Seasonal variation of the concentrations of nitrogenous species and their nitrogen isotopic ratios in aerosols at Gosan, Jeju Island: Implications for atmospheric processing and source changes of aerosols
Institute of Low Temperature Science, Hokkaido University, Sapporo, Japan
Graduate School of Environmental Science, Hokkaido University, Sapporo, Japan
 Atmospheric aerosol samples (n = 84) were collected at Gosan site, Jeju Island, South Korea between April 2003 and April 2004 for the measurements of total nitrogen (TN) and its isotopic ratio (δ15N) as well as nitrogen species (NH4+ and NO3−). Measurements were also conducted for remained nitrogen (remained N) and removed nitrogen (removed N) on HCl fume treatment. A pronounced seasonal variation was found in the δ15N of TN, remained N (mostly NH4+), and removed N (mostly NO3−). The highest mean δ15N values of TN (+16.9‰ ± 4.5‰) and remained N (+20.2‰ ± 5.2‰) are detected in summer (June–August) whereas the lowest mean δ15N values (+12.9‰ ± 3.4‰ and +11.3‰ ± 5.1‰, respectively) are in winter (December–February). These trends can partly be explained by an enhanced contribution of 15N-enriched emissions from agricultural straw burning in China in a harvest season (summer and autumn). The mean δ15N of removed N showed an opposite trend: the lowest (+8.9‰ ± 3.7‰) in warm season (March–August) and the highest (+14.1‰ ± 3.7‰) in cold season (September–February). These results can be explained by changes in source regions and emission strengths of nitrogenous species, as well as difference in secondary aerosol nitrogen formation between warm and cold seasons. Higher ratios of Ca2+/Na+ and the lowest ratios of Na+/(Cl− + NO3−) are associated with lower δ15N values of removed N as a result of less isotopic enrichment (ɛproduct-reactant) during the reaction between HNO3 and dust particles. This study proposes that 15N/14N ratio can be regarded as process tracer of nitrogenous species in the atmosphere.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Nitrogen compounds play key roles in the formation of smog, aerosol, and ozone and in the determination of acidity in precipitation [Marsh, 1978; Galloway and Likens, 1981]. Possible shift of NO3− and NH4+ to the coarse mode significantly enhances the rate of nitrogen deposition to the coastal ecosystems and hence marine productivity [Galloway et al., 1995; Jickells, 1998; Spokes et al., 2000; Duce et al., 2008]. It is therefore very important to know atmospheric processing of nitrogenous species as discussed earlier and their sources in the atmosphere.
 The present research has been conducted with the aim of evaluating the usefulness of 15N/14N ratio to interpret the complex atmospheric processing and sources of nitrogen species in the East Asian aerosols collected at Gosan site in Jeju Island, the East China Sea, for 1 year from 2003 to 2004. Here we present seasonal variations of δ15N of total nitrogen (TN), remained nitrogen (mostly composed of NH4+), and removed nitrogen (mostly composed of NO3−) in the aerosol samples. Then, we interpret the observed isotopic composition (δ15N) and its seasonal variations based on the chemistry of nitrogenous species in the atmosphere and air mass transport pathways with which we can backtrack their source regions. Although δ15N data have been reported previously as part of comprehensive study of organic and inorganic chemistry of Gosan aerosols [Kawamura et al., 2004], the present paper focuses on δ15N composition of nitrogenous species and discuss in more details on the chemical processes that govern the seasonal variations of the δ15N values in aerosols.
2.1. Site Description
 Gosan site is located on a cliff (∼71 m ASL) at the western tip of Jeju Island (33°29′N, 126°16′E). It is approximately 100 km off the south of Korean Peninsula, ∼500 km off the east of China (Jiangsu province or Shanghai), ∼200 km off the west of Kyushu Island, Japan, and ∼1000 km off the northeast of Taiwan (Figure 1). The site and its surroundings are covered with grasses but no trees. Because of its location and very limited local anthropogenic emissions [Kim et al., 1998], Gosan has been considered as an ideal site to monitor the impact on air quality of the western rim of the North Pacific due to the outflows from East Asia [Carmichael et al., 1996, 1997; Chen et al., 1997]. Gosan was used as a supersite of ACE-Asia campaign in 2001 [Huebert et al., 2003], and PEM-West A and PEM-West B programs conducted between 1991 and 1994 [Hoell et al., 1996, 1997]. It is now used as one of superstations of Atmospheric Brown Cloud (ABC) program [Lee et al., 2007]. More detailed description of this site is given elsewhere [Kim et al., 1998; Kawamura et al., 2004; Lee et al., 2007].
2.2. Aerosol Sampling
 Total suspended particles (TSP) in the atmosphere were collected at Gosan site over 2–7 days throughout the year from 2003 April to 2004 April. TSP samples (n = 84) were collected on precombusted quartz fiber filters using a high volume air sampler (Kimoto AS-810) installed on the roof of a trailer house (∼3 m above the ground). Before and after the sampling, filters were put in clean glass jars (150 mL) with Teflon-lined screw cap. Samples were transported to our laboratory in Sapporo and stored in a dark freezer room at −20°C until analysis. Field blank filters were collected every month.
