During our aerosol measurement program at Syowa Station, Antarctica, in 2004–2007, some low-visibility (haze) phenomena were observed during winter–spring under conditions with low winds and without drifting snow and fog. During “Antarctic haze” phenomena, the number concentration of aerosol particles and black carbon concentration increased by 1 to 2 orders higher relative to background conditions at Syowa Station, whereas surface O3 concentration dropped simultaneously, especially after polar sunrise. Chemical analysis showed that major aerosol constituents in the haze phenomena were sea salt (e.g., Na+, Cl−). Trajectory analysis and the Navy Aerosol Analysis and Prediction System model showed that plumes from biomass burning in South America and southern Africa were transported to Syowa Station, on the Antarctic coast, because of the eastward (occasionally westward) approach of cyclones in the Southern Ocean and subsequent poleward flow. This poleward flow from midlatitudes of the plume and injection of sea-salt particles during the transport might engender Antarctic haze phenomena at Syowa Station. Differences of O3 concentration between the background and the haze conditions tended to be larger in spring (after polar sunrise) than in winter. Enhancement of sea-salt particles in the haze events can serve important roles in providing additional sources of reactive halogen species.
 Haze phenomena have been observed in the Arctic during winter–spring as the so-called “Arctic haze.” Many observations over several decades have indicated that major aerosol constituents in the Arctic haze are anthropogenic species such as SO42−, NO3−, and black carbon (BC) [e.g., Quinn et al., 2007; Law and Stohl, 2007] transported from industrial regions, mainly in Eurasia [Shaw, 1990, 1995; Heintzenberg and Leck, 1994; Law and Stohl, 2007]. The Arctic haze affects visibility and the solar radiation balance through scattering and absorption by aerosol particles; it also changes the surface albedo by BC deposition onto the snow surface [Hansen and Nazarenko, 2004]. For instance, heating rates of 0.1–0.2 K/d were obtained from direct measurements of Arctic haze in AGASP-II (Arctic Gas and Aerosol Sampling Program) [Valero et al., 1989] and ASTAR 2000 (Arctic Study of Tropospheric Aerosol and Radiation) [Treffeisen et al., 2005].
 In contrast to the Arctic haze phenomena, it had been believed that haze phenomena do not appear in the Antarctic regions because of its isolation from combustion sources (e.g., industry and biomass burning) and mineral dust sources. Indeed, Antarctic ice core records show a negligible signal and trend of anthropogenic impact, in contrast to increasing trends of anthropogenic species, such as nitrates and sulfates, in Arctic ice core records since the Industrial Revolution [Legrand and Mayewski, 1997]. Nevertheless, results of recent investigations suggest that mineral particles in ice cores have been transported from South America (Patagonia region) [e.g., Gaiero, 2007; Delmonte et al., 2008] and Australia [Revel-Rolland et al., 2006], though the concentrations of crustal elements were not high. Despite isolation from strong combustion and dust sources, low-visibility conditions were observed during winter–spring at Syowa Station, Antarctica during our aerosol observation campaign in 2004–2006. In general, the haze phenomenon (low visibility) results from high aerosol concentrations through light scattering and absorption [Hyslop, 2009]. Although occurrence of drifting snow, precipitation, and fog can engender lower visibility in polar regions, low-visibility events suggesting “Antarctic haze” were observed even without drifting snow, precipitation, or fog. High aerosol number concentrations in weak winds were also observed at Syowa Station in August 1997 when the surface O3 concentration decreased simultaneously to <5 ppb [Esaki et al., 1998]. Furthermore, high aerosol optical depth was observed frequently using sun photometry at Syowa Station during the austral spring in regular meteorological observation [e.g., Esaki et al., 2000]. To our knowledge, however, similar events have not been reported for other Antarctic stations. This study investigates (1) physical and chemical properties of the Antarctic haze, (2) origin of the haze, and (3) atmospheric implications of the haze.
2. Observations and Analysis
2.1. Aerosol Measurements
 The aerosol measurement program was conducted at Syowa Station (69°00′S, 39°00′E), Antarctica as a part of Japanese Antarctic Research Expedition (JARE) during February 2004 to January 2007 (JARE45–47). Syowa Station is located on East Ongul Island, approximately 4 km distant from the Prince Olav coast (Antarctica). Aerosol measurements were conducted at a “clean air observatory” built in the windward side from the main area of the station. Details of the observatory, aerosol inlet and tubing for aerosol measurements were described by Hara et al. , and Osada et al. .
Table 1 presents instruments for aerosol measurements and sampling used for this study. We used two optical particle counters (OPCs), a scanning mobility particle sizer (SMPS), and a condensation particle counter (CPC) to obtain size distribution and number concentrations of aerosol particles. In addition, SMPS with a thermo-denuder (TSMPS) was operated simultaneously for continuous measurements of semivolatile and nonvolatile particles from February 2005 through December 2006. The thermo-denuder was installed in the sampling line before the differential mobility analyzer (DMA). In TSMPS measurements, the size distribution was scanned at three different temperatures, room temperature (approximately 20°C), 100°C, and 240°C, which were controlled by a programmable controller (E5AK; Omron Corp.) and a heater. The number concentration of aerosol particles larger than 10 nm diameter occasionally decreased by lower concentration than 50 cm−3 at Syowa Station during winter. Therefore, it took 1 h for scanning at each temperature and 280 min for one cycle of TSMPS measurement. In addition, BC measurements were made using a particle soot absorption photometer (PSAP) during JARE45 (2004–2005) and a multiwavelength (seven wavelength) aethalometer during JARE46–47 (2005–2007). Details of BC measurement at Syowa were described by Hara et al. . To obtain chemical compositions and the mixing states of aerosol particles, aerosol samples were collected using a midvolume impactor (MVI) for bulk analysis, and a low-volume impactor (LVI) for single particle analysis following the same procedure as described by Hara et al. [2003, 2004]. To avoid the intake of locally contaminated air, MVI and LVI samplings were operated under conditions with wind blowing from the clean air sector. The O3 concentration was measured at the clean air observatory using a UV photometer (Model 1100; Tokyo Dylec Corp.). Sampled air was taken from the sampling tower next the observatory using a Teflon tube with the same procedure described by Aoki . For screening of locally contaminated data, all data for OPCs, CPC, SMPS, TSMPS, BC, and O3 were filtered using wind data (wind speed and direction) and standard deviation of 10 min mean condensation nuclei (CN) data, as described by Hara et al. .
