the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 19 Nov 2025
Sea-ice ridges are a major component of Arctic sea-ice export through the Fram Strait
Abstract. This study presents seven years (2012–2019) of Arctic sea-ice draft observations from upward-looking sonars combined with coincident observations of ice drift velocity from four moorings located across the Arctic outflow in the Fram Strait at 78.83° N. The data set covers in total about 150 000 km of drifting Arctic sea ice, at a 1 m spatial resolution, providing one of the most extensive spatially referenced sea-ice draft records in the Arctic available today. Level ice makes up about 40–50 % of the ice cover, with modal ice thickness varying between 1 m and 2.5 m, and thicker level ice westward towards the east Greenland shelf. Using local level-ice thickness and a variable-threshold ridge detection algorithm, we identify and quantify the sizes of sea-ice ridges, including shallow ridges with keel drafts less than 5 m deep, often overlooked by traditional methods using a fixed threshold. The study highlights ridges as a significant component of the sea-ice cover, with keels covering some 20–30 % of the ice bottom and contributing 28–55 % of the total sea-ice volume. The typical spatial density varies from 6 to 9 individual ridges per kilometer of sea-ice, with approximately 3000 to 5500 ridges per month at each site. A westward increase in ridge frequency and coverage was associated with the differences in the origin of sea ice arriving at the mooring locations. Further, we show that shallow ridges comprise up to 80 % of all ridges and 35–45 % of the ridged ice volume, and thus play an important role in the sea-ice volume budget. Thus, shallow ridges deserve greater attention, especially given the ongoing changes in the Arctic sea-ice cover.
Competing interests: Sebastian Gerland is an editor in 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
(1240 KB) - Metadata XML
-
Supplement
(5411 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2025-5511', Alek Petty, 26 Jan 2026
-
RC2: 'Comment on egusphere-2025-5511', Anonymous Referee #2, 20 Aug 2026
Review of "Sea-ice ridges are a major component of Arctic sea-ice export through the Fram Strait"
Divine, D. V., Gerland, S., and Granskog, M. A. — egusphere-2025-5511
Summary of key findings and general review
This manuscript presents seven years (2012–2019) of spatially-referenced sea-ice draft observations from four moorings in the western Fram Strait, combining ULS draft with ADCP-derived ice velocity to reconstruct ~150,000 km of drifting sea-ice bottom topography at 1 m resolution. Using a variable-threshold (locally defined) ridge/level-ice detection scheme rather than the traditional fixed 2.5 m level-ice threshold, the authors quantify ridge frequency, keel-depth distributions, and areal/volumetric contributions of ridges — including shallow keels (<5 m) that are typically excluded by fixed-threshold methods. The central results — that ridges cover 20–30% of the ice bottom and contribute 28–55% of ice volume, that shallow ridges make up 50–80% of all ridges and 35–45% of ridged-ice volume, and that ridge frequency and volume increase westward — are valuable and, to my knowledge, represent one of the most complete spatially-resolved ridge climatologies available for this sector of the Arctic outflow.
The dataset itself is a major asset to the community: a nearly-decade-long, high-resolution, spatially-referenced record of this kind is rare, and the paper's core message (that shallow ridges are an underappreciated but volumetrically important component of the ice cover, especially as the pack thins) is well supported and important, particularly for ice–ocean drag parameterization and volume/freshwater budget work.
That said, I have a number of concerns that I believe should be addressed before publication, summarized as major comments below, with specific line-by-line points following. My main substantive concern is methodological: the paper does not adequately quantify how sensitive its main conclusions are to the definition of the residual "mixed/other" ice category, which is sizeable (order 25–40% of the record based on Fig. 3) and could plausibly contain ridged ice that is currently excluded from the ridge statistics. Beyond this, the paper is currently quite descriptive; several of its physical interpretations (source-region effects, atmospheric drivers of seasonality) are asserted rather than demonstrated, and closer engagement with the ice-ocean drag/ridge-geometry modelling literature would substantially increase its value to that user community. None of these are fatal flaws — the dataset and core findings stand on their own — but addressing them would considerably strengthen the paper.
