Linking open-ocean polynyas and deep convection in the Southern Ocean across CMIP6 models

Lu, Shunzi; Marinov, Irina; Molodtsov, Sergey

doi:10.5194/egusphere-2026-3060

Preprints

https://doi.org/10.5194/egusphere-2026-3060

Preprints

10 Jul 2026

| 10 Jul 2026

Status: this preprint is open for discussion and under review for The Cryosphere (TC).

Linking open-ocean polynyas and deep convection in the Southern Ocean across CMIP6 models

Shunzi Lu, Irina Marinov, and Sergey Molodtsov

Abstract. Open-ocean polynyas (OOPs) and deep convection in the Southern Ocean are critical features of the global climate system, however, their representation and mutual dependence in climate models remain poorly understood. This study investigates the occurrence and coupling of OOPs and deep convection across 49 CMIP6 models using long-term pre-industrial control simulations. Our results reveal that while most models simulate both phenomena, their spatial and temporal co-occurrence varies substantially. Although deep convection is typically associated with surface salinification and heat loss, it does not always result in detectable polynyas. We identify two distinct regimes of OOPs across the ensemble: "deep OOPs", which are directly coupled to deep ocean convection, and "shallow OOPs", which form independently of deep mixing, likely driven by surface forcing or sea-ice divergence. The representation of these regimes is strongly influenced by the choice of ocean model component. These findings highlight the importance of process-based diagnostics in evaluating Southern Ocean overturning and suggest that the connection between surface polynyas and deep water formation is more complex than traditionally assumed in climate models.

Received: 05 Jun 2026 – Discussion started: 10 Jul 2026

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Shunzi Lu, Irina Marinov, and Sergey Molodtsov

Status: open (until 01 Oct 2026)

Post a comment Subscribe to comment alert

RC1:
'Comment on egusphere-2026-3060', Anonymous Referee #1, 28 Jul 2026 reply
Review of Linking open-ocean polynyas and deep convection in the Southern Ocean across CMIP6 models
This study examines Southern Ocean open-ocean polynyas (OOPs) across CMIP6 models. The study is extensive, analysing 49 models from the CMIP6 piControl ensemble using the final 500 years of data from each model. The authors use a combination of deep convection diagnostics, polynya detection methods, and a standard set of Southern Ocean variables, presenting frequency maps and correlations with sea ice and ocean variables.
The authors find that, while deep convection is associated with salinification and heat loss, it is not always associated with the occurrence of OOPs. This was a particularly interesting result. They identify two distinct types of OOPs within the models, referred to as deep and shallow polynyas, which are associated with different mixing processes and are highly model dependent.
Overall, the manuscript is very well written, with a clear experimental design and results presented in a logical and accessible manner. The science is of high quality, including appropriate statistical analyses, and fits well within the scope of The Cryosphere. The methodology is sound and well justified, acknowledging the wide range of approaches used in the literature for detecting both polynyas and deep convection, while adapting these methods to the availability of variables across CMIP6 models. Additionally, I particularly appreciated the comprehensive assessment of each model group.
The results will be of interest to researchers studying polynyas in either polar region, the mechanisms driving their formation, and their representation across climate models. The manuscript builds upon the currently limited CMIP6 sea ice literature and contributes to the growing body of work assessing polynya formation. With a few very minor corrections, I believe this manuscript will be suitable for publication in The Cryosphere. I therefore recommend minor revisions.
General comments
Several references require revision. In a number of places, citations appear somewhat arbitrary or overly specific. For example, (Bromwich and Kurtz, 1984; Thompson et al., 2020) are both valid references, but it is unclear why Thompson et al. (2020) is cited specifically when numerous other studies have examined the same topic. Consider either replacing these with a more representative citation, using e.g., or providing justification for selecting that particular study. This issue appears throughout the manuscript and should be addressed consistently.

On the deep convection, I think it would be interesting to look at the driving forces at the surface and with depth, and whether temperature or salinity change drives the convection, specifically when associated with the polynyas, in different regions and models in more detail. I appreciate that the authors briefly touch upon this in Figure 5, 6, 8 and the Appendix C, and you mention briefly on line 579 that the direction of the relationship between these variables and polynyas remain unclear and of course that salninification is a driving force, This may be extensive beyond the scope of the paper, however, maybe a small discussion point could be included.

Specific comments
Line 52: The 2016/17 Polynya should not be described as a full WSP. It is more accurately classified as a large MRP (e.g. Campbell et al., 2019; Cheon and Gordon, 2019).

Line 79: The statement "However, OOPs remain prevalent in CMIP6 simulations and are often overrepresented relative to observations (Mohrmann et al., 2021)." should be reconsidered. The comparison is somewhat problematic given the relatively short observational record compared with the much longer climate model simulations. OOP occurrence is highly variable over time, and models may exhibit extended periods with clustered events followed by long intervals without occurrence. As such, the assertion that OOPs are "overrepresented relative to observations" should either be supported with additional context.

Line 210: Repeated ‘Specifically’.

