The Antarctic ozone hole during 2018 and 2019

Keywords: Antarctica, climate, meteorology, Montreal Protocol, ozone, ozone depletion, ozone hole metrics, southern hemisphere, stratosphere.

Assessing the behaviour of the annual Antarctic ozone hole continues to be an important aspect of monitoring the impact of Montreal Protocol controls on ozone depleting substances (ODS) and for observing the effects of Antarctic ozone on the interactions between components of the climate system (Goyal et al. 2019; Lickley et al. 2020; Solomon et al. 2020). While the Montreal Protocol is working to reduce stratospheric ozone loss (WMO 2018), recent increases in some ODS emissions have been identified (Montzka et al. 2018; Claxton et al. 2019; Fang et al. 2019; Rigby et al. 2019) with likely effects on recovery of Antarctic ozone (Dhomse et al. 2019). At the same time, the accuracy in evaluating ozone trends is influenced by variability in stratospheric conditions and their relationship to large-scale climate modes (Zhang et al. 2017; Chipperfield et al. 2018; Maycock et al. 2018; Ball et al. 2019).

The processes leading to ozone depletion from ODS primarily take place inside the Antarctic stratospheric polar vortex from late winter through to early summer, with peak ozone loss occurring in spring (Solomon et al. 1986). The overall level of ozone depletion is strongly dependent on temperatures inside the vortex, as well as the availability of chlorinated and brominated ODS (Langematz et al. 2018). This primarily arises because of the central role played by Polar Stratospheric Cloud (PSC) particles, which form in the lower polar stratosphere when temperatures fall below ~190 K, in the process of transforming ODS into forms that destroy ozone. Years when the vortex is cold and stable during winter and spring result in the greatest ozone loss for a given level of ODS (Zhang et al. 2017). Rapid warming of the vortex occurs when the dynamical barrier below or around the vortex is weakened by the upward propagation of atmospheric waves – primarily the planetary-scale Rossby waves. The origin of these disturbances may be from remote Antarctica, driven by large-scale deep convection or land–sea temperature contrasts and influenced by meanderings of sub-tropical and mid-latitude circulation (Peters and Waugh 2003). Weakening of the vortex leads to intrusion of warmer air from lower latitudes or removal of air that has been preconditioned for ozone depletion (i.e. chemically processed by heterogeneous reactions on PSC particles and other aerosols (Haigh and Roscoe 2009)). Both these effects can curtail the ozone depletion process.

Climate model simulations and observational data have shown robustly that Antarctic ozone depletion leads to cooling of the stratosphere, which promotes circulation changes that couple downward to the surface on time-scales of weeks to months, affecting conditions mainly in summer (Polvani et al. 2011; Bandoro et al. 2014; Li et al. 2016). This coupling affects not only the Antarctic surface but wide areas of the southern hemisphere (Thompson and Solomon 2002). It is manifested primarily as a tendency for the Southern Annular Mode (SAM) to be forced into a more positive state, with lower surface pressures at sub-polar latitudes (around 65°S), higher pressures at mid-latitudes (around 40°S) and a general strengthening and poleward contraction of the westerly wind belt over the Southern Ocean (Son et al. 2010; Thompson et al. 2011). The SAM is the primary mode of surface climate variability in the Antarctic and Southern Ocean region (Gong and Wang 1999). It not only responds to stratospheric ozone change but also to the effects of tropospheric global warming (Wang and Cai 2013) and other large-scale climate modes, including the El Niño Southern Oscillation (ENSO) and the Indian Ocean Dipole (IOD) (Fogt and Marshall 2020). Dennison et al. (2015) showed that the persistence and variability of the stratospheric SAM is enhanced by ozone depletion, and this leads to longer lasting coupling between the stratosphere and troposphere. Lim et al. (2019) found that weakening of the Antarctic vortex tended to enhance fire conditions in eastern Australia. A series of unprecedented climate extremes occurred in the southern hemisphere during late 2019 and early 2020, which were enhanced by stratospheric influences (Lim et al. 2020). The extremes included widespread heatwaves and severe bushfires in Australia (Davey and Sarre 2020) and South America (Blunden and Arndt 2020) and record high temperatures in Antarctica (Robinson et al. 2020). Associated with these conditions was a generally strong negative phase of the SAM and the most strongly positive IOD since 1997 (Blunden and Arndt 2019). In the Pacific, the ENSO was in a weak La Niña phase. Extreme conditions developed in the Antarctic stratosphere in September 2019, with a minor Sudden Stratospheric Warming (SSW; Lim et al. 2020; Newman et al. 2020; Rao et al. 2020) resulting in one of the smallest ozone holes since the 1980s (Kramarova et al. 2020; Wargan et al. 2020). Conditions that favoured the development of the SSW included the prevailing positive IOD phase, warm central Pacific sea surface temperatures, the easterly phase of the Quasi-Biennial Oscillation (QBO) and high latitude tropospheric patterns, including weak zonal wind anomalies in the Indian and Pacific Ocean sectors (Milinevsky et al. 2019; Lim et al. 2020; Rao et al. 2020; Shen et al. 2020). Downward coupling from the SSW is implicated in enhancing surface temperatures during summer in Antarctica (Scambos and Stammerjohn 2020) and Australia (Lim et al. 2020).

Southern hemisphere conditions during the latter half of 2018 and early 2019 were much less marked compared with the following year. While temperatures across parts of the southern hemisphere were above normal, heatwave conditions were only notable in eastern Australia (Blunden and Arndt 2019). Generally, ENSO-neutral conditions persisted until November and then moved into a weak La Niña. While there were significant fires in parts of Australia, overall fire activity in the southern hemisphere was below the 2003–2017 average. The IOD was in a positive phase for much of the year, although this was less marked than in 2019. The SAM was in a positive phase during spring and early summer, and this moderated the drying effect of the positive IOD phase in eastern Australia. The Antarctic ozone hole was close to average in size for this century, while the temperature of the stratospheric vortex was generally below average during spring (Kramarova et al. 2019).

