Impact of Dust Storms on the Organic Composition in the Ambient Aerosol

To better understand the impact of dust storm (DS) events on the ambient aerosol organic composition in Seoul, samples of solvent-extractable organic compounds in the atmospheric particulate matter with an aerodynamic diameter of less than or equal to a nominal 10 μm (PM10) were collected for 3 months (March–May) in 2007 and characterized. The results of the statistical analysis suggested that during the DS events the ambient PM10 concentration increased, the n-alkanes concentrations decreased, and the PAH concentrations show no trend, although they increased slightly. The distributions of n-alkanes and PAH species showed that among n-alkanes only C25 and C27 species in the non-DS samples increased compared to in the DS samples. Since C25 and C27 species are mainly emitted from plants, this suggests that the impact of biogenic emissions of n-alkanes was reduced during the DS periods. Among PAHs, only the concentration of Fluoranthene increased in a statistically significant manner during the DS period, and only the 1,3,5-Triphenylbenzene level was higher during the non-DS period. The backward trajectory analysis results showed that the air parcels over Seoul did not show much difference between DS and non-DS days. The average level of n-alkanes per unit mass of PM10 was statistically lower in the DS samples, while those of the PAHs showed no statistically significant differences between the DS and non-DS samples.


INTRODUCTION
Dust storm (DS) becomes a major socio-environmental issue in East Asia (Arimoto et al., 2006).Dust particles emitted from the deserts and loess areas of North and Northwest China, South Mongolia, and Kazakhstan are carried eastward along with the prevailing seasonal westerlies and southwesterlies and passed over downwind areas including East China, North and South Korea, and Japan.If meteorological conditions are favorable, relatively fine airborne dust particles can be carried much further across the Pacific Ocean which might affect the air quality in the west coast of North America (Duce et al., 1980).
Dust particles can affect both regional and global environment such as visibility reduction, change of radiative forcing, and detrimental effects on human health.Since intensified desertification in China has increased the frequency of DS occurrence along with its intensity and duration, it is expected that the impact of DS will persist for a foreseeable future (Kim, 2008).
Although DS mainly consists of ambient mineral particles, these particles can act as a long-range transporter of various ambient trace species such as secondary inorganic ions, Persistent Organic Pollutants (POPs), and trace elements including heavy metals (Ghim, 2011).There are several previous studies on DS which demonstrated that DS events have brought a large amount of dust particles along with increased concentrations of anthropogenic species such as sulfate, nitrate, black carbon (BC), gaseous pollutants (e.g., CO, SO 2 , and O 3 ), and even bioaerosols (Choi et al., 2001;Kim et al., 2009;Chang et al., 2010;Jeon et al., 2011).
According to Choi et al. (2001), at Seoul, the concentrations of the earth crustal elements like Mg 2+ , Ca 2+ , Na, Mg, Al, Ca, Fe, and Mn in total suspended particles (TSP) and PM 10 increased in accordance with the intensity of DS, while those of secondary inorganic ions like sulfate (SO 4 2-), nitrate (NO 3 -), and ammonium (NH 4 + ) salts and harmful heavy metals including Ni, Cu, Zn, Cd and Pb were lower during heavy DS periods than non-DS periods.Back trajectory results showed that the concentrations of anthropogenic pollutants were largely dependent on the transport pathways of DS.Kim et al. (2009) carried out a statistical analysis on the aerosol composition change between DS and non-DS periods in spring from March 1992 to May 2006 for the TSP data measured at Gosan, a background area in Korea.Among eight inorganic ions and twelve elements analyzed, the concentrations of all species except NH 4 + were statistically significantly higher during the DS period.This result was in contrast with another study using the same data base but shorter period; between 1992 and 1999 (Park et al., 2003).Park et al. (2003) found that among the anthropogenic species, the concentrations of Pb, non-sea salt (nss)-SO 4 2-, and NO 3 -were higher during the DS period.Kim et al. (2009) suggested that the drastic increase of primary energy consumption in Northeast Asia during the 2000s might affect the aerosol composition significantly.Jeon et al. (2011) collected 85 TSP samples from May 2008 to March 2010 at Seoul and applied various biological techniques to identify and quantify bacteria.They found that the concentration of culturable bacteria was significantly increased and the bacterial community structure was abruptly changed during the DS events.These might affect human health significantly during a DS event.