Major comments
-
Data availability. The data availability statement currently reads only "The data for the study will be made publicly available by the time of its publication." Given how much of the paper's value lies in the dataset itself (a unique, spatially-referenced, decade-scale ridge climatology), I consider public archiving of the processed draft/ridge dataset (and ideally the ridge detection code) a precondition for publication, not an optional add-on. Please provide a DOI/repository link, or at minimum a firm commitment and timeline, before this goes further in review.
-
Definition and treatment of the "mixed/other" ice category. The three-way classification (level ice / ridges / mixed) leaves a residual "mixed" class that is explicitly acknowledged to include shallow rubble fields and possibly dynamically-formed ice that "could not be associated with the other two categories" (P6L174-176), and the authors themselves note in the Conclusions that "the contribution from deformed ice might be even larger" because of this (P13L389-391). Given that this residual class appears to make up a substantial share of the record (see Fig. 3, "Other" category), and that the paper's headline volumetric numbers are all conditional on how this class is treated, I think this deserves much more than a passing acknowledgement. I would like to see: (a) a quantitative breakdown of what fraction of the "mixed" class is level-ice-like vs. clearly deformed but undetected as a distinct ridge; (b) a sensitivity analysis showing how the reported ridge density/areal/volume fractions would change under different treatments of this class (e.g., a bounding scenario where all "mixed" ice is deformed); and (c) example transects (see point 3) showing what "mixed" ice actually looks like in the draft profiles, so a reader can judge for themselves whether the classification is conservative or generous.
-
Show representative transects in the main text. Figure S1 (referenced at P4L144) apparently shows example ridge/level-ice classifications, but this is relegated to the supplement. Given how central the classification scheme is to every downstream result, I think at least one figure of real draft transects — annotated with level ice, ridge, and "mixed/other" segments as actually assigned by the algorithm — belongs in the main text. This would also let readers visually assess artefacts such as the running-average smoothing (P4L135-138), the local threshold interpolation, and the fixed-vs-variable threshold comparison (currently Fig. S2).
-
More technical detail on processing artefacts (bias, aliasing). The paper states ADCP velocities were sampled at 20-minute intervals and used in a "double-weighted double-quadratic interpolation scheme" to produce a spatially uniform 1 m series (P4L104-106), but given that individual ridges can be only a few metres wide, I would like more discussion of how velocity variability within a 20-minute window (particularly during high-drift-speed or highly variable-drift events) could alias into the reconstructed spatial series — e.g., smearing or mis-locating ridge features, or biasing the derived ridge width/spacing statistics. Similarly, the draft accuracy discussion (P4L94-100) gives an overall accuracy of ~0.1 m but does not discuss whether errors could be systematically different for ridge keels vs. level ice (e.g., due to beam-footprint or tilt effects at steep slopes), which would matter for the reported keel-depth PDFs.
-
Deeper interpretation of the observed spatial/temporal patterns. The east–west gradient in ridge density and the seasonal cycle in keel depth are both attributed qualitatively to differences in sea-ice source region / deformation history and to winter atmospheric forcing (e.g., P8L241-247, P10L297-307), citing Sumata et al. (2023) and Krumpen et al. (2025) rather than demonstrating the link with the authors' own data. Since ice motion products (e.g., Tschudi et al., already used for Fig. 1) and reanalysis wind fields are readily available, some quantitative back-trajectory or correlative analysis (e.g., relating monthly ridge density/keel depth to upstream drift speed, residence time, or wind stress) would move this from a plausible narrative to a demonstrated mechanism, and would substantially strengthen the paper's contribution beyond a descriptive climatology.