To further build on previous analyses, we provide here additional information on the Antarctic ozone holes of 2018 and 2019, and compare and contrast their behaviour using a variety of observations, including measurements made at sites in East Antarctica and the Southern Ocean operated by Australia and New Zealand researchers and collaborators. This continues a series of analyses of annual Antarctica ozone holes, which has covered the period since 2007 (Klekociuk et al. 2019 and references therein; Tully et al. 2019; Krummel et al. 2019b). The outline of this paper is as follows. In Section 2 we describe the data and methods used. In Section 3, we provide metrics for the 2018–2019 ozone holes and discuss the characteristics of the metrics in terms of prevailing meteorological factors and observed effects at specific measurement sites. A summary of our findings and conclusions are presented in Section 4. The methods used to obtain the ozone hole metrics are described in Klekociuk et al. 2019 and references therein; Tully et al. 2019; Appendix A.

To characterise the vertical distribution of ozone within the ozone hole, we used in situ measurements of ozone mixing ratio obtained from ozonesonde balloons that were flown approximately weekly at Davis Research Station, Antarctica (68.57°S, 77.97°E; Fig. 1) during 2018 and 2019. The equipment for each ozonesonde consisted of a Science Pump Corporation ECC6A electrochemical cell that were flown under a 1200 g balloon. The sensing cathode solution (3.0 cm³) consisted of 1% potassium iodide (weight per volume) with potassium bromate and buffers, and the anode solution (1.5 cm³) was a saturated potassium iodide solution (Johnson et al. 2002). These general specifications and operation the ozonesonde equipment (including the calibration method) have been maintained since the start of the program at Davis in February 2003, with the only exception being the type of radiosonde equipment used (RS80 prior to March 2006, RS92 until February 2019 and currently RS41). Here we used 10 s cadence measurements (obtained at approximately 50 m height intervals) that are archived at WOUDC (2020).

Fig. 1.  Map showing locations of Casey, Davis, Mawson, Macquarie Island, Scott Base and Arrival Heights.

The Australian Radiation Protection and Nuclear Safety Agency (ARPANSA) monitors the levels of solar ultraviolet radiation (UVR) at Australia’s Antarctic research stations (Fig. 1) – Casey (66.28°S, 110.53°E), Davis and Mawson (67.60°S, 62.87°E) – using broadband UVR detectors (UVBiometer Model 501, Solar Light Co., Philadelphia, USA). The spectral responses of these detectors are similar to that of human skin, described by the erythema spectral weighting function (CIE 2020). The recorded values of erythemally effective solar UVR were averaged over a period of 10 min and converted to UV Index values before being reported. One UV Index unit is defined to be equivalent to an erythemally effective irradiance of 25 mW m⁻² (WHO 2002). The maximum 10-min average UVR intensity value recorded each day was reported as the UV Index.

Prior to deployment, the broadband radiometers were calibrated at ARPANSA by comparison with simultaneous solar UVR measurements from a spectroradiometer (model DTMc300, Bentham Instruments Limited, Reading, UK). The spectroradiometer was calibrated for irradiance using a 1000 W quartz tungsten halogen lamp traceable to the National Measurement Institute (Australia) and its wavelength calibration was determined with reference to the UV spectral lines of a mercury lamp.

At Arrival Heights in Antarctica (77.9°S, 166.8°E), a mid-infrared (MIR) Fourier transform spectrometer (FTS) is used to obtain gas-phase total column abundances for a variety of species relevant to ozone chemistry from solar observations (Kohlhepp et al. 2012), including ozone (O₃), hydrochloric acid (HCl), chlorine nitrate (ClONO₂), nitric acid (HNO₃) and nitrogen dioxide (NO₂). FTS measurements have been made between late August and early April annually since 1996. Additionally, measurements of stratospheric chlorine monoxide (ClO) have been taken at Scott Base (77.85°S, 166.76°E) using a microwave radiometer since 1996 (Nedoluha et al. 2016). Both the FTS and microwave radiometer measurements are made under the auspices of the Network for the Detection of Atmospheric Composition Change (De Mazière et al. 2018; NDACC 2020).

Total column ozone measurements are made at Macquarie Island (54.50°S, 158.95°E) year-round, several times a day using a Dobson spectrophotometer, continuing a time series originating in 1957. Data are archived by WOUDC (2020) and NDACC (2020). Macquarie Island is normally situated outside the vortex near its northern edge. The only other active sites at a similar latitude measuring total column ozone are in South America: Ushuaia (54.8°S, 68.30°W), Punta Arenas (53.16°S, 70.92°E) and Rio Gallegos (51.62°S, 69.21°W). Here we only consider the Macquarie Island data in order to examine ozone variations in the Australian longitude sector.

We used daily global gridded (level 3) total column ozone data from three separate satellite-borne instruments to evaluate metrics of the ozone hole using the methods described in Appendix A: (a) the Total Ozone Mapping Spectrometer (TOMS), (b) the Ozone Monitoring Instrument (OMI) and (c) the Ozone Mapping and Profiler Suite (OMPS). These instruments provided/provide ozone information from sun-synchronous orbits using nadir measurements of backscattered (by the atmosphere) solar UV. The time intervals, spatial resolutions and versions of the total column ozone data used in our analysis are provided in Appendix A. TOMS instruments operated on three satellites: Nimbus 7 (October 1978–May 1993), Meteor 3 (August 1991–November 1994) and Earth Probe (July 1996–December 2005 (McPeters et al. 1998). The OMI instrument is on the Earth Observing Satellite Aura, operational from August 2004 (Kroon et al. 2008; McPeters et al. 2008). OMPS is a three-part instrument on the Suomi National Polar-orbiting Partnership satellite, operational from October 2011 (Bai et al. 2013). Specific issues have affected data quality from some of these instruments. Earth Probe TOMS suffered from a progressive degradation in measurement quality, which has been corrected empirically (Antón et al. 2010). The OMI retrievals account for an unintended obstruction in its field of view and performance issues in a particular viewing direction (Dobber and Braak 2010).

We used version 4.2 level 2 swath data from the Microwave Limb Sounder on the Aura satellite (Aura/MLS) for air temperature, geopotential height, nitrous oxide (N₂O) and ozone mixing ratios on pressure levels to assess temperature and composition anomalies over the Antarctic region. The Aura/MLS data were selected according to quality control thresholds (NASA 2020a). Daily temperature data were interpolated to a uniform geometric height grid with 1 km spacing and averaged over latitudes poleward of 60°S. The ozone and N₂O data (which are provided in the level 2 product with reference to an absolute molar scale based on calibrated in situ measurements) were interpolated to specific isentropic levels and spatially averaged within the ‘inner vortex edge’ region (Lee et al. 2001; Roscoe et al. 2012) defined in the MLS Derived Meteorological Product (DMP) dataset (NASA 2020b) with the Modern-Era Retrospective analysis for Research and Applications version 2 reanalysis (MERRA-2; Gelaro et al. 2017). As noted by Manney et al. (2007), identification of the vortex edge can be problematic under certain circumstances, particularly outside of winter, and mixing ratio estimates within this region may not necessarily accurately reflect the broad average of the inner vortex.