2.3. Analytical Methods
 In the present study, we follow a simple approach to approximate the nitrogen isotopic composition of both NH4+ and NO3− [Kawamura et al., 2004]. We expose the filter cuts of samples to HCl fuming in a desiccator in order to remove NO3−. On the basis of the ion chromatographic (IC) analyses of the samples before and after HCl treatment, we confirm that NO3− is completely removed, but NH4+ remains intact. Here we refer nitrogen that remained on the filters due to HCl fume treatment as remained nitrogen (remained N) whereas nitrogen flew away from the filters as removed nitrogen (removed N). Quality assurance of this procedure will be given later with evidence to prove that remained N is mainly composed of NH4+ whereas removed N is mainly composed of NO3−.
 For TN and remained N analyses, two small discs (area 2.54 cm2) were cut off from each filter sample. One disc was put in a tin cup and shaped into a rounded ball using a pair of flat-tipped tweezers. The samples were introduced into the elemental analyzer (EA; model: NA 1500 NCS, Carlo Erba Instruments) using an autosampler and were oxidized in a combustion column packed with chromium trioxide at 1020°C, in which the tin container burns (>1400°C) to promote the intensive oxidation of sample materials in an atmosphere of pure oxygen. The combustion products are transferred to a reduction column packed with metallic copper that was maintained at 650°C. Here excess oxygen is removed and nitrogen oxides coming from the combustion tube are reduced to molecular nitrogen (N2). The N2 was isolated from CO2 online using a gas chromatograph (GC) installed in EA and then measured with a thermal conductivity detector. Aliquots of N2 gas were then introduced into an isotope ratio mass spectrometer (ThermoQuest, Delta Plus) through a ConFlo II interface (ThermoQuest) to monitor 15N/14N ratios. The nitrogen isotopic composition was calculated using the following standard isotopic conversion equation.
 An external standard (acetanilide) was used to determine TN, remained N and their isotopic ratios. The δ15N value of acetanilide is 11.89‰.
 Another disc was placed in a 10 mL glass vial and exposed to fuming HCl overnight in a glass desiccator (40 cm wide) to remove NO3−, which facilitates only the measurement of remained N (mostly NH4+) on the filter sample and its nitrogen isotopic composition. Similar analytical procedure has been used in the previous study [Kawamura et al., 2004], in which the δ15N values of removed N were calculated using the following isotopic mass balance equation.
 “Ammonium” (remained N) and “nitrate” (removed N) were not explicitly measured but were measured using a chemical method of HCl fuming and EA. Hereafter, we use symbols NH4+* and NO3−*, which mean concentrations of remained N (“ammonium”) and removed N (“nitrate”), respectively.
 For the analyses of water-soluble inorganic ion, aliquots (area 2.54 cm2) of filter samples were placed in 50 mL glass vials, soaked using 10 mL of Milli Q water, and agitated over 15 min using an ultrasonicator. The extracts were filtered with prewashed Millex (Millipore Corporation, Bedford, USA, 0.45 μm) syringe filters. The concentrations of the ions in the aqueous extracts were determined using a Metrohm 761 IC (Metrohm, Herisau, Switzerland). Anions in the aqueous extracts were separated on a Shodex SI-90 4E column with 1.8 mM Na2CO3 and 1.7 mM NaHCO3 (Kanto Chemical, Japan) as eluent. Cations were isolated on a Shodex YK-421 column with 4 mM H3PO4 as eluent (Kanto Chemical, Japan). The injection loop volume was 200 μL. Cations and anions are quantified against a standard calibration curve.
2.4. Quality Assurance of Data
 Field blanks for TN showed very small peaks of nitrogen on GC chromatogram, which were equivalent to 0.4%–9% of real samples. The reported concentrations for both TN and NH4+*, and their δ15N values were corrected for the field blanks. The analyses of a few samples (n = 7) were replicated for TN and NH4+*, and their isotopic ratios. The precisions for TN and NH4+* measurements ranged 1.4%–5% (av. 3%) whereas those for the δ15N measurements ranged 0.03‰–0.36‰ (av. 0.23‰).
 In order to check whether any isotopic fractionation occurs during the exposure of filter samples to fuming HCl, we spiked different amounts of pure NH4+, NO3−, and their mixtures with other cations and anions on fresh quartz fiber filters to mimic the ionic composition of our filter samples. Then we measured nitrogen isotopic composition of the spiked samples before and after HCl fume treatment. The results are shown in Table 1. The δ15N values of spiked NH4+ samples are 6.6‰–6.9‰ with a mean of 6.8‰ and standard deviations of ±0.1‰, which are similar to those of HCl-treated spiked NH4+, and NH4+ plus NO3− samples. The isotopic mass balance equation shows that the δ15N values of eliminated NO3− from spiked NH4+ plus NO3− samples on HCl treatment are similar to those of spiked NO3− samples. The δ15N values of spiked samples with different levels of mixture of NH4+, NO3−, and other water-soluble cations and anions are also similar. These results suggest that difference in the concentration levels of NH4+ and NO3− does not cause any isotopic fractionation during HCl fume treatment. It is also important to note that after the HCl treatment on the spiked filter samples NO3− peak was not detected on the ion chromatograms, indicating that NO3− was completely removed from the filter.