-
Closer connection to ice–ocean/ice–atmosphere drag parameterization needs. The introduction motivates the study partly by its relevance to drag parameterizations (P2L49-51, citing Tsamados et al., 2014), but the results section does not return to this explicitly. Tsamados et al. (2014, and its supplementary material) identify specific unresolved parameters that limit current parameterizations — keel/sail spacing, keel-face angle or slope distribution, and macroporosity, among others. This paper already derives ridge density (i.e., spacing), keel-depth PDFs, and makes explicit macroporosity assumptions (P11L337-346); it would strengthen the paper considerably to explicitly report these in the form needed by drag parameterization schemes, and to discuss what is still missing (e.g., keel angle/shape, which cannot be derived from 1-D profiles) rather than leaving this only implicit in the ridge-orientation caveat (P4L145-147, P11L324-327).
-
Sail–keel comparison. The manuscript excludes ridge sails from the volumetric calculations and instead relies on a literature-derived assumption that sails contribute "not more than 10% of the total ice ridge volume or mass" (P11L354-357). I appreciate that co-located sail data are not available at these moorings and that this is a genuinely hard problem, but I would still like to see this discussed more critically — e.g., how sensitive would the total ridged-ice volume/mass budget be if the true sail fraction were, say, 15–20% instead of 10%? Even a short sensitivity discussion, or explicit flagging of this as a priority for future co-located (e.g., altimetry) validation, would be useful.
-
Overall, the results read as largely descriptive. Much of Section 3 presents time series and comparisons to prior literature without much synthesis beyond "our values are broadly consistent with study X." I don't think every one of my other major comments needs to be addressed to move the paper forward, but I would encourage the authors to select 1–2 of the above (my suggestions: the mixed-ice sensitivity analysis, and a more quantitative treatment of drivers) to elevate the paper from a valuable descriptive climatology to one that also demonstrates mechanism.
Specific comments
P1L20 — I find that the claim ridges are "arguably the least studied component of Arctic sea ice" is a strong statement to open on; a supporting citation (or softening to something like "a relatively understudied component") would be useful, especially since the following sentences already cite several ridge-focused studies.
P2L28-29 — I find that the oft-cited "~90% of sea-ice export" figure would benefit from a one-sentence caveat on the timescale/period it applies to (interannual variability in export fraction is itself substantial per Sumata et al. 2022, cited later).
P3L73-76 — I find that a compact site map inset with bathymetry contours (in addition to Figure 1) would help the reader relate the strong westward gradients reported throughout the results (e.g., water depth 2470 m at F11 vs 270 m at F14, Table 1) to the ridge statistics, since shelf proximity/water depth is a plausible additional driver of the west–east ridge gradient beyond source-region history.
P3L77-79 — I find that "minimal possible temporal gaps" is somewhat optimistic given Figure 2 shows quite substantial multi-month-to-year gaps at several sites (e.g., F13 2013–2015); I'd suggest rewording, or explicitly quantifying total data coverage as a percentage of the 7-year period per site.
P4L94-100 — I find that the accuracy discussion would benefit from explicitly stating whether the ~0.1 m accuracy figure is expected to hold at ridge keels (steep slopes, beam-footprint/tilt effects) as well as over level ice, since this bears directly on the keel-depth PDF fits in Section 2.3.
P4L104-106 — I find that more detail is needed on the "double-weighted double-quadratic interpolation scheme": what is being weighted, and over what window? Given the 20-minute ADCP sampling vs. 2 s draft sampling, some discussion of potential aliasing of ridge features into the reconstructed 1 m spatial series (see Major Comment 4) would be valuable here.
P4L107-111 — I find that Table 1's "km of ice" and "days of ice" columns partially address this, but it would help to also show, per site, what fraction of days had ULS-only vs. combined ULS+ADCP data, since this determines representativeness of the "effective" statistics reported later.
P4L118-122 — I find the description of the Rayleigh criterion clear, but a schematic (even reproducing a simplified version of the Ekeberg et al. 2015 figure referenced here) directly in the main text, alongside a real example transect (see Major Comment 3), would make the method much more accessible without requiring the reader to consult another paper.
P4L133-138 — I find that the specific choice of a 0.025 m⁻¹ gradient threshold and a 3 m running-average smoothing window are adopted from Wadhams and Horne (1980) without discussion of whether they are appropriate for this dataset's 1 m sampling and modern (generally thinner, more variable) ice conditions. A brief sensitivity check (e.g., how ridge counts change under ±50% variation in the smoothing window) would strengthen confidence in the method.