We use data from three reanalyses to provide information on the atmospheric state in the southern hemisphere during 2018: the European Centre for Medium-Range Weather Forecasts (ECMWF) fifth generation reanalysis ERA5 (Hersbach et al. 2020), MERRA-2 and the Multi-Sensor Reanalysis version 2 (MSR2; van der A et al. 2015). ERA5 was used to derive the meridional eddy heat flux T*u* as a function of latitude and pressure (where T and u are the air temperature and zonal wind speed, respectively, and * denotes the zonal anomaly), which is a proxy for the upward energy propagation by planetary-scale (Rossby) waves. We also used ERA5 to evaluate a proxy for the size of the stratospheric polar vortex as the area enclosed by contours of potential vorticity on particular isentropic surfaces (Braathen and Proffitt 2000). A widely used method of determining the extent of the vortex is that of Nash et al. (1996). While estimates of vortex area are problematic, and alternative methods have been developed (Roscoe et al. 2012; Smith and McDonald 2014), the area proxy used here shows similar seasonal and inter-annual behaviour to the that obtained with the Nash et al. (1996) method (not shown) and is more readily applied.

MERRA-2 was used to evaluate the daily climatology of ozone mixing ratio as a function of height for Davis over 2008–2017 to examine the ozone anomaly in the ozonesonde measurements during 2018 and 2019. The MSR2 data (KNMI 2020) consist of daily total column ozone estimates constructed from a data assimilation method applied to satellite and ground-based total column ozone measurements. These data, along with ERA5 total column ozone data, were used to evaluate the annual ozone hole metrics using the same methods that were applied to the satellite total column ozone data (Appendix A). For the period 2000–2006, the ERA5.1 special run (Adrian et al. 2020) was used for evaluation of the ozone hole metrics as discussed in Appendix A.

Both ERA5 and MERRA-2 are regarded as providing a suitable representation of temperatures and dynamics in the lower stratosphere of the southern hemisphere, particularly in the past decade when increased assimilation of observations, particularly from Global Positioning System occultation profiles, has reduced temperatures biases (Friedrich et al. 2017; Adrian et al. 2020). However strong stratospheric disturbances been shown to provide challenges for the representation of the vertical temperature structure and chemical mixing ratios in modern reanalyses, particularly in the upper levels (Manney et al. 2008; Manney et al. 2009).

Fig. 2a shows daily minimum total column ozone south of 35°S (Metric 4 in Table 1). In 2018, the first excursions below the 220 Dobson Unit (DU) contour, which provides a threshold to define the ozone hole (Uchino et al. 1999; Newman et al. 2004; Bodeker et al. 2005), occurred sporadically in the second half of July and again in early August, which was typical of recent years. These variations potentially relate to effects from variability in the size and shape of the ‘collar’ region of the vortex (Newman et al. 2004), particularly due to the mixing of air within the vortex with ozone-poor air at the vortex edge, where photolytic ozone destruction was only just beginning due to the return of sunlight at the end of the polar night. Additionally, variability in the size of the ozone hole during its formation can be influenced by chemical changes associated with the enhancement of PSCs by mesoscale waves, or small-scale effects from gravity waves (Alexander et al. 2013). The metric remained below 220 DU from the second week of August until the ozone hole closed in the first week of December. The tendency in this metric from August until mid-September was similar to the long-term (1979–2017) mean (Fig. 2a), after which the metric remained below average for the remainder of the ozone hole period. It reached a relatively deep minimum of around 102 DU (ranking 14th lowest since 1979) during early October. Up until early November, the time series of this metric was similar to that observed in 2015 and was at or near the lowest observed values from mid-October to early November.

Table 1.  Summary of ozone hole metrics
The description of each metric and the method used to produce it are described in Appendix A. Values for 2018 and 2019 are highlighted by blue and red bold font, respectively. Values produced using MSR2 data are highlighted in black bold font. Rank 1 = lowest ozone minimum, greatest area, greatest ozone loss etc.; rank 2 = second lowest ozone minimum, etc.

Fig. 2.  Daily ozone hole metric time series. (a) Daily minimum total column ozone poleward of 35°S (daily values of Metric 4 of Appendix A) for 2014–2019. (b) As for (a) but showing the daily average ozone hole area (daily values of Metric 1). (c) As for (a) but showing the daily ozone deficit (daily values of Metric 6).

Fig. 2b shows the daily Antarctic ozone hole area (the area within the 220 DU contour; Metric 2 in Table 1) during the 2018 season. The maximum daily area of the ozone hole (24.7 million km² (Mkm²) on 20 September) was the 17th largest in our analysis and ~86% of the largest area observed in 2000. The maximum 15-day average ozone hole area (Metric 1 of Table 1) for 2018 (24.1 Mkm²) was the 14th largest hole in our analysis, larger than the median value for 1979–2017 (22.1 Mkm²). These statistics indicate that the 2018 ozone hole was on the larger side of ‘middle of the pack’. Notably, it was also larger compared to the ozone holes of the preceding 5 years, except for 2015 (Table 1) when ozone depletion was aided by volcanic aerosols and the absence of strong dynamical disturbance of the polar vortex (Stone et al. 2017).

The ozone hole area time series for 2018 shown in Fig. 2b exhibited specific disturbed periods during the austral spring, which are discussed in Section 3.3 below. The ozone hole grew rapidly in size from mid-August and peaked in area on 20 September, about 2 weeks earlier than the timing of the lowest total column value (Fig. 2a). Late September and early October saw two small warming events which reduced the ozone hole area from around 24 Mkm² to around 22 Mkm². The ozone hole area remained in the 20–22 Mkm² range for the remainder of October and the first week of November. From early November, a generally stronger decline in area began, which was punctuated by brief increases around the 10, 20 and 28 November. The recovery during November took place during the overall warming of the mid- to lower stratosphere discussed in Section 3.3.

