Seasonality and scenario dependence of rapid Arctic sea ice loss events in CMIP6 simulations

Sticker, Annelies; Massonnet, François; Fichefet, Thierry; DeRepentigny, Patricia; Jahn, Alexandra; Docquier, David; Wyburn-Powell, Christopher; Quint, Daphne; Shivers, Erica; Ortiz, Makayla

doi:https://doi.org/10.5194/egusphere-2024-1873

Preprints

https://doi.org/10.5194/egusphere-2024-1873

Preprints

01 Jul 2024

| 01 Jul 2024

Seasonality and scenario dependence of rapid Arctic sea ice loss events in CMIP6 simulations

Annelies Sticker, François Massonnet, Thierry Fichefet, Patricia DeRepentigny, Alexandra Jahn, David Docquier, Christopher Wyburn-Powell, Daphne Quint, Erica Shivers, and Makayla Ortiz

Abstract. The end-of-summer Arctic Ocean is projected to face at least one occurrence of practically ice-free conditions (sea ice extent < 1 million km²) by the middle of the century under all Climate Model Intercomparison Project phase 6 (CMIP6) scenarios. Climate models indicate that this transition toward a nearly ice-free Arctic Ocean in late summer will be punctuated by rapid ice loss events (RILEs), i.e., reductions in sea ice extent that occur at a much faster rate than expected from the forced contribution. The extreme sea ice loss associated with RILEs in climate models is larger than what has been observed since the start of the satellite era (−0.28 million km² per year over 2001–2008). As such, it could lead to a much faster transition toward practically ice-free conditions than expected based on a linear trend. RILEs are not well understood and it is currently impossible to predict their occurrence a season to several years ahead. It is therefore essential to improve our understanding of these events. This study presents the first comprehensive analysis of RILEs in a diverse set of 26 CMIP6 models, including 5 large ensembles, following both low and high warming scenarios over the period from 1970 to 2100. Our analysis shows that RILEs are expected to occur year-round. However, the timing and duration of the events are found to be season-dependent, with less frequent but longer-lived RILEs in winter and spring, and more frequent but shorter-lived RILEs in summer and autumn under a high emission scenario. In addition, we find that the warming scenario has a greater influence on RILE characteristics in the winter/spring season than in the summer/autumn season. Our results also emphasize that model uncertainty is larger regarding the probability and characteristics of RILEs for winter/spring events compared to summer/autumn ones. Finally, while the initial sea ice extent at which RILEs are triggered depends on whether they occur in September or March, the initial sea ice volume is similar for both months, which emphasizes the critical role of sea ice thickness as a preconditioning factor for RILEs. Based on CMIP6 models, there is 63 % chance that at least one summer RILE starts before 2030 in September. The study of RILEs is particularly opportune as, after more than ten years of relatively stable conditions between 2012–2023, the current summer Arctic sea ice state has an increased probability to be on the verge of a rapid reduction for the coming decade.

Received: 19 Jun 2024 – Discussion started: 01 Jul 2024

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 preprint. The responsibility to include appropriate place names lies with the authors.

Download & links

Preprint (PDF, 3487 KB)

Supplement (2917 KB)

Download & links

Annelies Sticker, François Massonnet, Thierry Fichefet, Patricia DeRepentigny, Alexandra Jahn, David Docquier, Christopher Wyburn-Powell, Daphne Quint, Erica Shivers, and Makayla Ortiz

Status: final response (author comments only)

RC1:
'Comment on egusphere-2024-1873', Anonymous Referee #1, 12 Oct 2024
General comments
Sticker et al. analyse rapid Arctic sea ice loss events (RILEs)—year-to-year reductions in sea ice extent greatly exceeding that expected from the long-term trend—in CMIP6 and large-ensemble simulations. In contrast to previous studies, they examine RILEs in all seasons and assess the consistency across the CMIP6 ensemble. They find that RILEs do occur in winter/spring but less frequently than in summer/autumn and strongly dependent on future warming scenario. Interestingly, the sea ice volume (SIV) at which RILEs are typically triggered is about the same for March and September RILEs, despite differences in initial sea ice extent (SIE).
This work advances our understanding of large-scale sea ice variability on interannual timescales in the context of RILEs, which are likely to become more frequent in the coming decades as the authors explain and demonstrate. I found the result relating to pre-conditioning based on the initial SIV being similar for winter and summer RILEs particularly striking, although I think the discussion/interpretation of it needs a little expansion (see specific comments).
I see no issues with the methodology overall, except perhaps with using the CMIP6 results to infer to the likelihood of a RILE starting in the next few years. It is not clear where the rather specific probability of 63% stated in the abstract comes from, and I think there are caveats here which I don’t see that the authors have acknowledged (see specific comments).
The figures are presented well, and I appreciate the concise length of the manuscript. One weakness in terms of presentation and structure is the results section 3. I found myself having to jump back and forth between Figs. 2–5 too many times while reading the text, making it difficult to follow. I suggest the authors consider rearranging figures (and possibly some of the text) to improve readability here. I have given a few suggestions in the specific comments below.
Otherwise, subject to addressing these points and the other minor and technical points noted below, I believe this work should be published in The Cryosphere.
Specific comments
Major comments
Preconditioning/SIV result: this result is, at first thought, quite surprising, because one expects the typical SIV in winter to be larger than the typical SIV in summer (e.g., see PIOMAS time series). But it can be explained by either:
Winter RILEs mainly occurring later (i.e., when winter SIV of the mid–late 21^st century is presumably comparable to summer SIV of the late 20^th century)

