Indicators of Antarctic ozone depletion: 1979 to 2019

. The National Institute of Water and Atmospheric Research/Bodeker Scientiﬁc (NIWA–BS) total column ozone (TCO) database, and the associated BS–ﬁlled TCO database, have been updated to cover the period 1979 to 2019, bringing both to version 3.5.1 (V3.5.1). The BS–ﬁlled database builds on the NIWA–BS database by using a machine-learning algorithm to ﬁll spatial and temporal data gaps to provide gap-free TCO ﬁelds over Antarctic. These ﬁlled TCO ﬁelds then provide a more complete picture of winter-time changes in the ozone layer over Antarctica. The BS–ﬁlled database has been used to calculate 5 continuous, homogeneous time series of indicators of Antarctic ozone depletion from 1979 to 2019, including (i) daily values of the ozone mass deﬁcit based on TCO below a 220 DU threshold, (ii) daily measures of the area over Antarctica where TCO levels are below 150 DU , below 220 DU , more than 30 % below 1979 to 1981 climatological means, and more than 50 % below 1979 to 1981 climatological means, (iii) the date of disappearance of 150 DU TCO values, 220 DU TCO values, values 30 % or more below 1979 to 1981 climatological means, and values 50 % or more below 1979 to 1981 climatological means, 10 for each year, and (iv) daily minimum TCO values over the range 75 ◦ S to 90 ◦ S equivalent latitude. Since both the NIWA–BS and BS–ﬁlled databases provide uncertainties on every TCO value, the Antarctic ozone depletion metrics are provided, for the ﬁrst time, with fully traceable uncertainties. To gain insight into how the vertical distribution of ozone over Antarctica has changed over the past 36 years, ozone concentrations, combined and homogenized from several satellite-based ozone monitoring instruments as well as the global ozonesonde network, were also analysed. A robust attribution to changes in the 15 drivers of long-term secular variability in these metrics has not been performed in this analysis. As a result, statements about the recovery of Antarctic TCO from the effects of ozone depleting substances cannot be made. That said, there are clear indications of a change in trend in many of the metrics reported on here around the turn of the century, close to when Antarctic stratospheric concentrations of chlorine and bromine peaked.

tions, and modeling studies (Newman et al., 2009). The Montreal Protocol, enacted in 1987, with subsequent Amendments and Adjustments, committed countries to significantly reduce their production of ODSs. The Protocol has been labelled as one of the most successful global environmental treaties (Gonzalez et al., 2015;McKenzie et al., 2019). The reduction in stratospheric chlorine and bromine loading resulting from compliance with the Protocol has led to a recovery of ozone from the effects of ODSs in many regions of the atmosphere (e.g. Yang et al., 2008;Kuttippurath, 2013;Solomon et al., 2016). Nevertheless, an ozone hole continues to appear over Antarctica in each austral spring (Douglass et al., 2011). Projections of the future evolution 30 of the ozone layer over Antarctica using coupled chemistry-climate models suggest that with continued compliance with the Protocol, the ozone layer over Antarctica is expected to return to 1980 levels around 2060 (Dhomse et al., 2018;Amos et al., 2020).
Three metrics commonly used to define the Antarctic ozone hole are the area of the hole (adding the areas of cells falling below some threshold in a TCO field), the minimum TCO value within the hole, and the Antarctic ozone mass deficit (Uchino 35 et al., 1999;Huck et al., 2007). Bodeker et al. (2005) reported on these metrics, using four different criteria for ozone hole type values, viz. (i) TCO below 150 DU, (ii) TCO below 220 DU, (iii) TCO 30 % or more below the 1979 to 1981 climatological mean, and (iv) TCO 50 % or more below the 1979 to 1981 climatological mean. Time series of these metrics were were updated in subsequent publications (Müller et al., 2008;Struthers et al., 2009). Other studies to date have shown that all three metrics show a slowing of Antarctic ozone depletion, consistent with the first stage of ozone recovery from the effects of ODSs Interannual variability in Antarctic stratospheric dynamics, manifest most obviously in interannual variability in Antarctic stratospheric temperatures, drives significant interannual variability in the severity of Antarctic ozone depletion (Schoeberl et al., 1996;Newman and Nash , 2000;Newman et al., 2004Newman et al., , 2006 This study presents updated time series of metrics of the Antarctic ozone hole calculated from the long-term homogenised 2 Ozone databases This paper takes advantage of several features of the new version 3.4 (V3.4) NIWA-BS and BS-filled TCO databases (Bodeker et al., 2020a), updates them to the end of 2019 to create V3.5.1 of the databases, and uses the BS-filled database to define 60 continuous, homogeneous time series of several metrics describing key attributes of the Antarctic ozone hole. The databases are constructed using measurements from 17 different satellite-based instruments wherein offsets and drifts between (i) the groundbased Dobson and Brewer spectrophotometer networks and (ii) a subset of the satellite-based measurements, are removed and then used as the basis for homogenising the remaining TCO data sets. V3.4 and V3.5.1 of the BS-filled TCO databases comprise spatially filled TCO fields that use a machine-learning approach to infer missing data in regions and at times for 65 which measurements were not available (Bodeker et al., 2020a). This approach significantly improves on the 'over the pole' method described in Bodeker et al. (2001a) to create far more physically plausible renditions of the ozone fields in regions of missing data. The result is a continuous gap-free database of daily TCO fields at 1.25 • longitude by 1 • latitude resolution.
Unlike previous versions of the database, both V3.4 (1979-2016) and V3.5.1 (1979-2019) include uncertainties traceable to uncertainties in the TCO fields measured by the 17 different space-based instruments that are used to construct the database. To 70 propagate the uncertainties on the TCO fields to uncertainties in the Antarctic ozone depletion metrics, two additional databases were created, one in which the 1σ uncertainties were added to each TCO field, and another where the 1σ uncertainties were subtracted from each TCO field, i.e. assuming uniform over-estimation of TCO and under-estimation of ozone by ±1σ. By calculating the ozone depletion metrics across all three databases, estimates can be obtained for +1σ and -1σ uncertainties on all metric time series.