Table 1. Determination of the δ15N of Pure NH4+, NO3−, and Their Mixture With Water-Soluble Cations and Anions Spiked on Quartz Fiber Filter Disc (18 mm in Diameter) at Different Concentration Levelsa
Spiked Levels (μg)
Without HCl Treatment
With HCl Treatment
The δ15N of NH4+ + NO3− in the second column is for NH4+ only as a result of complete removal of NO3− due to HCl treatment. It is confirmed that NO3− is completely removed and NH4+ remains intact due to HCl treatment by analyzing the samples before and after HCl treatment using an ion chromatography instrument.
NH4+ + NO3−
1.3 + 1.6
2.6 + 3.2
5.2 + 6.4
NH4+ + NO3− with cations and anions
1.0 + 3.0
4.0 + 12.0
6.0 + 20.0
 Field blanks for ions showed very small peaks of NH4+ and NO3−. They were less than 0.1%–2% and 0.02%–0.6% of real samples, respectively. The concentrations of ions reported here are corrected for the field blanks. In order to examine whether 15 min ultrasonication in 10 mL water can completely extract NH4+ and NO3− ions from the filters, a few samples (n = 7) were ultrasonicated with the same amount of water for longer periods (>30 min). No significant differences in the concentrations of ions between the extraction periods were observed. A few samples (n = 7) were also subjected to replicate analyses. The coefficients of variations for NH4+ and NO3− measurements were about 2% and 6%, respectively.
 The nitrogen contents in NH4+ and NO3− were compared with TN measured by EA for data quality assurance. Figure 2 shows a relation in the concentrations between (NH4+ + NO3−)-nitrogen and TN. A very good correlation suggests that total nitrogen in our filter samples is mainly composed of NH4+ and NO3−, and nonwater soluble nitrogen (probably organic nitrogen) accounts for only 3% of TN.
 The exposure of aerosol samples to fuming HCl overnight removed 5%–90% of TN (av. 36%). Strong relations were also obtained between NH4+-N measured by IC and NH4+*-N measured by EA (r2 = 0.94) and between NO3−-N measured by IC and NO3−*-N measured by EA (r2 = 0.89) with slopes close to unity and intercepts of nearly zero (Figures 3a and 3b). On the basis of these observations, we can postulate that NH4+*-N measured by EA is mainly composed of NH4+ whereas NO3−*-N measured by EA is mainly composed of NO3−. Similar hypothesis was given by Kawamura et al. . In order to prove this hypothesis, we analyzed a number of samples (n = 10) for the determination of NH4+ and NO3− using the IC technique before and after HCl fume treatment. We found that HCl fume treatment completely removed NO3−, but the concentration of NH4+ remained almost unchanged (<5%). These results confirm that the above hypothesis is correct.
2.5. Potential Sampling Artifacts Associated With Quartz Fiber Filter
 Aerosol samples collected on quartz fiber filters may have two types of sampling artifacts: (1) adsorption of gaseous HNO3 and NH3 onto aerosols already collected within the sampler or onto the collection substrate [McMurray, 2000] and (2) dissociation of NH4+ salts such as NH4NO3 on the filter [Zhang and McMurray, 1992]. Simultaneous measurements of NH4+ and NO3− in aerosols, and their gaseous precursors (NH3 and HNO3, respectively) can provide the quantitative estimation of the first type of artifacts. This type of measurements is not available for our study. However, we can estimate this artifact based on the previous studies carried out at our sampling site. Kim and Seinfeld  reported that the ratio of particulate NO3− to total nitrate (gaseous HNO3 + particulate NO3−) was 0.82 in spring whereas the annual mean of particulate NO3− fraction at Gosan site was 0.74 with ±0.1 deviations for the majority of months [Hayami and Carmichael, 1998]. Their gas-aerosol equilibrium model observations suggest a strong tendency for HNO3 to preferentially exist in the aerosol phase. Harrison and Pio  observed that the adsorption of NH3 onto marine aerosols collected on the filter substrate was insignificant.
 Analysis of aerosol composition data measured at Gosan site from March 1992 to December 1994 using a gas-aerosol equilibrium model suggests that particulate NO3− and Cl− are mostly present in the coarse mode and NH4+ is in the fine mode to neutralize nss-sulfate [Hayami and Carmichael, 1997], being consistent with the size distribution of these species in aerosols collected on Teflon substrate from the same site during Asian dust and nondust storm events [Park et al., 2004]. This provides little formation of NH4NO3 and NH4Cl. On the other hand, we found a good correlation (r2 = 0.81) between NH4+ and SO42− in the aerosol samples studied here. The annual mean molar ratio of NH4+/SO42− was calculated as 1.1 with ±0.3 deviations. These results indicate a close association of NH4+ with SO42−, suggesting that the second artifact (e.g., dissociation of NH4NO3 on the filter) should not be important in our study. We therefore conclude that the presence of coarse mode NO3− and NH4+ collected on the quartz fiber filters cannot be explained by a sampling artifact and thus the reported concentrations and isotopic ratios of nitrogenous species are real.
