4.1. Seasonal Variation
 The reason for the high concentrations of NOy(g) and PM10NO3− observed in spring is considered to be due to air masses transported from continental China, which are associated with the eastward movement of low- and high-pressure systems from China. It is suggested that the transport of air pollutants is related to the movement of low- and high-pressure systems [Uno et al., 1998; Takami et al., 2006b]. Hatakeyama et al.  reported that highly polluted air was transported from central China to the East China Sea when a low-pressure system originating near Taiwan moved along Japan to the northwest Pacific Ocean. In addition, highly polluted air was observed when a high-pressure system moved from central eastern China to the northern East China Sea [Hatakeyama et al., 2004]. The travel of a cold front with a low-pressure system lifts the air pollutants ahead of the front and the continental outflow of precursors occurs behind the front. A high-pressure system confines air pollutants within the lower troposphere by the strong downward motion of air. These highly polluted air masses are transported eastward by westerly winds [Carmichael et al., 1998; Bey et al., 2001; Liang et al., 2004].
 Back trajectory analyses show that air masses were often transported from continental China in spring when the concentrations of NOy(g) and PM10NO3− were high (Figure 3a). In this case, a low- or high-pressure system originating in continental China moved toward the east or southeast and covered the whole of Okinawa Island, where CHAAMS is located. This indicates that the transport of air pollutants from China was frequently accompanied by movement of a low- or high-pressure system. The lidar results showed that dust plumes often reached CHAAMS in spring. In this case, too, it was suggested that air masses were transported from China on the basis of back trajectory analyses and weather patterns as described above. In events of China origin, the average concentrations of NOy, NOy(g), PMcNO3−, and PMfNO3− during dust events were higher than those of nondust events (Table 1). In particular, the rate of increase in PMcNO3− was highest and rose to about 1.7 and 2.0 times in comparison with the nondust events of China origin and the average for all observation periods, respectively. It is considered that alkaline dust particles can take up acids resulting in increased coarse mode NO3− [Song and Carmichael, 2001; Jordan et al., 2003]. Therefore, the increase of PMcNO3− at CHAAMS during dust events is closely related to the transport of dust from the continent. Thus, it can be concluded that the high concentrations of NOy(g) and PM10NO3− in spring were due to the frequent transport of air masses from continental China by low- and high-pressure systems including dust events.
 The low concentrations in summer are mainly due to cleaner air masses from the Pacific Ocean. Okinawa Island is usually covered by the Pacific high-pressure system in summer, in which air masses are frequently transported from the Pacific Ocean where no anthropogenic emission is expected. In this way, the seasonal variation is mainly controlled by the air mass history associated with the weather pattern for each season.
 The fraction of PM10NO3− in NOy in summer was 10%, which was the lowest among all seasons. This seems to be related to the air mass history and the time spent in the atmosphere. There are no anthropogenic sources in the Pacific Ocean, so less NOx and NH3 are available to produce particulate NO3−. NH4NO3 cannot easily be maintained because the average summer temperature is about 30°C at CHAAMS. Also, most of the PMcNO3− is deposited before reaching CHAAMS because air masses stay in the marine atmosphere for a long time. This is why the fraction of PM10NO3− in NOy was lowest in summer, and the major species of nitrogen compounds in air masses was NOy(g).
4.2. Chemical Transformation of Reactive Nitrogen During Transport
 When gas phase acids are emitted, they are oxidized by reaction with photochemically produced hydroxyl radical (OH), and then form secondary aerosols through the processes of nucleation, condensation/evaporation, and absorption of atmospheric inorganic species, which are partitioned into gaseous and condensed (particles) phases. Secondary aerosols can also form by heterogeneous reaction. For example, uptake of SO2 to existing aerosol produces H2SO4. The liquid phase reaction of SO2 and H2O can form SO42− in aerosols. As for reactive nitrogen, HNO3 is produced via reaction between NO2 and OH radical, and particulate NO3− is formed by adsorption of HNO3 in existing particles. In this section, we analyze the chemical transformation of particulate NO3− during transport from Qingdao, China to CHAAMS, Okinawa.