Fig. 2c shows the daily maximum ozone deficit in the Antarctic ozone hole with respect to 220 DU (Metric 6 of Table 1), which is a function of both ozone hole depth and area. The maximum daily ozone deficit in 2018 was 34.8 million tonnes (Mt) in late September, the 14th largest deficit. Integrated over the whole ozone hole season, the total ozone deficit (the sum of the daily ozone deficits) in 2018 based on OMPS data was approximately 1810 Mt. This is the 14th largest cumulative ozone deficit analysed and ~71% of the value for 2006, which is the maximum annual deficit in Table 1.

The formation of the 2019 ozone hole became apparent from mid-July to early August, similar to 2018. As seen in Fig. 2a, daily minimum total column ozone values (Metric 2) stayed consistently below the 220 DU threshold from late July. This was slightly earlier than in 2018. The metric rapidly dropped from early August and at times fell below the climatological range of the satellite record up to the third week of August. After reaching a minimum of 120 DU on 18 August, the metric started to trend back towards the climatological average. By mid-September the minimum column value was noticeably different from the behaviour in the recent years (Fig. 2a). In the second week of November, the metric crossed the 220 DU threshold, resulting in one of the earliest breakdown dates of the ozone hole in the observational record (Metric 8 of Table 1).

Coincident with falling ozone concentrations in the sunlit portion of the polar cap, the daily area metric (Fig. 2b) rapidly rose in early August and was above the climatological mean for most of the month. By early September the growth of the hole started to wane. However, the area reached a brief peak on 8 September of 15.7 Mkm², ranking Metric 2 the 9th smallest in our analysis. In the following two weeks, the overall size of the ozone hole decreased and reached a local minimum around 21 September. Subsequently there was a period of slight growth to early October. This was followed by a period of fairly stable size, although this was about half the climatological mean at that time. From the third week of October, there was an overall decline in the metric until breakdown.

Consistent with the growth in ozone depletion in early August, the daily ozone deficit (Fig. 2c; Metric 6) showed a rapid early rise to levels that were at the maximum of the climatological record up to 18 August. After this, there was only limited further growth. The peak deficit of 10 Mt was reached on 3 September, ranking this metric 34th in our analysis and comparable with values observed in the mid-1980s. The subsequent behaviour of Metric 6 tended to follow a similar variation to the area metric (Metric 4; Fig. 2b) with a period of relative stability from late September to mid-October, before a final general decline.

Overall, the behaviour of the 2019 ozone hole strongly contrasted with that of 2018. The early growth in 2019 was generally stronger than in 2018, and indeed the early size of the 2019 ozone hole was near the climatological maximum. However, disturbances to the polar vortex that began in mid-August, becoming exceptional in September (as discussed in Section 3.3 below), significantly curtailed the ozone depletion process. These disturbances generally abated by early October. By this time, the 2019 ozone hole was well below the climatological mean and was one of the smallest observed since the early years of the observational record. It is useful to note that the metrics for 2019 in Table 1 are closely ranked to some other years of the preceding decade, particularly 2017, which is discussed in detail by Klekociuk et al. (2019). We return to this point in Section 3.5.

The annual time series of Metrics 4, 1 and 7 of Table 1 are presented in Fig. 3a, b and c, respectively, which further highlights the differences between 2018 and 2019 and how exceptional 2019 was in comparison with earlier years. While there are obvious differences in the annual metrics calculated using the different data sets (the origin of which is discussed in Appendix A), particularly for 2019, the ozone hole in 2018 showed characteristics similar to 2007 and 2009 (Table 1) when ozone depletion was aided by relatively low stratospheric temperatures (Tully et al. 2008; Tully et al. 2011). In contrast, the 2019 ozone hole was similarly anomalous to 2002 and 1988 when the stratosphere was strongly disturbed (Newman and Nash 2005; Roscoe et al. 2005).

Fig. 3.  Annual ozone hole metric time series. (a) Minimum daily total column ozone (Metric 4) for MSR2 (MSR2; orange dots) and ERA5 (ERA5; yellow dots and line) reanalyses, and CSIRO analysis of satellite data (CSIRO; blue dots connected by black line). (b) As for (a) but showing the annual maximum daily average ozone hole area (Metric 1). (c) As for (a) but showing the annual total ozone deficit (Metric 7).

Table 2 presents monthly ERA5 polar cap temperature anomalies at three pressure levels (10 hPa – mid-stratosphere, 50 hPa – lower stratosphere and 100 hPa – near tropopause) for June–December. Anomalies for 2018 and 2019 are provided, and for the exceptional years of 2002 and 1988. The winter months (June–August) of 2018 and 2019 were not particularly outstanding compared with the interquartile range, although temperatures in the later part of the 2018 winter tended to be well below average. The exceptionally warm stratosphere in spring and early summer of 2019 is clear and shows a time–height progression similar to the anomalous behaviour in 2002 (Table 2). However, unlike 2019, the 2002 winter was anomalously warm for most of the period. The 2019 warming became prominent in September, particularly at 10 hPa, and progressed to strongly influence the lower levels by mid–late summer (Kramarova et al. 2020; Lim et al. 2020; Newman et al. 2020). As can be seen in Fig. B1, the potential extent and duration of PSCs in the polar cap was likely much reduced in 2019 (Fig. B1b) compared with 2018 (Fig. B1a), not only because of the strong warming in spring but because of above average temperatures in the early part of the winter.

Table 2.  ERA5 monthly mean temperature anomalies (K) with respect to the 1979–2019 climatology over 60–90°S for June–December for selected years at 10, 50 and 100 hPa
Red and blue fonts indicate positive and negative anomalies respectively. Blue and orange shading indicates 2018 and 2019, respectively

Fig. 4 shows the height-time behaviour of the daily Aura/MLS temperature anomaly over the southern polar cap for from January 2017 to July 2020. Obvious features in Fig. 4 are the warm and cool anomalies that descend with time, generally starting in the upper mesosphere (~90 km) in early autumn (March) and reaching the lower stratosphere by early summer (November). These anomalies have been discussed for specific years (e.g. 2017 is discussed in Klekociuk et al. 2019). As seen in 2018 (Fig. 4b), a band of below-average temperatures was situated underneath warm anomalies during spring, which contrasted with 2017 (Figs. 4a) and 2019 (Fig. 4c). The strength of the anomalies in the upper stratosphere in 2019 (towards the stratopause at ~50 km) up to mid-August was generally less marked, particularly in comparison to 2017. However, after this time, strong warming took place in the mid- and lower stratosphere. This appeared to persist through to March 2020 (Fig. 4d) and possibly longer near the tropopause (~10 km height). This anomaly sometimes reached the maximum daily extreme in the Aura/MLS climatology (cross-hatching in Fig. 4c). Indeed, the ERA5 monthly average temperature anomaly at 50 hPa in September was +15.9 K, which is the largest in the 1979–2019 climatology (Table 2). Above this warm anomaly was a broad region in the upper stratosphere and mesosphere of below-average temperatures, consistently at the lowest extremes of the climatology, which mostly persisted through to mid-summer.