Large interannual variability in winter and summer SIV, such that anomalously high summer SIVs are comparable to anomalously low winter SIVs

It seems like both are relevant, from Fig. 2 (for 1) and Fig. S4 (for 2), but the problem is that authors do not mention either (or anything else). Currently they just state that the initial SIV is the same for winter and summer RILEs based on Figs. 4c and 5c, and this therefore indicates a role of preconditioning (e.g., L255, 266, 311). I think these interpretations, and some comment on the extent to which one or the other is dominant, should be added to the discussion.
Chance of RILE by 2030: I’m uncertain about the authors’ claim in the abstract that the real Arctic has a “63% chance” of exhibiting a RILE by 2030. Firstly, this value of 63% is only present in the abstract (L19) and so it is not clear where it comes from. In any case, there are surely too many uncertainties with this estimate to state such a specific value, so it would be better to rephrase into a more general statement with approximate likelihood (e.g., “suggest about a 60% chance”). From a readability standpoint, it would also make more sense to put this sentence after the sentence which currently follows (i.e., make it, “The study of RILEs is particularly opportune […]”, then “Based on CMIP6 […]”).
The authors need to explain somewhere in the main text how they are deriving this estimate and mention the underlying assumptions. In particular, they need to note their estimate of the likelihood is going to be affected by the issue of model uncertainty/spread. I suspect their estimate could be biased towards certain models with stronger sea ice declines (e.g., CanESM5, EC-Earth3, from Fig. 1b). For example, EC-Earth3 appears to have quite a large left tail in the distribution of trends after stability (Fig. S6), meaning it probably contributes more to the estimate of imminent RILE likelihood. In a couple of places, the authors use current estimates of the real Arctic sea ice volume to help justify the mapping from CMIP6 simulations to what might happen in the real world (e.g., L253, L272). The problem here is that there remain large systematic biases in Arctic sea ice (trends) in CMIP6 compared to the real world, so even if you match the current trends and/or values of the SIV/SIE, the models may be simulating that with unrealistic global warming, for example (e.g., Rosenblum and Eisenman, 2017). This, then, casts doubt on how applicable the underlying statistics leading to the model estimate of RILE likelihood at a given SIV level/stability period is to the real world with the same SIV level/stability period.
Readability of section 3: this could mostly be addressed by rearranging the figures. Figure 4 panels e–g seem like they should be in a separate figure to panels a–d, and similarly for Fig. 5. Separating the e–g panels out of each and combining into one figure might be better. Panels e–g are referred to in section 3.1 before the a–d panels for both Figs. 4 and 5. Figure 6 is also mentioned in section 3.2 before Fig. 4a, which is first mentioned in section 3.3.
Minor comments
L5: I suggest making it clear that you are referring to year-to-year changes in total sea ice extent, as this description could equally apply to sub-seasonal time scale and/or regional scale sea ice loss. Indeed, there is a separate body of literature on such “very rapid ice loss events (VRILEs)”, which is obviously quite different to what you study. It is unfortunate that there is such a clash of terminology/acronyms, and while I suspect the term “RILE” will ultimately be more commonly used for the interannual events you are describing, it is better to be clear up front. Simply adding “year-to-year” in front of “reductions” and “total” in front of sea ice extent would be one way to address this point.
L58–65: As above, I suggest briefly noting somewhere in this paragraph the distinction from short timescale “very rapid ice loss events” (e.g., McGraw et al., 2022; Wang et al. 2020).
L84: Worth noting that Arctic sea ice stabilizes in SSP1-2.6 (e.g., IPCC AR6/TS). So, towards the latter part of the 21^st century RILES are occurring in the absence of a background trend for most models.
L100: Since you are considering total sea ice extent/area (SIE/SIA) on interannual time scales, I would not expect model resolution to matter too much. The difference is that a change in SIA can occur with relatively little change in SIE, so that RILEs defined in terms of SIA are (potentially) physically/fundamentally different to those defined in terms of SIE. I think it’s fine to use SIE (especially considering you examine SIV too), but might be worth noting this as another reason for checking the impact of using SIA.
L126–128: “not shown”: it is good that the authors include some model evaluation. While I understand they do not wish to clutter their manuscript with too much tangential material, I think here the authors could include a figure demonstrating this in their supplementary materials, or at least cite some other studies (surely the extreme November–June sea ice departures from observations in MIROC6 have been found already, for example?)