 We set a grid of 1.5° × 1.5° around Qingdao, and air masses that passed the grid were defined as being of Qingdao origin. The selected air masses reached CHAAMS via this region from 5 to 24 April 2006, and the air masses the meandered after passing through this region were excluded from the analysis. The selected sample number was 63 (about 6 days including 3 dust events) in this manner. For the selected air masses, the concentrations of NOx, PMfNO3−, and PMcNO3− at Qingdao were compared with those at CHAAMS. Figure 5 shows the variation in nitrogen oxide as the transport time increased. In this analysis, the transport time = 0 was set when air masses passed Qingdao, where NOy(g) was defined by NOx and NOy was defined by the sum of NOx and PM10NO3− (section 2.2). The transport time was defined as the time taken for air masses to travel from the Qingdao region to CHAAMS, which is the time spent over the sea and is calculated using HYSPLIT4.
Figure 5. Variation in nitrogen oxide and sulfur oxide with transport time for concentrations of (a) NOy(g) (solid circles) and PM10NO3− (open circles), (b) NOy(g)/NOy (solid circles) and PM10NO3−/NOy (open circles), (c) PMfNO3−/PM10NO3− (open squares) and PMcNO3−/PM10NO3− (solid squares), (d) concentrations of SO2 (solid triangles) and nss-SO42− (open triangles), and (e) SO2/SOy (solid triangles) and nss-SO42−/SOy (open triangles).
Download figure to PowerPoint
 The concentrations of NOy(g) and PM10NO3− continuously decreased as the transport time increased (Figure 5a). This is due to dry and/or wet deposition and dilution with other air masses during transport. Figure 5b shows that the variation in fractions of NOy(g) and PM10NO3− in NOy. As the transport time increased, the fraction of NOy(g) continuously decreased, whereas that of PM10NO3− continuously increased. This suggests that nitrogen oxides in the gas phase were converted to particulate NO3− during transport. As for the size of particulate NO3−, the fraction of PMcNO3− in PM10NO3− increased, while that of PMfNO3− decreased (Figure 5c). In other words, most of the converted particulate NO3− was PMcNO3−.
 The chemical transformation process of particulate NO3− during transport is considered as follows. In China, large amounts of NH3 are emitted through agricultural activities such as fertilizer application [Streets et al., 2003] in addition to NOx (NO + NO2) emission from fossil fuel combustion and biomass burning. HNO3, which is formed from the oxidation of NOx, produces NH4NO3 through the reaction with NH3. The existence of NH4NO3 indicates that NOy contains both NOy(g) and PMfNO3−. In addition, the fraction of PMfNO3− in PM10NO3− is higher than that of PMcNO3−. Thus it is considered that the main components of NOy were NOy(g) and NH4NO3 at the Qingdao area. On the other hand, the fraction of PMcNO3− in PM10NO3− was higher than that of PMfNO3− at CHAAMS. This is explained as follows. The decomposition of NH4NO3 to HNO3 and NH3 occurs while air masses are approaching CHAAMS, since it is located in the southern area where the temperature is higher. Air masses mix with marine air and/or dust plumes, whereupon HNO3 is absorbed by coarse particles such as sea salt and dust particles. Thus, the fraction of PMcNO3− in PM10NO3− was higher than that of PMfNO3− at CHAAMS. This is supported by the results obtained by Matsumoto et al. , who measured aerosols at Hahajima (140.10°E, 26.38°N) from 1994 to 1997 and reported that the fraction of PMfNO3− (Dp < 1.1 μm) and PMcNO3− (Dp > 1.1 μm) was 2.7 and 97.3%, respectively. The observation site is about 1300 km east of CHAAMS, suggesting that the longer the transport time is under relatively high temperature, the larger the fraction of PMcNO3− in PM10NO3− becomes [Shimohara et al., 2001].
