First, we analyze the impact of a hypothetical No-MPA scenario on ozone and the subsequent variations in stratospheric temperature and zonal winds due to ozone changes. In the second part, we investigate how the (ozone-driven) stratospheric circulation changes, as well as the impact of CFCs, are linked to the large-scale tropospheric circulation and the surface.

In a scenario where the MPA is not in place, the abundant CFCs in the atmosphere severely reduce the global total ozone column to only 60 DU by the end of the 21st century in both noMPA experiments (Fig. a). Note that both noMPA scenarios are lying on top of each other, suggesting that radiative effects of CFCs alone do not affect the global ozone content. This severe ozone reduction is consistent with findings from, e.g., , who reported a collapse of the global ozone layer with ozone columns below 100 DU in a No-MPA scenario in the mid-21st century (see also, e.g., ). Figure b–d show the zonal mean ozone of the world without a Montreal Protocol (noMPA, b) compared to the reference (c) and the difference (noMPA - ref) (d) by the end of the century for austral winter (JJA). The uncontrolled CFC emissions have increased the chlorine concentration by a factor of 20–80 compared to the reference by the end of the 21st century, which causes an ozone depletion by up to 90 % in the stratosphere, with the strongest reduction happening in the lower stratosphere. Gas-phase ozone destruction by chlorine is additionally accelerated by its heterogeneous activation on stratospheric aerosols and polar stratospheric clouds (PSCs), which also became much more widespread due to the temperature drop in the lower stratosphere (Fig. a, c). The cooling is especially prominent in the tropical lower stratosphere, where temperatures drop below the PSC Type 1 formation threshold of 195 K between 130 and 20 hPa (Fig. c) and Cl2 (Fig. b, d) accumulates. This was also observed in and , and we also see an increase in PSCs Type 1 (nitric acid trihydrate, NAT, and supercooled ternary solution, STS) (Fig. a) in the tropics. PSCs Type 2 (ice crystals, when temperatures fall below 190 K) are parameterized to only extend from 0–50∘ in each hemisphere in SOCOL (see also ).

In the troposphere, ozone is depleted by up to 60 %, consistent with the documented impacts of ODS on tropospheric ozone via, e.g., changes in stratosphere–troposphere exchange . At around 100 hPa in low latitudes, self-healing of the ozone layer occurs (Fig. d). With depleted ozone in high altitudes, UV radiation can penetrate further down and produce ozone there, as is also observed in , , and . In the No-MPA scenario by the end of the 21st century, ozone depletion is no longer subject to any season or restricted to the polar regions but is happening globally all year round (Fig. ). We particularly want to highlight here the severe ozone reduction in the tropical lower stratosphere, which introduces new dynamical consequences compared to the past and present-day ozone-depletion effects (see next Sect. ).

Here, we investigate the effect of the No-MPA scenario in JJA on zonal mean temperatures (Fig. c), zonal winds (Fig. f) and age of air (Fig. c) and go into the decomposition of the CFC chemical (Fig. a, d) and radiative effect (Fig. b, e) to investigate the processes and quantify their contributions to the total impact of No-MPA. Consistent with other world-avoided studies (e.g., ), the global ozone depletion by the end of the 21st century leads to a severe decrease in lower-stratospheric temperatures. Figure c shows the temperature response to the combined effect of the CFC chemical effect (ozone depletion), which mainly cools the stratosphere, and the CFC radiative effect, which warms the troposphere and parts of the stratosphere. Lower-stratospheric temperatures (100–20 hPa) drop by over 20 K and by over 30 K in the upper stratosphere (3–0.7 hPa) as shown in Fig. a, c. This is coherent with the pattern of ozone anomalies induced by CFCs (Fig. d), indicating that the cooling is mostly due to reduced ozone absorption of shortwave solar radiation, as well as of longwave terrestrial radiation (Fig. a, d). The cooling is especially prominent in the tropics (Fig. a, c), where heterogeneous chlorine activation enhances ozone destruction (Fig. a, b). This severe cooling is seen throughout all seasons (Fig. ). The area of reduced cooling between 20 and 3 hPa in the tropics and NH can be explained on the one hand by the maximum ozone concentration region at around 10 hPa (see Fig. b and c) and on the other hand by the increased absorption of infrared radiation at 9.6 µm, as missing ozone allows this radiation to penetrate higher up .