P4L142-143 — I find that the choice of hkmin = 2 m "to filter out ice rubble" is a fairly consequential threshold (it directly bounds the population of "shallow ridges" analyzed throughout) and deserves brief justification beyond the single sentence given — e.g., was this value tuned against the data, or taken from prior literature?
P4L145-147 — I find this an important and honest caveat (true keel widths cannot be inferred without ridge orientation), but as noted in Major Comment 6, I think its implications for the derived ridge-density/spacing statistics (which are used later as a proxy for model-relevant spacing parameters) should be discussed more explicitly rather than left as a single caveat sentence.
P6L174-181 — I find this section is where the "mixed" residual category is defined, and as detailed in Major Comment 2, this deserves substantially more quantitative treatment given its apparent size and its acknowledged potential to include undetected deformed ice.
P6L182-199 — I find the literature comparison of level-ice fractions useful, but quite long relative to its analytical payoff; a summary table (study, period, region, method, level-ice fraction) would let the reader absorb this comparison more efficiently than the current prose list.
P7L217-221 — I find this a strong, clearly supported point (fixed 2.5 m threshold rarely reached, so fixed-threshold ridge detection would undercount) — this is one of the better-substantiated claims in the paper and could be foregrounded more in the abstract/conclusions.
P8L224-227 — I find that the statement "the effective ridge density can be as high as 30 ridges per km" would benefit from context — is this a single extreme daily value, or a sustained condition? A brief note on the tail behaviour (e.g., what fraction of days exceed 20 ridges/km) would help.
P8L241-247 — I find this explanation of the westward increase in ridge density (attributed to source-region/deformation-history differences) plausible but not demonstrated with the authors' own data; see Major Comment 5 for a suggested quantitative approach (back-trajectories or correlation with available ice-motion/atmospheric products).
P9L257-264 — I find the comparison with Ekeberg et al. (2014)'s >5 m-keel counts convincing and a nice consistency check; I would encourage a similar quantitative cross-check for the shallow-ridge fraction, if any independent estimate exists, to validate that part of the method as well.
P9L262-264 — I find that reporting the decline as "statistically significant only at F12" without also reporting the p-values or trend magnitudes (with confidence intervals) for the other three sites makes it hard to judge how close they came to significance; consider adding these to Table 2 or the text.
P9L280-282 — I find the speculation that the deepest keels (25–30 m) could be "small icebergs from the Greenland ice sheet" interesting and worth retaining, but since this bears on the tail of the reported keel-depth distribution, I would like to see even a minimal discussion of what criteria (if any) could in principle distinguish an iceberg fragment from a genuine pressure ridge keel in a 1-D draft profile (e.g., width, shape, isolation from surrounding ice), even if the authors ultimately conclude this cannot be done with the current data.
P10L297-307 — I find this discussion of atmosphere-driven seasonal ridge formation plausible but, again, presented as narrative rather than demonstrated; see Major Comment 5.
P10L308-314 — I find the link to the well-documented 2018 outflow minimum (Sumata et al., 2022) a nice, specific, testable example of interannual coherence across sites — this is one of the more compelling pieces of physical interpretation in the paper and could be a model for how the source-region/seasonal arguments elsewhere could be similarly substantiated.
P11L324-327 — I find the "anisotropy" assumption used to justify treating linear-profile areal/volume fractions as orientation-invariant reasonable as a first-order argument, but it would help to state explicitly what averaging length scale is assumed sufficient for this to hold, and whether this has been tested (e.g., via bootstrap over sub-segments of the record) rather than simply asserted.
P11L337-346 — I find the treatment of macroporosity as an envelope (0–30%) a good, transparent way of expressing this uncertainty, and would suggest the same envelope approach be considered for the sail-fraction assumption (Major Comment 7) and possibly for the "mixed" ice class (Major Comment 2).