 Figure 6 shows the variation in ratio for each NOy component of China origin with respect to the transport time for all observation periods. In Figure 6, other gases are defined as gas phase nitrogen compounds, which are calculated by subtracting the sum of HNO3 and PM10NO3− from NOy. As the transport time increases, the ratio of PMcNO3− in NOy increased and accounted for half of NOy when the transport time exceeded 48 h (Table 2). The variation of HNO3 was similar to that of PMcNO3−. The ratio was about 10% when the transport time was within 24 h, and rose to 25% when the transport time exceeded 48 h. This suggests that the main components of NOy become HNO3 and PMcNO3− as air masses are transported from Qingdao to CHAAMS. As for other gases, the main component is NOx, which is the precursor of HNO3. The ratio decreased as the transport time increased, suggesting that NOx is converted to HNO3 and particulate NO3− during transport. Although PAN was not measured at CHAAMS, as it is important as a reservoir of NOx [Talbot et al., 2003], it is necessary to consider the temperature dependence of its stability. Miyazaki et al.  reported PAN concentrations and their fraction in NOy with latitudes over the western Pacific during the TRACE-P campaign. They evaluated that both PAN concentrations and the fraction were lower at 10–30°N (189 pptv, 0.20) than at 30–45°N (928 pptv, 0.33), consistent with its chemical stability at lower temperatures. It is considered that the warmer temperature in CHAAMS results in a release of NOx from decomposition of PAN as well as a release of HNO3 from NH4NO3. The average ratio of PMfNO3− in NOy was 4% due to the temperature effect on NH4NO3. In this way, the chemical transformation process of nitrogen compounds is influenced by the transport time of air masses, the geographical position of Okinawa and the temperature dependence of NH4NO3, particularly in particulate NO3−.
Table 2. Ratio of Each NOy Component of China Origin With Respect to Transport Time
|Transport Time, h||HNO3, ppbv||PMcNO3−, μg m−3||PMfNO3−, μg m−3||Other Gases, ppbv||Sample Number|
|0–24||0.11 ± 0.07||0.17 ± 0.05||0.04 ± 0.02||0.68 ± 0.08||31|
|25–36||0.15 ± 0.09||0.20 ± 0.09||0.04 ± 0.02||0.61 ± 0.12||140|
|37–48||0.20 ± 0.08||0.21 ± 0.07||0.03 ± 0.02||0.56 ± 0.10||27|
|49–60||0.25 ± 0.08||0.48 ± 0.09||0.04 ± 0.03||0.23 ± 0.11||22|
|Average||0.15 ± 0.09||0.20 ± 0.09||0.04 ± 0.02||0.60 ± 0.13||220|
 The variation in sulfur oxides is shown in Figures 5d and 5e since their chemical transformation process during transport is very different from that of nitrogen compounds. Here, SOy is defined as the sum of SO2 and nss-SO42−. One of the differences in conversion processes of sulfur oxide and nitrogen oxide is whether volatilization from particles can occur or not. As for nitrogen oxides, HNO3 can be volatilized from particles, which depends on the temperature, relative humidity and chemical composition of aerosols. The volatilized HNO3 is taken up by coarse particles such as sea salt and dust particles, forming particulate NO3− in coarse mode. In contrast, nss-SO42− is continuously formed by gas-to-particle conversion. As the transport time increased, the fraction of SO2 in SOy continuously decreased, whereas that of nss-SO42− continuously increased. This suggests that sulfur oxides in the gas phase are converted to nss-SO42− during transport. Moreover, the production of nss-SO42− is faster than that of particulate NO3−. For example, the fraction of nss-SO42− in SOy is about 60%, whereas the fraction of PM10NO3− in NOy is about 20% when the transport time is less than 24 h. This means that particulate NO3− forms a coarse mode during transport in contrast to nss-SO42− which is transported in a fine mode. Song and Carmichael  reported the aging process of gas phase anthropogenic pollutants and aerosols during long-range transport using a Simulating Composition of Atmospheric Particles at Equilibrium (SCAPE) model [Kim et al., 1993a, 1993b]. The results showed that nitrogen oxide shifts to a coarse mode from gas phase and sulfur oxide shifts to a fine mode from the gas phase, which is consistent with this study.