Fig. 4.  Daily time-height section of zonal average air temperature anomalies for January 2017–July 2020 over latitudes 65–85°S from Aura MLS quality-controlled version 4.2 data (NASA 2020a). The anomalies are evaluated relative to the base period of 8 August 2004 (the start of measurements) to 31 December 2018. Single diagonal hatching marks anomalies that are outside the inter-decile range (i.e. <10th percentile or >90th percentile) of measurements prior to 2019. Crossed diagonal hatching marks anomalies that exceed the daily maximum or minimum value for all measurements over the climatology. Measurements are unavailable within the white vertical bars.

Fig. 5 shows the eddy heat flux as a function of time and pressure for two latitude bands in 2018 and 2019. The latitude ranges were chosen to separately show the distinction between disturbances at mid-latitudes and how they relate to the high latitudes. By examining the full vertical range of the reanalysis, some indication of the ability of the waves to propagate vertically can be gleaned. Comparing the two years in the 35–55°S band (Figs 5a and 5b), there was more prominent poleward heat transport in 2019 (i.e. greater negative heat flux indicated by the stronger red contours), particularly for July–September. This was also the case in the 65–85°S band (Figs 5c and 5d). Indeed, climatological extreme negative flux values were set during August–September 2019 (cross-hatching in Fig. 5d). The higher latitude band showed more activity in the 2018 autumn (particularly April–May) than in 2019 and then became quieter once the vortex became established. The lower latitude band exhibited generally equatorward heat transport at higher altitudes during late September and most of October 2019 as a response to the disturbances that occurred. The strongest poleward heat flux pulses in 2018 (e.g. July–October) generally had less penetration to the higher latitudes and lower height levels than those present in 2019. This is an indication of a dynamically stronger vortex in 2018 than in 2019. The source of the contrast in activity between the two winters is further discussed in Section 3.5.

Fig. 5.  Daily eddy heat flux anomaly as a function of pressure evaluated from the ERA5 reanalysis averaged between latitudes of (a) 35–55°S for 2018, (b) 35–55°S for 2019, (c) 65–85°S for 2018 and (d) 65–85°S for 2019. Negative values (red colours) indicate poleward transport of heat. The base period of the daily climatology used to create the anomalies is 1979–2017. Single diagonal hatching marks anomalies outside the inter-decile range over the climatological period. Crossed diagonal hatching marks anomalies less than the minimum or greater than the maximum over the climatological period.

A measure of high latitude daily Rossby wave activity near the edge of polar vortex in the lower stratosphere (at 50 hPa and 60°S) is shown in Fig. 6 (Klekociuk et al. 2019). The amplitudes of geopotential height disturbances in winter–spring of 2019 were generally higher than in 2018. Of particular note were pulses of large-amplitude disturbance in early May, mid-June and for much of August–early October in 2019. These disturbances resulted in the strong warming of the lower stratosphere discussed above (Table 2). In September, the warming significantly reduced the zonal wind strength in the lower stratosphere (shown by the descending negative zonal anomaly for 2019 (Fig. B2b), which was in strong contrast to the situation at that time of year in 2018 (Fig. B2a)). This disturbance to the zonal circulation was the largest seen since the major SSW of 2002 (Newman et al. 2020). However, the warming was classified as minor as the zonal mean zonal wind at 10 hPa did not become negative. Fig. B3 shows south polar stereographic maps for temperature, geopotential height and ozone concentration at 10 hPa from Aura/MLS measurements spanning 14–22 September. The key date of the warming at 10 hPa was September 18 (Rao et al. 2020), when temperatures reached a peak across the polar cap. Around this time, the main part of the vortex (indicated by the region of low geopotential heights) was displaced towards the Antarctic Peninsula and the tip of South America and became weaker. The ozone distribution at this level showed a generally bipolar structure, suggesting separation of the upper vortex into two or more lobes. Following the SSW, the vortex became more centred on the pole and, although weaker, tended to stabilise through to late October (as indicated in the ozone hole boundary shown in the daily total column maps of Krummel et al. 2020).

Fig. 6.  Daily range of geopotential height (GPH) for latitude 60°S at 50 hPa (~18 km altitude) from ERA5 reanalysis. The time series for 2018 and 2019 are shown in blue and red, respectively. For the period 1979–2019, the climatological range is shown by the light grey shaded area; the inter-decile range spans the darker grey shaded area, and the mean is shown by the white time line.

In terms of heat transport, the activity in early June 2019 was notably strong at high latitudes in the lower stratosphere (outside the inter-decile range; Fig. 5d). This activity occurred at the time of a peak in the amplitude of quasi-stationary planetary wave number 1 near the vortex edge (Fig. 4 of Milinevsky et al. 2019). Aside from elevated activity around the beginning of June 2018 (coinciding with the warming at high latitudes – Fig. 5c), and subsequently around the beginning of October and early November, there was generally below-average activity during most of the winter and early spring of 2018 (see also Fig. 5 of Krummel et al. 2020).

The strong planetary wave activity and poleward heat transport in the winter and spring of 2019 resulted in marked erosion of the vortex (Fig. 7). In the upper level (Fig. 7a; 850 K potential temperature (θ) or ~30 km height), the vortex area began to trend below average from late July and declined most significantly from the end of August, when it fell below the 5th percentile of the climatology since 1979, until recovering towards a typical size for this height in mid-October. At the lower level (Fig. 7b; θ = 450 K or ~18 km height) the decline began in early September and was consistently below the 5th percentile for the remainder of period until the vortex fully broke-down at this height at the end of November. In contrast to 2019, the 2018 vortex showed generally above average size for August–October in the upper level (Fig. 7a) and from mid-June–November in the lower level (Fig. 7b). The overall smaller size of the 2019 vortex compared with 2018 is apparent for θ of 850 and 450 K in Fig. 8a and b, respectively. In particular, the average vortex area for May–December 2019 on the 450 K isentrope was notably small, approaching the size evaluated for 2002. Note also that the average vortex area for 2019 at 850 K was larger than in 2002; this appears to be due to the relatively colder temperatures at this level in 2019 (Table 2).

