| 08 Sep 2025
Constrained attribution of changes in winds over the Southern Ocean from 1950 to 2100
Abstract. Strong near-surface westerly winds drive the Southern Ocean circulation and play a key role in setting regional and global climate. In the latter half of the 20th century, depletion of stratospheric ozone over Antarctica has caused these winds to accelerate and move polewards, particularly in austral summer. However, the future evolution of these winds remains uncertain. We use reanalysis data and the UK Earth System Model (UKESM1), with full atmosphere interactive chemistry, to assess the drivers of the winds over the recent past and coming century. We first characterize the wind mean state, distribution, and trends over 1980–2020 in the most commonly used atmospheric reanalyses (ERA5, JRA3Q, MERRA2, and R1) to gain insights into observed wind behaviour in the past. We show that while the representation of the mean wind is similar among reanalyses, MERRA2 and R1 show stronger wind acceleration trends that persists year-round, while JRA3Q and ERA5 show weaker acceleration, primarily in austral summer. Using an observational Southern Annular Mode (SAM) index, we show that the weaker, summer-focused trends of JRA3Q and ERA5 are likely more accurate. UKESM1 represents historical trends in winds accurately compared to ERA5. Targeted model simulations show that ozone depletion is overwhelmingly responsible for the wind acceleration observed in 1980–2020, which occurs primarily in austral summer. The effect of ozone depletion on winds peaks in 1980–2000, when it is roughly double that for the entire 40-year period. Ozone recovery is then associated with a slowdown of winds from 2000 to 2050. Beyond 2050, the ozone effect becomes minimal and winds accelerate primarily due to greenhouse gas induced warming, with this trend more evenly distributed across seasons.
Summary: The authors assess the suitability of and differences between reanalysis products regarding winds over the Southern Ocean and the Southern Annular Mode index. They find considerable differences among four contemporary products. They then proceed to characterize the UKESM1 model w.r.t. how it simulates those winds and the SAM index, and what drives multidecadal trends in them. They find that ozone depletion dominates in summer and during the period of increasing ozone-depleting substances (i.e. the late 20th century) but in the 21st century and in other seasons, ongoing greenhouse gas increases drive weaker trends that in the 21st century are counteracting the impact of ozone recovery.
I don’t have any major issues with the paper but also struggle to see the new, fundamental insights presented by the authors regarding what drives trends in the SAM. The review of the literature does not acknowledge several papers, including foundational ones, that have generally made these points. The comparison of the re-analyses is of course useful, but again the properties found here have at least partly been found in the literature before. The point is made that UKESM1 well represents trends in the SAM index. That may be the case. However, UKESM1 has also been found to simulate quite a large ozone depletion (as acknowledged by the authors), meaning it simulates stronger ozone forcing than any of the few other CMIP6 models with interactive ozone that have been studied, and almost certainly stronger than observed (although observations prior to 1979 are scarce). I thus hypothesize that this fairly large ozone forcing is compensated by difficulties with correctly propagating the dynamical impacts of ozone depletion into the troposphere, i.e. only a muted response ensues in tropospheric pressure and winds. Perhaps the authors can elaborate on this point further.
Below are several papers that should be touched upon in the discussion:
Son S-W, Tandon N F, Polvani L M and Waugh D W 2009 Ozone hole and Southern Hemisphere climate change Geophys. Res. Lett. 36 L15705
Son S-W et al. 2010 The impact of stratospheric ozone on Southern Hemisphere circulation changes: a multimodel assessment J. Geophys. Res. 115 D00M07
Simpkin & Karpechko, 2012, doi:101007/s00382-011-1121-2
Eyring V et al. 2013a Long-term ozone changes and associated climate impacts in CMIP5 simulations J. Geophys. Res. Atmos. 118 5029–60
Gerber E P and Son S-W 2014 Quantifying the summertime response of the austral jet stream and Hadley cell to stratospheric ozone and greenhouse gases J. Clim. 27 5538–59
Seok-Woo Son et al 2018 Environ. Res. Lett. 13 054024, DOI 10.1088/1748-9326/aabf21
Morgenstern, O. (2021), JGRA, 126, https://doi.org/10.1029/2020JD034161
Maybe not all of these need to be discussed separately, but a more in-depth review of the literature would be in order. Morgenstern (2021) also reviewed some observational products for the SAM index and also found that R1 is an outlier, particularly during southern winter.