2.6. Cluster Analysis of Air Mass Backward Trajectories (April 2003 to April 2004)
 Three-dimensional 5 days-air mass backward trajectories arriving at Gosan were calculated with Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model [Draxler et al., 2006] using FNL meteorological data of NOAA/ARL (National Oceanic and Atmospheric Administration/Air Resources Laboratory). The backward trajectories were calculated four times a day at 0000, 6000, 1200, and 1800 UTC. We collected aerosol samples for 335 days from April 2003 to April 2004. Samples could not be collected for the other remaining days due to either technical failure of our sampler or adverse meteorological conditions. Aerosol samples were collected for 117 days in spring (n = 37), 73 days in summer (n = 13), 54 days in autumn (n = 10), and 91 days in winter (n = 24), where, “n” in parenthesis indicates the number of samples. Therefore, the total number of trajectories calculated in spring, summer, autumn, and winter are 468, 292, 216, and 364, respectively.
 All the trajectories in each season were separately subjected to cluster analysis using the cluster algorithm of HYSPLIT model. Given a set of trajectories beginning at one location, the cluster analysis will objectively result in subset of trajectories, called clusters that are each different from the other subsets. Figure 4 shows the results of cluster analyses for four seasons. Each trajectory is the mean of a subset of trajectories. The origins and trajectory pathways of mean clusters in different seasons have been summarized in Table 2. On the basis of the mean trajectories, we found that air masses significantly originated from northeastern China, eastern China, or Korean Peninsula. A few percents of air masses in all seasons, except in winter, originate from or pass over Japan, or southern China (Figure 4). The contributions of either convective or advective boundary layer (<2 km) air masses are 16% in spring (mean trajectory 4, Figure 4a), 74% in summer (mean trajectory 1, 3, 4, and 5; Figure 3b), 23% in autumn (mean trajectory 3 and 4; Figure 4c) and 37% in winter (mean trajectory 1; Figure 4d). All the remaining air masses are descended to the boundary layer at sampling site by convection or advection from the free troposphere (>2 km) (Figure 4). Cluster analysis also suggests that a long-range atmospheric transport not only from East Asia but also from Siberia should be important in determining the chemical and isotopic composition of aerosols at Gosan site.
Table 2. Summary of the Origins and Pathways of the Mean Backward Air Mass Trajectories for Different Seasons Shown in Figure 4a
Name of Mean Trajectory
Origin of Air Masses
Description of Trajectory Pathways
Trajectory Path 1
Trajectory Path 2
Trajectory Path 3
The name of places inscribed with superscripts a, b, and c are the provinces of China, except Beijing and Tianjin. These two cities fall within Hebei province.
All of the air masses are originated from either inland of East Asia or Siberian plateau of Russia. It is to be noted that a significant time (22%) in this season Gosan site remains under the influence of heavily polluted air masses (mean trajectory 2, slow moving air mass).
Significant fraction of air masses (55%) are originated or passed over polluted zone of China. The remaining air masses are oceanic in origin. Although air masses concerned with mean trajectory 4 are oceanic in origin, they passed over the polluted zone of China.
Only 40% of air masses are originated from China (mean trajectory 1) while the remaining 38% and 22% are from Siberia (mean trajectory 2 and 4) and Ocean (mean trajectory 3 and 5). Mean trajectories concerned with 1, 2 and 4 pass over the polluted area of China.
Majority of air masses (80%) are originated from Siberia (mean trajectory 1, 2, 4 and 5) but pass through polluted areas of China. 20% of air masses (mean trajectory 3) are originated in Mongolia and passed over the polluted area of China.
3.1. Seasonality of Total Nitrogen (TN), NH4+, and NO3−
Table 3 summarizes the analytical results of TN, NH4+*-N, and NO3−*-N measured by EA and NH4+ and NO3− along with other chemical species. Figure 5 shows the seasonal variations of TN, NH4+*-N, NO3−*-N, NH4+-N, and NO3−-N in aerosol samples. Generally, all of these chemical species show similar seasonal trend with spring and winter maxima, and summer and autumn minima. The highest TN concentration (9 μg m−3) was found in winter whereas the lowest concentration (0.21 μg m−3) was found in autumn. The average concentrations of TN and NH4+*-N were found to be the highest in spring (3.1 and 2.2 μg m−3, respectively) followed by winter (2.8 and 1.8 μg m−3), summer (2.4 and 1.8 μg m−3), and autumn (1.6 and 0.9 μg m−3) (Table 3). In contrast, the average concentration of NO3−*-N showed the highest in winter (1 μg m−3) followed by spring (0.9 μg m−3), autumn (0.7 μg m−3), and summer (0.6 μg m−3).
Table 3. Summary of Analytical Results of Atmospheric Aerosols Collected at Gosan Site for Four Seasonsa
NH4+*-N, NO3−*-N and their δ15N were not explicitly measured but were measured using a chemical method. See the text for details. The seasons are defined as March–May (spring), June–August (summer), September–November (autumn), and December–February (winter).
Numerical average and standard deviation (±1 SD).
One sample in each spring, summer and winter is excluded in calculating the average δ15N for TN and NO3−*. See the text for details.
 Seasonal variations in the concentrations of NH4+-N and NO3−-N measured by ion chromatography are shown in Figure 5. Their variations are quite consistent with NH4+*-N and NO3−*-N measured by EA. The highest value (6.1 μg m−3) of NH4+ was found in a spring sample. In contrast, NO3− showed the highest value (19.5 μg m−3) in a winter sample (Table 3). The highest mean concentration of NH4+ was found in spring (2.4 μg m−3) followed by winter (1.9 μg m−3) whereas the lowest mean concentrations were observed in summer (1.2 μg m−3) and autumn (1.2 μg m−3). In contrast, the highest mean concentration of NO3− (5.6 μg m−3) was found in winter followed by spring (5.3 μg m−3) and autumn (3.6 μg m−3) (Table 3). The lowest mean concentration (3.1 μg m−3) was recorded in summer.