Line 230: I think that this is an important result. There is a lot of discussion around the quantitative definition of a polynya in the literature and showing that no single metric is consistent between models is important. In Figure 8 you even show that sea ice thickness varies vastly between models, which as you rightly mention, has been used as a polynya detection method in a number of studies.

Line 239: Could you specify what you mean my the OOP region? Do you mean the larger scale Weddell Sea/ Ross Sea etc, or do you mean directly under the masked polynya area in Figure 1?

Line 250: The wording "only 33 models show OOPs" is potentially misleading. Thirty-three models represent approximately 75% of the models that produce OOPs, which is a substantial proportion. Consider rephrasing to avoid implying that this is a small number.

Reply
Citation: https://doi.org/10.5194/egusphere-2026-3060-RC1
RC2: 'Comment on egusphere-2026-3060', Anonymous Referee #2, 22 Aug 2026 reply

The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2026/egusphere-2026-3060/egusphere-2026-3060-RC2-supplement.pdf
Reply

Citation: https://doi.org/10.5194/egusphere-2026-3060-RC2
RC3:
'Comment on egusphere-2026-3060', Anonymous Referee #3, 04 Sep 2026 reply
The submitted manuscript addresses an important field of research: the representation of open-ocean polynyas (OOPs) and deep convection in the Southern Ocean in coupled climate models. The manuscript is well-written and presented. It includes a substantial methodological consideration on the question of how best to compare deep convection and OOPs in coupled climate models. The manuscript then applies these methodologies to CMIP6 models to identify the co-occurrence of deep convection and OOPs. In general, my main concern is the lack of well defined research aims and novel results.
My major concerns are as follows:
The manuscript is very close in nature to the study by Mohrmann et al. (2021) which analyses the presence, frequency and spatial distribution of polynyas in the Southern Ocean in 27 CMIP6 models. The main two differences between the two studies is 1) the methodological discussion in this manuscript 2) this study uses the pre-industrial control simulations, and Mohrmann et al. (2021) uses the historical ones. It is clear that part of the motivation for using the historical CMIP6 simulations is in order to compare the models to observations of OOPs. By focusing on pre-industrial control simulations, we lose this aspect of the motivation of the study; it is impossible to say whether these models are behaving realistically or not. In the conclusions, the authors state "However, our analysis indicates that some CMIP6 models still exhibit substantial open-ocean convection.", this has already been established by Mohrmann et al. (2021), but also it is generally true that coupled models convect more with pre-industrial control forcing, so this is also not necessarily a surprising result.

Regarding aim (ii) (L86), I think more could be done to understand shallow OOPs and their distinction from deep OOPs, this is a potentially interesting avenue for further work. Figure 7 appears to show that for a lot of models the overlap is small against all OOPs. This suggests that there are a lot of polynyas that happen which are not deep OOPs (a lot of shallow OOPs), but ultimately this is presented as a percentage of the area, so it is very hard for the reader to understand the actual area of these events. It is likely that there is a correlation between the area of a deep convection, and whether it is also an OOP (i.e. very small polynyas might not be coincident with deep convection). I think it would really improve the discussion of shallow OOPs to better understand these events, this would require doing some analysis on the individual events, e.g. size distribution of individual shallow OOPs and deep OOPs. I also think that the MLD within the individual shallow OOPs should be looked at carefully, ultimately the mechanism for some of the events might be the same (sensible heat from mixing to 700 m which brings up warm CDW).

Regarding aim (iii) "which large-scale mean-state properties and model configurations favor each regime": it is not clear to me how it is possible to answer with the current methodology. Disentangling cause from effect here is very difficult without looking at properties before/after convection events (e.g. composites, timeseries). The authors state: "Models with larger mean convective areas tend to have higher SST (both September and annual mean), deeper September MLD and greater September heat loss within convective regions", these appear to be responses to convection rather than mean-state properties that favour convection.

Similar to the above point, the central hypothesis (L391) could be better posed: "These findings collectively support our central hypotheses that: (a) saltier surface waters can destabilize the water column and trigger deep convection, resulting in OOPs; and (b) in deep OOPs, a deeper MLD and enhanced vertical mixing engage more of the warm, saline upper CDW layer than in shallow OOPs, which results in warmer and saltier surface waters and a more significant reduction in sea ice cover compared to shallow OOPs". Point (a) has been made in other modelling and observational studies (e.g. Campbell et al. 2019, Kurtakoti et al. 2021, Kjellson et al. 2015), and point (b) in its current wording appears to state that deep mixing OOPs have deeper MLDs and more mixing, which is self-evident. I think it would be better to phrase the central aim around (a) the methodology (b) the mechanisms that are unique to deep/shallow OOPs if there are distinct mechanisms.

Most of the results seem to be confined to September monthly means, but different models may well have different seasonal peaks to their convective periods, and some may convect for larger proportions of the year. E.g. a model that convects in a small area for 6 months of the year might have the same integrated convective area as one which convects for only one month but extensively. This equally applies to polynya areas, in the observational record, polynyas can form during the sea ice breakup or formation periods e.g. via a Maud Rise halo. If this is happening in some models but not others, this could be missed by the analysis.

