Since March 2003, measurements of surface ozone have been made at the British Antarctic Survey Clean Air Sector Laboratory (CASLab) at Halley station in coastal Antarctica. Detailed measurements of boundary layer meteorology, as well as standard meteorological parameters, are also measured at the CASLab. Combining these data allows us to probe the transport pathway of air masses during ozone depletion events (ODEs). ODEs were observed at Halley on several occasions during Antarctic spring 2003. On some occasions, extremely rapid loss of ozone was observed (loss of 16 ppbv in 1 min on one occasion), which was associated with regional-scale transport. For each such event during 2003, the air mass originated in the southern Weddell Sea, an area of vigorous sea-ice production. On other occasions the development of the event and its recovery were strongly associated with the build-up and decline of a stable boundary layer. In these cases, air masses had had recent contact with a nearby open water lead where sea-ice production is known to occur. The data presented here are entirely consistent with the idea that halogens responsible for ozone loss are derived during new sea-ice formation from an associated surface such as brine slush or frost flowers.
 The discovery of ozone depletion events (ODEs) within the polar marine boundary layer [Bottenheim et al., 1986; Barrie et al., 1988] came as a complete surprise. Ozone plays a pivotal role in the chemistry of the troposphere and the existence of ODEs pointed to significant weaknesses in our understanding of tropospheric ozone chemistry. Tropospheric ozone depletion is now known to occur in the Arctic, sub-Arctic and the Antarctic [Wessel et al., 1998], but has only been observed around the coast. The spatial extent of these episodes is not known. Ozone makes a significant contribution to radiative forcing [Lacis et al., 1990] and ODEs are postulated to have climate implications regionally. However, the significance will depend on the size of these “tropospheric ozone holes” as well as the height to which they have an influence.
 In the Arctic marine boundary layer in springtime, ozone depletion occurs from the surface up to ∼1.5 km [Bottenheim et al., 2002], and occurs over periods of days to weeks, with concentrations of ozone as low as 50 pptv at times (compared to normal values of around 30 ppbv). The depletion events are clearly associated with halogens, with Br/BrO playing a central role [e.g., Barrie et al., 1988; Foster et al., 2001; Hönniger and Platt, 2002; Platt and Hönniger, 2003]. Satellite observations of BrO, showing particularly high concentrations over sea-ice zones in springtime [Wagner et al., 2001], have reinforced current ideas about the general nature of the reactions leading to depletion, and suggested that sea ice is implicated. Modeling calculations suggest that the ozone-destroying halogens are derived from a sea salt surface via an autocatalytic mechanism, the so-called “Bromine Explosion” [Vogt et al., 1996; Lehrer et al., 2004].
 The precise source of halogens, however, remains an open question; suggestions have included sea salt deposited to the snowpack during polar winter [McConnell et al., 1992], sea-ice surfaces and/or sea salt aerosol [Frieß et al., 2004], and concentrated brines on new sea ice and in frost flowers [Rankin et al., 2002; Kaleschke et al., 2004]. Frost flowers are delicate crystalline structures that grow out of the concentrated brines on newly forming sea ice. They have a large surface area, but persist for only a few days, being destroyed either by wind or buried by falling or blowing snow. The episodic nature of frost flowers has been used as further evidence to link them to ODEs.
 In January 2003, a surface ozone monitor was installed at the Clean Air Sector Laboratory (CASLab) at the British Antarctic Survey base, Halley (75°31′S, 26°40′W) in coastal Antarctica. Surface meteorological observations and detailed boundary layer meteorological measurements are also carried out at Halley. In this paper we combine these data sets, in order to investigate in detail the role played by air mass origin and transport in observed ODEs in this region of Antarctica.
2. Halley and the Clean Air Sector Laboratory
 Halley base lies on the floating Brunt Ice Shelf in the Weddell Sea sector of Antarctica. The surrounding area is flat ice, and the nearest land outcrops are over 200 km away. The station is located in an ellipse of the coastline, such that it is roughly 15 km from the coast to the north, west, and south west (see Figure 1). Halley is subject predominantly to wind from the east, but with a notable number of weaker winds from the west at times [König-Langlo et al., 1998]. During the winter, the Weddell Sea ice extends up to the ice front, but the prevailing easterly winds keep a number of polynyas open, notably Precious Bay to the south west of Halley (see Figure 1). This open water is a significant source of water vapor during the winter/spring months [Anderson, 1993], and, notably, frost flowers forming on brine slush as new sea ice forms have been regularly observed in Precious Bay by Halley overwintering staff.