Table 1. List of Aerosol Measurements in This Study
Instrument or Sampling
Data or Analytical Method
Data Range or Unit
SMPS with a thermo-denuder measurements were stopped in July 2006 because of mechanical trouble.
Additional aerosol samples were taken in haze events and lower O3 episodes in addition to routine sampling.
TEM, transmission electron microscopy; SEM-EDX, scanning electron microscopy–energy dispersive X-ray spectrometry.
Optical particle counter
>0.3, >0.5, >1.0, >2.0, >5.0 μm
Monitoring since 1998
Optical particle counter
>0.08, >0.1, >0.2, >0.3, >0.5 μm
Feb 2005–Jan 2007
Condensation particle counter
Monitoring since 1998
Scanning mobility particle sizer
Feb 2004–Jan 2007
Scanning mobility particle sizer with a thermo-denuder
 Regular meteorological data were obtained from the Japan Meteorological Agency (JMA) near the meteorological observatory at the main area of Syowa Station, approximately 500 m distant from the clean air observatory. We used meteorological data (e.g., wind, temperature, relative humidity, and visibility) with 1 min resolution in this study. Visibility was observed visually by meteorologists from JMA every 3 h on the roof of meteorological observatory. To obtain the air mass history and transport pathway, 160 h forward trajectories were computed from 500 m using isentropic mode in the HYSPLIT model with NCEP reanalysis data (R. R. Draxler and G. D. Rolph, HYSPLIT—Hybrid Single-Particle Lagrangian Integrated Trajectory model, NOAA Air Resources Laboratory, Silver Spring, Maryland, 2003, available at http://www.arl.noaa.gov/ready/hysplit4.html).
2.3. Sample Analysis
 For single particle analysis, we used scanning electron microscopy (SEM)–energy dispersive X-ray spectrometry (EDX) (SEM-EDX, Quanta FEG-200F; FEI Co. or XL30; EDAX Inc.) for elemental analyses of particles. Procedures of single particle analysis were explained by Hara et al. [2003, 2005].
 For determination of water-soluble aerosol constituents, aerosol samples were analyzed using ion chromatography (DX-120; Dionex Corp.) after extraction by 14 mL of ultrapure water (18.2 MΩ). Analytical conditions and procedures were in accordance with Hara et al. .
3. Results and Discussion
3.1. Haze Phenomena at Syowa Station, Antarctica
 During our measurement period, low-visibility conditions such as “haze” were observed occasionally at Syowa Station. Although visibility in polar regions usually falls under conditions with drifting snow, precipitation and fog, the haze episodes were observed even without fog, precipitation, or drifting snow. Figure 1 presents a typical example of the haze phenomenon observed at Syowa during 10–12 August 2005. Under the haze condition (11 August 2005, Figure 1a), we were able to identify only those icebergs that were close to Syowa station, and did not see distant icebergs or the coast line of the continent right next to the island, which is marked by a red arrow, or the ridgeline of the glacier on the continent. Figure 1a shows that fog and drifting snow did not occur at all during this phenomenon. Although snowfall was observed slightly on 11 August 2005, no such precipitation was observed at the time when the photograph (Figure 1a) was taken. Moreover, the sky appeared to be slightly brownish yellow during the haze event. Such a haze condition has never been reported in cases of local contamination from Syowa Station during weak winds. Moreover, surface winds were from the direction of prevailing winds, without local contamination sources. Consequently, haze phenomena cannot be explained by local contamination.
 By contrast, distant icebergs and the ridgeline (right side of the island marked by red arrows in Figure 1) were clearly identifiable after the haze event ceased on 12 August, as shown in Figure 1b. Considering the distance to the icebergs and the ridgeline on the continent, visibility might have been 6–7 km during the haze on 10–12 August 2005, and >15 km after the haze event. According to regular visibility observation every 3 h, visibility dropped to approximately 10 km on 11 August 2005 [Japan Meteorological Agency, 2007]. Because of occasional slight snowfall on 11 August 2005 at Syowa, we cannot reject the likelihood of overestimation of visibility (6–7 km) during the haze event. Similar episodes, however, were observed occasionally at Syowa Station under conditions without drifting snow, precipitation, or fog. One example is a haze event occurring in June 2004. Consequently, low visibility in the haze phenomena is expected to result from an increase of aerosol number concentration.
 To pick up the haze events during our measurement (February 2004–January 2007), we attempt to identify haze and haze-like events using wind data and the aerosol number concentration. Although aerosol number concentration and BC concentration increased under the storm conditions (e.g., blizzard condition), low visibility due to precipitation and drifting snow prevents identification of haze and haze-like events. Therefore, likelihood of haze and haze-like events during the storm conditions were excluded from analysis and discussion in the present study. In this study, haze and haze-like events were divided into cases without fog and drifting snow and with larger number concentration (>10 cm−3 in diameter larger than 0.3 μm), lower wind speed (<5m s−1 for haze events or 5–15 m s−1 for haze-like events), lower visibility (<10 km for haze events or >10 km for haze-like events) and duration longer than several hours. In the cases of lower visibility and strong winds (>5 m s−1), we identified as “haze like” in the present study. For example, OPC data with the criteria (divided into “haze and haze-like episodes”) were only 0.73% of all filtered data in 2005 at Syowa Station. Therefore, the haze and haze-like episodes should be identified as peculiar phenomena at Syowa. Table 2 presents dates of haze and haze-like events at Syowa Station during our measurement period. These events were observed from late May to late September (winter–spring), but they were not observed during summer. In the following sections, we will discuss characteristics and air mass origins of the haze events at Syowa.
