Arctic Ozone Depletion in 2019/20: Roles of Chemistry, Dynamics and the Montreal Protocol

We use a three‐dimensional chemical transport model and satellite observations to investigate Arctic ozone depletion in winter/spring 2019/20 and compare with earlier years. Persistently, low temperatures caused extensive chlorine activation through to March. March‐mean polar‐cap‐mean modeled chemical column ozone loss reached 78 DU (local maximum loss of ∼108 DU in the vortex), similar to that in 2011. However, weak dynamical replenishment of only 59 DU from December to March was key to producing very low (<220 DU) column ozone values. The only other winter to exhibit such weak transport in the past 20 years was 2010/11, so this process is fundamental to causing such low ozone values. A model simulation with peak observed stratospheric total chlorine and bromine loading (from the mid‐1990s) shows that gradual recovery of the ozone layer over the past 2 decades ameliorated the polar cap ozone depletion in March 2020 by ∼20 DU.

recovery (or healing) in the upper stratosphere (e.g., Newchurch et al., 2003) and in the Antarctic springtime lower stratosphere (e.g., Solomon et al., 2016). Some recovery is also expected in Arctic ozone but the large observed interannual variability has so far precluded its detection (Chipperfield et al., 2017).
Arctic winter 2019/20 experienced a sustained period of low temperatures in the lower stratosphere and a stable vortex that persisted into late March . These conditions were conducive to an unprecedented extent of PSC area (DeLand et al., 2020), large levels of ozone depletion of up to 2.8 parts per million by volume (ppmv) ) and subsequently small total column values Wohltmann et al., 2020). This large depletion rivaled or even exceeded that observed in 2010/11, the previous Arctic winter with record ozone depletion (Manney et al., 2011).
In this paper, we use a detailed atmospheric three-dimensional (3-D) chemical transport model (CTM), evaluated using satellite data, to investigate Arctic ozone depletion in winter/spring 2019/20. A multidecadal model run is used to compare this winter with others over the past few decades, in particular years with large ozone depletion. We use the model to distinguish between the roles of chemistry and transport in causing the low ozone values. We also use the model to quantify the extent of the ozone recovery signal in the Arctic.

TOMCAT 3-D CTM
We have performed a series of experiments with the TOMCAT/SLIMCAT (hereafter TOMCAT) 3-D CTM (Chipperfield, 2006). The model contains a detailed description of stratospheric chemistry, including heterogeneous reactions on sulfate aerosols and PSCs. The model was forced using European Center for Medium-Range Weather Forecasts (ECMWF) ERA5 winds and temperatures (Hersbach et al., 2020) and run with a resolution of 2.8° × 2.8° with 32 levels from the surface to ∼60 km following Dhomse et al. (2019). The surface mixing ratios of long-lived source gases (e.g., CFCs, HCFCs, CH 4 , N 2 O) were taken from WMO (2018) scenario A1. The solar cycle was included using time-varying solar flux data (1995-2019) from the Naval Research Laboratory (NRL) solar variability model, referred to as NRLSSI2 (update of Coddington et al., 2016Coddington et al., , 2019. Stratospheric sulfate aerosol surface density (SAD) data for 1995-2016 were obtained from ftp:// iacftp.ethz.ch/pub_read/luo/CMIP6/ (Arfeuille et al., 2013;Dhomse et al., 2015). As year-to-year solar flux variations (and their effects on ozone) are small (e.g., Dhomse et al., 2016), solar fluxes from December 2019 are used to extend the simulation until April 2020. Similarly, SAD values are not yet available for the whole period; thus for 2017-2020 the monthly mean SAD values were repeated from 2016. The model has a passive ozone tracer for diagnosing polar chemical ozone loss which is initialized from the chemical ozone tracer every December 1 and June 1 (e.g., Feng et al., 2007).
We performed a total of three multidecadal model simulations. The control run (CNTL) was spun up from 1977 and integrated until April 2020 including all of the processes described above. Sensitivity run ODS95 was initialized from CNTL in 1995 and integrated until 2020 using constant surface mixing ratios of halogenated ODSs at 1995 levels. Sensitivity run World Avoided (WA) was initialized from CNTL in 1987 and integrated to 2020 using an ODS scenario which assumes no controls from the Montreal Protocol but rather a continuing 3% yr −1 growth in emissions. This follows on from Chipperfield et al. (2015) who studied the Arctic winter 2010/11 with a similar simulation; results are discussed in the Supplementary Material.

Satellite Data Sets
To compare to our CNTL model simulation, we use observations from the Ozone Monitoring Instrument (OMI; McPeters et al., 2008) level 3 (OMTO3d) total column data. The OMTO3d is a daily gridded data set, generated by gridding and merging only high-quality level 2 measurements (based on a Total Ozone Mapping Spectrometer (TOMS)-like algorithm) for a given day. Data is available from October 1, 2004 until present at 0.25° × 0.25° resolution and is obtained via https://search.earthdata.nasa.gov/search?q=OMDOAO3e_003.