 In order to examine the influence of the transformation process on the transported amount of pollutants, NOy and SOy concentrations at Qingdao were compared with those at CHAAMS (Table 3). Both NOy and SOy concentrations at CHAAMS are assumed to be from the same air masses that passed over the Qingdao area in April 2006 on the basis of back trajectory analyses. The ratios of NOy and SOy transported are defined as
where Cc is the concentration at CHAAMS, Cq is the concentration at Qingdao and Cb is the concentration of Pacific Ocean origin, which is defined as a background concentration. The results showed that RNOy and RSOy were 10.6 and 26.1%, respectively, suggesting that the fraction of NOy transported from Qingdao to CHAAMS was less than that of SOy. This evaluation focused on the effect of dry deposition, but most of the selected air masses did not experience precipitation, and if any, experienced precipitation of less than 0.5 mm h−1 according to HYSPLIT4 and WXT510. Therefore, it is considered that the effect of wet deposition is negligible in this analysis.
Table 3. Average Concentrations at Qingdao and CHAAMS From 5 to 24 April 2006a
|Place||Gas, ppbv||Aerosol, μg m−3||NOy, ppbv||SOy, ppbv|
|CHAAMS||1.44 ± 0.41||0.74 ± 0.44||0.88 ± 0.34||0.20 ± 0.17||5.36 ± 4.19||1.98 ± 0.55||2.82 ± 1.79|
|Qingdao||11.21 ± 5.12||5.66 ± 3.55||1.34 ± 1.38||2.70 ± 1.65||4.04 ± 2.31||12.67 ± 5.42||7.55 ± 4.24|
 The difference in the fraction transported from China between NOy and SOy is likely caused by the difference in chemical transformation process during transport. The main components of sulfur oxide are nss-SO42− and those of nitrogen oxide are HNO3 and PMcNO3−. Zhang et al.  reported that the dry deposition velocities of SO2, H2SO4 and HNO3 for water surface are 2.1, 2.0 and 2.6 cm s−1, respectively. Therefore, the lifetime is evaluated by the following equation [Morino et al., 2006]:
where τx is the lifetime (τx; s) for species x, and Vx is the dry deposition velocity (cm s−1). H is the boundary layer depth (m) and is assumed here to be 1000 m as the marine boundary layer. τSO2, τH2SO4 and τHNO3 were calculated to be 13.2, 13.9 and 10.7 h, respectively. The dry deposition velocity of fine particles and coarse particles was evaluated by the difference in diameter and was assumed to be 0.03 and 0.22 cm s−1, respectively [Matsuda et al., 2001]. τfine and τcoarse were 925.9 and 126.3 h, respectively.
 The lifetimes of HNO3 and PMcNO3− are less than 0.5 and about 5 days, respectively, whereas that of nss-SO42− is more than a month. SO2 and H2SO4 are considered to be the main deposition form of sulfur oxide for the transport time of air masses from the Qingdao area to CHAAMS (average 24 h). However, the effects of the deposition process of SO2 and H2SO4 on the loss of SOy are minimal because both species are converted to nss-SO42− at the source region and/or during transport. As shown in Figure 5d, it is for this reason that nss-SO42− concentration did not decrease as the transport time increased. On the other hand, the lifetimes of HNO3 and PMcNO3− are shorter than that of nss-SO42−. In addition, decomposition of NH4NO3 in fine particles occurs during transport, resulting in the conversion to HNO3 and PMcNO3−. Thus, the loss of NOy is strongly affected by the chemical forms of HNO3 and PMcNO3−. In this way, the chemical transformation process during transport largely influences the fraction of NOy transported from Qingdao to CHAAMS in addition to the dry/wet deposition and the dilution processes during transport.