Table 2. Date and Meteorological Conditions of Each Haze Episode at Syowa Station During the Aerosol Measurement Program
3.2.1. Physical Properties of Aerosols in the Haze: August 2005
Figure 2 depicts short-term variations of the number concentrations and BC concentration during the haze event at Syowa Station of 10–12 August 2005. This event was one of the heaviest haze events at Syowa Station during our measurement period. This haze event in August 2005 occurred immediately after storm conditions of 9–10 August. On the evening of 10 August, the wind speed dropped suddenly from >20 m s−1 to <10 m s−1. Simultaneously, the number concentrations of Dp > 10 nm and Dp > 0.3 μm increased drastically from ≤100 cm−3 and 5 cm−3 to 717 cm−3 and 65 cm−3, respectively. In addition, the number concentration of coarse particles with size of Dp > 2.0 μm increased by approximately 0.9 cm−3 during the haze event. Furthermore, the BC concentration increased from <10 ng m−3 to approximately 60 ng m−3. Compared to the lower concentrations in the background conditions at Syowa Station in August (e.g., CN, ≤100 cm−3; BC, <5 ng m−3), these higher concentrations are inferred as unusual levels. The surface O3 concentration clearly dropped from 35.0 ppb to 28.0 ppb during the haze condition. The haze event with higher aerosol number concentration and BC concentration continued until the night of 11 August 2005. At 0000 UT, 12 August 2005, the aerosol number concentration and BC concentration decreased drastically to background levels. The haze duration was approximately 32 h in this case.
 During the hazy period, relative humidity decreased gradually from approximately 90% to less than 80% (Figure 2c). Because of the lower relative humidity, the haze event was demonstrably not caused by fog. Figure 2d shows that surface wind blew from the direction of 0–90° during the haze event. This wind sector corresponds to the prevailing wind direction at Syowa, without local contamination sources, because of katabatic wind. Aerosol number density started increasing on 10 August, when winds blew from prevailing wind (clean air sector) and wind speed was still >10 m s−1. We therefore infer that local anthropogenic activity at Syowa did not cause the haze phenomenon. Furthermore, the nearest station (Molodezhnaya, 67.41°S 45.51°E) is located approximately 400 km distant from Syowa Station. According to Hagler et al. , the atmospheric BC concentration decreased rapidly to <1 ng m−3 in the first 30 km from BC sources in the polar region. Consequently, the anthropogenic impact from the other stations was too small to make a meaningful contribution to the haze event at Syowa. Therefore, the haze phenomenon is expected to result from the transport of an air mass with higher aerosol concentration. The transport pathway and air mass history are discussed in sections 3–4.
Figure 3 portrays the size distribution of aerosol particles during and after the haze event of August 2005. The SMPS measurement exhibits that ultrafine particles, particularly those larger than 20 nm, were markedly enhanced during the haze event. Surprisingly, the size distribution of aerosol particles exposed to 240°C in TSMPS was almost equal (slightly lower) to that in SMPS operated at room temperature (approximately 20°C). Considering that sulfate (mostly H2SO4) particles can be vaporized at 240°C scan at Syowa Station during the summer [Hara et al., 2007], ultrafine and fine particles in the haze conditions might not be composed of sulfate particles but rather of nonvolatilized species such as sea salt, soot (BC), and minerals. Although the BC concentration increased drastically during the haze event, as depicted in Figure 2a, the BC concentration was too low to account for the increase of the number concentration in ultra fine and fine particles in the haze. Similarly, mineral particles must be transported to Syowa Station to make a meaningful contribution because of the negligible source strength of minerals in the Antarctic region. Consequently, sea salts are expected to be the most plausible species for nonvolatile species in ultrafine and fine particles.
3.2.2. Physical Properties of Aerosols in the Haze: June 2004
 Similar haze episodes were observed in June 2004. Figure 4 depicts variations of the aerosol number concentration and BC concentration on 16–19 June 2004. The related CN (Dp > 10 nm) measurements suggest that the haze event started during the afternoon of 17 June and continued until strong winds blew in the evening of 18 June. Different from the haze in August 2005, no precipitation was observed in this case. In spite of the lack of precipitation and weak winds, visibility dropped to 10 km [Japan Meteorological Agency, 2006]. Visibility was greater than 30 km under clear conditions, even in June. Duration of the haze in June 2004 was approximately 33 h. As presented in Figure 4d, the wind direction ranged dominantly from 45° to 90° during the haze event, although the wind direction occasionally changed to 90–180°. Without local contamination sources in the wind sector, the haze event in June 2004 was also not caused by local contamination. Because relative humidity was lower than 70% during the haze, fog did not contribute to the lower visibility.
 As during August 2005 as described above, the aerosol number concentrations of CN (Dp > 10 nm) and particles with Dp > 0.3 μm, respectively, increased markedly to 836 cm−3 and 91 cm−3 compared to the background level (CN, <70 cm−3; Dp > 0.3 μm, <1 cm−3). In addition, the BC concentration changed from <5 ng m−3 to 60.6 ng m−3. The surface O3 concentration decreased from 34.2 ppb to 33.3 ppb in this case. Although the haze event in August 2005 was observed immediately after storm conditions, such storm conditions were not observed immediately before the haze.
 Fortunately, balloon-borne OPC measurement was made as a part of aerosol measurement program on 18 June 2004 at Syowa [Japan Meteorological Agency, 2006]. That measurement showed that high number concentrations (10 cm−3) of aerosol particles were obtained with size of Dp > 0.3 μm from the surface to approximately 2200 m above sea level (asl). In addition, aerosol measurements made using aircraft [Yamanouchi et al., 1999] showed that higher aerosol concentrations were observed from the surface to approximately 2300 m asl on 30 August 1997 when surface ozone depletion occurred simultaneously with the aerosol enhancement [Esaki et al., 1998]. Therefore, Antarctic haze might have reached a thickness of 2200–2300 m at Syowa Station during this event.