The drastic temperature changes in the stratosphere alter the lapse rate (Fig. ) in the world-avoided scenario, lifting the tropopause in the tropics, which was also observed by . The upward shift in the tropical tropopause (reaching 50 hPa) is almost entirely due to ozone depletion, with CFC radiation barely affecting it, as seen in Figs. b and . This effect is similar to the tropopause rise from well-mixed GHG . Above the tropopause, the stratospheric temperatures strongly increase up to the inflection point at 3 hPa, where they start to decrease again, suggesting that the stratopause drops to lower altitudes in the No-MPA experiments, shrinking the stratosphere compared to the reference.

Interestingly, the Antarctic stratosphere exhibits a warming of over 3 K at around 10 hPa (similar signal for Arctic stratosphere in Fig. d and f). explained it by an increased downwelling due to the Brewer–Dobson circulation (BDC) speedup. However, CO changes shown in Fig. d and f indicate reduced vertical transport from the mesosphere. CO can be interpreted as a dynamical tracer from its production region (CO2 photolysis) in the mesosphere (e.g., ) and especially for the polar stratosphere (e.g., ). showed a very efficient CO descent in the mesosphere and stratosphere in the NH polar vortex during winter. However, when the vortex gets perturbed from a sudden stratospheric warming (SSW), this descent reduces. We argue that with the weaker vortex under No-MPA conditions, we have similar weak SH vortex conditions, which are reflected in the reduced polar CO in Fig. d and f due to the weaker mesospheric vertical transport. The warming at 10 hPa could then also be partly explained by the weaker vortex, allowing warmer air from the midlatitudes to be mixed into the polar stratosphere more easily. A similar warming has also been observed under ozone hole conditions by (see their Fig. 5) and (see their Fig. 1).

CFCs by themselves (i.e., without considering their effects on ozone; Fig. b) strongly warm (by up to 5 K) the troposphere, consistent with previous studies ; we will examine this feature, along with surface temperature, in more detail in Sect. . The CFC-induced warming also extends into the lower stratosphere up to 20 hPa, which is consistent with the recent findings of , indicating that this is a direct (radiative) effect without any influence of dynamical changes in this region. Upper-stratospheric warming at high latitudes in Fig. b most likely stems from the BDC speedup (see later in this section).

Next, we aim at understanding in more detail the dynamical impacts of a No-MPA scenario. In Fig. f the zonal wind response to the overall CFC effect is depicted. The wintertime polar vortex speed reduces substantially, whereas the subtropical jets (STJs) shift up and accelerate in both hemispheres. Furthermore, the BDC also speeds up, as the age of air gets younger in the entire stratosphere (Fig. c).

As seen in Fig. c, the missing shortwave absorption from depleted ozone starts to reduce tropical temperatures in the lower stratosphere by 2030 and causes them to drop to 185 K by 2090. This severe cooling in the tropics reduces the Equator-to-pole temperature gradient (Fig. a). This reduction in the gradient is largest in the SH winter and starts to weaken the polar cap zonal wind at 10 hPa by 2040 (blue and red lines in Fig. ). Consequently, the polar vortex slows down due to the severe cooling of the tropical lower stratosphere (around 50 hPa in Fig. a) from the CFC chemical effect, i.e., ozone depletion. By the end of the 21st century, the polar vortex in the SH has significantly slowed down by up to 25 m s-1 at 10 hPa and 40 m s-1 at 1 hPa (Fig. d). In the NH, we find a slowdown of the vortex by up to 15 m s-1 at 10 hPa, although this signal is limited to individual seasons such as fall, winter and spring (Fig. ). observe a similar weakening of the polar vortices. For summer, we observe the opposite effect in both hemispheres. The stratospheric winds strengthen (Fig. d for SH and Fig. d for NH summer) due to stronger polar cooling than in winter, which increases the Equator-to-pole gradient again (Fig. d for SH and Fig. a for NH summer). Seasonally decomposing the vortex response for ozone depletion shows that the vortex reacts in a similar way to recent ozone-depletion trends in the summertime (strengthening of stratospheric westerlies), while it acts in the opposite direction in wintertime (weakening of stratospheric westerlies).