Polar Processing
Figures 1a-1d shows the anomaly in the monthly mean Arctic mean (63-90°N) mixing ratios of N 2 O, HNO 3 , HCl, and ClO at 480 K from MLS and model run CNTL (2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019)(2020). The equivalent direct comparisons of Arctic mean mixing ratios are given in supporting information Figure S1. Due to the degradation of the MLS N 2 O observations we do not show its observed anomaly. The Arctic winter 2019/20 stands out as extreme in the record of many of these species . Modeled N 2 O, which compares well with MLS observations early in the record ( Figure S1a), indicates strong descent in spring 2020. The HNO 3 observations tend to show large negative anomalies in cold Arctic winters such as 2011, 2016, and 2020 and positive anomalies in warm winters such as 2015. Prior to 2020, the model captures this behavior well, including the extreme 2015 and 2016 cases. However, in 2020 the model overestimates the negative anomaly (i.e., the model overestimates denitrification) compared to MLS measurements, which indicate that the winter is not as extreme. Together HCl and ClO indicate the extent of PSC processing and chlorine activation which, for example, produces negative HCl anomalies and positive ClO anomalies in cold years (e.g., 2005, 2008, 2011, and 2016). For these species, 2020 stands out as significant in terms of chlorine activation; the activation began earlier and lasted longer in 2019/20 than in the previous record winter 2010/11 (see also Manney et al., 2020). The model captures these variations in chlorine species well. Figure 1e shows the evolution of the monthly mean Arctic mean ozone anomaly at 480 K from MLS observations and model run CNTL. The largest observed anomalies occur in the springtime and vary between years with strong negative values (e.g., 2011 and 2016) and strong positive values (e.g., 2019). These variations are captured well by the model. Within this time series, 2020 stands out in both the observations and model as having the largest negative anomaly of ∼35%-40%.

Ozone
Arctic winter/spring ozone levels are maintained by a balance of dynamics and chemical depletion, with both processes making large and variable contributions to the column amount in any year. Figure 2a shows the mean March Arctic column ozone from OMI observations versus model run CNTL. The OMI observations clearly show 2020 (315 DU) and 2011 (329 DU) as the 2 years with extremely low column ozone with, by this metric, slightly lower values in 2020. The chemical ozone tracer from model run CNTL captures the overall variation and the two extreme years, very well. Results from the model run can be used to separate the contributions of dynamics and transport. The modeled passive ozone shows values between 306 DU (2015) and 355 DU (2018) in December, with little interannual variability (mean 328 DU, standard deviation (σ) 14 DU). Descent over winter typically increases passive ozone in March to 481 ± 42 DU (1σ), a mean increase of 153 DU with much larger interannual variability. However, both 2011 and 2020 stand out as significant anomalies with March mean passive ozone columns of 396 DU (increase of 64 DU since December) and 376 DU (increase of 59 DU), respectively. This shows that a relatively small increase over the winter due to weak transport contributed significantly to the overall low ozone columns in these years (see also Isaksen et al., 2012;Wohltmann et al., 2020). The model further suggests that the contribution of transport would have led to slightly lower column ozone in early spring 2020 than in 2011.
The difference between modeled active and passive tracers quantifies the seasonal chemical ozone loss (lower panel of Figure 2a). The estimated seasonal loss varies between ∼40 DU (in warm winter 2018/19) and ∼80 DU (in 2015/16). Note that this metric, over this wide geographical area which combines inside and outside vortex regions, smooths out the larger variations in chemical ozone loss which occur in the vortex core. Nevertheless, 2019/20 does stand out as a year with large chemical ozone loss (∼78 DU), which is comparable to that in the other cold winters of 2004/05, 2010/11, and 2015/16. However, the model results show that anomalously weak transport played a decisive role in causing the overall low column ozone in winter 2019/20.