3.3. Aerosol Chemical Properties During the Haze Events
3.3.1. Bulk Aerosol Constituents in June 2004 and August 2005
Figure 5 presents the mass fraction (percentage) of water-soluble aerosol constituents during the haze event in June 2004 and August 2005. Sea-salt constituents such as Na+ and Cl− accounted for more than 97% (June 2004, Sample A) and 95% (August 2005, Sample B) in both cation and anion. During the haze (Sample A and B), total concentrations of Na+ increased by 187.7 nmol m−3 in June 2004 and 176.7 nmol m−3 in August 2005. Sea-salt (Na+ and Cl−) concentration tends to increase under the storm conditions at Syowa [Hara et al., 2004]. Such higher concentrations of Na+ and Cl−, however, were observed in a few samples a year in routine aerosol sampling with 3 d resolution at Syowa using the MVI system in 2003–2007 (K. Hara et al., unpublished data). Although SO42− was obtained among the anions in Samples A–C, the SO42− concentration corresponded to sea salt SO42−. In addition to sea salt (Na+ and Cl−), the concentrations of NO3−, CH3COO−, and HCOO− increased during haze event in August 2005 relative to the background levels. Compared to monthly mean and median concentrations of NO3− (mean, 0.66 nmol m−3; median, 0.62 nmol m−3), HCOO− (mean, 0.14 nmol m−3; median, 0.09 nmol m−3), and CH3COO− (mean, 0.14 nmol m−3; median, 0.09 nmol m−3) in August at Syowa Station (K. Hara et al., unpublished data), the concentrations of NO3−, HCOO− and CH3COO− were enhanced to 2.03, 0.47 and 0.19 nmol m−3 in Sample B, respectively. NO3− was enhanced in August 2005, whereas the total NO3− concentration (0.12 nmol m−3) on 18 June 2004 corresponded to the monthly mean in June (approximately 0.20 nmol m−3, K. Hara et al., unpublished data). Although oxalate concentration ranged below the detection limit in the haze episode in June 2004 and August 2005, particulate NO3−, oxalate, and sea salt (e.g., Na+ and Cl−) were also enhanced during low O3 concentration at Syowa Station in October 1996 [Osada et al., 1998].
 As portrayed in Figure 5c, sea-salt particles (Na+) were distributed also in fine (Dp: 0.2–2.0 μm) and ultrafine particles (Dp: <0.2 μm) during the haze events. In particular, mass fractions of ultrafine and fine sea-salt particles were greater than 90% in June 2004 and 60% in August 2005. Even in the ultra fine particles (backup filter), the percentage of Na+ concentration reached 53% (June 2004) and 35% (August 2005) among cations in Sample A and B. Consequently, most aerosol particles in ultrafine and fine modes might be sea-salt particles during the haze, which agrees very well with TSMPS measurement, as presented in Figure 3. This suggests that sea-salt particles made an important contribution to lower visibility in the haze of August 2005, although aerosol constituents (e.g., NO3− and BC) other than sea salt increased markedly. Results of recent tandem DMA measurements in the marine boundary layer [Clarke et al., 2006] and laboratory measurements [Mårtensson et al., 2003; Massel, 2007] suggest that ultrafine sea-salt particles were released from the ocean surface by breaking waves and bursting bubbles. Although sea salt and sea ice are important sources of sea-salt particles in polar regions, emission processes of ultrafine sea-salt particles in polar regions remain poorly understood.
3.3.2. Aerosol Constituents of Individual Particles During the Haze in June 2004 and August 2005
 From single particle analysis using SEM-EDX, most aerosol particles contained Na, Cl, and Mg. These particles might be identified as sea-salt particles. In EDX analysis, no characteristic X-ray of elements (Z > 11, Na) was observed from a few particles. We used carbon-coated collodion thin film on a Ni micro–grid as sampling substrates. Therefore, the carbon signal from aerosol particles cannot be divided. Consequently, we can identify external mixing state of carbonaceous particles according to their unique morphology (aggregated structure) and no characteristic X-ray signal.
Figure 6 depicts typical examples of EDX spectra of aerosol particles collected during the haze event on 18 June 2004 and August 2005. As shown in Figure 6a, S, Ca, Si, and Fe were detected. Considering the relative atomic ratio of Ca (44.8%) and S (44.1%), this particle might be mainly composed of gypsum (CaSO4). Consequently, this particle is identifiable as a mineral particle. In most aerosol particles, sea-salt constituents such as Na, Mg, and Cl are observed in EDX analysis (Figure 6b). As shown in Figure 6c, the internal mixtures between sea salts (e.g., Na and Cl) and minerals (e.g., Si, and Ca) were observed occasionally. Furthermore, some carbonaceous particles were identified, as shown in Figure 6d and Figure 7. No characteristic X-rays other than background peaks (C, O, and Ni) derived from collodion thin film and Ni microgrid were obtained using EDX analyses. In addition, these particles had the unique aggregate structures (Figure 7). Thus, these particles might be identified as carbonaceous particles in the present study.
 For quantitative discussion, the relative abundance (percentage) was estimated from EDX analyses (Table 3). Single particle analysis using SEM-EDX exhibits a dominance of sea-salt particles, as presented in Table 3. Although mineral particles were not obtained in the haze of August 2005, a few mineral particles were identified in the haze of June 2004. The presence of mineral particles and carbonaceous particles strongly suggests poleward transport from the midlatitudes and the continents because of very few source strengths of minerals and carbonaceous particles in the Antarctic regions during winter.
Table 3. Relative Abundance of Aerosol Particles Collected During the Haze of June 2004 and August 2005
 As found also in our previous work [Hara et al., 2005], Mg tended to be enriched relative to the bulk seawater ratio [Wilson, 1975]. Ternary plots presented in Figures 8 show variations of sea-salt constituents in each sea-salt particle obtained using SEM-EDX analysis. When Cl liberation from sea-salt particles occurs by heterogeneous processes with acids, each sea-salt particle is distributed from a black star (seawater ratio) to white star (wholly Cl depleted) along with the stoichiometric line in Figure 8. Most sea-salt particles in coarse mode were distributed around 40% of Cl close to the stoichiometric line (Figures 8b and 8d) and showed slight Mg enrichment, although some sea-salt particles showed remarkable Mg enrichment (>50%; Figure 8d). In contrast to coarse sea-salt particles, more Mg-rich sea salt particles were observed in fine sea-salt particles. According to Hara et al. , the Mg-rich sea-salt particles were derived from sea ice through sea-salt fractionation during the winter. The presence of Mg-rich (fractionated) sea-salt particles suggests that sea-salt particles involved in haze phenomena had been released not only from the open sea surface but also from sea ice surface through wind blowing during transport. Because sea-salt particles were dominated during the haze events, the unusually high concentrations of sea-salt particles might cause the Antarctic haze conditions. This was in good agreement with TSMPS measurements as shown in Figure 3. Sea-salt emissions from sea surface and sea ice, however, cannot explain the enhancement of BC and NO3− during the haze: other processes are probably related to the haze event. As suggested by Hara et al. , BC must be transported from midlatitudes and low latitudes to Syowa Station (Antarctica). To elucidate the haze event well, the air mass history and meteorological field are discussed next.
