The measurement results for the whole experiment are shown in Figure 2, in which all measurements were hourly averaged. Most of gas and aerosol species were noticeably enhanced in several events. Throughout the campaign, two dust events accompanied by pollution plumes were observed during 17–18 and 28–29 March. There was another incident identified as an anthropogenic plume on the 22–23 March that was not affected by dust storm. These three episodes shaded in Figure 2, were all associated with frontal passages, during which both gas and aerosol concentrations were remarkably enhanced. In this study, it is noteworthy that dust plumes were clearly distinguished from pollution plumes by Ca2+ concentrations measured every hour. In addition to these major events, concentrations of major species were shown to be elevated by the influence of nearby continents on 15–16 and 21 March, even though these were not accompanied with frontal system. Thus the chemical characteristics of aerosols during the ABC-EAREX 2005 were mainly determined by these events. During the whole course of experiment from 13 to 30 March, the mean concentrations of major species were 56 ppbv for O3, 297 ppbv for CO, 3.2 ppbv for SO2 and 2.2 ppbv for NOx. The average mass concentration of PM2.5 was 27.3 μg/m3. Among PM2.5 constituents, OC and nss-SO42− were the most abundant, of which means were 4.22 μg/m3 and 3.34 μg/m3, respectively. Ammonium and nitrate concentrations were less than that of nss-SO42− and their mean concentrations were 1.38 μg/m3 and 1.14 μg/m3, respectively. In addition, the mean of EC concentrations was 1.00 μg/m3, which was about a factor of four lower than that of OC. Other soluble ions such as Na+, Cl−, Ca2+, Mg2+, sea-salt SO42−, and K+ were below 1.00 μg/m3 on average.
3.1. PM 2.5 Chemical Characteristics
 To examine the general characteristics of aerosol compositions during the ABC-EAREX 2005, aerosol constitutions were divided into four groups: soluble ions, organic matter (OM), EC, and the unidentified. Soluble ions include SO42−, NO3−, NH4+, Na+, Cl−, K+, Ca2+, and Mg2+. The OM content was estimated to be OC concentration times 2.1, which was a conversion factor for nonurban aerosols [Turpin and Lim, 2001]. The “unidentified” is the rest of PM2.5 that was not included in three categories. The concentrations of these four groups were daily averaged from 0 to 23 h and presented in Figure 3a with PM2.5 mass concentrations. The daily mean of PM2.5 varied from 12.56 to 56.50 μg/m3 and represented well episodic events associated with pollution and dust plumes. The amounts of OM and soluble ions were comparable and the unidentified varied greatly day by day. In general, the greater PM2.5 mass was, the greater amounts were left unidentified. To compare the relative importance of these constituents, the mass fractions of each subgroup against PM2.5 mass were calculated and shown in Figure 3b. The average mass fractions of soluble ions, OM, EC, and the unidentified were 31.3%, 39.7%, 3.9%, and 25.1%, respectively. The soluble ions and organic matters were about equal in contribution to PM2.5 mass concentrations. Since OM contents was driven from OC concentrations using the conversion factor, the fraction of OM to PM2.5 shown in Figure 3b had better be considered as an approximation.
Figure 3. (a) Daily averaged (0–23 h) concentrations of aerosol mass (PM2.5) and its composition divided into four groups: soluble ions, OC, EC, and unknowns. (b) Daily variations of mass fractions against PM2.5 total mass for four subgroups. (c) Daily averaged meteorological parameters: temperature, relative humidity, wind speed, and precipitation measured at Gosan station.
Download figure to PowerPoint
 The variations of PM2.5 mass and their compositions were observed to be closely related to meteorological conditions. If averaged daily, meteorological parameters including temperature, relative humidity, wind speed, and the amount of precipitation exhibited a characteristic variation during the three weeks of experiment, reflecting the passage of cold fronts (Figure 3c). Details about meteorological setting during the campaign can be found in this issue. In Figure 3, PM2.5 and meteorological parameters show some similarity in daily variations, which exemplified the role of frontal system in altering chemical constituents of the atmosphere by carrying out continental outflows.
 PM2.5 mass reached the maximum on 24 March when wind speed was very high and its hourly mean went up to about 30 m/s. While concentrations of precursor gases or other major aerosol constituents stayed low, Cl− and Na+ concentrations were noticeably enhanced. Accordingly, the increase in unknown fraction was apparent, leaving about a half of PM2.5 unidentified. Other high concentrations of PM2.5 were mostly associated with either pollution or dust events with high unidentified fractions. The unknown portions were significantly reduced on 19, 20, 27, and 30 March with low PM2.5 concentrations and low wind speed. In contrast, OM fractions were sharply increased.