Likewise, dust particles may also absorb and convey diverse organic compounds such as polycyclic aromatic hydrocarbons (PAHs) (Kim et al., 2012).Even though organic compounds frequently make up the most significant fraction of aerosols, especially fine particles mass concentration (Jacobson et al., 2000;Seinfeld and Pankow, 2003), complete characterization of the organic fraction of ambient aerosol is difficult due to the complexity of their physical and chemical properties like solubility and reactivity (Jacobson et al., 2000).As a result, relatively few studies have been reported on the organic compositions of aerosols during DS.
Normal alkanes (n-alkanes) are an important class of organic compounds in the ambient aerosols, and their homologue distribution may indicate different pollution sources as some of alkane species have been successfully used as molecular tracers in source apportionment of atmospheric particulate pollution (Li et al., 2010).n-Alkanes also do have negative impacts on human health.For example, the n-alkanes with more than 16-carbon number are known to have a detrimental effect on the skin and even engender skin cancer (Li et al., 2010).
PAHs are one of the most widespread atmospheric organic pollutants mainly derived from incomplete combustion processes of carbon-containing fuels like biomass, wood, coal and oils.Therefore, PAH emission concentrations and patterns are strongly correlated with combustion sources and conditions (Schnelle-Kreis et al., 2007).PAHs are released into the atmosphere in both gaseous and aerosol phases.Approximately 90% of PAH emissions in urban and industrial atmospheres are estimated to be anthropogenic in origin (Fang et al., 2010).Some of PAHs and their derivatives which mainly exist in the particle phase in the atmosphere are of concern because they have been identified as strong human carcinogens (Lee and Kim, 2007).Therefore, longrange transport of PAHs over hundreds to thousands of kilometers has also been of interest in various locations in the world.
There have been a few studies investigating the relationship between DS events and the concentration of PAHs.At various locations in East Asia, increased concentrations of particle-associated PAHs were observed during DS events (Guo et al., 2003;Fang et al., 2005;Tamamura et al., 2007).Guo et al. (2003)  In order to estimate adverse effects on the environment and human health, a study of organic aerosols in PM 10 during DS events should be conducted since long-range transported dust particles are typically in the coarse particle size range with aerodynamic diameters of 1-10 μm (Ghim, 2011;KMA, 2011;Stone et al., 2011).According to Chun et al. (2001), the particles in the range of 1.35-10 μm in diameter were increased significantly both at Seoul and Anmyon island during a heavy DS in spring, 1998.The PM 10 which prevailed in the Seoul samples showed highly diurnal variation regardless of DS events and thus, were thought to be mainly resulted from local emission.
Seoul, the capital of the Republic of Korea, is one of the largest cities in East Asia with a population of approximately 10 million and an area of 605.33 km².It is located in the mid-western part of Korean Peninsula and has a long history of early spring time dust storm blown mainly from the Taklimakan Desert, the Gobi desert and the Loess Plateau of China (Ghim, 2011).
The purposes of this research are; (1) to analyze and compare organic compositions and concentrations (18 nalkanes and 16 PAHs) in PM 10 between the DS and non-DS period at Seoul, Korea from March to May, 2007, (2) to identify the effect of DS on the levels and patterns of ambient n-alkanes and PAHs, and (3) to discuss on the possibility of the role of dust particle as a long range transporter.

Aerosol Sampling
Atmospheric aerosol samples were collected on the roof of the School of Public Health building at the Seoul National University (127°001'E, 37°51'N; 17 m above ground level) between March and May, 2007.Total of 15 samples was obtained and six DS events were observed during sampling period.We collected 24-h PM 10 samples on pre-baked quarts fiber filters (Whatman Quartz Microfibre Filter, QM-A, 20.3 × 25.4 cm 2 ) using a high-volume air sampler (Kimoto, model 121 series) equipped with PM 10 size-selective inlets in a flow rate at 700 L/min.Quartz fiber filters were prebaked in an oven at 450°C for 8 hr.Each sampling began at about 9:00 am local time each day.