3.4. Meteorological Conditions and Air Mass History of the Haze Events
3.4.1. Haze Event in August 2005
 According to our previous study [Hara et al., 2008], high BC concentration during winter–spring at Syowa was caused mostly by poleward transport related to the eastward (occasionally westward) approach of cyclones. Because strong wind conditions were observed immediately before the haze event in August 2005, it is expected that cyclone movements are related to haze events. Figure 9 depicts the geopotential height distribution on 9–12 August 2005 and a surface weather chart for 10 August 2005. On 9 August, a large cyclone located at 20°E, 65°S engendered storm conditions at Syowa Station. The cyclone moved to 25°E, 67°S on 10 August and continued its strong influence of the weather at Syowa. The cyclone, however, did not move eastward; it suddenly weakened on 11 August. Although passing of cyclones to the north of Syowa Station caused higher BC concentration by poleward flow [Hara et al., 2008], the cyclone did not pass in this case. The surface weather chart for 10 August shows that a cold front was close to Syowa. For that reason, a sudden drop of wind speed and increase of aerosol concentrations on August 10 might be coincident with passage of cold front and the decline of the cyclone.
 According to the NAAPS model (http://www.nrlmry.navy.mil/aerosol/), smoke from biomass burning was transported from South America and southern Africa to the Atlantic and Indian oceans on 6–7 August 2005 before the haze event at Syowa (Figure 10). The smoke was observed clearly in Moderate Resolution Imaging Spectroradiometer (MODIS) images on 4 August in South America (Figure 10a) and 7 August in southern Africa (not shown). Because biomass burning emits BC and precursors of particulate NO3− and organic acids, long-range transport of the plume to Syowa can lead to increase of the concentrations of BC and NO3− at Syowa. To discuss the relation between the outflow from biomass burning from the continents and the haze event at Syowa, we attempted to compute the 160 h forward trajectory from the outflow plume in South America and southern Africa.
 In South America, many fire spots were distributed in low-latitude and midlatitude areas of Brazil (Amazon) as shown in Figures 10b and 10c. The plume from biomass burning spread to southern Brazil and northern Argentina–Patagonia; then it flowed out toward the Atlantic Ocean from 1 to 2 August 2005 in the NAAPS model. Forward trajectories from the outflow in South America (Figure 11) show that part of the plume was transported eastward through the lower troposphere and approached the Antarctic coast at 20–40°E at 1200 UT on 10 August 2005, as shown by square marks in Figure 11. This transport pathway overlapped with a storm track in the Southern Ocean [Hoskins and Hodges, 2005]. Moreover, it shows good agreement with moisture transport processes to Syowa Station during the winter [Suzuki et al., 2008]. In contrast to the outflow from South America, plumes from southern Africa were transported eastward to the Indian and Pacific oceans. Air masses from southern Africa at 1200 UT on 10 August were quite distant from Syowa Station. Therefore, the haze event in August 2005 at Syowa is likely to have been influenced by the plume from biomass burning in South America. Indeed, long-range transport of continental air to Antarctic coasts was occasionally observed as a radon storm at Syowa and over the Southern Ocean [Polian et al., 1986; Balkanski and Jacob, 1990; Ui et al., 1998]. Considering the forward trajectory and the NAAPS model, the biomass burning plume might be transported from fire spots at low-latitudes and midlatitudes of South America (Brazil) to Syowa Station for approximately one week. This result is very consistent with enhancement of BC and NO3−, which can be emitted as a result of biomass burning [e.g., Maenhaut et al., 1996; Andreae and Merlet, 2001] during the haze event. Then, sea-salt particles might be mixed with the air mass containing atmospheric species originated from biomass burning during transport to Syowa Station through the storm track and sea ice area. Although nss-K has been used as a tracer of biomass burning [e.g., Andreae, 1983; Quinn et al., 2002; Guazzotti et al., 2003; Okada et al., 2008], nss-K (K-rich particles) from biomass burning was hardly divided in single particle analysis because of lower concentrations of nss-K by significant deposition during the transport and K enrichment by sea-salt fractionation on sea ice [Hara et al., 2004].
3.4.2. Haze Event in June 2004
Figure 12 presents the distribution of geopotential height from 16 to 19 June 2004 and a surface weather chart on 18 June 2004. A strong cyclone was located east (70°E, 65°S) of Syowa Station on 16 June. Nevertheless, the wind speed was influenced only negligibly by the cyclone on 16 June. The distribution of geopotential height suggests that the airflow (wind field) to Syowa on 17 June remained influenced by the cyclone (60°E, 60°S). On 18 June (haze condition), Syowa gradually came under high-pressure conditions. As in August 2005, the cyclone weakened suddenly before the haze event at Syowa.
 As presented in Figure 13, the plume outflow from biomass burning in southern Africa (e.g., Angola, Zambia, and Zaire) to the Indian Ocean was observed on 13 June 2004 in the NAAPS model. Furthermore, the plume off South Africa was obtained obviously in SeaWifs images (not shown here). To elucidate the relation between the outflow and the haze event at Syowa Station in June 2004, forward trajectories (Figure 14) were calculated from the plume in the same manner as in the case of the haze of August 2005. The air mass of the outflow from southern Africa was transported southeastward to 80°E, and then westward to the Antarctic coast and inland areas. Although most air mass went to inland areas close to Dome Fuji Station (77.19°S, 39.42°E), Figure 14 shows that part of the plume approached Syowa Station on 18–19 June. In addition, the air mass height at 0000 UT on 18 and 19 June 2004 was consistent with balloon-borne OPC measurement at Syowa Station on 18 June 2004 (as described above). Consequently, the air mass of the haze event in June 2004 might be mixed with the plume from the biomass burning taking place in southern Africa. The air mass from Africa passed under the strong wind condition over the Southern Ocean, as suggested by the distribution geopotential height (Figure 12). Therefore, sea-salt particles might be injected in oceanic and coastal areas during transport. The transport time from Africa to Syowa was estimated as approximately 5 d for the haze event of June 2004.