 According to one-way ANOVA (analysis of variance), the statistical significances for the measured nitrogenous aerosol components (TN, NH4+*-N, NO3−*-N, NH4+ and NO3−) between the seasons range from 0.03 to 0.08. However, these results do not tell us which seasons are responsible for these differences. Therefore, we carried out two post hoc (Tukey and Gabriel) tests. According to these tests, significant (p < 0.05) differences were observed in the mean values of TN, NH4+*-N, and NH4+ between spring and autumn. There are also significant differences in the mean values of NO3−*-N and NO3− between spring, winter and other seasons.
 TN concentrations (0.2–9 μg m−3, av. 2.5 μg m−3) of this study are comparable with those (0.6–16 μg m−3, av. 3.1 μg m−3) reported for aerosol samples collected from April 2001 to March 2002 at Gosan site [Kawamura et al., 2004]. The contributions of TN to aerosol mass range from 0.7% to 7.3% with an annual mean of 3.2% (Table 3). The concentrations of NH4+ range from 0.1 to 6.1 μg m−3 (av. 1.7 μg m−3), whereas those of NO3− range from 0.7 to 19.5 μg m−3 (av. 4.4 μg m−3) (Table 3). These results are similar to those (NH4+: 0.2–13 μg m−3, av. 2.1 μg m−3 and NO3−: 0.2–22 μg m−3, av. 5.3 μg m−3) reported in aerosols collected between April 2001 and March 2002 at Gosan [Kawamura et al., 2004]. The concentrations of NH4+ and NO3− are reported to be 1.4–3.4 μg m−3 and 1.9–4.2 μg m−3, with annual means of 2.3 and 3 μg m−3, respectively, in Gosan aerosols collected during PEM-West B campaign in 1993 (NASA Langley Atmospheric Sciences Data Center, 2010, http://eosweb.larc.nasa.gov). Although NH4+ data of this study as well as the study of Kawamura et al.  are similar to those of the PEM-West data, the concentrations of NO3− increased from 1993 to 2004 by 60%–77%. This increase in the concentration of NO3− over a decade can be interpreted by an increased emission of NOx (main precursors of NO3− in aerosol) in East Asia, particularly in China, from which a major fraction of air masses are transported to Gosan site (Figure 4). This result is consistent with an increased column concentration of NO2 in eastern China (e.g., Shangdong and Jiangsu) observed by satellite [Richter et al., 2005].
 Higher concentrations of TN and NH4+ in spring than in summer are consistent with the fact that almost all of the air masses in spring are originated from the free troposphere and then transported from the heavily polluted regions in East Asia (Figure 4a and Table 2). The lower concentrations in summer are due to an increased clean air mass transport from the Sea of Philippine, Pacific Ocean, and South China Sea (Figure 4b and Table 2). Lower concentrations in autumn than in summer are associated with the enhanced free tropospheric transport of air masses from long distances (Figure 4c and Table 2). Although major fraction of air masses in winter are descended to the boundary layer and then transported far away from East Asia, they passed over northeastern provinces of China (Figure 4d and Table 2), where emissions from coal and other fossil- and bio-fuel combustion increase significantly in winter (Beijing Statistical Bureau, http://www.bjstats.gov.cn). That is why we have observed the highest mean concentration of NO3− in winter although the concentrations of other species showed the lower values in winter than in spring.
3.2. Seasonality of the δ15N for Total Nitrogen (TN), NH4+*, and NO3−*
 The δ15N values of TN, NH4+*, and NO3−* showed a wide variability among the samples (Figure 5). The δ15N of TN range from +8.3‰ to +23.7‰ (av. +14.7‰) in spring, +12‰ to +26.9‰ (av. +16.9‰) in summer, +9.9‰ to +18.1‰ (av. +15.8‰) in autumn, and +6.8‰ to +19.4‰ (av. +12.9‰) in winter (Table 3). In calculating the average δ15N of TN, we excluded three samples KOS 178, 197, and 220 in spring, summer, and winter, respectively, as they showed unexpectedly higher values, which have never been reported in literature. A significant statistical difference (p < 0.05) was observed in the mean δ15N values between summer and winter.
 The lowest mean δ15N of TN in winter can be partly interpreted by more contribution of TN from coal burning because the particles produced by the combustion of coal are generally more depleted in 15N (δ15N < 0) in comparison to other sources (gasoline, natural gas and waste incinerators) (Figure 6a, top) [Widory, 2007]. The consumption of coal is significantly enhanced in winter in East Asia, especially in China (http://www.bjstats.gov.cn) and air mass transport patterns also suggest that most of the air masses in winter are transported to Gosan site from northeastern China (Figure 4d). But the δ15N values of TN and NH4+* observed are significantly higher in winter than those of primarily emitted aerosol nitrogen from the combustion of coal and other fossil fuels, except for bio-fuels (Figure 6a). This suggests that the combustion-derived nitrogen is diluted by other sources and/or relevant atmospheric chemical processes.