GC/MSD Analysis
Detailed methods for extraction and analysis are described elsewhere (Lee et al. (2011b) for n-alkanes and Lee et al. (2006Lee et al. ( , 2011a) ) for PAHs).In brief, a half of filters were ultrasonicated two times for 30 min each with a mixture of dichloromethane and methanol (3:1; v/v).Before extraction, two deuterated n-alkanes and three deuterated PAHs standards were added to each sample as a surrogated internal standard.Total extracts were filtered using a syringe (Whatman PVDF, ID 25 mm, pore size 0.45 μm) and then blown down to 500 mL using a concentrator (ZymarkTurboVap 500) under a pure nitrogen stream at 40°C.
The chemical compositions were identified and quantified by a Hewlett Packard 7890A gas chromatograph equipped with a 5975 mass selective detector (GC-MSD, Agilent Technologies) in the synchronous selected ion monitoring (SIM)/Scan mode.The names of used internal standards and their target compounds are given in Table 1.
The GC-MSD operating conditions were as follows; temperature hold at 60°C for 1 min, increase from 60°C to 310°C at a rate of 6 °C/min with final isothermal hold at 310°C for 15 min.Helium was used as a carrier gas.The sample was injected on a splitless mode at an injector temperature of 300°C.The mass spectrometer was operated on electron impact (EI) mode at 70 eV and scanned from 50 to 650 Da.Data was acquired and processed with the Chemstation software.Individual compounds were identified by comparison of mass spectra with literature and library data, comparison with authentic standards, and interpretation of mass spectrometric fragmentation patterns.The data reported here were corrected for the recoveries.Three field blanks were analyzed and there were no detection of PAHs in the field blanks except Fluoranthene and Pyrene.Fluoranthene and Pyrene were detected in only one field blank and the amount of these compounds in the blank was minor (under 0.1% compared to the amounts in the samples).
For n-alkanes, the compounds from C22 to C29 were detected in the three field blanks, however, the amounts were under 1% of the amounts in the samples.Therefore, no correction of field blank was applied in the reported data.

Statistical Analysis
To assess the concentration differences between DS and non-DS aerosol samples, 'independent samples (or two samples) t-test' was performed for the data.Theoretically, when small samples (n < 30) are used in the hypothesis testing, certain assumptions should be met if the test is to work properly and yield reliable results (Schefler, 1988).One of them is that the variances between two groups should be homogeneous.
In this study, the f-test was carried out to decide whether or not the variances between two groups are homogeneous so that the above assumption can be satisfied.For most cases, the variances between two sets were homogeneous except the PM 10 mass concentrations and some of alkane species.However, according to Boneau (1960), even relatively severe departures from homogeneity of variances did not have serious consequences.Thus, we still carried out the t-test for those species.
In present study, the f-test and t-test are conducted at the 90% (p < 0.1) and 95% (p < 0.05) confidence level.

Air Mass Backward Trajectory Analysis
In order to track the transport pathways of dust clouds emitted from source regions, backward trajectories were calculated by the NOAA Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT4) model (Draxler and Rolph, 2003) with starting height at 1500 m considering the main transport routes of DS.
The location of the backward trajectory start point was the School of Public Health building at the Seoul National University (127°1'E, 37°51'N).The run time of every trajectory was 72 h (3 days), with 6-h time resolution.The Final Run (FNL) meteorological data, which was six hourly archive data from NCEP's GDAS (Global Data Assimilation System), were used for the trajectory calculation.