 Since trace metal and silicate contents were not measured in this study, a considerable fraction of the unidentified constituents was possibly to be metals and silica. The major source of trace elements is the Earth's crust and accordingly, their concentrations in aerosols used to be highly elevated during dust events [e.g., Choi et al., 2001]. In their study, metal content was as high as 17% of the PM10 mass for heavy dust incident, even though Si, the most abundant crustal material, was not counted. As the median size of dust particles are about 2–3 μm [Chun et al., 2001], mineral aerosols are thought to have less significant influence on mass concentrations of PM2.5 than PM10. During this experiment, the increase in PM2.5 mass was more pronounced in pollution plumes, whereas PM10 concentrations were much higher in association with dust plumes [this issue]. In recent study conducted in Hong Kong area [Ho et al., 2006], the contribution of crustal elements to PM2.5 was not negligible, even though their relative amount were higher in PM10 than in PM2.5. Taking unmeasured trace metals into account, the unidentified fraction are in reasonable agreement with other studies [Ho et al., 2006; Rees et al., 2004]. It is also worthy to mention that the fraction of soluble ions were remarkably decreased on 18, 24, and 29 March, when the wind speeds were very high or Asian dust events were observed. These results illustrated that the chemical composition of fine aerosols were tightly couple with meteorological conditions.
 Since aerosol compositions were determined hourly, their diurnal variations were available to investigate characteristic changes of aerosol composition by time, particularly between day and night. Daily variations of major species and meteorological parameters are shown in Figure 4. While PM2.5 mass reached the maximum in the afternoon, its major constituents showed slightly different patterns in diurnal variations (Figure 4a). The variation of OC was in phase with that of PM2.5. On the other hand, both sulfate ion and EC were enhanced at night. These results highlight the role of OC in determining PM2.5 mass concentrations as the most abundant constituent. It is also interesting to compare diurnal patterns of gases with those of aerosol species (Figure 4b). While variations of both SO42− and EC were similar to that of CO with maximum concentrations at night, variations of O3 and OC looked alike with afternoon peaks. It is noteworthy that the enhancement of SO42− and OC was found in different times, which could have significance in the generation of fine aerosols of the research region. In addition, SO2 and CO showed higher concentrations at night. For NOx, concentrations were increased in the morning that was typical for vehicle source, along with evening peak. Particularly, the concentrations of gases except O3 were enhanced at night, which was concomitant with the shift of wind direction (Figure 4c) and increase in PM2.5 mass and all aerosol species including SO42− and EC. It signifies the importance of transport to controlling PM2.5 mass concentrations.
Figure 4. Diurnal variations of aerosol composition, gaseous species, and meteorological parameters averaged hourly for the whole measurement. (a) PM2.5, nss SO42−, OC, and EC; (b) O3, CO, SO2, and NOx; and (c) temperature, relative humidity, wind direction and speed.
Download figure to PowerPoint
3.2. PM2.5 Chemical Tracers
 Figure 4 illuminated the relationship among precursor gases and aerosol species. For example, nss-sulfate and OC track the variations of CO and O3, respectively. Hence the correlations between compounds including aerosol and gas were investigated. Any good correlation could be useful to identify source or to trace transport path, on which chemical compositions of fine aerosols were largely dependent. CO has been most widely used as a tracer for ozone production in pollution plumes owing to its role as a precursor and relatively long lifetime in the atmosphere. In this study, CO was well correlated with EC for all measurements (Figure 5), which was understood in terms of common source of combustion processes [de Laat et al., 2001]. EC and CO were often found to covary [Chen et al., 2002; Lim and Turpin, 2002]. One of the striking results of this study is a good correlation of a major secondary aerosol such as nss-SO42− with CO, which was comparable to that of EC with CO. It implies that CO can be used to trace the formation of secondary aerosols in anthropogenic plumes.
 In previous study, the ratio of OC to CO was suggested as a useful tracer to identify different sources for fine aerosols [Maria et al., 2003]. The relationship between OC and CO from this study was illustrated in Figure 6, where OC was plotted against CO. Since the lifetime of OC is much shorter than CO, the ratio of CO/OC would decrease as the formation of secondary aerosols, while increasing with being transported away from the source. Although these two processes affect the ratio in opposite directions, the CO/OC ratios were useful to distinguish different sources of fine aerosols over the northeast Asia [Maria et al., 2003]. In this study the average ratio of CO (ppbv) to OC (μg/m3) was estimated to be 35 from the linear regression. Although the correlation of OC with CO was not as good as those of EC or nss-SO42− with CO, the emission ratios of OC to CO are better distinguished among adjacent countries. On the basis of emission inventories of Asia in the year of 2000 [Streets et al., 2003], the ratio of CO to OC is 27 for China and 81 for South Korea, which are presented as dotted lines in Figure 6. All data are within the envelope of these two lines, while the majority of data are scattered along the line of CO/OC ratio for China. Accordingly, the high OC associated with relatively low CO shown in the lower right of Figure 6 were obtained in air transported from the southern part of China. However, the extremely high CO values shown in the upper right near the line representing Korean emission were observed in pollution plumes carried by cold front system from the northwest. These facts address the limitation of applying averaged ratios to single out specific source, even though overall, the measured ratios of CO/OC conveyed the signatures from possible sources of fine aerosols. In addition, it is worthy to mention that the background concentration of surface CO during the ABC-EAREX2005 was about 144 ppbv, estimated from the linear fit shown in Figure 6, which was higher than ∼100 ppbv of the estimated from airborne measurements during the ACE-Asia [Maria et al., 2003].