For the CFC radiative effect, we see that the polar vortices are slightly stronger by up to 10 m s-1 in the SH (Fig. e) and NH during winter (Fig. e) compared to the CFC chemical effect. This enhancement originates from the CFC warming in the tropical troposphere and lower stratosphere. Therefore the Equator-to-pole temperature gradient in the lower stratosphere is larger when the CFC warming scenario is included. This is also seen in the SH polar cap wind evolution (purple line Fig. ).

Additionally, we observe a strengthening of the upward flank of the subtropical jets near the tropopause and poleward shift (around 3∘ N and S) of the STJs throughout all seasons and scenarios (Figs. d–f and ). Polar lower-stratospheric cooling during summertime further contributes to these dynamical changes, acting in the same way as Antarctic ozone hole conditions (e.g., ).

As a consequence of the weaker vortices and the stronger STJs, planetary waves can more efficiently propagate to the stratosphere. There they induce an acceleration of the BDC, leading to a decrease in age of air (AoA) in the global stratosphere (Fig. a–c), consistent with previous studies . Here, we find that this strengthening is almost entirely due to CFC-induced ozone depletion (Fig. a), similar to what occurred in the recent past . The strongest effect is on the shallow branch of the BDC, where the air gets younger by up to 0.8 years. The radiative heating by CFC further contributes to the speedup of the BDC (reduction of 0.3 years, mainly the deep branch; Fig. b), leading to a total AoA decline of 0.5 years in the deep branch and more than a year in the shallow branch (Fig. c).

In summary, the severe cooling from missing ozone has substantial implications for the stratospheric circulation, which, depending on the season, are the opposite to the historical ozone-depletion period (winter) or show the same sign (summer).

To better understand the stratospheric implications of the No-MPA scenarios on the tropospheric variability modes, we focus on the SH polar vortex and its implications for tropospheric SAM. The SAM is a large-scale climate pattern in the SH with implications for temperature and precipitation. Figure a–c show the seasonal cycle of zonal wind changes between 40 and 70∘ S. As described in Sect. , the wintertime polar vortex substantially slows down due to the CFC chemical effect (Fig. a, c). This weaker vortex in turn becomes more variable in winter and beginning of spring (Fig. d), as wave propagation into the stratosphere is facilitated, which increases the likelihood of sudden stratospheric warming (SSW). showed that SOCOL is among the models that can generate SSWs in the SH. The weaker vortex in winter and spring manifests in the troposphere by pushing the tropospheric SAM to a more negative phase (Fig. a for winter, Fig. a for spring). This finding is a novelty to the current understanding of how ozone depletion can affect the SAM: in the extreme ozone-depletion scenario, the SAM shows a pronounced wintertime negative SAM phase and no longer strengthens in spring.

In contrast, winds in the stratosphere strengthen in summer and fall, effectively extending the lifetime of the stratospheric polar vortex  and shifting the tropospheric SAM to a more positive phase (Figs. d, d). This is consistent with our current understanding of how ozone depletion affects the stratospheric circulation and, in turn, how different vortex states affect the tropospheric circulation (e.g., ).