General Characteristics
The daily average PM 10 concentrations for the 15 samples are summarized in Table 2 with the meteorological conditions and shown in Fig. 1(a).According to the Korean Meteorological Administration (KMA), 6 days of DS storm events were observed at Seoul during the sampling days as indicated in Table 2.The PM 10 mass concentrations during the DS events ranged from 34.5-1004.2(average: 308.6 ± 360.75) μg/m 3 , while that during the non-DS events were 36.3-95.2(average: 61.8 ± 21.69) μg/m 3 .Though it was reported as a DS day on April 2, 2007 (D4 in Table 2), the DS was rather weak and the PM 10 concentration was low as will be discussed later in this section.During the DS events, the concentrations of PM 10 were higher than the non-DS events significantly (p < 0.05, one-tailed t-test) when equal variances are assumed.F-test results, however, showed that the variances in two groups are not significantly similar for the PM 10 concentration (p < 0.1).As mentioned in section 2.3, when the number of samples (n value) is less than 30, the variances in the two groups should be similar for the t-test results to be statistically meaningful.Still, based on the finding by Boneau (1960), we carried out the t-test for the PM 10 concentrations.When 'unequal variances assumed t-test' was applied, the PM 10 concentrations were higher during the DS events only at p < 0.1 level (one-tailed t-test).
Previous studies have reported that PM 10 concentration can be significantly increased during a DS event.Han et al. (2005) found that the mean concentration of ambient PM 10 was about eight times higher during the DS periods in spring 2001 at Gosan.In this study, it is thought that the average PM 10 concentration during non-DS were rather high due to stable atmospheric condition resulted from frequent haze and mist events, while that of DS might be low due to rain and snow events during the DS events (Table 2).Statistical data from the Ministry of Environment (ME), Republic of Korea shows that the average concentration of PM 10 at Seoul during the spring season including DS periods (from March to May) in 2007 was about 77.7 μg/m 3 .
During the DS events, the maximum hourly values of PM 10 particles ranged 63-249 (average: 192.3 ± 66.79) μg m -3 were recorded by National Institute of Environmental Research (NIER), Republic of Korea.An unusual daily average concentration of PM 10 which exceeded the Korea's 24-h ambient air quality standard of 150 μg/m 3 was observed three times during the DS events on March 31 (D2), April 1 (D3) and May 25 (D6), 2007.On April 1, in particular, Seoul site was covered with extremely high level of aerosol concentrations (daily average concentration of PM 10 exceeding 1000 μg/m 3 ).This severe dust storm event originated from North and Northwestern China.However, on April 2, the level of PM 10 moved down to the lowest of all samples within a day.It seems like it was because April 2 was the last day of three-day dust storm event (March 31-April 2).D1 is regarded as a relatively weak dust storm event which originated from the forest regions in Siberia/ northeast China.
As shown in Table 3, 18 species of n-alkanes (C17-C34) were determined.The alkane data on N8 sample is not reported due to the low recoveries of internal standards which might be resulted from experimental errors.Also, total n-alkanes concentrations for each sampling days were shown in Fig. 1(b).The average total concentration of the n-alkanes was 66.87 ng/m 3 .The average total concentrations of n-alkanes for the DS samples was 54.13 ± 24.78 ng/m 3 with the range of 15.28-84.23 ng/m 3 , while that for the non-DS samples was 76.42 ± 25.94 ng/m 3 with the range of 44.49-122.73ng/m 3 .F-test results demonstrated the homogeneous variances between two groups and thus, equal variances assumed t-test was performed.The average concentration of n-alkanes during DS was even lower statistically than that of non-DS at p < 0.1.
Sixteen species of PAHs and their abbreviation forms are provided in Table 4 along with the concentration data.PAHs species of low molecular weight with 2 or 3 rings (e.g., Nap, Ace, Acy, and Flu) were not analyzed here as they exist predominantly in the gas phase in the normal atmospheric conditions because of their high volatility.
The average total concentration of the ambient particulate PAHs in our Seoul samples was 11.88 ± 5.55 ng/m 3 .The average concentration of the sum of 16 individual PAHs at Seoul between August, 2002 and December, 2003 was 26.6 ± 28.4 ng/m 3 and it showed winter maxima and summer minimum mainly due to the increased fuel consumption in winter both at Seoul and Northeast Asia (Lee et al., 2011a).The average total particulate PAHs concentration for the DS samples (13.44 ± 6.03 ng/m 3 ) was slightly higher than those for the non-DS samples (10.84 ± 5.31 ng/m 3 ) (Fig. most DS samples (except one) were also slightly higher than those of non-DS samples (Table 4).However, t-test result showed that these differences are statistically not meaningful (p < 0.05).