Figure 6. Correlation of OC with CO for the whole measurements. The solid line in center represents the best fit for measurements (CO/OC slope ≈ 35). The dotted lines denote CO/OC ratios estimated from emission inventory for Asia in the year of 2000 by Streets et al. . The top line corresponds to the emission ratio of South Korea (CO/OC slope ≈ 81), and the line on the bottom corresponds to China (CO/OC slope ≈ 27).
Download figure to PowerPoint
 As a major light absorbing component of the atmospheric aerosols, EC is known to be formed during the combustion of hydrocarbons [Jones and Harrison, 2005; Mader et al., 2002]. OC may derive from a wide variety of sources; combustion processes including fossil fuel combustion and biomass burning, and natural sources such as ocean and terrestrial vegetation [Duce et al., 1983; Matsumoto et al., 2003]. Several studies have used measurements of the OC to EC slope as an indicator of aerosol source types and evidence of secondary organic aerosol formation [Kadowaki, 1990; Kim et al., 2000; Lim and Turpin, 2002]. These analyses utilize the fact that EC is produced only in combustion processes as a primary pollutant while OC is both emitted from sources and produced by atmospheric reactions from gaseous precursors.
 As the first data set obtained at Gosan station, the average OC and EC concentrations of PM2.5 were 4.22 μg/m3 and 1.00 μg/m3, respectively. During ACE-Asia program conducted in March-April 2001, OC and EC contents were measured for TSP (total suspended particles) at Gosan station and for PM1.0 on the ship, which are available for comparison. For TSP collected at Gosan, the mean OC and EC concentrations were 7.00 μg/m3 and 1.29 μg/m3, respectively [Simoneit et al., 2004]. In shipboard measurement, 4.00 μg/m3 of OC and 1.32 μg/m3 of EC were observed near Korea in the East Sea [Lim et al., 2003]. All existing measurements resulted in higher OC/EC ratios than the assessed from emission inventories, which are 1.27 for South Korea and 3.23 for China [Streets et al., 2003].
 It was already shown in Figure 4 that there was similarity found between distributions of OC and O3. If daily averaged, the ratios of OC to EC were in good agreement with O3/CO ratios for both daily and diurnal variations (Figure 7). As the latter represents the extent of atmospheric oxidation in gaseous composition, the former could be the one for aerosol composition. Although the diurnal trend of OC/EC was not as smooth as that of O3/CO, their maximum values were found in the afternoon. These findings imply the formation of organic carbons from chemical transformation, as reported from previous studies [Lim and Turpin, 2002; Simoneit et al., 2004]. The OC/EC ratio was very high on 27 March, which was mostly caused by very low EC concentrations with the minimum PM2.5 mass concentration.
3.3. High PM2.5 Episodes
 During the ABC-EAREX 2005, several cold fronts passed by Gosan station, which not only was clearly depicted in the variations of meteorological parameters (Figure 3c), but also had a great influence on the compositions of major chemical constituents (Figure 3a). This frontal passage is a common meteorological feature over the northeast Asia in spring and effective in exporting Asian pollutants as well as mineral dust to the north Pacific. In this section, the high-concentration episodes associated with frontal system were examined in detail. The first case is a pollution plume coupled with Asian dust observed during 17–18 March. In about 10 d, a similar but rather weak pollution plus dust incident took place on 28–29 March. Between these two, another pollution plume that was not accompanied by dust storm was detected during 22–23 March. For three pollution and two dust plumes during the high-concentration episodes, the daily means of key species were summarized in Table 1. It shows that the relative enhancement of major species varied among plumes, which could have resulted from different source, transport path, and chemical processing of air masses.