 The Clean Air Sector Laboratory was commissioned at Halley in January 2003. It is dedicated to the study of the Antarctic troposphere, and the exchange of trace gases between the snow and the overlying boundary layer. To this effect, it is located 1 km south of the station generators, in an area that receives minimal interference from base exhausts. Routine access to the laboratory is by foot or ski. The boundary layer meteorology mast is situated 50 m to the south of the laboratory, surrounded by an open and unimpeded fetch restricted only minimally by the CASLab itself to the north.
3.1. Ozone Measurements
 The surface ozone measurements presented here were carried out at the Halley CASLab. Air was sampled from the main inlet stack, which sampled at a height roughly 10 m above the snow surface. A commercially available instrument (2B Technologies Inc.), which operates using a UV absorption technique, was used for these measurements. Data are presented as 1-min averages of 10 second data; this high resolution being necessary to capture the very rapid onset of certain ODEs. The data presented here have an accuracy and precision of ±1.5 ppbv.
3.2. Sodar Measurements
 A single axis, vertically pointing acoustic sounder (sodar) provided qualitative time-height images of the intensity of atmospheric turbulence at the site. The sodar data are displayed as time series of log(echo strength) profiles, with one profile every 10 s. After adjustment for range, echo strength is proportional to CT2, the temperature structure function, which is a measure of the local variability in the temperature field. The sodar data thus display information about the temperature structure within the boundary layer.
 At this time, CT2 profiles derived from acoustic radars provide little in the way of quantitative measurements. The profile images indicate where the atmosphere is nonhomogenous at a given length scale, but this inhomogeneity may be due to a number of mechanisms such as strong wind shear within weak temperature stratification (type A), a large step in temperature with weak shear at the boundary between different air masses (type B), or convection (type C), to name the most well known examples. Hence the mere presence of a strong echo cannot itself identify which specific mechanism is generating temperature variability.
 The overall image can be interpreted qualitatively, however. For instance, type A appears whenever wind flows over a cooling surface. A continuous horizontal layer aloft implies type B, and under normal conditions may be identified as the top of an inversion. When seen as a sloping horizontal layer, type B may be assumed again, implying the passage of a microfront. Under normal conditions and environments, vertical stripes would invariably imply ongoing convection (i.e., type C). Convective plumes take some while to dissipate, and such decaying “fossil” convection can also be observed on the sodar.
 Weak, shallow plume signature is occasionally observed in the Halley sodar profiles, and this is relevant to a number of the studies presented below. Culf  presents a climatology of Halley plume or “spiky echo” data, and shows that they correlate with westerly winds at the site. These infrequent westerlies place Halley downwind of Precious Bay, an area off the coast with persistent open water even in winter [Anderson, 1993]. Fine resolution temperature soundings during spiky echo events have been presented by Anderson  and show that the plumes are fossil convection, that is, the air has experienced vigorous convection at some recent time in the past, but is no longer being driven by surface warming. In such a way, the sodar data can tell us whether the air mass at Halley has recently experienced convection, which must be caused by passage of cold air over open water.
 The sodar data are also used to identify the depth of the well-mixed boundary layer. Variation in this depth affects the efficiency of trace gas transfer from the surface into the overlying free atmosphere. A well-defined boundary layer effectively caps vertical transport of air, such that any trace gases that are emitted from the surface can become concentrated within the lowest layer of the atmosphere.
3.3. Standard Meteorological Measurements
 Measurements of wind speed and wind direction are carried out at the main station, some 1 km from the CASLab. The anemometers and vanes are located at a height roughly 10 m above the snow surface, and have an accuracy of about 0.5 m/s for wind speed and 5° for wind direction [König-Langlo et al., 1998]. Data are output every 5 min, as averages of the previous 10 min sampling time; i.e., there is some inherent smoothing in the data presented here.
3.4. Meteorological Charts
 Six-hourly surface pressure charts from the European Centre for Medium-range Weather Forecasts (ECMWF) operational numerical weather prediction system were available for the period of interest. These are of high quality even in the Antarctic coastal zone [King, 2003] because of the availability of large amounts of satellite sounder data. The charts allowed the regional (mesoscale) meteorological situation to be examined around the times of the low ozone events. Not surprisingly, good agreement was found between the surface meteorological observations from Halley (which went into the ECMWF analysis system) and the surface pressure charts. However, the ECMWF fields do allow the Halley observations to be set into a broader context and indicate the origins of air masses arriving at the station.