Comparison of n-alkanes/PAHs Distribution Pattern
The above results revealed that the concentration of PM 10 tends to be increased by DS but the trend of the concentrations of organic classes during the sampling period varied differently.Generally, the average concentration of n-alkanes decreased during dust storm period (p < 0.1) and that of PAHs seemed to increase a little, though that increase was not statistically meaningful at p < 0.05.
To further understand the behavior of n-alkanes and PAHs, their compositional distribution pattern between the DS and non-DS samples are studied.Since some organic species are considered as organic molecular markers (or tracers) as they are emitted preferentially by some unique sources, several studies have used these organic compounds for source apportionment (Schnelle-Kreis et al., 2007;Li et al., 2010).Therefore, different concentration trends for nalkanes and PAHs might give a clue to their emission and transport patterns during DS.
The distribution patterns for the each species of n-alkanes and PAHs between DS and non-DS samples are shown in Fig. 2. The distribution of n-alkanes, expressed as the relative abundance of each homologue in total alkanes can help identifying the likely emission sources (Schnelle-Kreis et al., 2007;Li et al., 2010).The high concentrations of biogenic hydrocarbons (C 27 , C 29 , C 31 and C 33 ) when accompanied with low concentrations of the intervening even carbon number n-alkanes as shown in Fig. 2 suggest high plant wax release into the atmosphere (Cass, 1998;Li et al., 2010).The high average concentrations of C 25 , C 26 and C 27 in the non-DS samples seem to be mostly resulted from those of the most alkane-polluted day, N4 (Table 3).Statistically, the concentrations of C 25 and C 27 species in the non-DS samples were higher than those from the DS samples (at p < 0.1 for C 25 ; at p < 0.05 for C 27 ).However, in case of C 26 , there was no statistical change between the DS and non-DS samples.
Among the PAHs analyzed, only the concentration of Fluoranthene increased statistically during the DS period (p < 0.1) and those of 1,3,5-Triphenylbenzene alone was higher during the non-DS period (p < 0.05).1,3,5-Triphenylbenzene is a useful urban tracer since it has been detected in particles from solid waste incinerators and from burning of plastics (Simoneit et al., 2004).Excluding these two, there was no statistically significant concentration change during the DS events.When we compare the concentrations of each species, f-tests were carried out beforehand and most species showed homogeneous variances but four species showed non-homogeneous variances (C 17 , C 18 , C 25 and C 26 alkane species at p < 0.05).For these species, as explained before, we carried out unequal variance assumed t-test.
Apart from the DS event, transport from outside of Korea, especially from China, is one of the major sources of air pollutants in Korea during spring (Park et al., 2003(Park et al., , 2004)).According to the three-days of backward trajectory analysis, almost all air parcel trajectories for the samples were from Russia, North and Northeast China and Inner Mongolia region regardless of DS which means the sources of longrange transported organic pollutants would be quite similar between DS and non-DS event.These trajectories confirmed that Korea is influenced mostly by the westerly during spring (Park et al., 2004).In some cases, two samples with similar trajectories but different levels and distribution patterns of n-alkanes and PAHs were found.
Similar result was observed for the particulate ionic species at Gosan (Kim, et al., 2008).This variation may reflect additional local input or intervention of other complex environmental factors.Based on the previous studies, other factors, e.g., source types, transport pathways and mechanisms, meteorological conditions, local emission and atmospheric chemical compositions are also expected to affect the level and distribution of ambient organic species significantly in combination (Park et al., 2004;Lee et al., 2011a).