Fig. 7.  Proxy area of the southern hemisphere stratospheric polar vortex evaluated on the potential temperature (θ) surfaces of (a) 850 K (~30 km height) and (b) 450 K (~18 km height) using ERA5 6-hourly potential vorticity data at 2.5° horizontal gridding. The values have been evaluated as the daily average surface area enclosed by the contour of –600 potential vorticity units (PVU; 1 PVU = 10⁻⁶ m² s⁻¹ K kg⁻¹) in (a) and –30 PVU in (b), as provided by Braathen and Proffitt (2000). Shown is the daily average time series for 2018 (thick black line) and 2019 (thick magenta line) together with the mean for 1979–2019 (solid blue line) and the 5th and 95th percentiles (lower and upper solid red lines, respectively) for 1979–2019.

Fig. 8.  Average proxy area of the southern hemisphere stratospheric polar vortex for years 1979–2019 obtained from ERA5 6-hourly potential vorticity data. (a) Area on the 850 K isentropic surface (~30 km height) averaged over April–November. (b) Area on the 450 K isentropic surface (~18 km height) averaged for May–December. The values has been evaluated from the daily average surface area enclosed by the contour of –600 potential vorticity units (PVU; 1 PVU = 10⁻⁶ m² s⁻¹ K kg⁻¹) in (a) and –30 PVU in (b), as provided by Braathen and Proffitt (2000).

Monthly averages of vortex-wide mean N₂O and ozone mixing ratios for θ = 450 K are shown in Fig. 9a and b, respectively. The mixing ratio of N₂O in the Antarctic vortex is influenced by the phase of the QBO approximately 1 year earlier (Strahan et al. 2015; Klekociuk et al. 2019). Through a compact anti-correlation with inorganic chlorine (Cl_y), the N₂O concentration provides a proxy for the potential influence of tropical processes on ozone depletion in the polar vortex (Strahan et al. 2014; Strahan and Douglass 2018). Despite the unusually strong disturbance to the polar vortex in September 2019, both the N₂O and ozone mixing ratios for this month were not particular anomalous with respect to other years shown.

Fig. 9.  Vortex-averaged monthly mean mixing ratios on the 450 K isentrope (~18 km height) for (a) nitrous oxide (N₂O) in parts per billion by volume (ppbv) and (b) ozone (O₃) in parts per million by volume (ppmv). Values are shown for July (orange line, filled circles), August (blue line, filled circles) and September (grey line, filled circles) for 2004–2019; vertical bars: ±1 standard deviation in the monthly daily means. Aura/MLS measurements were interpolated to the isentropic surface and spatially averaged within the region enclosed by the ‘inner vortex edge’ using the MLS Derived Meteorological Product dataset based on MERRA-2 reanalysis.

Time series of the UVR Index around the coast of East Antarctica from measurements at Casey, Davis and Mawson in 2018–2019 are shown in Fig. 10. UVR levels were relatively high during spring in 2018 compared with the climatological average, particularly over late October and most of November when new extremes were sometimes set at each site. This is consistent with the combination of the relatively strong ozone depletion and large vortex over this period (as inferred from , and also evident from OMI overpass data, not shown, for each of the sites available from ESA 2020). In contrast, UVR levels were below the climatological mean for most of the period up to early December 2019 as a consequence of increased total column ozone levels (ESA 2020). New minimum observed values were set in some periods at each station. ERA5 indicates that cloud cover at the sites was generally within climatological norms for the time series in Fig. 10 (based on examination of the monthly mean cloud fraction for the nearest grid point to the sites with respect to the 2007–2019 climatology; not shown). The exceptions to this were values greater than the 90th percentile at Casey for September–November 2018, at Davis in November 2019 and at Mawson in November 2018; local UVR levels were potentially more affected by cloud during these periods, although a clear indication of this is not apparent (Fig. 9). Ozonesonde measurements at Davis (Fig. 11) showed markedly stronger depletion in the lower stratosphere for September–November 2018 than during the same months in 2019 (Fig. 11a and b, respectively). Ozone concentrations were lower than the 10th percentile of the 2008–2017 MERRA-2 climatology for Davis over most of these months in 2018 for heights 13–20 km (Fig. 11c). In contrast, concentrations were greater than the 90th percentile at times for May–November 2019, particularly above 20 km height (Fig. 11d). Generally, the appearance of increased ozone concentrations in the lower stratosphere from late November 2018 took place as the vortex edge shrank poleward of Davis and UVR levels decreased to be closer to the long-term mean (Fig. 10b). In 2019, the edge of the ozone hole moved poleward from early September. This produced the strong ozone enhancement apparent in Figure 11d (see figures in Appendix A of Krummel et al. (2019a)) and the lack of a clear influence of the ozone hole, unlike 2018 (Fig. 11c).

Fig. 10.  Daily maximum UVR Index measured at (a) and (b) Casey in 2018 and 2019, respectively, (c) and (d) Davis in 2018 and 2019, respectively, and (e) and (f) Mawson in 2018 and 2019, respectively. Also shown are the daily range (grey shaded region) and mean (black line) for the period 2007–2018.

Fig. 11.  Time-height Davis ozonesonde measurements. Ozone partial pressure for (a) 2018 and (b) 2019. Relative difference of ozone mixing ratio with respect to the MERRA-2 climatology for Davis over 2008–2017 for (c) 2018 and (d) 2019. The individual profiles have been gridded at a resolution of 7 days and 500 m. Vertical lines at the top of each panel show the times of the actual measurements. The solid black dots/black line marks the height of the thermal tropopause. In the lower panel, diagonal hatching shows where values are less than the 10th percentile or greater than the 90th percentile of the climatology; the grey and white contours are at anomalies of 0% and –25%, respectively.

For June–August 2019, Davis was generally inside the vortex edge in the lower stratosphere (not shown); this was also the case during these months in 2018. In these months, generally lower mixing ratios were measured in 2018, with higher values in 2019. This behaviour is consistent with vortex-wide mixing ratio time series shown in Fig. 9b. Over the first half of October 2018, ozone destruction was close to complete at Davis in the height range 14–17 km (Fig. 12a). By contrast in the same time period in 2019, there is no indication of ozone loss within this altitude range (Fig. 12b) and estimates of total column ozone derived from the profiles were approximately twice the 2018 values.