 During the months of August and September 2003, ozone depletion events were observed on several occasions (see Figure 2). Although our data set is limited, the character of ozone loss during this year was quite specific, allowing the events to be categorized into two types: type I with extremely rapid ozone loss, and type II with much more gradual loss. The ozone data, together with all supporting observations, are presented here according to these categories.
4.1. Type I
 These ODEs switch on extremely rapidly, and are associated with very significant ozone loss. Of note is that the onset of two of these events (A and B) occurred during darkness at Halley; the chemical reactions leading to ozone destruction require sunlight, so the fact that these ODEs switched on very rapidly at night, suggests that the ozone loss had not occurred locally. Rather, the air must already have been depleted in ozone, and this ozone-poor air was then advected to the observing site.
4.1.1. Event A: 19 August 2003
 In the days prior to this ODE, Halley had been subject to a violent springtime storm that had persisted for 2 days. Easterly winds had peaked at over 25 m/s. During this time, surface ozone mixing ratios were steady at between 25 and 30 ppbv. However, at 0530 on the morning of 19 August, ozone mixing ratios plummeted from this steady background to around 10 ppbv (see Figure 3, top panel). Indeed, close examination of the data shows that a drop of 16 ppbv was recorded between two successive 1-min readings. The local meteorology data show that, prior to the ODE onset, wind speeds were dropping steadily, and had reached roughly 7 m/s by 0530 (Figure 3, middle panel). Of particular note, however, is the rapid switch in wind direction from east to northwest (Figure 3, lower panel) which is exactly concurrent with the observed drop in ozone.
 A look at the regional charts gives detailed insight into the synoptic situation (see Figure 4), and clues as to the cause of the ozone reduction. An initial low, L1 evident at 1800 on 18 August, existed to the southwest of Halley over the southern Weddell Sea, with a trough that extended up the coast of Dronning Maud Land (Figure 4a). This trough was west of an area of high pressure over Dronning Maud Land, resulting in a marked pressure gradient and the very strong northeasterlies observed at Halley which brought continental air to the station. A secondary low, L2, then developed in the trough offshore of Halley (evident at 0600 on 19 August, Figure 4b). The result was that, rather than receiving continental air, air was drawn from deep within the southern Weddell Sea region and delivered to Halley. Winds at Halley backed and dropped in strength, and the rapid and severe drop in ozone was observed on the cusp of the changing air mass.
4.1.2. Event B: 28 September 2003
 The picture here is very similar to that described above. During the whole of 27 September, Halley had been subject to a storm, with easterly winds of over 15 m/s. Ozone mixing ratios were steady between 25 and 30 ppbv. Toward the end of the day on 27 September, wind speeds began to drop, and essentially reached zero at around 0230 on 28 September (see Figure 5). The drop in wind speed was accompanied by a change in wind direction, with the winds backing from easterly through to north westerly (Figure 5) (note that with a zero wind speed, measured wind direction becomes somewhat questionable). Surface ozone mixing ratios fell rapidly from 30 ppbv at roughly 0230, and reached 2 to 3 ppbv an hour later (Figure 5). A small recovery in mixing ratio was observed around 1030, coincident with a slight increase in wind speed, which can be attributed to mixing in of air rich in ozone. Full recovery is completed around 1900, when the winds had picked up once again to a stormy 18 m/s.
 The regional surface pressure charts for 27 September 2003 show a pronounced northerly airflow delivering air directly to the Antarctic coast around 3° W (Figure 6a). This air had thus had some considerable contact with sea ice on its path to the continent. The air approached Halley from the east during the storm of 27 September because the station was located to the south of the major high pressure system that was dominating surface pressure patterns of the Weddell Sea. Of note, here, is that although the air had previously passed over sea ice, no reduction in ozone was observed. Late on 27 September, a trough/ridge system approached Halley from the west (indicated by the arrows on Figure 6a). As the trough crossed Halley, the winds backed to the north, and then the southwest, drawing air from over the southern Weddell Sea. As the trough moved to the east, the flow over Halley became dominated by the ridge, causing the wind to veer to the north east, which then again brought air to Halley from the central Weddell Sea area (Figure 6b). Once again, air that was depleted in ozone had originated in the southern Weddell Sea region.