 Similar seasonal trends for the δ15N values of TN were reported for aerosols collected from urban Paris and rural Brazil, pointing out to the similar atmospheric processing of nitrogenous species in the atmosphere. The δ15N values of TN in aerosols ranged from +7‰ to +18.7‰ at Piracicaba in São Paulo State and from +8.4‰ to +17.4‰ at Santarém in Pará State, Brazil [Martinelli et al., 2002], where the upper limits are lower than our data. In these sites, the δ15N values are higher than those of vegetation tissues and soil organic matter, which are the potential sources of nitrogen in aerosol particles. The δ15N values in Paris aerosols also show the lower range than ours: +5.3‰ to +16.1‰ with mean values of 10.8‰ ± 3.4‰ in summer and 10‰ ± 3.4‰ in winter [Widory, 2007]. Our Gosan aerosols showed much broader δ15N values (+6.8‰ to +26.9‰) than Brazil and Paris aerosols, possibly due to more contributions of anthropogenic activities and atmospheric processing of nitrogenous species during a long-range transport of pollutants from East Asia to Gosan site.
 The δ15N values of NH4+* are +9.9‰ to +32.2‰ (av. +17.2‰) in spring, +13.7‰ to +29.8‰ (av. +20.2‰) in summer, +9.5‰ to +25.3‰ (av. +20.7‰) in autumn, and +4‰ to +19.9‰ (av. +11.3‰) in winter (Table 3). There is a significant difference (p < 0.05) in the mean δ15N values of NH4+* between winter and other seasons. As seen in Figures 6a and 6b, the mean δ15N values of TN and NH4+* are higher in summer and autumn than in two other seasons. Higher δ15N in summer and autumn can partly be explained by an enhanced contribution of biomass burning aerosols, which are enriched with 15N (Figure 6a). Actually, summer and autumn are the harvest seasons in China and significant fraction of agricultural straws are burnt [Wang et al., 2009]. Figures 4b and 4c also suggest that 55% (clusters 1 and 2) and 78% (clusters 1, 2 and 4) of total air masses are transported over China in summer and autumn, respectively.
 The δ15N of NO3−* ranged from +1.8‰ to +15.7‰ (av. +9.5‰) in spring, +1.7‰ to +13.6‰ (av. +8.3‰) in summer, +5.6‰ to +18.1‰ (av. +12.2‰) in autumn, and +9.7‰ to +21.6‰ in winter (av. +15.9‰) (Table 3). There exists a significant difference (p < 0.05) in the mean δ15N values between winter and the other seasons. Being opposite to the δ15N of NH4+*, NO3−* showed the lowest mean δ15N values in spring and summer (Figure 6c). These results suggest dominant contributions of NOx from gasoline and diesel combustion in spring and summer and from coal combustion in winter. The δ15N of NOx from gasoline and diesel combustion is lower than that from coal combustion (Figure 6c). The seasonal trend of δ15N at Gosan site is similar with the previously reported studies as shown in Figure 6c. For example, the δ15N values of NO3− that were reported for coastal aerosol samples from Weybourne (UK) [Yeatman et al., 2001b] and moderately polluted aerosol samples from Jülich (Germany) [Freyer, 1991] showed lower mean values in summer than in winter. Freyer  explained the seasonal trend in Jülich aerosol considering equilibrium isotope exchange reactions between oxinitrogen species including NO and NO2, and NO2 and N2O5 in the atmosphere. Their explanation was strongly supported by the higher δ15N of NO3− (range: −2‰ to +18‰) than that of NO2 (range: −15‰ to −5‰).
Freyer et al.  observed higher δ15N of NO2 in winter than in summer in the presence of high concentrations of NOx (∼10–45 ppbv) and moderate concentrations of O3 (∼4 to 40 ppbv) in the atmosphere. This trend was explained by the fact that in this condition in winter, an enhanced equilibrium isotopic exchange occurs between NO2 and N2O5 due to longer nights and reduced photochemical activity. Therefore, the dissociation of N2O5 to NO and NO2 in the daytime produces NO2 with higher δ15N in winter.
 In our study site at Gosan, NO3− is transported from China, Korea and Japan (Figure 9), where NOx levels can exceed ozone levels. Hence, equilibrium isotope exchange could occur between reactive nitrogen species in the atmosphere, which drives the seasonal cycle of δ15N of NO2 and subsequently δ15N of NO3− in Gosan aerosols. Moreover, NOx concentration generally maximizes and ozone concentration minimizes in winter in East Asian locations [Han et al., 2009; Li et al., 2007] from China, Korea and Japan. This condition favors higher δ15N of NO2 and NO3−. Freyer et al.  also observed the higher δ15N of NO2 when NOx concentrations exceeded those of O3.