75
To demonstrate how the vertical structure of the ozone layer over Antarctica has changed from 1985 to 2019, ozone concentrations were extracted from the Bodeker Scientific vertically resolved ozone database (BSVertOzone, Hassler et al., 2018), and mapped onto an equivalent latitude (φ eq ) coordinate system so that only values well inside the ozone hole (φ eq poleward of 75 • S) could be selected. BSVertOzone combines measurements from several satellite-based instruments and ozone profile measurements from the global ozonesonde network to create sparse fields of ozone concentrations on 70 altitude levels from 1 80 to 70 km. Offsets and drifts between each satellite-based ozone data set and a selected standard (SAGE-II in the stratosphere and ozonesondes in the troposphere) was used to create a single homogeneous database of ozone concentrations. Similar to the TCO database, measurement uncertainties and uncertainties from other sources (e.g., applied offset and bias corrections) are propagated through to the final product, i.e. every ozone concentration has an associated uncertainty. The development of the BSVertOzone database is described in detail in Hassler et al. (2018). For this study, the database was extended to cover the 85 period 1979 to 2019.

Ozone mass deficit
As in Bodeker and Scourfield (1995), the Antarctic vortex period (AVP; day 200-335; 19 July-1 December) mean ozone mass deficit has been calculated for each year and is plotted together with an estimate of the EEASC in Fig. 1. The mass deficit quantifies the mass of ozone that would need to be added to the atmosphere to return TCO values over Antarctica to above 220 DU (1 DU = 2.69 × 10 16 molecules/cm 2 ). The scale selected for the EEASC curve (right Y axis) maximizes the correlation between the AVP mean ozone mass deficit and EEASC over the period 1979 to 2000, just before EEASC peaks. The fact that after 2000 more of the data points fall below the EEASC curve than above it suggests that factors other than the decline in halogen loading of the Antarctic stratosphere is driving the return of the Antarctic ozone layer to pre-1980 levels.
The anomalously low AVP mean ozone mass deficits in 1988, 2002 and 2019 all result from sudden stratospheric warmings 95 (SSWs), large in 1988 (Kanzawa and Kawaguchi, 1990), major in 2002 (Newman and Nash, 2005), and minor in 2019 (Wargan et al., 2020) that elevated Antarctic stratospheric temperatures and curtailed the heterogeneous chemical processes driving polar ozone destruction. The SSW in 1988 led to an Antarctic ozone hole that was shallow in depth and small in area (Kanzawa and Kawaguchi, 1990;Schoeberl et al., 1989;Krueger et al., 1989). In 2002, unusually large planetary wave activity caused a major SSW that weakened and warmed the polar vortex, and resulted in reduced ozone depletion over Antarctica (Allen et al., 2003; 100 Glatthor et al., 2004;Konopka et al., 2005;Manney et al., 2005;Ricaud et al., 2005). The minor SSW in September 2019 resulted in significantly higher than usual polar TCO (Wargan et al., 2020;Safieddine et al., 2020)  source of the local minima in late October/early November in the time series for TCO 50 % or more below the 1979-1981 mean is discussed in Bodeker et al. (2005).
Annual maximum ozone hole areas, and the dates on which they occur, are shown for all four threshold conditions in Fig. 3.
The annual maxima in the daily values of the Antarctic ozone hole areas for the four ozone hole area criteria peak around the turn of the century, close to when EEASC peaks (see Fig. 1). In contrast, the date when the maximum occurs shows a steady 125 drift towards earlier dates over the 41-year period (linear trends equivalent to changes of between 15 and 19 days earlier, depending on metric). The cause of this drift towards consistently earlier dates of annual maximum ozone hole area has not been diagnosed here. The error bars, which are often asymmetric, are generally small.
The dates of disappearance of TCO values flagged as being within the Antarctic ozone hole by the four different criteria are shown in Fig. 4. After drifting to later in the year over the first 1.5 decades of the data series, since the early to mid-1990s the To that end we have calculated 6-hourly profiles of the 550 K meridional impermeability (κ) against equivalent latitude (Bodeker et al., 2002)

