the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 23 Feb 2026
A Reanalysis of the Arctic sea ice cover over the satellite era utilising summertime observations of SIT
Abstract. Climate change has significantly affected the Arctic over the satellite era, with sea ice undergoing a substantial decline. While changes in sea ice concentration (SIC) and sea ice extent (SIE) have been widely studied, sea ice thickness (SIT) and volume (SIV) remain less well constrained due to limited observations. Quantifying SIV trends is particularly important for understanding sea ice changes in response to climate change. Here we present three reanalyses that assimilate different combinations of SIC and SIT products, including year-round SIT observations, into the CPOM-CICE sea ice model, which incorporates advanced parameterisations for melt ponds, form drag, and rheology. Assimilating NASA Team SIC together with year-round SIT substantially improves SIT estimates, and year-round SIT assimilation outperforms winter-only SIT assimilation, even at the end of winter, by better initialising the growth season. Comparison with four independent observational datasets and PIOMAS identifies the best-performing reanalysis, which we analyse for 2010–2020 to diagnose model deficiencies. The model exhibits a seasonally compensating bias cycle: excessive freeze-up and overly thick, consolidated ice in autumn and winter lead to elevated extent and thickness and a suppressed marginal ice zone in spring, while enhanced late-summer melt offsets these errors, yielding September extents close to observations but with anomalously thick ice packed against the Canadian Arctic Archipelago. This suggests that misrepresentations of ice growth, lead formation and refreezing, marginal ice zone dynamics, mechanical redistribution, and melt timing interact to obscure errors in concentration and thickness. Additionally, our best performing reanalysis also shows the thickest ice is less thick and more evenly distributed across the central Arctic in the 2010s. This reanalysis provides new insight into recent Arctic sea ice change and its underlying processes, as well as identifying key deficiencies in the sea ice model physics which can be a focus for future model development.
Competing interests: At least one of the (co-)authors is a member of the editorial board of The Cryosphere.
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.- Preprint
(6665 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-742', Anonymous Referee #1, 08 Jun 2026
-
RC2: 'Comment on egusphere-2026-742', Anonymous Referee #2, 27 Aug 2026
General comments
The manuscript “A reanalysis of the Arctic sea ice cover over the satellite era utilising summertime observations of SIT” presents an interesting sensitivity study with the potential to demonstrate the value of assimilating year-round Sea Ice Thickness (SIT) observations to improve sea ice reanalyses and better understand sea ice model biases.
The intended purpose of the reanalyses presented in the paper could, however, be clarified. Is the primary objective to assess the impact of different assimilation strategies and support the development of future sea ice reanalyses, or to provide a dataset for broader scientific applications such as climate change and Arctic variability studies?
The manuscript would also benefit from a stronger focus on CASIRA-n. In particular, the behaviour of CASIRA-b prior to 2010 is not sufficiently explained, making it difficult to interpret some of the comparisons presented later in the paper.
Results and evaluation
The Results section could be better organized to strengthen the overall narrative. More specifically, the evaluation of the reanalysis should not rely solely on comparisons with in situ SIT observations but should incorporate the broader set of variables and diagnostics presented throughout the Results section.
One possible structure could be:
- Demonstrate that assimilating year-round SIT observations, independently of SIC assimilation, improves the behaviour for all parameters and eventually improves the agreement with independent in situ SIT measurements.
- Show how the improved reanalysis helps identify and understand the underlying model deficiencies, thereby providing guidance for future developments of sea ice reanalyses, potentially including fully coupled ocean-sea ice systems.
- Demonstrate that the combined assessment of observation fit and model behaviour contributes to a more comprehensive evaluation of the reanalysis and increases confidence for future users.
Forcing and validation datasets
The choice of the forcing and validation datasets deserves further justification.
- Why were additional available in situ observations not included in the validation?
- Why were independent datasets, such as ICESat-2 observations, not used for evaluation?
- A spatial density map of the OIB observations would also be useful to assess the representativeness of the validation dataset.
Similarly, the choice of ocean forcing is not sufficiently justified. The rationale for using this dataset should be explained in more detail. In addition, I was unable to locate the reference corresponding to Ferry et al. (2011), and this should be verified.
Model and reanalysis description
The description of the model configuration and reanalysis setup could be made more precise.