3.3. Aerosol Chemistry and Its Relation to the δ15N of Nitrogenous Species in Aerosols
 On the basis of the Ca2+/Na+ ratios, we divide our 84 samples into three categories: (1) Na+ enriched samples, (2) moderately Na+ enriched samples, and (3) Ca2+ enriched samples. NO3− predominantly exists in the coarse mode due to the reactions of HNO3 with sea salt and dust particles. We observed that the mean δ15N of NO3−* is higher by 3.6‰ in Na+ enriched samples than in Ca2+ enriched samples, suggesting that 15N-depleted HNO3 is preferentially taken by Ca2+ that may originally exist as CaCO3 and hence NO3− produced from this reaction becomes depleted in 15N (Figure 7a). It could also be related with a fact that nitrate, with a different loading of Ca2+ and Na+ in aerosols, is produced from different sources. Nitrogen isotopic discrimination, during NO3− formation in the aerosol phase via the reactions between dust/sea-salt particles and HNO3, has not been reported previously. There is evidence that the δ15N (NO3−) (+4.2‰ to +8‰) in the fine aerosols (<3.5 μm) is higher than that (0.6‰–5.5‰) in the coarse aerosols (>3.5 μm) [Freyer, 1991]. However, Morin et al.  did not observe any significant difference between the δ15N (NO3−) from different size classes of aerosols. Therefore, further studies are required to better understand the nitrogen isotopic fractionation during NO3− production in the aerosol phase.
 Na+/(Cl− + NO3−) ratio in the aerosols produced from sea spray is 0.85 [Wall et al., 1988; Zhuang et al., 1999]. Any deviation from this ratio is an indication of additional sources and/or atmospheric reactions of Na+, NO3−, and Cl−. Some coarse Na+ may originate from soil or combustion sources, and some coarse Cl− or NO3− may evaporate, resulting in ratios >0.85. On the other hand, Na+ should not evaporate under ambient atmospheric conditions. Thus, the lower ratios observed in our samples are possibly due to the reaction of HNO3 with dust particles containing Ca2+. The mean δ15N of NO3−* for samples with the lowest Na+/(Cl− + NO3−) ratios (0.2–0.5) is lower by 2.6‰ than that for samples with higher Na+/(Cl− + NO3−) ratios (0.5–0.85) (Figure 7b). This is consistent with the relation between δ15N of NO3−* and higher Ca2+/Na+ ratios (Figure 7a). Na+/(Cl− + NO3−) ratios in a number of samples are higher than 0.85, which can be caused by the removal of Cl− as HCl due to the reactions between H2SO4 and NaCl assuming that Na+ does not have other sources. The average δ15N of NO3−* for this sample group is 9.9‰, which is equivalent to the sample group with the lowest Na+/(Cl− + NO3−) ratios (Figure 7b).
 The δ15N values of NH4+* were found to decrease with an increase in NH4+*-N/TN concentration ratios in aerosols (Figure 7c). In contrast, the δ15N values of NO3−* increased with an increase in NO3−*-N/TN concentration ratios (Figure 7d). A test of Pearson's correlation (α = 0.05) shows that the correlations between NH4+*-N/TN and the δ15N are significant in spring (r2 = 0.35, n = 37), autumn (r2 = 0.58, n = 10), and winter (r2 = 0.52, n = 24), except in summer (r2 = 0.10, n = 13). Similar statistical test for the correlations between NO3−*-N/TN and the δ15N are significant in spring (r2 = 0.43) and summer (r2 = 0.67), but not in autumn (r2 = 0.24) and winter (r2 = 0.06). The trends observed in Figures 7c and 7d may suggest that there exists a difference in the isotopic shift between NH4+ and NO3− formations. These results may also suggest that the isotopic shift during the formation of secondary aerosol nitrogen depends upon the seasons. Although an inversed relation was observed between TN and its δ15N in urban Paris aerosols in winter, there was a positive relation in summer [Widory, 2007]. These trends were explained by season specific secondary aerosol nitrogen formation.
3.4. Unusual Values of the δ15N of TN and NO3−
 The δ15N values of TN for samples KOS 178 in spring, KOS 197 in summer and KOS 220 in winter are +48.9‰, +48.4‰, and +183.6‰, respectively. Although such higher values have not been reported so far, we have confirmed that these values are real based on repeated analyses (4 times). Interestingly, such high δ15N values were not observed for NH4+* in these samples, suggesting that removed NO3− is heavily enriched with 15N. We calculated the δ15N of NO3−* in these samples using isotopic mass balance equation. Unexpectedly high values (KOS 178: +132‰, KOS 197: +244‰ and KOS 220: +464‰) were derived for these samples. Probably, “flash” (really fast) reaction such as denitrification of NO3− in atmospheric aerosols, to some extent, could explain such type of higher values of NO3−.
 Nitrate in groundwater of the western Kalahari of southern Africa is converted to N2 by bacterial denitrification [Heaton, 1985]. The isotopic separation/enrichment factor (ɛproduct-reactant) for denitrification was found to be −35‰ [Vogel et al., 1981]. Cline and Kaplan  reported similar values of −40‰ to −30‰ for oceanic denitrification. Laboratory experiments have produced values from −33‰ to −10‰ depending on the experimental conditions influencing bacterial metabolism [Mariotti et al., 1982]. Because of the denitrification, remaining NO3− in aerosols becomes progressively enriched in 15N. But, this requires anaerobic condition. Therefore, we need to assume that aerosols in those samples could be coated with hydrophobic materials that create a barrier of microlayer to prevent the penetration of oxygen from the air into aerosols. Further studies are required to explain these unexpectedly high δ15N values in the future.