4.3. Variation in Particulate NO3− When Dust Plumes Reached CHAAMS
 On 19 March 2006, the concentrations of NOy(g) and PM10NO3− were the highest. Figure 7 shows the variation in concentrations of NOy(g), PMcNO3−, and PMfNO3− from 16 to 20 March 2006. Also shown are the weather conditions (relative humidity and wind direction), PM2.5 mass concentration measured by TEOM and time-height cross section of the extinction coefficient of mineral dust and spherical aerosols by lidar during the same period. Frontal systems associated with a low-pressure system passed over CHAAMS twice during this period. The first one passed at around 0500 UT on 16 March when the wind direction changed from south to north and the relative humidity suddenly decreased. The lidar results show that dust plumes reached CHAAMS around the same time (Figure 7b). As the front was approaching, NOy(g) concentration increased, whereas the concentration of PMcNO3− gradually decreased. The concentration of PMfNO3− increased when the front passed, showing that the variation in PMfNO3− was similar to that of NOy(g). In this case, it is supposed that air pollutants reached CHAAMS earlier than dust plumes. Back trajectory analysis shows that the origin of air pollutants was Shanghai. Similar time lags for air pollutants and dust plumes were also reported by Uematsu et al. , Matsumoto et al. , Hatakeyama et al. , and Takami et al. .
Figure 7. (a) The 1-h average concentrations of NOy(g) (blue), PMcNO3− (red), and PMfNO3− (green). The 1-h average PM2.5 mass concentration (black), 1-h average relative humidity (gray), and 1-h average wind direction (squares) from 16 to 20 March 2006. The passage of the front was identified from relative humidity, wind direction, and surface weather charts and is marked by two dashed lines. (b) Time-height cross section of the extinction coefficient of mineral dust and spherical aerosols observed with a polarization lidar at CHAAMS.
Download figure to PowerPoint
 Another front passed at around 0900 UT on 18 March when the wind direction changed from south to north and the relative humidity suddenly decreased as in the first case. The lidar results show that dust plumes reached CHAAMS at around the same time, too. As the front approached, the concentrations of NOy(g), PMcNO3−, and PMfNO3− simultaneously increased and peaked at 1600 UT on 18 March. In this regard, Hatakeyama et al.  reported that dust plumes and air pollutants were found in different layers in the marine areas between China and Japan. Therefore, in this case it is considered that dust plumes behind cold fronts of a low-pressure system intermingled, at least to some extent, with aged air pollutants. However, observational data are not yet available to help determine how they intermingle in the marine atmosphere. The synchronized increase of coarse and fine particles was also reported by Zhang et al. . After that, the concentrations of NOy(g), PMcNO3− and PMfNO3− temporarily decreased. Back trajectory analyses show that there was rainfall during transport.
 Although air pollutants and dust plumes showed different patterns and times taken to reach CHAAMS, the presence of dust affected PMcNO3− concentration and the distribution of NOy in the same manner. In the 16 March event, the fraction of PMcNO3− in NOy was 0.08 ± 0.05 for 6 h before the fronts passed, which was not observed with the dust plumes at CHAAMS. On the other hand, the fraction rose to 0.18 ± 0.01 when the fronts passed over CHAAMS and dust plumes began to be observed, after which the fraction changed to 0.20–0.35 until another front passed. These variations also appeared in the 18 March event. The fraction of PMcNO3− in NOy was 0.17 ± 0.03 for 3 h before the fronts passed and 0.28 ± 0.01 after they passed. This suggests that the presence of dust produces PMcNO3− and increases the fraction of PMcNO3− in NOy.
 After the concentrations of NOy(g), PMcNO3−, and PMfNO3− simultaneously peaked at 1600 UT on 18 March and temporarily decreased by rainfall during the transport, they peaked at 0100 UT on 19 March again. The difference between these two peak events was the distribution of PMcNO3− and PMfNO3−. The fraction of PMcNO3− in PM10NO3− around about 1 h of the first and second peak time was 0.67 ± 0.01 and 0.89 ± 0.03, respectively. As compared with the first peak, the second peak shows that PMcNO3− concentration was lower, and its fraction in PM10NO3− was higher. This is considered to be due to the higher dust levels. The height distribution of dust might influence the distribution of PMcNO3− and PMfNO3− in PM10NO3−, but there are insufficient observation data to evaluate how they are associated with the formation of particulate NO3−.