L174: I agree with comparison between and interpretation of Figs. 4e,f,g and 5e,f,g, but would suggest there is a bit more to say here about the (e) panels in particular as not all large ensembles look the same. The CanESM5 large ensemble has a fairly uniform distribution, and looks like CMIP6 multi-model ensemble for SSP5-8.5, whereas the MIROC6 large ensemble’s distribution looks more like the CMIP6 distribution for SSP1-2.6, even though SSP5-8.5 simulations are used for all large ensembles in Fig. 5e. If I understand correctly, this is because MIROC6 has a relatively weak long-term trend and so looks like the CMIP6 average for SSP1-2.6 even when simulating SSP5-8.5. Either way, I think the interpretation should be described in the text (rather than just, “the characteristics are found to be very similar”).
L178: I think this also follows (more obviously?) from Fig. 3 rather than Fig. 5 (see also point above about readability in this section)
L185: I see you have defined “consistently September ice free” in the caption of Fig. 2, but this should be stated in the main text (either here or somewhere in section 2). I also think a reference should be provided for this (e.g., Senftleben et al., 2020).
L211: “No SRILE occurs after consistently ice-free conditions occur in September”: is this not obvious? You need sea ice in September to have a September RILE. Is this what you meant to write?
Technical corrections
L29: “more vulnerable to atmospheric and oceanic variability”: “variability” → “forcing” (or “more vulnerable to variability in atmospheric and ocean forcing”)?
L32: “a interannual” → “an interannual”
L49: “during one or several” → “during one or over several”
L140: I suggest not introducing the acronym “RICE” here, since it is only used in this line and nowhere else in the rest of the manuscript.
L220: “onset to” → “onset on”
L255: “preconditionning” → “preconditioning”
L264: “had” → “has”
L270: repeated reference (I think it’s clear that you are still describing results from Döscher and Koenigk, 2013, cited on L266).
L272: “in mid-2020s” → “in the mid-2020s”
L295: Unclear; re-phrase? (E.g., to “…the percentage of members with at least one RILE per year ranges from 62–96%, and every model experiences at least one RILE during the analysis period”)
L306: “ice” → “sea ice”
L308: Remove or rephrase “nicely” (e.g., “This result is most clearly illustrated by EC-Earth3, for example, …”)
L320: I commend the authors for including the specific data citations for all CMIP6 models and simulations—a lot of other studies, particularly multi-model studies, do not bother with these. However, I do suggest moving them from the supplementary materials to the main text (e.g., add them to Table 1?). Otherwise, I don’t think that cross referencing systems will detect the citations, which is important for tracking usage of CMIP data (and it would then have been a waste of your time to include them in the first place!)
L322: This NSIDC dataset has a proper citation with DOI; I suggest adding this rather than a URL.
Figure 1, caption: “NISDC” → “NSIDC”
Figure 6, caption: “using SSP5-8.5 scenario” → “using the SSP5-8.5 scenario”
References
McGraw, M. C., Blanchard-Wrigglesworth, E., Clancy, R. P., and Bitz, C. M.: Understanding the Forecast Skill of Rapid Arctic Sea Ice Loss on Subseasonal Time Scales, J. Climate, 35, 1179–1196, https://doi.org/10.1175/JCLI-D-21-0301.1, 2022
Rosenblum, E. and Eisenman, I.: Sea Ice Trends in Climate Models Only Accurate in Runs with Biased Global Warming, J. Climate, 30, 6265–6278, https://doi.org/10.1175/JCLI-D-16-0455.1, 2017
Senftleben, D., Lauer, A., and Karpechko, A.: Constraining Uncertainties in CMIP5 Projections of September Arctic Sea Ice Extent with Observations, J. Climate, 33, 1487–1503, https://doi.org/10.1175/JCLI-D-19-0075.1, 2020
Wang, Z., Walsh, J., Szymborski, S., and Peng, M.: Rapid Arctic Sea Ice Loss on the Synoptic Time Scale and Related Atmospheric Circulation Anomalies, J. Climate, 33, 1597–1617, https://doi.org/10.1175/JCLI-D-19-0528.1, 2020
Citation: https://doi.org/10.5194/egusphere-2024-1873-RC1
RC2:
'Comment on egusphere-2024-1873', Anonymous Referee #2, 12 Nov 2024
The authors of this study investigate Arctic Rapid Ice Loss Events (RILEs) in the CMIP6 ensemble, including their frequency through the year, dependence on emission scenario, and possible preconditioning conditions. The manuscript is well written, the figures high quality, and the methodology sound and well described. I think this paper will be an important contribution to the literature and have only minor suggestions before it should be published.
Specific Comments:
Variability results: (Line 259 and elsewhere). The impact of variability on the frequency or probability of RILEs is a major result, yet there are no figures showing variability. The value of the large ensembles in this analysis is the ability to look at variability for a single model and how that relates to RILEs, something you can’t do with the CMIP6 multi-model ensemble. A figure like 5b and 5c but with variability at RILE start would be beneficial. Additionally, Figure S5 may be relevant for the main manuscript and something you could add as a panel to Figure 1. I think you should spend a bit more time on this result since in your conclusions you list it (Line 307-309) but no main figures show larger variability in a model leads to more RILES.

