Sampling site description

The measurement site was located on the western tip of Cheju Island (33°17′N, 126°10′E, Fig. 1). A trailer containing all the gas analyzers and a particle sampler was situated about 10 m inside from a cliff of about 70 m above mean sea level. Both sampling inlets were located about 6 m above the ground. The Cheju upper-air meteorological station is located 100 m northeast of the trailer and upper-air meteorological parameters have been measured twice a day and other meteorological parameters measured routinely.

The emissions of SO₂ and NO_x from Cheju Island are quite low, amounting to only 0.7% and 1.1% of the total emissions from Korea, respectively, or about one-tenth of the emissions of both species from Seoul (MOE 1994). Cheju Island is popular for tourism; however, the western side of the island is less developed than other parts of the island and, therefore, has less traffic. The nearest village is about 1 km east of the site, and the nearest highway is about 2 km east of the site, with traffic flow of about 3000 vehicles per day. The effects of local emissions on gaseous measurements are discussed below.

Measurements

Both aerosol and gaseous air pollutants were measured. An automatic high-volume sampler (Kimoto 195A) was used to collect the TSPs. Particles were collected on a Teflon tape filter surface that was located at a sampling probe for either 6 or 24 h and then a new clean filter surface was moved to the sampling probe on which particles were collected. The flow rate was about 170 L min⁻¹. After 6 h of data collection, daily average values were calculated. The filters were analyzed for water soluble ions: SO²⁻₄, NO⁻₃, Cl⁻, NH⁺₄, Na⁺, K⁺, Ca²⁺, and Mg²⁺. Anions were analyzed by an ion chromatography (Dionex DX-100), NH⁺₄ by the indophenol colorimetric method (Diode Array 8541A), and other cations by atomic absorption spectroscopy (Philips SP9-800). Gaseous species were measured by continuous analyzers: O₃ by a UV photometric (TECO 49), SO₂ by a UV fluorescence (TECO 43S), and NO_x by chemiluminescence (TECO 42). In the NO_x analyzer, nitrogen oxides are converted to nitrogen monoxide. Some fractions of nitric acid and other organic nitrates are also converted to nitrogen monoxide; however, the conversion efficiencies for those species were not established. Therefore, the measured nitrogen oxides (NO_x) include some other species that make up NO_y. Data were stored both in a computer and on strip chart paper. Calibrations were performed at least once a week by a calibrator (TECO 146) and a zero air generator (TECO 111) with high-purity standard gases.

Aerosol

The overall mean values, standard deviations, and ranges of daily average concentrations of water soluble ions in TSPs are given in Table 1. Also shown in Table 1 are the data for nss ions. The nss concentrations are estimated based on sodium ion and seawater composition by Horne (1969) was used for estimation.

The aerosol collected in the marine environment are subject to sampling artifacts and may not accurately represent the actual composition. One example is the evaporation of Cl⁻ to HCl and the resultant negative concentration of nss Cl⁻; this could occur in the ambient air (Möller 1990) or on the filter due to the interactions among Cl⁻, acidic gases, and alkaline sea salts. Other volatile species, such as NO⁻₃ and NH⁺₄ ions, also can experience the same kinds of sampling artifacts. In the case of nitrate, the measured NO⁻₃ concentration is generally higher than the actual one because of interactions between gaseous HNO₃ and sea salt in the particle phase (Kaneyasu et al. 1995). Ammonium in the particles may evaporate due to the high alkalinity of sea salt [the average pH of surface seawater is about 8.2 (Horne 1969)]. Kim et al. (1995) applied a gas/particle equilibrium model, Simulating Composition of Atmospheric Particles at Equilibrium, to the measurements at Kosan between 11 and 17 March and predicted higher NH⁺₄ and lower NO⁻₃ and Cl⁻ concentrations than the measured values. Therefore, the Cl⁻, NO⁻₃, and NH⁺₄ concentrations of aerosol may contain sampling artifacts owing to reactions with sea salts.

Sulfate was the largest constituent of the water soluble ions in TSPs. The overall mean value was 8.79 μg m⁻³, which represented 52% of the total water soluble ion concentration by mass. Chloride and sodium were the next largest constituents, reflecting the fact that the site is near the sea. The arithmetic mean value of the nss sulfate concentration was 8.30 μg m⁻³, which represented 94% of the total sulfate concentration. Similar values had been observed at the same site by other investigators. For example, Lee et al. (1995) observed nss sulfate concentrations of 8–10 μg m⁻³ by the same sampler between March and April 1992, while obtaining the overall mean value of 6.53 μg m⁻³ for the period of March and December 1992.