Up to normalization, the SAM index presented here is the “Gong & Wang” index (Gong, D., & Wang, S. (1999). Definition of Antarctic oscillation index. Geophysical Research Letters, 26(4), 459–462. https://doi.org/10.1029/1999GL900003). It forms the basis for the Marshall index which as noted is derived from 6 stations, i.e. it is not a zonal mean. This should be acknowledged.
As for the behavior of the UKESM1 model, Morgenstern et al. (2020, https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2020GL088295, and 2022, https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2022JD037123 ) and Keeble et al. (2021) show that Antarctic ozone loss in this model is stronger than in observations and other CMIP6 models. That should be emphasized more strongly, I think. I’m not sure that the trends at 60-90 degreesS are more realistic than elsewhere (not according to Morgenstern et al. 2020). The amplitude of the trend is just larger than elsewhere.
Minor points:
P2L41: I don’t understand how the polar vortex can be “accelerated”. How about “deepens”? There is evidence that the polar vortex is more long-lived with ozone depletion than without, with everythin else in a model remaining unchanged.
P2L70: There could be effects other than the wind-driven ocean circulation that can limit the usability of a model such as UKESM1 for simulating the uptake and transport into the abyss of CO2 by the Southern Ocean, such as the quality of simulation of Antarctic bottom water formation. I’m not sure about UKESM1, but in other models this has certainly been found deficient, with impacts on the carbon cycle. So a more careful formulation here might be in order, such as that atmospheric drivers of ocean uptake in this model are fine.
P5L133: I suggest rather than comparing the Marshall index to the Gong & Wang index, to simply calculate the index in the same way as Marshall, i.e. using the 6 locations. Then you compare “oranges to oranges”. Alternatively, you can show that the two derivations yield nearly the same results.
P5L161: Rather than discussing biases here (which is fine), the interesting question is how trends in TCO compare to observations, because trends in TCO are linked to climate change. Here I’m not convinced (following Morgenstern et al. 2020) that UKESM1 is more suitable because the bias over the Antarctic is smaller than elsewhere.
P6L169: Surely you are not actually “modifying emissions” but rather surface mixing ratios of ODSs (which are globally uniform and follow a prescribed scenario). I would not use the word “emissions” in this context.
P6L175: In 1950 there were about 0.6 ppbv of Cl in the atmosphere due to CH3Cl, plus traces of CFCs that had been invented in the 1930s. So it’s not quite correct to say that “no ODSs” reached the stratosphere. Just the large increases since 1950 do not reach the stratosphere in this experiment.
P6L180: The assumption of linearity (needed to attribute any differences between ozone-hist and ozone-1950 to GHGs) is a little dangerous because of nonlinear couplings between GHGs and ODSs, such as reflected in the roles of NOx and HOx (both affected by GHGs) in interfering with halogen-catalyzed ozone depletion. It would be better to have a simulation GHG-1950 in which the leading GHGs (CO2, CH4, N2O) are held constant and ODSs follow their usual trajectory. Not sure why this does not exist in AerChemMIP, but if you can produce these simulations, that would certainly help allay this concern.
P12L373: Reproducing the trend in winds is not sufficient to conclude that the model successfully reproduces ozone depletion. Based on other literature, I think it somewhat exaggerates ozone depletion, but then struggles to fully transport this dynamical driver into the troposphere. I suggest to rephrase this, allowing for this cancellation of errors.
P13L423: I would replace “poleward jet trend” with “poleward progression of the jet latitude” or similar.
P14L429-430: Winds don’t “accelerate” or “slow down”, they “increase” or “decrease”.
P13L445: A more robust approach would use multiple models (there are more full-chemistry models in the CMIP6 archive) and use an emergent-constraint analysis (making use of historical biases in these models) to produce a multi-model projection of future winds over the Southern Ocean.
P14L448-457: This is a surprising, ocean-centric conclusion to this paper, given that the Southern Ocean plays almost no role in the rest of the paper. Perhaps you can refocus this to reflect more on the actual findings of this paper?
Figure 1: Which period and which level do the plots refer to? This information should be in the caption. Ditto figures 2, 3, and table 2 (the level needs to be indicated).
Tables 2 and 3: Here it would be fairly straightforward to complement this with data for selected CMIP6 models that are comparable to UKESM1, such as CESM2-WACCM, EC-Earth-AerChem, CNRM-ESM, or more. Given the large ozone depletion in UKESM1, contrasting it with models with weaker ozone depletion (the weakest would be found in MRI-ESM2), perhaps something useful can be learnt about the role of ozone forcing.
Figure 4: Again here the definition of the mean jet position should be included, especially at which level the winds are evaluated.