4.1.3. Event C: 30 September to 1 October 2003
 The situation for surface ozone and meteorology measured at Halley during this event is shown in Figure 7. Again, strong winds of over 15 m/s, this time from the northeast, dropped in strength just prior to the ODE. At 0850, three things then happened; ozone mixing ratios dropped by 10 ppbv, wind speeds dropped from 9 to 4 m/s, and the winds backed from north easterly to westerly. Wind speeds then recovered slightly around 1300, allowing the ozone loss to be replenished with background air. Full recovery of ozone mixing ratios occurred around 0700 on 1 October, at which time wind speeds were once again around 15 m/s and easterly.
 On the regional scale, a very deep low over the southern Weddell Sea was responsible for the strong north easterly winds arriving at Halley (see Figure 8a). However, a key feature here was a very short-scale ridge (shown by the arrows on Figure 8a) that propagated southeastward through the central Weddell Sea and temporarily drew air from the area of the southern Weddell Sea toward Halley (Figure 8b). The winds observed at the station backed for a very short period of time, coincident with observed ozone loss, before returning to the climatological norm at which point ozone levels also recovered.
4.1.4. Summary of Type I
 In each case presented here, the climatological large-scale flow bringing background ozone over Halley, was disrupted by transient small-scale features. These interrupted this climatological flow, and in each case, drew air from the southern Weddell Sea area and delivered it to Halley. When this happened, ozone levels dropped extremely rapidly, indicative of very sharp gradients in ozone between adjacent air masses. Of particular interest is that the southern Weddell Sea is known to be an area of extensive sea-ice production. Climatological southerly airflow from off the Ronne Ice Shelf results in the regular formation of coastal polynyas [Renfrew et al., 2002]. With the cold temperatures of winter/spring, sea-ice production in this region is vigorous, accounting for over 9% of sea-ice production in the entire Weddell Sea in some years [Renfrew et al., 2002]. The clear association presented here between air mass origin and changes in observed levels of ozone is very strong evidence that new sea-ice formation in the southern Weddell Sea is highly significant in driving ozone destruction processes.
 These type I ODEs were, in each case, associated with strong easterlies (greater then 15 m/s) backing to weak westerlies. The unsubtle nature of this association makes it possible to consider the null scenario; i.e., does this apparent association always hold true? By looking at the complete data set of wind speed and wind direction during spring 2003, there was no occasion when such a meteorological situation did not lead to an ODE. From the one years data present here, it thus appears that it should be possible to forecast type I ODEs in future years.
4.2. Type II
 These ODEs appear much more gradually, and are not as deep as those described above. Of particular interest are the clear associations between ozone loss and structures within the boundary layer; ozone mixing ratios declined as layers within the boundary layer formed, and recovered as these layers disappeared. In both of the cases presented here, the onset of ozone depletion occurred during daylight hours, allowing for the possibility that photochemical reactions had occurred locally. Further, the air depleted in ozone was not fast moving and had had contact with offshore leads local to Halley. It is known that new sea ice forms in these leads, together with associated frost flowers growing from surface brine slush.
4.2.1. Event D: 25–27 August 2003
Figure 9 shows the behavior of surface ozone during this depletion event. Ozone loss was much less severe than described above, and the loss occurred much more gradually. Looking at wind speeds, the very striking change in wind speed seen for type I ODEs was not evident here. The winds remained below 5 m/s throughout this period, although they were at their lowest during the ODE. Wind directions, however, again show a shift during the onset of the ODE, but now a more gradual change from south and east through to north and west.
 The quiescent regional meteorological conditions allowed the formation of complex structures within the boundary layer, as shown by the sodar data (Figure 9, lowest panel). Already some structure was evident at the start of 25 August between the surface and roughly 100 m height. Later in the day, several layers propagated upward, commencing around 1330. This coincided with the onset of ozone depletion. The layers remained in evidence throughout the day, and collapsed around 0330 on the morning of 27 August. This was the time that ozone mixing ratios recovered to their former values.
 Complex structure is frequently observed in the Halley acoustic echo data, when winds are light and the whole of the boundary layer is stratified [Anderson, 2003]. Such structure implies a deep stable boundary layer with weak vertical mixing. Under such conditions air within the boundary layer would be contained with little transfer of trace gases upward into the overlying free troposphere. Consequently, any surface emissions would build up and subsequent chemistry would be contained.