Fig. 12.  Davis ozonesonde profiles for three dates in early October (a) 2018 and (b) 2019.

The much earlier weakening and reduction in size of the polar vortex in 2019 (Figs 5−7) resulted in enhanced transport of ozone-rich air from lower latitudes above 20 km (see also Fig. 11). Comparison of the evolution of specific partial ozone columns above Davis is shown in Fig. 13. Over the 15–18 km height range (Fig. 13a), 2018 was comparable with 2006, when the strongest depletion was observed at Davis, with values near zero at the beginning of October in both years. Similar seasonal variations in 2018 and 2006 occurred for the 18–21 km partial column (Fig. 13b), at least up to late October (~ day 300). However, the 18–21 km values in 2018 did not reach the extreme lows (less than 5 DU) measured in 2006 or 2009 (not shown). Similarly, the 12–15 km partial ozone column at Davis in 2018 reached low values that were comparable to the most extreme years in the record (Fig. 13c). An interesting contrast is provided with 2015 in which ozone loss in the lower stratosphere was sustained into December due to enhanced levels of aerosols resulting from the Calbuco volcanic eruption and the relative stability of the polar vortex (Stone et al. 2017; Tully et al. 2019). In 2019, the rate of ozone loss up to the end of August in the 18–21 km height range was consistent with 2018, but then ceased following the stratospheric warming discussed earlier (Fig. 13d).

Fig. 13.  Partial column ozone from Davis ozonesonde flights compared to the 2003–2017 record: (a) 15–18 km partial column 2018 (red) and 2006 (orange); (b) 18–21 km partial column 2018 (red) and 2006 (orange); (c) 12–15 km partial column 2018 (red) and 2015 (grey); (d) 18–21 km partial column 2018 (red) and 2019 (blue).

At sub-Antarctic Macquarie Island, which generally does not lie within the polar vortex, total ozone values were well above the long-term (1987–2017) range throughout August–September in 2019 (Fig. 14), reflecting enhanced meridional transport. Daily mean values in excess of 450 DU were recorded on 10–12 September, with the highest individual Dobson measurement being 481 DU on 11 September 2019. In 2018, total ozone values at Macquarie Island were generally within the usual range, with no values below 250 DU that were associated with the large 2015 ozone hole (Tully et al. 2019). Relatively low daily mean values of 295 DU (30 July) and 301 DU (31 August) appeared to be influenced by the proximity of the ozone hole.

Fig. 14.  Daily average total column Dobson spectrophotometer ozone measurements. Macquarie Island measurements (red line) during July–December are shown in (a) for 2018 and (b) for 2019 along with the 1987–2017 range (black line: 30-day moving average; dark blue and light blue bands show the 10–90th and 30–70th percentile, respectively). (c) Arrival Heights measurements during 2018 (blue dots) and 2019 (red dots). The thick black line is the 10-day mean over 1988–2019; the thin lines are the standard deviation in each 10-day interval. Individual measurements over 1988–2017 are shown as black dots.

Further poleward, at Arrival Heights, Dobson, spectrophotometer total column ozone measurements (Fig. 14c) during the spring (days 245–305) were markedly less in 2018 (blue dots) compared with 2019 (red dots). For the FTS measurements at the site, the total column ozone retrieval at the same time of year (Fig. 15a) for 2018 (orange dots) was in lower range observed since measurements began in 1996, while values for 2019 (red dots) were at the upper observed range. The contrast between the years is apparent from October (day 275) with values near the lowest observed on particular days of the year in 2018 and at the maximum extreme at times in 2019. This panel, and the panels for other species in Fig. 15, also highlights 2006 and 2017, when the ozone hole was large (Klekociuk et al. 2019), and 2016, which was relatively small due to warm stratospheric conditions (Tully et al. 2019). Generally, the other species shown in Fig. 15 exhibited a similar contrast in the two years. The potential for ozone loss during spring depends on the availability of the chlorine reservoir compounds HCl and ClONO₂, which are present due to the breakdown of halogenated source gases. Through heterogeneous reactions involving PSC particles and the background aerosols during the winter, the reservoir compounds are activated forming ClO, chlorine gas (Cl₂), chlorine dioxide (OClO), and hypochlorous acid (HOCl). The partitioning of chlorine compounds into active forms, together with other forms involving activated bromine compounds, deplete ozone in catalytic cycles in the presence of sunlight (Solomon et al. 1986). Fig. 15b shows the daily difference between daytime and nighttime ClO column from measurements at Scott Base (near Arrival Heights), which indicates the amount of chlorine activation (Nedoluha et al. 2016). At night, ClO in the lower stratosphere rapidly converts to dichlorine dioxide (Cl₂O₂), and the measured values are small. The generally reduced ClO column in the 2019 spring compared with the other years points to lower availability of active chlorine to destroy ozone, which is consistent with the vortex being relatively warm in 2019 (see also fig. 6.14(c) of Kramarova et al. 2020). Fig. 15c and d indicate that the column abundances of both HCl and ClONO₂ were generally lower in the spring of 2018 compared with 2019. This suggests that greater activation of these compounds by heterogeneous reactions occurred in 2018, which is consistent with the prevailing lower temperatures (Fig. B1a) which would favour greater available PSC surface area. An interval of reduced HCl in late October–early November 2019 (approximately days 300–320) was caused by the measurement site being situated equatorward of the ozone hole edge (see ozone maps in Appendix A of Krummel et al. 2020). The amount of gas-phase HNO₃ is an indicator of PSC volume, with lower concentrations due to uptake on PSC particles. As shown in Fig. 15e, HNO₃ abundance was lower in 2018 compared with 2019, which is consistent with the expected difference in PSC abundance in the two years discussed above. A further indicator of PSC processes is the abundance of gas-phase NO₂ (Fig. 15f), which is reduced due to uptake by PSC particles and their sedimentation. Again, the concentration was less in 2018 compared with 2019, which is consistent with the expected difference in PSC abundance between these years.

Fig. 15.  Daily total ozone column abundance and related chemical species from measurements at Arrival Heights and Scott Base, Antarctica. (a) Ozone (O₃). (b) Chlorine monoxide (ClO), shown as day minus night, from the Scott Base Microwave Radiometer (MWR; not operated in 2018). (c) Hydrochloric acid (HCl). (d) Chlorine nitrate (ClONO₂). (e) Nitric acid (HNO₃). (f) Nitrogen dioxide (NO₂). All measurements except those shown in panel (b) were obtained using the mid-infrared (MIR) Fourier Transform Infrared Spectrometer (FTIR) at Arrival Heights. Measurements for individual years are shown in different colours, legend in panel (a).