3.5. Characteristics of Other Antarctic Haze Events at Syowa Station
 As described above, the haze events in June 2004 and August 2005 were influenced by mixing of sea-salt particles during transport and by biomass burning plumes extending from southern Africa and South America, respectively. For better understanding of the characteristics and air mass origins in the other haze episodes at Syowa, aerosol properties and characteristics during the haze are presented in Table 4. The sudden increase and decrease of aerosol number concentration and BC concentration were obtained clearly also in the other haze episodes. On the basis of marked changes of the CN and BC concentrations, the duration of the haze events was estimated as 19–73 h (mean, 35 h). Longer duration than 74 h of the haze events was not observed at Syowa Station during our measurements, probably because of frequent approaches of cyclones during winter–spring. Horizontal scales of the Antarctic haze events were estimated using surface wind speeds. As presented in Table 4, the horizontal scales were 159–659 km in haze episodes (mean wind speed, ≤5 m s−1) and 1328–1770 km in haze-like episodes (mean wind speed, 5–15 m s−1).
Table 4. Aerosol Properties During the Haze and Haze-like Events at Syowa Station
Maximum CN Concentration (cm−3)
Maximum Concentration of Dp > 0.3 μm (cm−3)
Maximum BC Concentration (ng m−3)
Air Mass Origin
Traveling Time to Syowa (d)
Horizontal Scale (km)
28 Sep–1 Oct
 Moreover, BC was enhanced to 7–61 ng m−3 (mean, 41.9 ng m−3) in the other haze episodes. The range of BC concentration in the haze and haze-like events was similar to that under the storm conditions [Hara et al., 2008]. Our previous investigation [Hara et al., 2008] showed that higher BC concentration was often observed under the strong wind (storm) conditions. This was associated with poleward flow of BC-rich air masses from Atlantic and Indian Oceans by cyclone approach and blocking events [Hara et al., 2008]. This transport pathway was very similar to that in haze and haze-like events as shown in Figures 11 and 14. Storm conditions can cause aerosol injection from sea ice and sea surface to BC-rich air mass (from midlatitudes) by wind blowing. Without a sudden decline of the cyclone, however, the highly aerosol-enhanced air mass passed rapidly over Syowa in storm conditions [Hara et al., 2008]. In most cases, cyclone passed in the northern side of Syowa, so that many BC peaks were observed under the storm conditions by poleward flow as suggested by Hara et al. . In contrast, aerosol enhancement might remain around Antarctic coast for longer time after sudden decline of cyclone. Therefore, the approach and sudden decline of the cyclone might be an important meteorological process prompting the occurrence of the haze and haze-like phenomena at Syowa Station.
 Both SMPS and TSMPS measurements indicate the dominance of nonvolatile particles in the haze events during 2005–2006 as in the case of August 2005. Furthermore, additional aerosol samples for the haze events suggest that major aerosol constituents were sea salt (Na+ and Cl−) in haze events such as those described above. Moreover, NO3−, CH3COO−, and HCOO− were enhanced in the other haze events.
 Haze and haze-like events were observed from May through early October during the aerosol measurement program for 3 years. As described above, haze phenomena were associated with outflow of biomass burning from southern Africa and South America and the sudden decline of cyclones. To obtain better knowledge about the relation between the plume outflow and the haze events, we attempted a comparison to seasonal features of the frequency of plume outflow from southern Africa and South America. According to the NAAPS model (Figures 10 and 13), the plume outflow, which can be transported to Syowa Station, was observed in South America and southern Africa. When air mass with aerosol optical thickness larger than 0.1 in the NAAPS model was dispersed from the continents to Oceans as shown in Figures 10 and 13, we identified the occurrence of “outflow” in this study. As presented in Figure 15, air masses with relatively higher aerosol optical depth often flew out from southern Africa during April–December, whereas outflow from South America was observed during July–September. The seasonal variation of the outflow was coincident with higher fire counts in May–October (Africa) and June–October (South America), respectively [Edwards et al., 2006]. Surprisingly, the outflow was very frequent in July–August. Therefore, the plume from biomass burning is expected to flow out subsequently from South America and southern Africa to the Atlantic, Indian and Southern oceans during May–October. The highly frequent outflow suggests atmospheric substances (e.g., BC) from continents were dispersed along storm track in Atlantic and Indian Oceans by cyclone activity. When cyclone approached to northern side of Syowa, BC-rich air mass may be traveled to Syowa by poleward flow. This showed good agreement with many BC peaks at Syowa during storm conditions because of poleward flow by the cyclone's approach [Hara et al., 2008]. During the austral summer, no haze or haze-like phenomena was observed at Syowa. This might be associated with less frequent blizzards at Syowa [Sato and Hirasawa, 2007] and less outflow from southern Africa and South America.
 To understand the travel time of the outflow plume from southern Africa and South America, we attempted to estimate the days using trajectory data, as described above. For instance, 3–4 d were necessary to transport from South America to Syowa in the case of very rapid cyclone approach on 21–28 July 2004, whereas the traveling days were 6–7 d from South America in most cases. In contrast, it took 4–5 d from southern Africa in most cases. The estimated periods corresponded well with estimates from a circulation model by Krinner and Genthon  and a trajectory analysis [Suzuki et al., 2008]. On average, the travel time from southern Africa was shorter than that from South America. This difference of the travel time might reflect concentrations of continental origin species such as BC and minerals. Indeed, mineral particles were observed in the outflow from southern Africa (June 2004), although they were not observed from South America (August 2005), as listed in Table 3.