4.2.2. Event E: 2–4 September 2003
 This was another event where ozone loss was fairly deep, declining from a background of roughly 30 ppbv to roughly 10 ppbv (see Figure 10), but again, the decline occurred gradually. Wind speeds were low, varying between zero and 5 m/s. The wind direction had been easterly since the start of the month, and during the morning on 2 September, showed no clear direction, the variability meaning very little at these low wind speeds. As the wind speed picked up slightly, around 1600, the observed direction was westerly, and remained so until speeds dropped once again, around 0130 on 3 September. Once wind speeds had recovered slightly, the wind direction was once again easterly. So once again, the switch from easterly to westerly, and then back to easterly, seems to be in evidence, but this time, at very low wind speeds. The regional surface pressure charts show a high pressure system was controlling the quiescent conditions at this time.
 In order to understand what was driving the ODE, it is helpful to turn again to the sodar data. Figure 10 (lowest panel) shows the boundary layer structure over this period. By comparing with the ozone mixing ratios over the same time period, a very interesting association becomes apparent. The main feature of the sodar echogram is a significant layer rising from the surface around 1900 on 2 September to an altitude of 500 m. This layer was maintained throughout the following day, then gradually descended again reaching the surface at around 0200 on 4 September. The lifetime of this layer coincided very closely with the onset and termination of the ODE, which took off at exactly the same time on 2 September, was maintained throughout 3 September, and then recovered rather fast around 0230 on the morning of 4 September.
 Another aspect of the echogram is also very significant. Beneath the elevated layer, the echo is “spiky”, that is, showing weak plume signature. Such a signature is evidence of fossil convection and implies recent surface heating. The most plausible source for such convective heating at this time of year in Antarctica is passage of cold air (<−20°C) over open water (>−2°C). When the upper level of the spiky layer re-attached to the surface, the local instruments sampled air with no history of recent passage over water. Indeed, the ozone mixing ratios then returned to background levels.
4.2.3. Summary of Type II
 Quiescent regional meteorological conditions allowed stable layers within the boundary layer to form. Such layers contained air within the boundary layer such that any surface emissions would build up and the signature of subsequent chemical reactions would be amplified. Once these layers became unstable, mixing occurred with background air and the ozone-depleted air was enriched. Certainly one of the type II events had a clear signature in the sodar data showing recent contact with open water in nearby Precious Bay, an area also known for new sea-ice formation.
 The data presented here are of interest for a variety of reasons. First, they clearly show that the meteorological conditions are instrumental in determining whether an ODE is observed at a particular site or not. From that point of view, the ODE is merely an episodic “event” because depleted air masses pass over the fixed measurement location for only a relatively short period of time. The data here suggest that there is a pool of ozone depleted air sitting over the southern Weddell Sea, tongues of which will be observed at Halley when the mesoscale flow brings air from this region of the sea-ice zone in to the measurement station.
 Secondly, these observations contribute to the debate about the source of ozone-destroying halogens. The precise source of halogens for polar tropospheric ozone depletion has been regularly debated. There is no doubt that it is associated with the surface of sea ice, as unambiguously demonstrated by the trajectory study of Frieß et al. . However, a better understanding of the nature of the sea ice (e.g., first year ice, brine slush, frost flowers etc.) becomes important for numerical modeling calculations that seek to simulate observed ODEs, and investigate their potential wider implications.
 The results presented here show that, in 2003, for each of the 3 occasions on which very rapid ozone loss was observed at Halley, the air masses had originated in the southern Weddell Sea region, an area that is known to produce very large amounts of sea ice. This association points strongly to an association between the process of forming new sea ice (e.g., the brine slush or the frost flowers) and ozone depletion mechanisms. On the occasions when ozone loss appeared more gradually, the signature of the sodar data implied that the air was significantly influenced by the local polynya in nearby Precious Bay. Here, frost flowers growing from the brine slush on newly forming sea ice are often observed by the Halley overwinter staff. The inference here is that, for type II ODEs, local processing associated with newly forming sea ice is the driver for ozone destruction. This is investigated in further detail below.
5.1. Back Trajectories and Sea-Ice Conditions
 Radar scatterometer instruments aboard satellites can give useful information about polar ice conditions. The strength of the return beam reflects surface roughness, such that in polar regions, areas of newly forming sea ice (e.g., brine slush or frost flowers) give a distinct return. Rankin et al.  successfully used images from the QSCAT instrument in tandem with back trajectory analyses (calculated using the British Atmospheric Data Centre trajectory package) to track air parcel movement relative to areas of newly forming sea ice. We use a similar approach here, to assess whether air that is depleted in ozone has had recent contact with newly forming sea ice. This approach is particularly useful for type II events when a specific local source is under investigation. Following trajectories back for only 1 or 2 days gives a high degree of reliability in the trajectory path.