For the CFC radiative effects, the opposite happens: the vortex strengthens in all seasons, causing a shift in the SAM to a more positive phase (SAM+) all year round (Figs. b, e, b, e). Additionally, the vortex variability decreases (Fig. e). Combining both CFC effects shows (Figs. c, f, c, f) that the SAM+ response is dominated by the CFC radiative effect. It is only slightly reduced where the SAM is in a negative phase due to the CFC chemical effect (summer and spring) but reinforced where chemical and radiative CFC effects contribute to the positive phase (winter and fall). This partial cancellation of effects (for summer and spring) is similar to what found in terms of how ozone-mediated impacts of GHGs on the summer SAM offset the direct (radiative) effects of ODS and GHGs.

Overall, the results show that the way ozone depletion affects the large-scale tropospheric circulation would have substantially changed in a future without the MPA as compared to how ozone depletion has affected the circulation in the recent past. Until today, the CFC chemical effect has been presumably dominating the SAM response as it is driven by polar ozone depletion and recovery. In the No-MPA scenario, the polar ozone depletion gets nearly saturated towards the end of the 21st century. The polar vortex is then mostly affected by the tropical ozone depletion, weakening the meridional temperature gradient and thus changing the sign of the stratospheric SAM and NAM anomalies. In addition we show that in the troposphere the CFC radiative effect becomes dominant, particularly towards the end of the century (see also Sect. ). However, the overall SAM response is still strongly modulated by the stratospheric changes, which strengthen or weaken the radiative effect, depending on the season.

Our findings are consistent with changes in wind at 500 hPa, which we use as a proxy for the eddy-driven jet (Fig. ). Most remarkably, the strongest response in the eddy-driven jet is seen in austral summer (December–February, DJF), when the jet strongly contracts poleward in the SH (Fig. c); this is due to the fact that during this season, chemical and radiative effects of CFCs act in the same direction, in much the same way that GHGs and the ozone hole affected the westerly winds in the recent past (e.g., ). The poleward contraction and strengthening of the westerly winds, and in particular the shift to a more positive SAM phase, have wide repercussions on regional weather regimes and precipitation patterns across the Southern Hemisphere . show in their Fig. 7b the 2090–2100 annual mean precipitation differences in the combined chemical and radiative CFC effects (total effect). Precipitation increases over the Southern Ocean and Antarctica, while it decreases over South America and off the coasts of South Africa and southern Australia, consistent with the behavior of a positive SAM . These regional precipitation changes are similar to what is expected in a changing climate for the SH .

The previous results show that the absence of the MPA and subsequent changes in ozone lead to drastic changes in the stratospheric circulation and also influence the tropospheric circulation in the SH. Similar to the SH, large-scale climate variability in the NH can be described by the Northern Annular Mode (NAM) (Fig.  for winter, Fig.  for the other seasons). For the European and the North Atlantic sector, it is often referred to as the North Atlantic Oscillation (NAO) . The weakening of the polar vortex in the NH due to the CFC chemical effect, i.e., ozone depletion, is reflected in the decrease in the meridional near-surface pressure gradient in Fig. a. In boreal winter, we find a pressure increase at the NH pole and a decrease in the midlatitudes, which is reminiscent of a shift in the NAM to a more negative phase (NAM-). The sea-level pressure gradient forms a tripole-like pattern between the Atlantic and the Pacific. Since the pressure increase in the high latitudes is mainly in the Atlantic sector, we refer here to the NAM- as NAO-. The NAO- pattern is also strongly reflected in the surface temperature response in Fig. b (DJF and March–May, MAM), with a warming over eastern Canada and a severe cooling over northern Eurasia.

For the radiative CFC effect, the NAM is zonally more symmetric and shows the opposite signal all year round with varying strength. It shifts to a more positive phase, i.e., the near-surface pressure gradient between the middle and high latitudes increases (Fig. b; for the other seasons Fig.  middle column). It is most likely that this response does not originate from the stratosphere, as stratospheric changes induced by CFCs are small (see Fig. b and e), but arises from the CFC-induced warming in the troposphere and in particular the upper tropical troposphere. This CFC GHG effect is similar to what is observed in a future changing climate (see, e.g., ). The combined (total) CFC effects cancel each other out in winter in the NH Atlantic region (Fig. c); i.e., the NAO is unaffected by the collapse of the ozone layer. However, the Pacific sector shows a strengthening of the meridional pressure gradient as both chemical and radiative CFC effects reinforce each other there. The stronger pressure gradient over the North Pacific is associated with changes in regional precipitation patterns, such as enhanced precipitation over the North Pacific, Alaska, Canada and parts of the Arctic (Fig. 7b in ). With climate change, Arctic latitudes are projected to receive more precipitation . Hence, the absence of the MPA would amplify the effects of climate change on the hydroclimate of these regions.