3.5. Variability of Chemical and Isotopic Composition Among Trajectory Types
 To characterize the observed NH4+ and NO3− in the different air masses encountered during the collection of samples, the mean air mass back trajectory for each sample was calculated. The most of the air mass trajectories are generally tagged with higher concentrations of NH4+ in spring, summer and winter (Figure 8). For example, only 19%, 46%, and 33% of samples in spring, summer and winter, respectively, showed lower concentrations of NH4+ than the observed mean concentration of NH4+ in autumn. Most of the air mass trajectories in autumn, except for 3, originated from Siberia with a transport pathway over northeast China. This result suggests that NH4+ transport from northeast China to Gosan is significantly reduced in autumn.
 Although the variability in NO3− concentrations among the trajectories is similar to NH4+ in spring and winter, there is a difference in summer and autumn (Figure 9). Majority of the air masses are characterized by lower concentrations of NO3− in summer than in autumn. Only 27% samples in spring and 33% samples in winter showed the concentrations lower than the observed mean concentration in summer. Although air mass transport patterns in winter and autumn are very similar, only a few samples in winter showed the concentrations of NO3− to be lower than the mean concentration of NO3− in autumn. Thus, the present study indicates that northeast China as well as Siberia is an important source region for NO3− in winter aerosols.
 Higher δ15N values of NH4+* in warm season (March–August) are characterized by the trajectories arriving at Gosan site from the east, southeast, south, southwest, and by the stagnant air masses that remain in the vicinity of Jeju Island (hereafter referred to as S/SE/SW transport) (Figure 10). In contrast, lower δ15N in cold season (September–February) is characterized by the rapid transport of air parcels from northeastern China and Siberia to the west and northwest of Gosan site due to strong westerlies (hereafter referred to as W/NW transport).
 The mean δ15N of NH4+* during S/SE/SW transport in warm season (March–August) is higher by 7.4‰ than that during W/NW transport in the later half of cold season (winter: November–February) (Table 3). We expected to find lower δ15N values of NH4+* for S/SE/SW transport because of the preferential emission of 14NH3 over 15NH3 in warm season from soil, fertilizer, and animal excreta [Russell et al., 1998], as depicted in Figure 6b for lower δ15N values of NH4+ in aerosol and rain samples of Pretoria, South Africa. However, higher δ15N values for the samples of S/SE/SW transport and for the samples of the first half of W/NW (autumn) transport were found in Gosan site (Figures 10a, 10b, and 10c), again suggesting that an increased emissions from biomass burning and some relevant atmospheric chemical processes may determine the nitrogen isotopic ratios of the NH4+. The isotopic fractionation under equilibrium conversion of NH3(gas) to NH4+(aqueous) may likely occur, which was estimated to be in the range of +25‰ to +35‰ [Mariotti et al., 1984; Heaton, 1987]. This estimation can be supported by higher δ15N of NH4+ in atmospheric aerosols collected from barnyard [Yeatman et al., 2001b].
 The mean δ15N value of NO3−* during S/SE/SW transport in warm season is lower by 5.6‰ than that during W/NW transport in cold season (Table 3 and Figure 11). Higher δ15N values of NO3−* in W/NW transport can be explained by an enhanced contribution of NOx from coal (NOx from coal combustion is enriched with 15N, see Figure 6c) and by the difference in NOx scavenging mechanism between warm and cold seasons. The enhanced contribution of coal NOx is associated with its increased consumption; e.g., total residential energy consumption in Beijing during cold season is 54% higher than that of warm season in 2006 (http://www.bjstats.gov.cn). The NO3− produced via an enhanced reaction between NOx and OH radicals in warm season is depleted with 15N whereas it is enriched with 15N produced via an enhanced reaction between NOx and NO3 radicals in cold season [Freyer et al., 1993].
 We found that total nitrogen (TN) concentrations in atmospheric aerosols collected from Gosan site, Jeju Island, are high (0.2–9 μg m−3) with an average of 2.5 μg m−3, in which TN is mainly composed of NH4+ (annual mean: 56.3%) followed by NO3− (42.3%). Significant seasonal variations of the δ15N of TN, NH4+*, and NO3−* were found in the Gosan aerosols, which we interpreted by seasonal differences in the sources and chemical processing of nitrogenous species in the atmosphere. It appears that lower δ15N values of NO3−* are caused by less isotopic enrichment (ɛproduct-reactant) during the reactions between HNO3 and dust particles. This study also demonstrates that a significant isotopic enrichment of 15N occurs in aerosols during the gas-to-particle conversion (NH3 → NH4+) and the subsequent gas/aerosol partitioning of nitrogenous species. We propose that the δ15N of the different species in aerosols is useful tool to interpret the secondary production of aerosol nitrogen in the atmosphere and to better understand the changes in the sources and source strengths of nitrogenous species in the atmosphere with seasons.
 This study was in part supported by the Japanese Ministry of Education, Science, Sport and Culture (MEXT) through grants-in-aid 14204025 and 17340166 and the Environment Research and Technology Development Fund (B-0903) of the Ministry of Environment, Japan. S.K. acknowledges financial support from the MEXT. The authors are grateful to NOAA Air Resources Laboratory (ARL) for allowing the installation of registered version of Windows-based HYSPLIT model. We are also grateful to NASA Langley Atmospheric Sciences Data Center for supplying the data in a CD-ROM for PEM-West A and PEM-West B campaigns.