed to the elevated δ 13 C TC during the strong AD + LTP episode (KOS603).Thus, if we exclude the δ 13 C TC in the KOS603 samples, no correlation was obtained between the OC1 fraction and δ 13 C TC .In order to investigate the dependence of δ 13 C TC on thermally evolved OC fractions, the fractions of OC evolved at each temperature step in total OC during the pollen episodes are plotted as a function of δ 13 C TC in Fig. 13a, b.The OC1 and OC2 fractions normalized by total OC mass showed good correlations with δ 13 C TC with R 2 of 0.81 and 0.73, respectively, during the pollen episodes (Fig. 13a) whereas a moderate correlation was observed between OC2 mass concentration and δ 13 C TC with R 2 = 0.48 (Fig. 13b).However, almost no correlation was observed between the OC1 mass concentration and δ 13 C TC (Fig. 13b).
Because the Pollen Gosan had low δ 13 C TC of −28.0 ‰ (Table 3), the negative correlations of the normalized OC2 mass fraction and OC2 mass concentration with δ 13 C TC indicated that a large fraction of the pollen-enriched TSP samples evolved at the OC2 temperature step (300-450 • C).Almost no correlation between OC1 mass concentrations and δ 13 C TC indicated that the positive correlation between the normalized OC1 fraction and δ 13 C TC was mainly attributed to a relative decrease in OC1 fraction to total OC as OC2 fraction increases during the pollen episodes.A positive correlation between mass concentrations of OC1 and OC2 (Fig. 12) and a negative correlation between the normalized OC1 and OC2 fractions (Fig. 13) during the pollen episodes imply that a small fraction of pollen also evolved in the OC1 temperature step but dominant fractions of them evolved in the OC2 temperature step.

Summary and conclusion
Satellite remote sensing, air mass backward trajectories, and particulate chemical and stable carbon isotopic composition analyses allowed us to categorize the carbon emission episodes observed at the Gosan background super-site (33.17 • N, 126.10 • E) in East Asia during spring of 2007 and 2008 as long-range transported anthropogenic pollutants (LTP) from the Asian continent, Asian dust accompanying LTP (AD + LTP), and local pollen episodes.Carbon episodes were defined in this study as mass concentration of TC > 10 µgC m −1 .Subdivision of the carbon episodes was conducted as follows.Pollen episodes were identified by the elevated concentrations of citric acid and pollen in the TSP samples as described by Jung and Kawamura (2011).LTP episodes were identified by the elevated concentrations of nitrate and sulfate during the carbon episodes.AD + LTP episodes were identified by the elevated concentrations of calcium ion (Ca 2+ ) and low aerosol Ångström exponent as well as the elevated concentrations of nitrate and sulfate during the carbon episodes.The carbon episodes caused by the pollen episodes were characterized by the lowest δ 13 C of TC (δ 13 C TC ) (avg.−25.2 ± 0.9 ‰), followed by the LTP episodes (avg.−23.3 ± 0.3 ‰) and the AD + LTP episodes (avg.−21.8 ± 2.0 ‰).Using the HCl fume treatment on the dust-enriched samples, we found that ∼8 % in total carbon (TC) during the strong AD + LTP episode (KOS603 sample) was attributed to carbonate carbon that was not removed via the reaction with the acidic gases such as www.atmos-chem-phys.net/11/10911/2011/nitrogen dioxide, nitric acid, and sulfur dioxide during the long-range atmospheric transport, resulting in higher δ 13 C TC .
Thermal evolution patterns of OC during the pollen episodes were clearly characterized by higher OC evolution in the OC2 temperature step (450 • C).Different evolution patterns of OC obtained for the LTP and AD + LTP episodes can be explained by different formation mechanisms of secondary organic aerosols and the effect of aging of organic aerosols during the long-range atmospheric transport.Different sources of organic aerosols from the Asian continent may contribute to the different thermal evolution patterns of OC.
Based on the carbon isotope mass balance equation, we found that during the pollen episodes ∼42 % of TC was attributed to airborne pollen emitted from Japanese cedar trees planted around tangerine farms in Jeju Island.The negative correlation between the citric acid-carbon/TC ratios and δ 13 C TC and similar δ 13 C values of citric acid between the airborne pollens (−26.3 ‰) collected at the Gosan site and tangerine fruit (−27.4 ‰) imply that citric acid emitted from tangerine fruit may be adsorbed on the airborne pollen and then transported to the Gosan site.

Table 1 .
Specifics of sampling, PM 10 mass concentration, non-sea-salt calcium ion (nss-Ca 2+ ), and Ångström exponent at the Gosan site during the carbon episodes in spring of2007 and 2008.
1 AD + LPT: Asian dust plus long-range transported pollution episode, Pollen: pollen episode. 2 Average (Maximum).3Ångströmexponents during the KOS606, 608, and 751 sampling periods were obtained from the Gwangju AERONET site which is located at ∼200 km north of the Gosan site.

Table 2 .
Concentrations of chemical compositions and stable carbon isotopic composition (δ 13 C) in the total suspended particle (TSP) samples collected at the Gosan site during the carbon episodes.

Table 4 .
Average mass fractions of OC evolved at each temperature step in total OC and OC2/OC1 mass ratios according to categorized carbon episodes.OC3, and OC4 represent the carbon evolved at a temperature of 300 • C, 450 • C, 600 • C, and 650 • C, respectively. 2 represents the pyrolized organic carbon during the OC analysis mode in an inert atmosphere (pure Helium).3LTPcase 1 includes KOS606, 614, and 619 samples with relatively high Ångström exponent >1.2. 4 LTP case 2 includes KOS621 and KOS623 samples with relatively low Ångström exponent <1.0.