Overall under the extreme scenario of the No-MPA, we observe significant changes in the sea-level pressure anomalies both in the Atlantic and in the Pacific sector. However, the overall changes in the Atlantic sector cancel each other out due to the opposing signs of the individual effects. This is in contrast to today's knowledge of the effects of historical ozone-depletion trends in the Arctic, which are deemed unlikely to have induced any trends in the NAM . We also want to emphasize here that the forcing applied from this extreme No-MPA scenario exceeds by far any forcing from the historical ozone-depletion period.

As a consequence of the abundance of CFCs and their large greenhouse gas potential, the global surface temperature rises in the noMPA experiment by almost 3.8 K compared to 1980 and is around 1.7 K higher than the reference experiment by the end of the century (red line in Fig. a). This is similar to the warming by the end of the century obtained with the SSP5-8.5 scenario (see Fig. 4 in ), as well as what show for their world-avoided study based on the RCP6.0 scenario. However, we are not taking into account the additional warming from the additional release of biospheric carbon in the No-MPA scenario as they did. Since the surface temperature changes are small in the 1980–2020 period (see also Fig. 4 in ), we decided to only show it from 2020 to 2099 to zoom in on the period around 2050 when the temperature curves start to diverge due to the radiative CFC effect (see later this section). When the CFC warming is not considered (noMPA_CFCRadOff), the ozone depletion leads to a decrease in surface temperature by the end of the century (blue line in Fig. a) by 0.6 K compared to the reference (orange line in Fig. a). This temperature decrease is also depicted in Fig. b. For the boreal winter, we obtain the strongest cooling exceeding -3 K over northern Eurasia, whereas northern Canada experiences a warming of up to 3 K. This warming is most likely part of the dynamical response and the resulting negative NAO phase due to the weakening of the stratospheric polar vortex discussed above. These regional temperature changes are due to, e.g., reduced advection of mild air over Eurasia (e.g., ). The CFC radiative effect increases the surface temperature by around 2.3 K globally, with the strongest signal being an over 12 K increase in the northern polar regions in DJF (Fig. c), which leads to a net warming of 1.7 K globally (Fig. d) compared to the reference. The most pronounced effect is seen over northern Canada where the warming from the ozone depletion adds to the CFC radiative effect, leading to an overall warming of over 13 K by the end of the 21st century compared to the reference. To put this warming into perspective, for the Arctic region, the IPCC AR6 projects a warming of 10 K over the period from 1995–2014 to 2081–2100 under the highest-emission scenario SSP5-8.5. Hence, the warming in a mid-level-emission scenario (SSP2-4.5) without MPA would even surpass the warming in a high-emission scenario (SSP5-8.5) with the MPA being in place.

As seen from the temperature evolution in Fig. a, the CFC warming effect starts to overpower the cooling from ozone at around 2055. We consider this point in time to be the shift in regimes when the surface response to the No-MPA scenario is no longer dominated by the CFC chemical effect. The radiative effect of CFCs takes over and continues to modulate the surface climate, as was indicated by, e.g., . Taken together, we find that the avoided warming due to the MPA is substantially modulated – at the regional scale – by the large-scale circulation changes induced by ozone and alterations in stratosphere–troposphere coupling. In addition, we find that globally, only a minor fraction (26 %) of the surface heating due to CFCs (via longwave trapping) is offset by the cooling due to the resulting ozone depletion , consistent with recent work examining the radiative forcing .