Role of the DS Particles as a Long Range Transport Carrier
Our initial assumption was that coarse crustal DS particles may absorb and convey various organic compounds, and therefore cause high organic aerosol concentration during a DS due to the increased PM 10 concentration.However, as summarized in section 3.2, a large number of coarse particles in DS might not be an important carrier of organic substances considering that the levels of n-alkanes and PAHs were not increased as much as those of PM 10 during the DS period.More precisely, in this study, the absolute concentration of n-alkanes even decreased during DS and no statistical difference was found in PAHs level between the DS and non-DS periods.
A similar conclusion was made by Choi et al. (2012).Choi et al. (2012) measured PAHs in both gaseous and particulate phases at Gosan for 15 days (March 29-April 12, 2002) and found out that (1) the levels of particulate PAHs during the DS period were close to or lower than those during the non-DS period and (2) they showed even higher value when PM 10 was substantially low, concluding the influence of coarse particles of DS on the levels of PAHs was probably not significant.The levels of n-alkanes and PAHs per unit mass of PM 10 , or the enrichment factor (Kim et al., 2009) were shown in Fig. 2. It seems that, in general, relative importance of organic species per unit mass of PM 10 during DS event would be not high.The average level of n-alkanes per unit mass of PM 10 was statistically lower in the DS samples than that in the non-DS samples, while that of PAHs per unit mass of PM 10 showed no statistical difference between the DS and non-DS periods (p < 0.05).N-alkanes/PM 10 concentration ratio followed a similar pattern to that of PAHs/PM 10 with the correlation coefficient of 0.71 (n = 14) when excluding the N8 sample.Considering the absolute concentration of n-alkanes also showed decreasing tendency during the DS period (p < 0.1), it can be estimated that PM 10 in the non-DS samples contained more n-alkanes, especially C 25 -C 27 species, than in the DS samples.
The tendency of PAHs/PM 10 ratios during sampling period was different to n-alkanes/PM 10 ratios (Fig. 3).The variation of PAHs/PM 10 ratios was small during sampling period, while, the ratios of n-alkanes/PM 10 were fluctuate widely.PAHs and n-alkanes have different emission sources and strength.Thus, different emission sources and chemical types of these compounds might give influence to the different distribution in the dust particles.Furthermore, it is possible that the two organic classes have different interactions with dust particles.However, at this stage, it is hard to conclude that DS event give different influence to PAHs and n-alkanes.
As shown in Fig. 2, the average concentrations of n-alkanes species were higher during the non-DS period, while those of PAHs were higher during the DS period.It can be explained with the above suggestions: Since PM 10 in dust particles contain less amount of n-alkanes than non dust particles, the overall concentration of n-alkane species went down during DS even with the increased PM 10 concentration of the DS period.However, the concentration of PAHs species increased during DS (even though it was not statistically meaningful) due to the increase of dust particles (PM 10 ) which contain as much PAHs as normal aerosols do.

SUMMARY
Dust particles may absorb and convey diverse organic compounds.Though organic carbon (OC) compounds are a major fraction of ambient aerosols, complete characterization of the organic fraction of ambient aerosol is difficult, especially, for dust particles.As a result, relatively few studies have been reported on the organic compositions of aerosols during dust storm (DS).
To better understand the impact of the DS event on the ambient aerosol organic composition at Seoul, the solventextractable organic compounds in the PM 10 samples The results revealed that the concentration of PM 10 tends to be increased by DS event while the absolute concentrations of n-alkanes showed decreasing tendency (p < 0.1) and those of PAHs were not changed statistically significantly (p < 0.05).
The distributions of n-alkanes and PAHs species showed that among n-alkanes, only C 25 and C 27 species in the non-DS samples increased in their level compared to those in the DS samples (at p < 0.1 for C 25 ; at p < 0.05 for C 27 ).Since C 25 and C 27 species are mainly emitted from plant, it suggested that the impact of biogenic emissions of n-alkanes was reduced during the DS periods.Among PAHs, only the concentration of Fluoranthene increased statistically during the DS period (p < 0.1) and only 1,3,5-Triphenylbenzene level was higher during the non-DS period (p < 0.05).
The average levels of n-alkanes per unit mass of PM 10 was statistically lower in the DS samples, while those of PAHs showed no statistical difference between the DS and non-DS samples suggesting; (1) non-DS particles carried more n-alkanes than DS particles and (2) comparable amount of PAHs was involved in the DS and non-DS particles.

Fig. 2 .
Fig. 2. Component distribution pattern of (a) n-alkanes and (b) PAHs for non-dust storm aerosol and dust storm aerosol samples.

Table 1 .
Internal standards used in this study and their target compounds.

Table 2 .
Summary on the sampling conditions and mass concentrations.
a N: Non-dust storm day sample, b D: Dust storm day sample.

Table 3 .
Mass concentrations (ng/m 3 ) of n-alkanes analyzed in 6 dust storm samples and 9 non-dust storm aerosol samples.

Table 4 .
Mass concentrations (ng/m 3 ) of 16 quantified PAH compounds analyzed in 6 dust storm samples and 9 non-dust storm aerosol samples.