These values were far higher than the average nss sulfate concentration measured at various other places in the Pacific Ocean [0.2–1.5 μg m⁻³ (Fitzgerald 1991)] and that of PM2 particles at remote marine regions [about 1 μg m⁻³ (Heintzenberg 1989)]. The range of sulfate concentrations measured in Bermuda by Wolff et al. (1986) and Hastie et al. (1988) was between 2.69 and 3.96 μg m⁻³. The sulfate concentrations in Japanese remote sampling sites were 3.59 μg m⁻³ in Oki Island for TSP (Mukai et al. 1990), and between 0.2 and 4.2 μg m⁻³ in the Pacific (Okita et al. 1986). In comparison, the average concentration of sulfate in Seoul, the capital of Korea, with about 10 million inhabitants, is between 5 and 10 μg m⁻³, which is comparable to this result (Baik et al. 1996).

The concentrations of crustal species, Ca²⁺, K⁺, and Mg²⁺ ions measured during this period were also higher than those measured at Oki Island (Mukai et al. 1990). However, their contribution to the total ion concentration were small. The average nss Ca²⁺ and nss K⁺ concentrations were 3.7% and 2.6% of the total ion concentration, respectively. Their mean concentrations were comparable to those reported by Carmichael et al. (1996) at the same site during the springs of 1992 and 1993.

Because the sampling site is in the marine environment, the contribution to nss sulfate concentration from biogenic sources should be assessed. The sulfur species from the biogenic sources were not measured during this study. During the “PEM-West A” measurement, however, methane sulfonic acid and nss sulfate concentrations in aerosol were measured at the Kosan site between April 1992 and February 1993 (Arimoto et al. 1996). They determined that the biogenic fraction of the nss sulfate was about 11%. Also, Luria et al. (1989) concluded that in the Gulf of Mexico area, dimethyl sulfide (DMS) emission can account for about 1 μg m⁻³ of sulfate. Furthermore, they concluded that in the Bermuda area less than 50% of the observed sulfate (1.9 μg m⁻³ for the TSPs), or 1 μg m⁻³ of sulfate, can be related to the natural sulfur cycle. Therefore, it was a reasonable estimation that less than 1 μg m⁻³ of the measured nss sulfate has originated from the biogenic sulfur cycle.

In summary, the observed water soluble ion concentrations during this period were higher than those measured at remote marine areas but were comparable to those at Seoul, a highly urbanized area. Since the local emission rates of air pollutants are quite low, it is likely that the observed high values of aerosol ion concentrations can be accounted for by the transport of air pollutants from outside Cheju Island, and this will be discussed later.

Gaseous species

The overall mean values, standard deviations, and ranges of daily average concentrations of gaseous species, such as SO₂, O₃, NO, and NO_x are also given in Table 1. The NO concentration during the period was mostly below the detection limit of the analyzer. The overall mean value of NO_x was 3.5 ppb. The mean value of NO_x was far higher than that at Oki Island (Jaffe et al. 1996).

The maximum hourly concentration of SO₂ during the period was less than 10 ppb, the maximum daily average value was 3.5 ppb, and the overall mean value was 0.97 ppb. These values were far lower than those observed at Seoul. For example, the yearly mean concentration of SO₂ in Seoul during 1993 was 23 ppb. Similar or slightly higher values for SO₂ were observed by other researchers at the same site. During March and April 1992, the average concentration of SO₂ was about 1 ppb (SERI 1993) and between February and December 1992 it was 1.42 ppb (Suh et al. 1995). The SO₂ concentration at Kosan, however, was higher than other clean areas. The SO₂ concentration in Oki Island in November 1992 was about 0.2 ppb (Hatakeyama 1994) and in Bermuda in spring of 1985 it was 0.27 ppb (Hastie et al. 1988).