145
While some previous studies have reported on annual minimum TCO values within the Antarctic vortex as a metric for tracking Antarctic ozone depletion (including Bodeker et al., 2005), Müller et al. (2008) showed that the utility of examining the minimum in daily TCO poleward of a threshold latitude was debatable, insofar as it relies on a single measurement. Müller et al. (2008) found that, for Arctic conditions, the minimum value often occurs in air outside the polar vortex, both in the observations and in a chemistry-climate model and that the minimum value does not show a good correlation with chemical 150 ozone loss in the vortex deduced from observations. They recommended that the minima, relying on a single measurement, should not be used as a metric of polar ozone depletion. Following that recommendation, we consider rather daily TCO zonal means calculated against equivalent latitude on the 550 K surface (Bodeker et al., 2001b). Examples of such zonal mean TCO profiles by equivalent latitude for 1 October of each year are shown in Fig. 6. The meridional profiles by equivalent latitude are characterised by very steep gradients across the dynamical polar vortex edge, typically around 62 • S equivalent latitude 155 (Bodeker et al., 2002), and very shallow gradients through the core of the dynamical vortex poleward of 75 • S equivalent     vertical extent, centred on 12 to 21 km altitude, are shown for the Southern Hemisphere spring in Fig. 9. Again, the anomalous warm winters of 1988, 2002 and 2019 resulting in more ozone are clearly visible. Across all spring months partial ozone columns have increased since the late 1990s. The percentage contribution of each 1 km thick layer to the monthly mean, polar cap mean partial ozone column between 11.5 and 21.5 km is shown in Fig. 10. It is not clear whether the significant shifts in 175 ozone between layers in September in the mid-1990s result from sampling biases in the measurements available (noting the screening of SAGE-II data below 23 km in the few years following the Mt. Pinatubo volcanic eruption in June 1991) or whether the vertical redistribution reflects a physical response to the eruption. During October and November the general sense is that from 1985 to around the turn of the century, ozone in the 11.5 and 21.5 km column is concentrated more in the upper part of the column (20 to 21 km) and less in the lower part of the column (13 to 18 km) as a result of the heterogeneous chemistry in 180 the Antarctic being most active between 16 and 18 km (Hofmann et al., 1987;Johnson et al., 1992). This trend reverses after the turn of the century with ozone showing a more equitable distribution across the 10 layers by 2019.

Conclusions
Several metrics describing the evolution of the Antarctic ozone hole over the 41-year period 1979 to 2019 are reported on above.
These analyses were only possible through the availability of a complete, homogeneous climate data record of daily TCO fields.

185
As detailed in Bodeker et al. (2020a), significant effort is required to homogenize the ozone fields from the 17 different spacebased sensors measuring ozone that comprise the BS-filled TCO database, as well as to infer missing data through the polar night and in other regions where the operational parameters of the satellites result in data gaps. The requirements of the GCOS (Global Climate Observing System; GCOS-138, 2010;Bojinski et al., 2014) for climate data records, and in particular the need for all data to have traceable uncertainties, has led to the most recent versions of the NIWA-BS and BS-filled TCO databases 190 (V3.4 and V3.5.1) including estimates of the uncertainties on every TCO value as described in Bodeker et al. (2020a). This has allowed, for the first time, uncertainties to be included on the Antarctic ozone depletion metrics, showing which metrics are sensitive to uncertainties in the source TCO fields and which are not.
While a formal attribution of changes in the metrics shown above to changes in Antarctic stratospheric halogen loading has not been made and, as a result, statements about the recovery of the Antarctic ozone layer from the effects of ODSs cannot 195 be made, all of the metrics directly related to ozone levels over Antarctica, i.e. AVP mean depleted mass, annual maximum Krueger, A.J.; Stolarski, R.S. and Schoeberl, M.R., Formation of the 1988 Antarctic ozone hole, Geophys. Res. Lett., 16, 381-384, 1989. Krzyścin, J.W., Jaroslawski, J., and Rajewska-Wiech, B.: Beginning of the ozone recovery over Europe? -Analysis of the total ozone data from the ground-based observations, 1964, Ann. Geophys., 23, 1685-1695, 2005.