3.6. Implications of the Antarctic Haze for Atmospheric Chemistry and Climate
3.6.1. Relation Between Haze Episodes and Lower O3 Concentration
Figure 2 shows that the decrease of O3 concentration occurred simultaneously with aerosol enhancement during the haze episode in August 2005. During the haze period, the O3 concentration decreased from the background level (35.0 ppb) to 28.0 ppb. The difference (ΔO3) between the background concentration and the depleted concentration in the haze event was 7.0 ppb. In addition, the aerosol number concentration in June 2004 (Figure 4) was enhanced considerably to a similar level to that in August 2005, whereas ΔO3 was only 0.9 ppb. This difference of ΔO3 is necessary for understanding of the relation between the haze episodes and lower O3 concentration (or O3 depletion). Table 4 presents ΔO3 in the haze episodes at Syowa during our measurement in 2004–2007. The O3 concentration decreased in the most haze episodes except during May 2005. Furthermore, higher ΔO3 concentrations were measured during August–September. In addition, very high ΔO3 (≈32 ppb) was observed in surface O3 depletion simultaneously with high aerosol enhancement, which is identifiable as a “haze phenomenon” in the classification used for this study, at Syowa on 28–29 August 1997 [Esaki et al., 1998]. In contrast, ΔO3 concentrations were less than 2.4 ppb during May–July. Because direct solar radiation is absent from June until mid-July at Syowa, the variation of ΔO3 might be related to photochemical processes. The following possibilities are considered for lower O3 concentration or O3 depletion in the haze episodes: (1) long-range transport of air mass with lower O3 concentrations from midlatitudes; and (2) O3 depletion by halogen catalytic processes.
 Our previous investigations [Hara et al., 2004, 2008] indicated that air masses under storm (blizzard) conditions were transported from midlatitudes via the lower troposphere. Regarding August 2005, the O3 concentration was 35.0 ppb under storm conditions on 9–10 August 2005 immediately before the haze episode. The O3 concentration before haze corresponded to the background O3 concentration without O3 depletion at Syowa in August [Murayama et al., 1992; Aoki, 1997]. Regarding the other haze events, ΔO3 concentrations in blizzards before or after haze events were 0–5.4 ppb. For example, the ΔO3 concentration in the blizzard condition on 26 September 2005 was only 5.4 ppb in spite of ΔO3 = 17.2 in the haze event on 27–28 September 2005. As described above, a part of the biomass burning plume was transported to Syowa Station during haze episodes. Atmospheric species from biomass burning, however, can engender photochemical O3 production [Kondo et al., 2004; Boian and Kirchhoff, 2005]. Although high ΔO3 (≈32 ppb) was under the conditions with high aerosol enhancement at Syowa in August 1997 [Esaki et al., 1998], transport from marine boundary layer in the midlatitudes to Syowa hardly caused such a lower O3 concentration (or O3 depletion). Consequently, lower O3 concentration in the haze phenomena might result from O3 depletion rather than transport from midlatitudes.
 As described above, sea-salt particles were dominant among aerosol particles during haze episodes. Figure 8 shows that many fractionated sea-salt particles originating from sea ice were distributed in the atmosphere during the haze. Sea-salt fractionation can result in Br enrichment on the sea ice surface [Piot and von Glasow, 2008, and references therein]. When Br enriched sea-salt particles and fresh sea-salt particles are released from sea ice and ocean surface by strong wind, the fractionated sea-salt particles and the fresh sea-salt particles can act as an important source of reactive halogen species by heterogeneous reactions and then have significant potential inducing surface O3 depletion under solar radiation. Although tropospheric halogen chemistry coupled with O3 chemistry in polar regions have been simulated together with frost flower and carbonate precipitation on sea ice [e.g., Sander et al., 2006; Piot and von Glasow, 2008, and references therein], the unusual high concentration of sea-salt particles containing Br− in ozone depletion has not been considered. To understand the atmospheric chemistry related to halogen and sea-salt chemistry in the Antarctic coasts and implications by the Antarctic haze phenomena, more field observations and model calculations are needed. Details of the relation between aerosol enhancement and O3 depletion will be discussed elsewhere.
3.6.2. Contribution of the Haze Episodes to the Antarctic Climate
 Atmospheric aerosols can influence climate through the radiation budget (known as direct effect) when the concentration of aerosol particles is extremely high [Intergovernmental Panel on Climate Change, 2007]. For instance, results of some investigations suggest that aerosol particles in the Arctic haze can exert a direct effect [Valero et al., 1989, Treffeisen et al., 2005]. In contrast to the Arctic haze, which is widespread and suspended for longer time in the Arctic Circles in the winter–spring, duration and horizontal scale in the Antarctic haze episodes were only in 19–73 h and 191–1770 km, respectively. Because the Antarctic haze episodes occurred only by injection of sea-salt particles emitted from the surface of the ocean and sea ice, and long-range transport of the plume from biomass burning in South America and southern Africa after transport for several days to one week, the Antarctic haze is expected to be an event with temporal scale that is smaller than that of the Arctic haze.
 During Antarctic haze events, BC concentrations and aerosol number concentrations of Dp > 0.3 μm increased to 120 ng m−3 and 9.1 × 104 L−1. This aerosol number concentration was similar to the order of 10 cm−3 (Dp > 0.3 μm) in the Arctic haze [Dreiling and Freiderich, 1997; Yamanouchi et al., 2005]. On the other hand, the BC concentration in the Arctic during winter–spring was 1 to 2 orders higher than that in the Antarctic haze at Syowa Station, and occasionally exceeded 1000 ng m−3 [Sharma et al., 2006; Quinn et al., 2007]. Although high concentrations of light-absorbing particles such as BC cause a warming effect (0.1–0.2 K d−1) in the Arctic haze [Valero et al., 1989; Pueschel and Kinne, 1995; Treffeisen et al., 2005], BC concentrations might be too low to make a meaningful contribution to a warming effect even in the Antarctic haze. Indeed, single scattering albedo (SSA) of >0.98 in the Antarctic haze is caused by lower BC concentration than that in the Arctic haze (M. Yabuki et al., manuscript in preparation for publication, 2010), whereas SSA in Arctic haze was 0.77–0.93 (mean 0.88) [Clarke et al., 1984] and 0.75–0.95 [Yamanouchi et al., 2005]. In contrast to the dominance of anthropogenic species such as sulfates and BC in the Arctic haze [Heintzenberg and Leck, 1994; Hara et al., 2003; Quinn et al., 2007], aerosol particles in the Antarctic haze were mostly sea salts. Moreover, SSA was greater than 0.98 in the Antarctic haze episodes (M. Yabuki et al., manuscript in preparation for publication, 2010), so that the Antarctic haze phenomena can somehow engender a cooling effect rather than warming effect. Considering the short duration and small horizontal scale (191–1770 km in coastal areas), as presented in Table 4, direct effects from Antarctic haze to climate in the Antarctic regions might be negligible at the moment.