On the contrary, the ozone concentration at Cheju was quite high. The overall mean value of O₃ was 55 ppb and the maximum daily average concentration was 73 ppb. The O₃ concentration at this site is one of the highest in Korea. Similarly, high ozone concentrations had also been observed in Japanese mountainous regions (Sunwoo et al. 1994) and in the Atlantic Ocean with the altitude of somewhere between 35° and 40°N (Thompson 1994) during the springtime. In Oki Island, Japan, lower values than this study, the mean values of 36.1 ppb in September and 42.2 ppb in October 1991, were observed (Akimoto et al. 1996). This discrepancy between two sites might be caused by two factors. First, Kosan is nearer to the Asian continent than Oki Island and thus more affected by photochemical reactions among air pollutants from it. Second, ozone concentrations in the region are highest during the springtime due to the mixing of the stratospheric ozone to the troposphere caused by the jet stream (Sunwoo et al. 1994). Since the measurements at Oki (Akimoto et al. 1996; Jaffe et al. 1996) were carried out during the fall, while this study was carried out during the spring, the ozone concentrations at the two sites might be different.

Since the effects of local emissions of SO₂ and NO_x to the concentrations of gaseous species have not been studied yet, those should be analyzed to interpret the trend of the gaseous species concentrations. During the period, three distinctive patterns of O₃ and NO_x concentration variations have been observed, as shown in Fig. 2. A typical case showing variations over less than 1 h is shown in Fig. 2a. During part of the period, the NO_x concentration increased sharply, while the O₃ concentration decreased accordingly. This can happen when a locally emitted NO plume is titrated with high O₃ concentration (Finlayson-Pitts and Pitts 1986; Sexton and Westberg 1979). Since this kind of phenomenon occurs independent of the time of day and SO₂ concentration did not change, it can be concluded that this is caused by a local NO source located very near to the site, probably automobile traffic.

Figure 2b shows another type of variation that is similar to the previous case, but it lasted a few hours longer. The SO₂ concentration over this period sometimes increased, but generally remained constant. It is likely that this phenomenon is influenced by NO emissions from other areas of Cheju. Figure 2c shows a case of both NO_x and O₃ concentrations increasing simultaneously with a time span of about 10 h. This case occurred mainly during the daytime and can be interpreted as the transport of polluted air mass from the outside of the Cheju Island. The SO₂ concentration shows no apparent trends in concentration during the period.

In summary, NO_x concentrations are affected by local emissions and, thus, are higher than that at Oki Island. The effects of the local emissions to SO₂ concentration were small. The aerosol ion concentrations, especially sulfate, at Kosan were high, comparable to those at Seoul, while the concentrations of SO₂ and NO_x at Kosan were much lower than those at Seoul. It is a typical characteristic of long-range transport of air pollutants, and a similar observation was made in remote islands in Japan (Akimoto et al. 1996).

Backward trajectory analysis

Since the air pollutant concentrations strongly suggest these are transported from outside Cheju Island, it is important to determine their transport path. Four-day backward trajectories of air parcels arriving at Kosan at 2100 LST were prepared using the Gridded Atmospheric Multilevel Backward Isobaric Trajectories model of NOAA (National Oceanic and Atmospheric Administration) (Harris 1982). The meteorological data used for this analysis were gridpoint values from the Japanese Meteorological Agency. The reason that trajectories at 2100 LST were calculated was that they corresponded approximately to the midpoint of the filter sampling period. Trajectories at the 700-, 850-, and 1000-hPa levels were estimated.

The northeastern Asia region was divided into five sections, as shown in Fig. 3. Section I corresponds to air parcels traveling from the Korean peninsula, and section II is from northern China. Section III corresponds to middle and southern China, including Taiwan;section IV is from the Pacific Ocean. Finally, section V is from Japan. In Table 2, the number of days and the location of air parcels represented by the 850-hPa trajectories at one, two, three, and four days before arriving at Kosan at each section are given. About 55% of air parcels at 850 hPa have traveled from section II, 30% from section III, 10% from section I, 7% from section IV, and none from section V. There was a strong time dependence of air parcel paths. During March, more than 80% of air parcels moved from section II. In April, about half of the air parcels moved from section III, while about 20% of air parcels moved from section II. Dominance of air parcels from sections II and III (more than 80% of the total days) is due to the westerlies, the prevailing winds for this region in spring. Thus, high concentrations of anthropogenic air pollutants, such as nss sulfate, can be attributed to the transport of air pollutants from the Asian continent.