Figure 4: The order of how you refer to the panels in the text is confusing and you should consider re-ordering them.

Figure 5: I do not understand what 5f and 5g are showing and 5f is not referenced in text anywhere. Maybe this figure or panels from this figure could be in the supplementary material.

Figure 6: If RILEs initiate more often after a period of stable SIE trends, does this imply that the SIV would still have a negative trend during this period. So the ice is thinning but not changing extent? Can you add some text about what’s going on with SIV during these periods?

Line 2 and Line 8: “practically ice free” and “nearly ice free” is awkward. Just define “ice free” and go with it.

Line 6-7: This is a confusing sentence because if you’re looking from start of satellite era, why are you listing 2000-2008 rates?

Line 7: what is “it”? Maybe say “As such, there could be a much faster transition…”

Lines 101-104: This description of these conversions is confusing. Maybe use equations instead of text? Also, shouldn’t grid cell area come in here?

Line 138-139: What do you mean a RILE can manifest as one year event or several years? This sentence sounds like you mean the metric by the Döscher and Koenigk since that’s the last you discussed, and that’s a one-year definition.

Line 140: Is a RICE substantively different than a RILE? Why not include this in the other metrics described?

Line 203: What is “a period”? Is it a single year where the 10-year previous trend was positive or zero? This sentence needs clarification.

Line 295: The sentence where you say the percentage goes from 62-96% is confusing and should be clarified.
Citation: https://doi.org/10.5194/egusphere-2024-1873-RC2

Annelies Sticker, François Massonnet, Thierry Fichefet, Patricia DeRepentigny, Alexandra Jahn, David Docquier, Christopher Wyburn-Powell, Daphne Quint, Erica Shivers, and Makayla Ortiz

Supplement

https://doi.org/10.5194/egusphere-2024-1873-supplement

Annelies Sticker, François Massonnet, Thierry Fichefet, Patricia DeRepentigny, Alexandra Jahn, David Docquier, Christopher Wyburn-Powell, Daphne Quint, Erica Shivers, and Makayla Ortiz

Viewed

Total article views: 458 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
299	83	76	458	42	16	20

HTML: 299
PDF: 83
XML: 76
Total: 458
Supplement: 42
BibTeX: 16
EndNote: 20

Views and downloads (calculated since 01 Jul 2024)

Month	HTML	PDF	XML	Total
Jul 2024	137	38	12	187
Aug 2024	38	10	3	51
Sep 2024	21	10	1	32
Oct 2024	46	8	2	56
Nov 2024	35	5	4	44
Dec 2024	22	12	54	88

Cumulative views and downloads (calculated since 01 Jul 2024)

Month	HTML	PDF	XML	Total
Jul 2024	137	38	12	187
Aug 2024	38	10	3	51
Sep 2024	21	10	1	32
Oct 2024	46	8	2	56
Nov 2024	35	5	4	44
Dec 2024	22	12	54	88

Viewed (geographical distribution)

Total article views: 453 (including HTML, PDF, and XML) Thereof 453 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 23 Dec 2024

Short summary

Our study analyses rapid Arctic sea ice loss events (RILEs), which are significant reductions in sea ice extent. RILEs are expected throughout the year, varying in frequency and duration with the seasons. Our research gives a year-round analysis of their characteristics in climate models and suggests that summer RILEs could begin before the mid-century. Understanding these events is crucial as they can have profound impacts on the Arctic environment.


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