 Recent investigations point out that BC can cause significant reduction of the surface albedo, leading to positive radiative forcing after deposition onto snow surfaces [Aoki et al., 1998; Hansen and Nazarenko, 2004; Intergovernmental Panel on Climate Change, 2007]. The surface albedo of snow surfaces is extremely sensitive to BC concentration [e.g., Aoki et al., 1997]. In fact, BC can be transported to Antarctic regions by cyclones and blocking events [Hara et al., 2008] and then might be deposited onto the snow surface by precipitation. Because of lower BC concentration in the Antarctic troposphere [Bodhaine, 1995; Wolff and Cachier, 1998; Pereira et al., 2006; Hara et al., 2008], the contribution of BC to the change of surface albedo might be unimportant in Antarctic regions at the moment. With increased BC emission from biomass burning and human activity in the southern hemisphere in the future, however, the contribution of BC may be expected to become increasingly important, even in Antarctic coastal regions. Long-term BC measurements in the atmosphere and snow in Antarctica are necessary for future studies. Figure 14 shows that air masses containing aerosols (e.g., BC) from biomass burning can be transported to the interior of the Antarctic continent. Antarctic haze phenomena and long-range transport of species derived from biomass burning are likely to be preserved as a record in Antarctic ice cores. Chemical analyses of aerosol species from biomass burning (e.g., BC) in ice cores are expected to be useful and important for better understanding of long-range transport, spatial distribution and climate impact of aerosols from biomass burning in South America and southern Africa.
 During our aerosol measurement program at Syowa Station, Antarctica from February 2004 through January 2007, low-visibility episodes were observed during late May to early October despite the absence of drifting snow and fog. In the heaviest haze events, visibility dropped to ≤5 km. The duration and horizontal scale of the haze (or haze-like) phenomena were estimated to be on the order of 19–73 h and 191–1770 km, respectively. The haze phenomena might be induced by the eastward (occasionally westward) approach and sudden decline of cyclone.
 The aerosol number concentration and BC concentration increased considerably by 1 and 2 orders in haze and haze-like episodes compared to background conditions. Furthermore, O3 concentrations dropped from the background O3 level during haze episodes occurring after the polar sunrise. Chemical analysis and TSMPS measurements suggest that aerosol particles in all size range (ultrafine ∼coarse particles) were dominantly composed of sea-salt particles. Unusual high concentrations of sea-salt particles might make significant contribution to haze events. Results of single particle analyses by SEM-EDX suggest that Mg was enriched markedly in some fractionated sea-salt particles (Figure 8). Because sea-salt fractionation proceeds on sea ice in the seawater freezing during the winter [Hara et al., 2004], sea-salt particles are expected to originate not only from the sea surface but also from the sea ice surface.
 During the haze episodes, the decreased surface O3 concentration occurred simultaneously with aerosol enhancement. The difference (ΔO3) between O3 concentration in the haze episodes and the background O3 concentration tends to be greater after polar sunrise (August–September), whereas ΔO3 was only <2.4 ppb in periods of polar night and lower solar radiation (May–July). The feature of ΔO3 is likely to be associated with photochemical halogen catalytic processes that occur during the polar sunrise. Sea-salt particles were dominant in the haze episodes; most of the Br− was liberated from sea-salt particles through heterogeneous processes in the haze episodes. Therefore, we propose hypothesis that the haze phenomena might engender a considerable release of reactive bromine species from sea salt, and subsequent surface O3 depletion on the Antarctic coast during the polar sunrise.
 MODIS satellite images and NAAPS model portrayed that smoke (plume) from biomass burning flowed from South America or southern Africa into the Atlantic and Indian oceans before the haze phenomena at Syowa Station. The forward trajectory from the plume outflow suggests that air masses containing atmospheric substances from biomass burning reached the Antarctic coast close to Syowa Station. This transport pathway was associated closely with the eastward (occasionally westward) approach of cyclones. Consequently, BC might be derived from biomass burning in South America and southern Africa; then sea-salt particles might be mixed with the BC-rich air mass through sea-salt release from the open sea and sea ice by wind blowing during transport. Furthermore, the travel time to Syowa Station was estimated as about 7 d from South America and about 5 d from southern Africa. As shown in Figure 14, air masses containing atmospheric substances from South America and southern Africa were transported into inland area. Considering that aerosol particles can be deposited gradually onto snow surface during the transport, some signals of “Antarctic haze (or transport of biomass-burning-origin species)” might be recorded in snow/ice. Thus, Antarctic ice cores are expected to be potential archive for past biomass burning in the southern hemisphere.
 We thank the members of the 45, 46, and 47th Japanese Antarctic Research Expedition for assistance with aerosol measurements at Syowa. We thank G. Shaw of the University of Alaska, Fairbanks, for useful and helpful discussion. The authors thank K. Suzuki for help drawing the plots of geopotential height. This study was supported by “Observation project of global atmospheric change in the Antarctic” for JARE 43–47. This work was also supported by a Grant-in-Aid (16253001, PI: T. Yamanouchi, and 15310012, PI: K. Osada) from the Ministry of Education, Culture, Sports, Science and Technology of Japan. The authors gratefully acknowledge the NOAA Air Resources Laboratory (ARL) for provision of the HYSPLIT transport and dispersion model and/or READY Web site (http://www.arl.noaa.gov/ready.html) used in this publication. High-resolution meteorological data (1 min) were provided from Japan Meteorological Agency.