The average concentrations of nss SO²⁻₄, nss Ca²⁺, and O₃ classified by backward trajectory sections two days before the arrival at the sampling site are shown in Table 3. It is apparent that the average concentrations of nss Ca²⁺, nss K⁺, and O₃ for the air parcels from section II, northern China, were slightly higher than those from southern China and other areas. The average concentration of nss Mg²⁺ for the air parcels from section II was also higher than that in section III. The high crustal species level for the air parcels from section II is due to the transport of dust particles from northwestern China that includes arid and semiarid areas from which dust particles are originated. The higher average ozone concentration for air parcels from northern China is probably due to the photochemical reactions on surface of dust particles (Zhang et al. 1994). To the contrary, the average concentration of nss SO²⁻₄ from southern China was slightly higher than that from northern China, due to less natural sources like dust particles in southern China than in northern China.

Relationship between air pollutant concentrations

The correlation coefficients between the ions and gaseous species concentrations are shown in Table 4. The nss sulfate, which is considered an indicator of anthropogenic air pollutants in the region (Carmichael et al. 1996), showed a strong relationship with nss K⁺ (r = 0.873), NH⁺₄ (r = 0.794), O₃ (r = 0.703), and SO₂ (r = 0.661). On the other hand, nss Ca²⁺, which is considered an indicator of natural dust in this region (Arimoto et al. 1996; Carmichael et al. 1996), showed a strong relationship with nss K⁺ (r = 0.790). Thus, it is likely that NH⁺₄, O₃, and SO₂ are transported along with the nss sulfate.

In this study nss SO²⁻₄ and NO⁻₃ showed no apparent relationship, as shown in Table 4. This is in contrast to Arimoto et al.’s result (1996) at the same site between 1992 and 1993. They reported that nss SO²⁻₄ and NO⁻₃ showed a close relationship (r = 0.78). The reason for this discrepancy is not clear.

It is generally assumed that main sources of K⁺ in the region are natural sources, mostly dust particles (Carmichael et al. 1996). The nss K⁺ in this study showed a strong relationship with both nss SO²⁻₄ (r = 0.873) and nss Ca²⁺ (r = 0.790). Also, nss SO²⁻₄ and nss Ca²⁺ show a strong relationship with the correlation coefficient value of 0.742. There are two possibilities that the relationship between nss SO²⁻₄ and nss Ca²⁺ might be high despite their different sources. One possibility is that they have a common air path, and the other is the reactions between dust particles and SO₂ (Carmichael et al. 1996) and a resultant close relationship between nss SO²⁻₄ and nss Ca²⁺.

The trends of nss SO²⁻₄, nss Ca²⁺, and nss K⁺ are shown in Fig. 4. The nss Ca²⁺, nss K⁺, nss Ca²⁺, and nss SO²⁻₄ generally showed good correlation in March when majority of air parcels moved from northern China with both anthropogenic and crustal sources. But they showed some discrepancy in April when the majority of air parcels moved from southern China with dominantly anthropogenic sources. It is further confirmed when correlation coefficients among three species based on air parcel paths are calculated (Table 5). In Table 5, the correlation coefficients between nss SO²⁻₄ and nss K⁺, and nss Ca²⁺ and nss K⁺ for sections II and III are shown. When air parcels moved from northern China, the relationship between nss Ca²⁺ and nss K⁺ was higher, while the relationship between nss SO²⁻₄ and nss K⁺ was higher when air parcels moved from southern China. It implies that nss K⁺ might have two originations:one anthropogenic and the other natural dust. One example of anthropogenic source for nss K⁺ is biomass burning (Artaxo et al. 1990).

To estimate the contributions from anthropogenic and natural sources to the concentration of nss K⁺, scattergrams of nss K⁺ and nss Ca²⁺, and nss K⁺ and nss SO²⁻₄, as shown in Fig. 5, were prepared. From the scattergrams, it can be seen that the nss K⁺ intercepts when nss Ca²⁺ and nss SO²⁻₄ are zero are 0.14 μg m⁻³ and 0.02 μg m⁻³, respectively. It means that the contribution from anthropogenic sources (when nss Ca²⁺ becomes zero) are higher than that from natural dust sources (when nss SO²⁻₄ becomes zero) by a factor of 7. Evidently this kind of analysis is a first estimation from measurements during only 40 days and more measurement data are necessary to make a more accurate estimation. It shows, however, that nss K⁺ collected at the site has more dominant anthropogenic sources than natural ones.