P11L347-353 — I find the shallow-ridge volume contribution (~40% for keels <5 m, rising toward F11) one of the paper's most important and useful findings; I would suggest it (along with the P7L217-221 finding) be given more prominence in the abstract, since together they make the strongest case for the paper's central "shallow ridges matter" argument.
P11L354-357 — I find the 10% sail-volume assumption under-discussed relative to its role in the volume budget; see Major Comment 7.
P12L358-361 — I find this paragraph (fractional coverage/volume decline, significant only at F11) would benefit from the same treatment suggested for P9L262-264 — report trend magnitudes/CIs for all four sites, not just flag which one is significant.
P13L389-391 — I find this sentence ("the contribution from deformed ice might be even larger...") is really the crux of Major Comment 2, and I think it deserves to be a quantified sensitivity analysis in the Results rather than a single caveat sentence in the Conclusions.
Table 1 — I find the inclusion of both "days" and "effective days of ice" (weighted by concentration) very useful; it might help to also add a column for the fraction of the record with combined ULS+ADCP coverage (see P4L107-111 comment above).
Table 2 — I find this table, summarizing per-site ridge count, coverage, volume, density, and keel depth, to be one of the most useful elements of the paper for other researchers (e.g., for model parameterization, Major Comment 6); I would suggest adding the shallow-ridge (<5 m) fraction and the assumed sail-volume contribution explicitly as additional columns here, consolidating scattered results into a single reference table.
Figures 3, 5, 7, 9, 10 — I find these clear and effective, but note that none of them show an actual example draft transect with classified segments (Major Comment 3); I think one additional figure of this kind, in the main text, would substantially aid interpretability of everything that follows.
Citation: https://doi.org/10.5194/egusphere-2025-5511-RC2 -
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 1,841 | 990 | 156 | 2,987 | 365 | 138 | 152 |
- HTML: 1,841
- PDF: 990
- XML: 156
- Total: 2,987
- Supplement: 365
- BibTeX: 138
- EndNote: 152
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This paper presents seven years (2012 to 2019) of spatially referenced sea‑ice draft observations from upward‑looking sonars, combined with ADCP‑derived ice‑drift measurements at four moorings in Fram Strait. Using these data, the authors reconstruct meter‑scale spatial profiles of sea‑ice draft and quantify ridge frequency, keel‑depth distributions, and the areal and volumetric contribution of ridged ice to Arctic sea‑ice export.
Overall, the long‑term mooring record combined with spatial reconstruction is particularly valuable, and the focus on shallow ridges is timely given ongoing Arctic sea‑ice thinning. The main results are convincing and relevant for sea‑ice mass‑balance studies, ice-ocean drag parameterizations, and satellite-retrieval insights.
Main Comments
The title and main stated objective doesn't quite feel right to me. The real strength of the study to me lies in quantifying the contribution of sea‑ice ridges to regional sea‑ice mass balance in the region, which is articulated well at the beginning of the conclusion. Given this, I would encourage the authors to make this contribution/advance more clear in the title and framing.
In addition, this dataset represents an important modern complement to historical Fram Strait observations (largely from submarine sonar). Clearer and more quantitative comparisons with these prior estimates would significantly strengthen the manuscript, the introduction of proper campaigns and estimates is introduced quite sporadically and vaguely.
The manuscript often refers to “ice thickness,” yet the analysis is based entirely on ice draft I think! This is confusing and should be corrected throughout. If thickness is inferred or discussed conceptually, the assumptions and limitations should be clearly stated.
While the authors appropriately acknowledge the limited record length and data gaps, discussion of declining ridge numbers and shallower keel depths should be framed more explicitly as suggestive, rather than a robust trend analysis of any kind, given the significant interannual variability.
Adaptive approaches for estimating level‑ice surfaces are not really new, despite the framing of this paper. Relevant examples include methods based on low elevation change over a given along‑track distance (e.g., Wadhams and Horne, 1980) and modal elevation detection (e.g., Williams et al., 2015; sail analysis in Petty et al., 2016). I think it's better to make that much clearer.
