the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 31 Aug 2026
Sea-ice thickness initialization improves summertime sea ice prediction in the Barents-Kara Sea
Abstract. Arctic sea ice has substantially declined over the past four decades. Given the projected decrease under global warming, reliable prediction of summer sea ice is becoming increasingly important. Yet most climate models still show limited low skill in predicting summer sea ice from spring, a limitation commonly referred to as the spring predictability barrier. Here we develop a seasonal forecast system based on an eddy-permitting coupled model and assess how different ocean and sea ice initialization strategies affect prediction skill. Results show that the summer sea ice in the Barents-Kara Sea is predicted most skillfully when sea ice and snow conditions in the model are initialized. Sea ice initialization improves the representation of the initial sea ice thickness, the subsequent ice-albedo feedback, and the meridional sea ice advection from the higher latitudes. These findings suggest significance of sea ice thickness initialization for skillful summer prediction despite the spring predictability barrier.
- Preprint
(3303 KB) - Metadata XML
-
Supplement
(2023 KB) - BibTeX
- EndNote
Status: open (until 21 Oct 2026)
- RC1: 'Comment on egusphere-2026-5070', Anonymous Referee #1, 15 Sep 2026 reply
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 126 | 50 | 25 | 201 | 28 | 26 | 22 |
- HTML: 126
- PDF: 50
- XML: 25
- Total: 201
- Supplement: 28
- BibTeX: 26
- EndNote: 22
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The paper presents a sea ice assimilation study comparing two fully coupled ocean-sea-ice-atmosphere models at different resolutions (SINTEX-F2 and SINTEX-F3) with the available fully coupled assimilation systems from the Copernicus Climate Change Service (C3S). The study finds that the C3S multi-model mean seasonal prediction of the sea ice edge (SIE) outperforms the SINTEX-F2 and SINTEX-F3 predictions. Furthermore, the study finds that sea ice thickness (SIT) initialization in SINTEX-F3 leads to the best sea ice edge predictions in SINTEX-F3.
While the topic is timely and relevant, I regret to say that I do not currently find the description of the methods or the argumentation supporting the main findings convincing, nor their findings novel. In addition, the authors make several unconventional methodological choices, for example, assimilating data from reanalysis, not including a reference run in their analysis, and using the term 'ensemble' for model runs employing different assimilation techniques, without discussing or motivating these choices. As a result, I find it difficult to recommend the manuscript for publication in its present form, mainly due to the following points:
1) Novelty
The title and summary suggest that the main novelty of this study lies in identifying SIT as the most important parameter for skillfully predicting the summer Barents-Kara sea sea ice extent. However, several previous studies (including some cited by the authors themselves (e.g., Day 2014b)) have reported similar results for the entire Arctic. At the moment, the study lacks a discussion section that relates the findings to the existing literature and clarifies how they advance beyond what earlier studies have already shown.
The statistical analysis of the physical drivers is a valuable contribution. However, the methods are unfortunately not described in sufficient detail to allow the reader to fully assess the underlying argumentation, and the discussion does not provide insight into the methodological limitations.
2) Unclear Methods
The methods are not described clearly enough to ensure reproducibility, or follow the authors argumentation.
For example, it is not clear which sea ice data are assimilated. The link to the data leads to an error message, and the reference could refer to either of two reanalyses - the one including sea ice parameters covers a different period than the study period of this manuscript. Furthermore, the statistical analyses presented in Figures 3-8 are not sufficiently described in the methods section. It is also not entirely clear what SINTEX-F3 refers to: in some places it seems to denote the general model setup, while in others (e.g., Figure 3) it appears to refer to an ensemble mean. A related lack of clarity concerns how the seasonal forecasts F2-SST, F3-SST, F3-3DVAR, and F3-SI were created. A table summarizing the models and assimilation techniques, ensemble sizes, and assimilated datasets would be very helpful here. In addition, several questions remain, such as: How was the re-forecasting in Section 2.3 performed? Which observations were used for the ACC calculation in Section 2.4? How was the regression analysis performed - over which time period, variables, datasets, etc.? How is the SIE index calculated?
3) Supporting the main findings
The main finding of the paper is that the initialization of SIT in SINTEX-F3 leads to the best sea ice extent predictions in the Kara-Barents Sea. However, the study also finds that the C3S re-forecast yields better SIE predictions in the Kara-Barents Sea than the SINTEX-F3 forecast. According to Table 1, none of the C3S models assimilate sea ice parameters, while the SIT, SIC, and snow data assimilated in SINTEX-F3 are taken from yet another reanalysis product. If a model without sea ice assimilation predicts the SIE better than a model with assimilation, the reason for performing the assimilation in the first place is not evident and the conclusion that SIT initialization has the most impact on SIE forecasts can not be drawn. In addition, the SINTEX-F3 model appears to be missing from Figure 3, which compares the different re-forecasts with the assimilation run. Furthermore, a reference run is missing from the analysis. All results supporting the findings are benchmarked solely against a persistence prediction. Including a SINTEX-F3 reference run (free run, without assimilation) should show that the assimilation does not degrade the model result, which can currently not be excluded.