- Spatial resolution: the reference to the ORCA1 grid may not be sufficiently clear for all readers. A brief explanation and/or an appropriate reference would be helpful.
- Temporal resolution: the temporal frequency of the reanalysis products should be stated explicitly (e.g., monthly means or another output frequency).
Data assimilation methodology
The data assimilation methodology should be described in greater detail.
- The system description could provide more information on the control vector and on the volume conservation constraints applied when only SIC is assimilated.
- Section 2.3 appears to mix the description of the observations with that of the observation operators, which makes the workflow difficult to follow.
- For example, SIC appears to be assimilated daily, whereas SIT is assimilated monthly. Separating the descriptions of the assimilated observations and the observation operators would provide a clearer understanding of the methodology and the associated processing steps.
Presentation of the figures and results
The way the figures are discussed in the text could be improved.
It may be easier for the reader to first describe the characteristics and behaviour of the observations and then discuss how the reanalysis reproduces, explains, or deviates from these observational features. A stronger focus on CASIRA-n throughout the Results section would also help improve the clarity of the narrative.
Ocean coupling
The absence of coupling with the ocean should be discussed more explicitly.
The current framework does not account for variations in ocean temperature and heat transport associated with ocean circulation, which may influence sea ice evolution. This limitation, together with its potential implications for the interpretation of the results, should be clearly acknowledged in the Discussion section.
Detailed Comments
- L125: The cited reference is no longer accessible, making it difficult to identify the ocean reanalysis that was used. Please clarify which product was employed and explain why more recent reanalyses (e.g. TOPAZ https://data.marine.copernicus.eu/product/ARCTIC_MULTIYEAR_PHY_002_003/description, GLORYS12, GREP) were not considered.
- L293: WIN-CS2 does not appear to be displayed in Figure 1. Please verify or clarify.
- L294: The text refers to a linear fit, but this metric is not reported in Table 2. The table includes a linear regression indicator, which is not described in the methodology or results. In addition, the best-performing metrics are highlighted in bold. Does this also apply to the regression metric? Both the text and Figure 1 suggest that CPOM-CICE provides the best fit, but this is not immediately apparent from Table 2.
- L295: What is the impact of assimilating SITD in addition to SIT? Introducing SITD at this stage tends to blur the main message of the paper. It would be helpful to show SIT distributions for all experiments alongside the AYR observations to better assess the respective impacts.
- L296: The reported correlation and RMSD values appear to vary depending on sea-ice thickness categories (e.g. >3 m and <3 m). These differences should be quantified to support the discussion.
- L324: The statement referring to “11 out of 14 cases” is unclear. I am unable to identify how this conclusion is derived from Table 2.
- Figure 2: Can the authors provide an explanation for the behaviour of CASIRA-b? The reanalysis shows a substantial positive bias relative to CPOM-CICE before 2010. If this behaviour is related to the assimilation of Bootstrap SIC, this connection should be stated explicitly and discussed when commenting Fig2.
- L359: This statement appears to repeat the discussion already provided at L346 and could probably be shortened or consolidated.
- L370: Trends have already been discussed in relation to Figure 4. Consider specifying that this paragraph focuses on trends in the seasonal cycle of SIE to avoid repetition.
- L377: The discussion refers to years 1988 and 1981, but these years are not clearly identified in the Figure 5 legend. While the 1988 curve can reasonably be inferred from the abrupt January-February change, the 1981 curve is difficult to distinguish. Please indicate more clearly that the 1981 feature is visible in CASIRA-b.
- L388: The pronounced increase in SIT observed in CASIRA-b, which is also evident in Figure 2, is not sufficiently explained. This behaviour should be discussed when introducing Figure 2, rather than only being mentioned later.
- L388: Typographical error: “an” instead of the current wording.
- L405: CPOM-CICE is an model experiment, not a reanalysis. Consider replacing “in all reanalyses” with “in all model experiments” (or equivalent wording).
- L497: The sentence may be missing a verb.
- L500: Is this interpretation supported by ice transport diagnostics? If such diagnostics are available, they should be presented or referenced to support the argument.
- L507: Does the December SIT distribution of CASIRA-n correspond to the SITD signal derived from CryoSat-2? Clarifying this relationship would help interpret the results.