My minor concerns follow:
The manuscript would benefit from more clearly substantiating in-line claims. For example, line 262: "Similarly, the higher-resolution HadGEM3-MM shows reduced convection and polynya extent relative to the lower-resolution HadGEM3-LL (1◦)". This is not obvious from Figure 3, or perhaps 'reduced convection' is not precise enough, -MM appears to have a large regions of high frequency convection in the Weddell Sea. -LL has a higher 'total convection area' (Figure 7) but a lower 'mean convection area' (Figure C1), which in my understanding means that -MM convects more than -LL, but that -MM usually convects in the same location.

Line 337: "For example, CAS-ESM2-0 (No.15), FGOALS-g3 (No.16), HadGEM3-GC31-MM (No.22), and CNRM-CM6-1 (No.30) exhibit relatively small mean convection areas but high heat flux loss, suggesting that additional mechanisms are involved". This could be a result of the particular regions that are convecting. Some regions have more heat to vent to the atmosphere than others, and if the same region convects recurrently, its may already have depleted heat. I would recommend including the spatial distribution of heat content if this point is being pursued.

Line 339 "Models with larger mean convective areas also show lower September SIC south of 55◦S (Fig. 6c). ". Is this not a result of convection causing melt of sea ice? I would recommend exploring this point by looking at the sea ice coverage within the convection area (the cells actively convecting, rather than a multi-annual mean).

Line 611 is an anticipation of what will happen in future projections, but nothing in this manuscript can provide evidence to this point.

In my opinion this manuscript would benefit from being re-framed and resubmitted. There is a lot of useful discussion surrounding the challenge and importance of identifying OOPs and deep convection events in model output, this makes up approximately a third of the manuscript and could be re-organised into a methodology-focused paper. Recently, a similar paper focused on coastal polynyas was published (Landrum et al., 2026). This would require additional work, as framing as a methodological paper would require some more robust statistics, and wider sensitivity studies (e.g. repeating analysis for other months of the year). I also think that further exploration into the occurence of shallow OOPs is an interesting research avenue, but I think that the manuscript has not done enough to characterise these events. Looking at the size distribution, spatial distribution, MLD in and around individual events would provide a better insight into what drive these phenomena.
Campbell, E. C., Wilson, E. A., Moore, G. W. K., Riser, S. C., Brayton, C. E., Mazloff, M. R., and Talley, L. D.: Antarctic offshore polynyas linked to Southern Hemisphere climate anomalies, Nature, 570, 319–325, https://doi.org/10.1038/s41586-019-1294-0, 2019.
Kjellsson, J., Holland, P. R., Marshall, G. J., Mathiot, P., Aksenov, Y., Coward, A. C., Bacon, S., Megann, A. P., and Ridley, J.: Model sensitivity of the Weddell and Ross seas, Antarctica, to vertical mixing and freshwater forcing, Ocean Modell., 94, 141–152, https://doi.org/10.1016/j.ocemod.2015.08.003, 2015.
Kurtakoti, P., Veneziani, M., Stössel, A., Weijer, W., and Maltrud, M.: On the generation of Weddell Sea polynyas in a high-resolution Earth system model, J. Climate, 34, 2491–2510, https://doi.org/10.1175/JCLI-D-20-0229.1, 2021.
Landrum, L. L., DuVivier, A. K., Holland, M. M., Krumhardt, K., and Sylvester, Z.: Challenges in identifying Antarctic coastal polynyas in satellite observations and climate model output to support ecological climate change research, The Cryosphere, 20, 1815–1840, https://doi.org/10.5194/tc-20-1815-2026, 2026.
Mohrmann, M., Heuzé, C., and Swart, S.: Southern Ocean polynyas in CMIP6 models, The Cryosphere, 15, 4281–4313, https://doi.org/10.5194/tc-15-4281-2021, 2021.

Reply
Citation: https://doi.org/10.5194/egusphere-2026-3060-RC3

Shunzi Lu, Irina Marinov, and Sergey Molodtsov

Viewed

Total article views: 350 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
217	108	25	350	33	25

HTML: 217
PDF: 108
XML: 25
Total: 350
BibTeX: 33
EndNote: 25

Views and downloads (calculated since 10 Jul 2026)

Month	HTML	PDF	XML	Total
Jul 2026	165	60	15	240
Aug 2026	27	31	3	61
Sep 2026	25	17	7	49

Cumulative views and downloads (calculated since 10 Jul 2026)

Month	HTML	PDF	XML	Total
Jul 2026	165	60	15	240
Aug 2026	27	31	3	61
Sep 2026	25	17	7	49

Viewed (geographical distribution)

Total article views: 201 (including HTML, PDF, and XML) Thereof 201 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 15 Sep 2026

Short summary

We studied how CMIP6 models represent Southern Ocean open water areas within sea ice, known as open-ocean polynyas, and the mixing of deep ocean waters, known as deep convection, and found that these two processes are not always connected, contrary to common assumptions. While some open-ocean polynyas are associated with episodes of convection, others form through different mechanisms. These results show that polynyas should not be used as a direct proxy for deep convection in climate models.


Total:	0
HTML:	0
PDF:	0
XML:	0