Figure 11 shows the back trajectory for air reaching Halley at 1400 on 25 August, the time when the type II ozone depletion event D commenced. The dots on the trajectory mark the air parcel position every hour. The trajectory is superimposed on the QSCAT image for that day. A dark patch on the QSCAT image to the south west of Halley, just offshore and partly obscured by the trajectory itself, indicates the presence of brine slush/frost flowers.
 Such back trajectories were calculated for each hour from 1100 on 25 August. The early trajectories show the air approaching Halley from aloft over the Brunt Ice Shelf, with the final approach to Halley close to the ground across the shelf ice to the south east of Halley. Over subsequent hours, the trajectory progressively changed, such that the final approach to Halley was closer and closer to the coast, but always close to the ground. The trajectory shown in Figure 11 arrived at Halley at 1400 and is the first of the series to clip the coast, and to cross the dark patch of newly forming sea ice shown on the QSCAT image to the south west of Halley. The ODE on this day began just after 1300, and by 1400 was clearly established.
 Of interest are the timescales involved here; the air parcel was moving very slowly, and spent of the order 1 to 2 hours in contact with the newly forming sea ice. It then took 7 to 8 hours to reach Halley. The background ozone mixing ratio prior to the ODE was 30 ppbv, but by 1400 it had fallen to 23 ppbv. This suggests that, if there was a local source of bromine offshore of Halley and that ozone destruction continues during transport to Halley, then the rate of ozone loss was of the order 1 ppbv/hour.
Figure 12 shows the QSCAT image for 2 September 2003, with an area of newly forming sea ice to the south west of Halley. Superimposed on the QSCAT image is the calculated trajectory for air arriving at Halley at the start of ozone depletion event E. Prior to the start of the event, the air arriving at Halley had remained over the land, tracking the coastline and arriving at Halley from the north east. After 1500 there was progressively more and more contact with the newly forming sea ice in Precious Bay. Again, the air was moving slowly, and the approach to Halley was at ground level. The trajectory arriving at Halley at 2200, when the ODE reached its nadir, showed that the air had spent roughly 4 hours directly over the newly forming sea ice, and then took roughly 5 hours to reach Halley. Ozone mixing ratios had decreased from a background of ∼30 ppbv to ∼10 ppbv by 2200.
 The coincidence of ozone depletion at Halley and the passage of air parcels over nearby newly forming sea ice strongly suggest that the depletion observed during type II events is locally driven, and sustained by the stable boundary layer.
5.2. Evidence for Frost Flower Link From Halley Aerosol Data
 Some additional evidence to address this issue appears in the form of the high resolution Halley aerosol record. In 1991 and 1992 a program of daily aerosol sampling was carried out at Halley [Hall and Wolff, 1998]. These data showed the very wide variation in sea salt aerosol concentration from day to day which was a feature from late autumn to early spring. By considering local meteorology, it was shown that the high concentrations of sea salt aerosol occurred under specific meteorological conditions; moderately strong offshore easterlies occurred before the high sea salt event; a drop in wind speed and a change in direction to westerly then occurred either on the day before the event or on the day of the event itself. The source of this aerosol was investigated and concluded to be associated with newly forming sea ice. The strongest piece of evidence for this was the fact that the aerosol was depleted in sulphate relative to other seawater ions. Such fractionation is known to occur during the formation of new sea ice and in particular frost flowers are significantly fractionated relative to standard mean ocean water [Rankin et al., 2002]. The high sea salt aerosol events during specific meteorological conditions could thus be linked directly to newly forming sea ice and associated brine slush and frost flowers. The fact that we observe ODEs under essentially the same meteorological conditions provides additional evidence that air depleted in ozone has passed over areas of newly forming sea ice, brine slush and frost flowers. If this is the case, it suggests that some chemical processing could occur in situ, but further, given that aerosol is mobilized under such conditions, it also suggests that processing could occur on the airborne aerosol during air parcel transport.
 At Halley during austral spring 2003, observed tropospheric ozone depletion events could be categorized into two types. Type I ODEs were associated with extremely rapid and significant loss in ozone, and the air masses on each occasion had an origin in the southern Weddell Sea. Weaker ODEs (type II), which displayed a much more gradual onset than the type Is, were strongly modulated by the behavior of the boundary layer. For both type I and type II, however, there is very strong evidence that the observed ozone depletion was driven by air mass contact with newly forming sea ice. It has been postulated previously that either brine slush or frost flowers that are generated during new sea-ice formation could be a source of ozone-destroying halogens. The data presented here provide compelling observational evidence that this is indeed the case.