Citation: https://doi.org/10.5194/egusphere-2026-742-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 985 | 637 | 71 | 1,693 | 121 | 161 |
- HTML: 985
- PDF: 637
- XML: 71
- Total: 1,693
- BibTeX: 121
- EndNote: 161
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The study “A Reanalysis of the Arctic sea ice cover over the satellite era utilising summertime observations of SIT” presents four different model experiments, which are compared to different observations, satellite products and other model runs, with the aim of identifying the models physical deficiencies. To me it appears like the authors are trying to do too many things at once. I could see that there are two studies in the current paper. One focusing on assimilation technic, looking at the thickness distribution and the assimilation period and one looking at the physical drivers of the marginal ice zone. Both studies however would need some additional model experiments and some heavy restructuring of the text. There are four major points that I think need to be address in relation to this:
1) Formulate a clear question you want to answer
Currently there is no clear question stated, nor answered. From the topics that are touched app on in the introduction I would assume that the paper answers some question in regards to assimilation technic, in that case I think the choice of model runs is questionable (see point 2), or that the study aims at finding physical drivers that are missing in the model set up to simulate a correct marginal ice zone (MIZ). For the second point I am also missing some experiments (see point 3), but also a clear definition of the MIZ and an ocean ice coupled model. Many effects in the MIZ are driven by the ocean and most state of the art sea ice modeling systems are using at least sea-ice-ocean coupled models, which the authors seem to be aware of, since they are citing them in the introduction. So, to add to the current discourse it would be crucial to run a sea-ice-ocean coupled model.
2) Choose the model runs accordingly
First of all, most studies cited as state of the art sea ice assimilation studies in the introduction are running at least sea-ice-ocean coupled models. I am aware that a fully coupled model is heavy to run, but by running a stand alone sea ice model, forced with a ocean climatology I wonder what the missing ocean forcing does to the sea ice state. This is not discussed nor even mentioned in the discussion. Further do the model runs used in the study appear a bit random. What does CASIRA-b add to the study? Even the authors them self seem not too interested in it, as they don’t mention it at all in the discussion. Finally I am wondering why the assimilation technics vary between CASIRA-n and CASIRA-d. Using the same assimilation technic would allow to actually discuss the influence of summer vs. only winter assimilation, but like the study is set up right now the differences might originate from either the data, the assimilation technic, or the assimilation period.
3) Support your findings with the results from the model runs
The study of the MIZ is interesting and highly relevant, however I am currently missing the support for the findings. The discussion for example states that the SIV overestimation is due to too little snow (line 472). But I can not see any results supporting this findings. I could also see other factors than the snow thickness effecting the SIV (for example the drag parametrization, or the ocean forcing). I would expect some sensitivity study or a like. As the study is build up right now the physical drivers might be a good guess, but I am not seeing any evidence for the findings in the study. Furthermore I would expect the ocean drivers to be quite important for the formation of the MIZ, but I am currently missing a thorough discussion of what the missing ocean forcing might do to the modeled MIZ.
4) Choice of data
I think it would be beneficial to actually use more independent data for the validation. The BGEP data is commonly used, but also spatial limited. An addition of, for example the Fram Strait observatory ULS data, would add a lot to the discussion. And also for the SIE analysis I would recommend another more independent data set, as for example the MASIE sea ice extent, which could be used to calculate the integrated ice edge error from the different experiments.
Finally I am missing a discussion of the short come of the used data sets. Which benefits/short comes do ULS vs. air born data have? Which questions can we answer with them, which not? How reliable is PIOMAS? What are the differences between AYR-CS2 and WIN-CS2? What effect might this differences have on the assimilation?
After mentioning my main concerns I would also like to mention that I do think that there is some potential in the paper, as already stated in the introduction paragraph. Furthermore, I liked the approach of using assimilation studies to evaluate sea ice models, appreciated the detailed description of the validation methods and the fact that the authors both evaluated the model sea ice volume and thickness.
Minor comments:
How is the assimilation state vector set up? Are they similar for all assimilated runs?
Color bars: Make them colorblind friendly
Figure 2g: It looks like the data sets are averaged over different periods and you compare the much shorter ULS with the longer model runs. This is misleading.
Figure 8: Why is WIN-CS2 not included?