As noted in Section 1, during 2018 and 2019 there were distinctly different patterns of the large-scale climate modes. The patterns of tropical warming in 2019 are implicated in allowing the penetration or ducting of Rossby waves into the polar cap (Milinevsky et al. 2019; Newman et al. 2020; Rao et al. 2020), and the prevailing patterns appeared much less effective in 2018. Here we examine Rossby wave propagation conditions in the two years and compare them with conditions in 2002.

In Fig. 16 we show the probability of blocking of stationary Rossby waves as a function of pressure and latitude for September 2018, 2019 and 2002. Here we have used the method of Karami et al. (2016) to evaluate the blocking probability for waves with zonal (k) and meridional (l) wavenumbers of (k,l) = (1,2) and (2,1). These waves are ubiquitous in the southern hemisphere during winter (Hartmann 1977) and are set up by contrasts in land and ocean temperatures that persist over large areas. In the troposphere and lower stratosphere, wave 1 (2) transports momentum poleward (equatorward). In 2018 (Fig. 16a, b) there was evidently greater blocking in the stratosphere (above 100 hPa) poleward of ~60°S compared with the other years. This view is supported by Fig. 5, which shows less strong poleward heat transport into the polar cap during spring compared with 2019. Also apparent is a channel of low blocking probability emerging from the upper troposphere near ~25°S and penetrating the upper stratosphere at mid- and low latitudes. We suggest that this channel favoured the vertical propagation of tropical waves, making them less able to interact with higher latitudes. In 2019 (Fig. 16c, d) and 2002 (Fig. 16e, f), the upper levels of the polar stratosphere appeared relatively transparent to wave propagation. There was also relatively weaker blocking below ~500 hPa over latitudes 30–50°S compared with 2018. In the case of zonal wavenumber 2, this wave appeared to be more confined to the lower stratosphere in those years, where it likely had less ability to propagate away from the vortex edge. These features tend to point to the greater ability of extratopical waves originating near the surface to influence the high latitudes during the spring in 2019 and 2002. Of further note is the warming of the polar cap above 10 hPa in April–May 2018 (Fig. 5c). While this is in contrast to the situation in 2019, the warming evidently did not have an appreciable influence in preconditioning the vortex for further disturbance during the subsequent winter.

Fig. 16.  September average probability (as a fraction of 1) for the blocking of stationary Rossby waves (planetary waves; PW) as a function of pressure level and latitude from ERA5 reanalysis. For 2018: (a) for zonal (k) and meridional (l) wavenumbers of (k,l) = (1,2); (b) for (k,l) = 2,1). For 2019: (c) (k,l) = (1,2); (d) for (k,l) = 2,1). For 2002: (e) (k,l) = (1,2); (f) for (k,l) = 2,1). Hatching indicates where values are outside the interquartile range over the 1979–2017 climatology.

As discussed by Newman and Nash (2005), a feature of the 2002 southern hemisphere SSW was the apparent preconditioning (warming) of the stratosphere during the winter, and the occurrence of stationary and travelling wavenumber 2 Rossby waves in the spring (Klekociuk et al. 2019), which appeared to reinforce the prevailing quasi-stationary wave 1 around the time of the vortex split in late September. The influence of travelling wave 2 around the time of a large warming event in 2017 has also been noted (Klekociuk et al. 2019), although the wave amplitudes in this case were much less than in 2002. In 2019, winter temperatures in the polar upper stratosphere were colder than in 2002 (Table 2), and travelling wavenumber 2 appears to have been relatively weak (Milinevsky et al. 2019). Comparison of Fig. 16d and e suggests that propagation of at least stationary wavenumber 2 waves to high latitudes was more likely in 2002 compared with 2019, and this may have been a factor that prevented an SSW from taking place in 2019.

As is apparent from Table 1 and Fig. 3, other years over the past decade are ranked near to 2019, being 2017, 2012 and 2010. At the other extreme, 2015 and 2011 were years of large ozone holes. As highlighted by Grytsai et al. (2017), the large (small) ozone hole years tend to correspond to an eastward (westward) shift in the zonal asymmetry of the total column ozone field. The inter-annual variability in these patterns bears a relationship with tropospheric waves 1 and 3 as well as the SAM. Of note is that years of weak ozone depletion tended to show tropical warming in the central Pacific. This was the case in 2019 and 2002, with El Niño conditions prevailing, although the strong positive IOD in 2019 was likely an additional important feature (Milinevsky et al. 2019). A question still to be resolved is how emerging changes in tropical forcing are influencing stratospheric variability. A feature of 2019 was that the warming tended to take place relatively suddenly in early spring, without the preconditioning warming that occurred in 2002. In investigating the source of these differences, the timing of tropical–extratropical connections is likely to be an important consideration.

We have used ground-based, in situ, satellite and reanalysis data to review characteristics of the 2018–2019 Antarctic ozone holes. Strongly contrasting behaviour was observed in Antarctic ozone depletion between these two years. During 2018, the Antarctic stratospheric vortex was relatively stable and cold, and this produced an ozone hole that was relatively large and persistent. The overall size of the 2018 ozone hole was in the upper-tercile of the 41 ozone holes that we characterised and ranked using a set of metrics based on total column ozone data. Ozone concentrations in the 15–18 km height range measured by ozonesondes at Davis were comparable with the extreme low values observed during 2006 and 2009, which were years of the greatest ozone depletion in the 16-year Davis record.

In contrast, the ozone hole of 2019 was smaller by several metrics and similar to ozone holes in the 1980s, being in the lower-tercile of our ranked metrics. Strong disturbances to the Antarctic stratosphere during August–October were a key factor that disrupted the ozone loss process. In particular, planetary wave activity in September produced strong deceleration of the zonal flow at 10 hPa that significantly eroded the vortex and resulted in a minor SSW. The strengths of the 2019 disturbances were comparable to the activity that produced a major SSW in 2002. However, the temperature of the vortex in the 2019 winter was generally colder than in 2002, and this may have been a determining factor preventing a major warming event. The strength of the wave activity produced extremely high total ozone values at sub-Antarctic Macquarie Island in August–September.

The authors declare no conflicts of interest.

This research did not receive any specific funding.