The distinction between ice rubble and ridges (L143) is introduced rather briefly, despite being a potentially important conceptual and methodological point. A clearer explanation is needed I think!
The discussion in Section 3.4 is overly qualitative and vague. This section would benefit from clearer, more quantitative comparisons with existing studies of ridge fraction and ridge volume, ideally incorporated directly into figures rather than solely discussed in the text, to better contextualize these data.
The manuscript reports both ridge count and ridge density. The rationale for including both metrics and how they differ in interpretation should be explained explicitly or dropped, I didn't quite understand why ridge count was included.
The macroporosity discussion is confusing to me. Macroporosity affects ridge density rather than geometric volume I think. While excluding air pockets may be justified in terms of actual ice volume, I don't really understand if this was needed here..
Minor Comments
Figure 2: Since longitude is fixed for each mooring, a table might be more efficient. “2019-2019” should read “2012-2019.”
Figure 3: Clarify in the caption that fractions are calculated relative to ice‑covered periods only (open water excluded). Clarify the “mixed” category in the legend.
Figure 4 is difficult to interpret. Violin plots may provide a clearer representation of the thickness/draft PDFs. The meaning of the black line is unclear, and the phrase “maxima in modal” is confusing.
Figure 5: A similar analysis for daily ridge density within each year could be useful. While there are hints of seasonality, it is difficult to assess this from the current presentation. The added value of showing daily data as plotted should be clarified. Note that daily ridge densities are calculated over short drift distances (typically <20 km), contributing to high variability.
Figure 6: The explanation of symbols and intervals is unclear. The caption should clearly define that circles and horizontal lines show median ridge densities and the 67% prediction intervals?
Figure 10 caption: Clearly state which macroporosity assumption is used for the reported mean values and trends.
L11-12: This statement is speculative (though reasonable) and likely too strong for inclusion in the abstract.
L20: The claim that ridges are the “least studied component” of Arctic sea ice is too strong given extensive submarine, airborne, and growing satellite-based papers. Consider softening or qualifying.
L24: This sentence appears unnecessary given the breadth and maturity of the existing literature.
L29: Please verify this statement.
L33: When stating that the dataset is “one of the most extensive,” briefly clarify the basis for comparison (e.g., time span, spatial coverage).
L42: Please reword for clarity. More explicitly explain that ice velocity is required to convert time‑series draft measurements into spatial profiles.
L57: Too vague, please clarify whether this refers to ridging studies in general or specifically to keel measurements.
L74: Clarify whether the quoted ULS footprint diameter refers to the nominal instrument depth or varies with actual mooring depth.
L87: If ASL is a company, this detail may be unnecessary. Focus instead on how the processing is done and how the data are provided.
L89: Consider reporting ridge numbers per day or per hour, which may be more informative.
L91: A citation is needed; the stated value seems quite low to me.
L95-100: The discussion of draft uncertainty conflates instrument precision and interpretive uncertainty. Clarify that ~10 cm represents an estimated total uncertainty, including both random and systematic components.
L123-129: Could the same effect be achieved by simply lowering the threshold?
L200-205: When comparing level‑ice fractions to historical submarine studies, briefly note differences in season or sampling that limit direct comparison.
L272: “This analysis” is not an effective sentence opening, please rephrase.
References:
Petty, A. A., M. C. Tsamados, N. T. Kurtz, S. L. Farrell, T. Newman, J. P. Harbeck, D. L. Feltham, and J. A. Richter-Menge (2016), Characterizing Arctic sea ice topography using high-resolution IceBridge data, The Cryosphere, 10 (3), 1161–1179, doi:10.5194/tc-10-1161-2016.
Wadhams, P. and Horne, R. J.: An analysis of ice profiles obtained by submarine sonar in the Beaufort Sea, J. Glaciol., 25, 401–424, 1980.
Williams, G., Maksym, T., Wilkinson, J., Kunz, C., Murphy, C., Kimball, P., and Singh, H.: Thick and deformed Antarctic sea ice mapped with autonomous underwater vehicles, Nature Geosci., 8, 61–67, doi:10.1038/ngeo2299, 2015