4.4. Effect of Volcanic Activity on Particulate NO3−
 On the basis of back trajectory analyses, we selected the air masses that passed the Sakurajima area (section 2.3) and analyzed the impacts of sulfur oxide caused by volcanic activity on nitrogen oxide. The sample numbers of Sakurajima and Japan origin were 78 and 103, respectively.
 The average concentrations of SO2 and nss-SO42− of Sakurajima origin were 1.74 ± 1.42 ppbv and 16.32 ± 11.73 μg m−3, respectively, whereas those of Japan origin were 0.60 ± 0.47 ppbv and 3.95 ± 0.40 μg m−3, respectively. This suggests that the air masses of Sakurajima origin reached CHAAMS under the influence of volcanic activity. The concentrations of NOy, HNO3, PMcNO3−, and PMfNO3− of Sakurajima origin were 1.11 ± 0.40 ppbv, 0.25 ± 0.19 ppbv, 0.43 ± 0.21 μg m−3 and 0.09 ± 0.05 μg m−3, respectively. On the other hand, those of Japan origin were 1.28 ± 0.45 ppbv, 0.17 ± 0.08 ppbv, 0.51 ± 0.22 μg m−3, and 0.12 ± 0.02 μg m−3, respectively. It is found that a slight increase of HNO3 was observed when air masses passed the Sakurajima area before reaching CHAAMS (Table 4), while NOy, PMcNO3−, and PMfNO3− slightly decreased though the standard deviation is relatively large. We consider this to be the effect of sulfur dioxide with volcanic activity.
Table 4. Average Concentrations of Sakurajima and Japan Origin
|Origin||Gas, ppbv||Aerosol, μg m−3||NOy, ppbv||Sample Number|
|Sakurajima||0.73 ± 0.33||0.25 ± 0.19||1.74 ± 1.42||0.43 ± 0.21||0.09 ± 0.05||16.32 ± 11.73||1.11 ± 0.40||78|
|Japan||0.98 ± 0.47||0.17 ± 0.08||0.60 ± 0.47||0.51 ± 0.22||0.12 ± 0.02||3.95 ± 0.40||1.28 ± 0.45||103|
 The air masses that passed the Sakurajima area originated from areas of human activity in Japan and contained a high fraction of PMfNO3− in PM10NO3−. However, PMfNO3− is converted to PMcNO3− through decomposition and adsorption by sea salt during transport. As the air masses passed the Sakurajima area, sulfur oxide was supplied to them. HNO3, which was adsorbed by sea salt, was substituted by H2SO4 (2NaNO3 + H2SO4 Na2SO4 + 2HNO3). The increase in HNO3 and decrease in PMcNO3− resulted from this reaction. PMfNO3− can also be substituted by H2SO4 as well as PMcNO3−. Undecomposed NH4NO3 was substituted by H2SO4 (2NH4NO3 + H2SO4 (NH4)2SO4 + 2HNO3), resulting in the lower concentration of PMfNO3− of Sakurajima origin compared to that of Japan origin.
 The lower NOy concentration of Sakurajima origin compared to that of Japan origin is considered to be due to the increase in HNO3. The lifetime of HNO3 was shorter than the transport time of air masses from the Sakurajima area to CHAAMS (about 30 h). This suggests that the substituted HNO3 was deposited during transport from the Sakurajima area to CHAAMS. In other words, the ratio of HNO3 to NOy became high when the air masses passed Sakurajima, resulting in the decrease in NOy. Kajino et al.  conducted model analysis and reported that the deposition of NO3− was accelerated by the supply of volcanic sulfate, which is consistent with our observation.