Table 1. Measurement Summary for High Concentration Episodesa
| ||PM2.5||nss- SO42−||NO3−||NH4+||Ca2+||K+||OC||EC||O3||CO||NOx||SO2||ΔnssSO42−/ΔCOb||ΔEC/ΔCOb||NO3/NOXc||nssSO42/SO2c||OC/ECc|
|17 Mar (24 h)|
|Median||26.85||5.76||0.53||2.31||0.29||0.51||3.46||0.99||65.90||367.52||3.94||4.40|| || || || || |
|18 Mar (24 h)|
|Mean||30.82||2.42||0.71||0.94||0.43||0.36||4.35||1.02||49.89||251.07||2.11||1.59|| || ||0.33||1.52||4.27|
|Median||23.05||1.64||0.34||0.51||0.32||0.34||4.29||0.99||50.60||247.01||1.70||0.79|| || || || || |
|22–23 Mar (48 h)|
|Median||22.60||3.35||0.86||0.63||0.15||0.25||4.95||0.98||60.65||303.05||2.03||2.20|| || || || || |
|28 Mar (24 h)|
|Median||16.7||2.09||1.12||1.68||0.20||0.19||3.37||0.45||52.85||248.51||1.27||0.88|| || || || || |
|29 Mar (24 h)|
|Mean||24.43||2.25||0.76||0.63||0.42||0.23||4.53||0.71||52.46||239.60||0.99||1.56|| || ||0.77||1.44||6.38|
|Median||20.55||2.25||0.74||0.53||0.31||0.22||4.53||0.72||52.2||234.83||0.93||1.47|| || || || || |
|Median||21.05||2.49||0.79||1.05||0.19||0.32||4.03||0.74||55.9||272.60||1.75||2.02|| || || || || |
|Maximum||161.90||20.08||10.91||8.49||1.13||0.91||10.85||4.15||73.60||1058.54||9.76||20.78|| || || || || |
 First, to get a hint of the origin and transport path of plumes, the trajectory analysis was performed using HYSPLIT-4 model that was developed by the National Oceanic and Atmospheric Administration (http://www.arl.noaa.gov/ready/hysplit4.html). For five plumes during three episodes, backward trajectories arriving at 1000 m above sea level were calculated for 5 d, assuming that air moved along the isentropic surface. The results are presented in Figure 8. For the first and third episodes, pollution (1 and 4) and subsequent dust (2 and 5) plumes were distinguished in trajectories. The air carrying large amount of pollutants came from the northwest, whereas Asian dust was delivered directly from the north.
Figure 8. Five-day backward trajectories at 1000 m altitude for high-concentration episodes. “1” is trajectory for 17 March at 1400 UTC, and “2” is for 18 March at 0200 UTC (dotted lines). “3” denotes the trajectory of air for 22 March at 1500 UTC (thin solid line). “4” is for air observed on 28 March at 1200 UTC, and “5” is for 29 March at 0300 UTC (bold solid lines). The local time of Gosan (Korea) is 9 h ahead of UTC.
Download figure to PowerPoint
 In this study, the dust-containing plume was unambiguously identified by high Ca2+ concentrations (Figure 2 and Table 1). All the other major species were raised ahead of Ca2+ only in pollution plumes. Particularly on 17 March, the enhancement of both gas and aerosol species were the most pronounced (Table 1). For the pollution event detected on 23 March, the air originated from the northern part of China but arrived at Gosan after passing by the Korean Peninsular. Hence Korean influence was likely superimposed on preexisting plume from China. Particularly, this episode was characterized by high EC and OC concentrations in daily mean (Table 1).
 These plumes were also distinguished by relative enhancement of nss-SO42− and EC to CO. Their correlations were much improved for individual pollution episode (Figure 9). The enhancement ratio of ΔSO42−/ΔCO and ΔEC/ΔCO was obtained from linear regression and given on the plot. In dust plumes, on the other hand, no significant relationship was found between nss-SO42− and CO or EC and CO. Moreover, the concentrations of other species than Ca2+ were lower than mean values of the whole measurements (Table 1). Hence the effect of dust was thought to be relatively less important in determining chemical compositions of PM2.5. Among three pollution episodes, both Δnss-SO42−/ΔCO and ΔEC/ΔCO were greatest on 28 March but least on 17 March. During 22–23 March, ΔEC/ΔCO was comparable to that of 28 March. Considering the trajectories and the nature of chemicals, it is likely that ΔEC/ΔCO tended to be modified by sources, while Δnss-SO42−/ΔCO reflected the extent of chemical processing during the transport. In addition, the ratios of mean concentrations for key species, including NO3−/NOx, nss-SO42−/SO2, and OC/EC were particularly high on 28 March (Table 1), when the fraction of soluble ions was the highest (Figure 3b). These results demonstrate that the chemical characters of fine aerosols were well distinguished by relative enhancement of major species including precursor gases and aerosol constituents. Thus these ratios can be useful indicators to identify sources and processes determining the chemical properties of fine aerosols.
Figure 9. Correlations of (a) EC and CO and (b) nss-SO42− and CO for three individual pollution plumes (17, 22–23, and 28 March). Lines stand for linear regression fittings, for which slope and r2 values are given.
Download figure to PowerPoint