Impact of Ozone Recovery
Although the chemical ozone depletion in Arctic winter 2019/20 has been shown to be large (  N2O anomaly (%) recent decreases in stratospheric halogen levels due to the Montreal Protocol. The differences in column ozone between runs CNTL and ODS95, which uses constant tropospheric ODS values from 1995, quantify the increase in ozone due to decreasing (from their peak) stratospheric halogens, often taken as a measure of recovery (Figure 2b). The increasing impact of decreasing halogens with time, especially in the polar regions, can clearly be seen. Depletion in the Antarctic ozone hole in 2019 is ∼30 DU less severe than it would have been under conditions of peak stratospheric halogen loading. For the Arctic the impact varies, but the increasing influence of halogen recovery and the favorable conditions for ozone loss produce the largest effect in 2020. This increasing recovery signal for March is also seen in Figure 2a; reductions in stratospheric halogens have resulted in mean column ozone depletion being ∼20 DU less severe than it would have been at peak loading.
The mean behavior of ozone in the polar region masks the variations within the vortex and local extreme values. Figure 3a shows OMI column ozone on March 18, 2020. This is during the phase of active PSCs (DeLand et al., 2020) and ongoing ozone loss, but it corresponds to the day of the lowest ozone column in the OMI record of 208 DU. This is well below the threshold of 220 DU which is commonly used to denote the boundary of the Antarctic ozone hole. Simulation CNTL (Figure 3b) gives a good representation of the spatial distribution of column ozone but produces larger regions with values below 220 DU. Figure 3c shows, however, that transport alone (between December and March) would have led to relatively low column values inside the vortex. These low columns are exacerbated by chemical depletion of up to 108 DU in the vortex (Figure 3d) to produce the modeled column in Figure 3b. Figure 3e and 3f show results from run ODS95. While the mean ozone recovery signal is ∼20 DU for the wider Arctic area (Figure 2), the differences peak at ∼35 DU in the core of the vortex. Supporting information Figure S3 shows the equivalent plots for March 30, 2020, 12 days later at the end of the ozone depletion phase. Chipperfield et al. (2015) used the TOMCAT 3-D CTM to quantify the benefits already achieved by the Montreal Protocol at the time of the large observed Arctic ozone depletion in 2010/11. They assumed a continuing scenario of 3% annual growth in ODS emissions after 1987. It is unlikely that we would have reached 2020 without some controls on the use of ODSs given the environmental damage that would have become apparent. However, we can use the model to investigate the impact on ozone by extending a similar "WA" experiment until winter 2019/20. Supporting information Figure S4 shows that with the assumed continued growth in stratospheric chlorine and bromine, Arctic ozone loss would by now have already become extremely severe, with March vortex columns of less than 85 DU.

Dynamical Influence on Polar Ozone
Planetary wave driving of the wintertime polar stratosphere is typically stronger and more variable in the Northern Hemisphere (NH) compared to the Southern Hemisphere (SH), leading to a warmer Arctic polar vortex and less chemical ozone depletion. In contrast, the Antarctic polar vortex is much less disturbed by wave forcing and temperatures are almost always low enough for extensive springtime chemical ozone depletion (Solomon et al., 2014;WMO, 2018). Weber et al. (2011) summarized the interannual variability and interhemispheric differences by demonstrating a compact linear relationship between the mean winter eddy heat flux at 100 hPa and the spring-to-autumn high-latitude ozone ratio. This is shown in Figure 4a   pletion was correspondingly less. Hence these points do not fall on the correlation lines for the three subsequent decades. It is interesting how little these lines differ, despite the decrease in stratospheric halogens since 1995. The impact of ozone recovery on this correlation is shown in Figure 4c, which shows results from the most recent decade for runs CNTL and ODS95. The larger halogen loading in run ODS95 does lead to lower ozone, especially in the Antarctic, but the effect on the slope is relatively small. As stratospheric halogens decay further, and recovery continues, chemical depletion will return to 1980s levels and the compact correlation can be expected to change significantly.

Discussion and Conclusions
We have shown that by many metrics the Arctic winter/spring 2019/20 exhibited extreme behavior within the record of the past 2 decades. Our 3-D TOMCAT/SLIMCAT CTM captures well the observed persistent low temperatures and strong chlorine activation in the lower stratosphere and shows that the extremely low column ozone abundances arose through a combination of chemical loss and weak replenishment through transport. Despite large chemical depletion, the model shows that ozone recovery from the effects of halogens since their peak stratospheric loading ameliorated the loss by ∼20 DU in winter/spring 2019/20. Without the Montreal Protocol at all, the ozone loss would have been extremely large, with Arctic March column ozone values of only 85 DU under the assumption of continued 3% yr −1 growth in ODS emissions. The unusual dynamics of Arctic winter 2019/20 fit well to the previously established correlation of spring/ autumn ozone column and wintertime eddy heat flux for both polar regions.
Stratospheric chlorine and bromine loadings are decreasing and signs of polar ozone recovery have been detected (e.g., Solomon et al., 2016). Nevertheless, winter 2019/20 has shown that the Arctic is still susceptible to very large (even record) ozone depletion under suitable meteorological conditions. Due to the Montreal Protocol, the potential for halogen-catalyzed polar ozone depletion will gradually decrease. However, the potential for weak dynamical events to cause low column ozone will remain and so there is a need for continued monitoring and process understanding of this part of the atmosphere.