Summary

This paper presents a new technique for understanding fractional sea ice coverage in the Arctic, by developing a Linear Ice Fraction (LIF) product from ICESat-2 ATL07 data. It’s great to see the high-resolution capabilities of ICESat-2 being used for this novel application. The paper was well-structured and enjoyable to read, and I have just a few comments to address prior to publication.

Comments

 L6-7: The statement comparing winter and summer biases is a little misleading. Without the further context provided in the paper, it reads as if summer biases are consistency smaller, rather than skewed by the NT algorithm. It would be useful to highlight here that in most cases, summer biases are larger. See also my comments on Section 3.2.

L12: “…measurements of the sea ice surface **with PM data** to enhance…”. IS2 LIF is still dependent on PM SIC data.

L26: Quantify “narrow”, because it’s an important point for justifying why LIF are useful 

L41: I disagree with the introduction of LIF being an independent measure of sea ice presence. The LIF is developed using IS2 data that rely on a PM concentration product to determine sea ice presence, so LIF is more complimentary than independent. Please make this clear through the paper.

L50: “**Then,** using…”

Table 1: This might be an EGU issue, but the date formatting in the table wasn’t great to read

Table 1, row 2, column 6: Remove “–“

Table 1, row 5, column 6: Do you mean 450 and 430?

L76: “…advanced **over the satellite period**…”

L115: “instrument” > “instruments”

L120: “utilizing” > “utilizes”

L135: The OIB acronym hasn’t been defined

L142-143: What do the authors mean by "outliers", and why this becomes more of an issue when MPF is greater than 50%?

L149: Remove “(box)” ?

L152: “these products” > “the PM products”

L153: Should the “(2)” say “(Figure 2a)” ?

L157-158: I couldn’t make much sense of this sentence. What do the authors mean by “strong similarity in patterns” ?

Section 3.2: The results here are particularly interesting, and I’d like a bit more information on why PM products exhibit a positive SIC bias in summer, and why it’s larger than winter. In Section 1 the authors explain that melt ponds on the sea ice appear radiometrically similar to open water, so if anything I’d expect PM to underestimate SIC compared with imagery. It would be great to add some brief text relevant to this in the abstract and Section 1 too.

L168: “Figure 2b **and Table 2**”

L172: The NT2 acronym hasn’t been defined

L172-173: Could the authors explain why they find this interesting? Because the changes to NT2 weren’t intended to account for ponding.

L182: An IS2 footprint of 10 m was stated in Section 1, and 11 m here

L187-188: What is meant by “likely recorded”? And what impact would this have on the IS2 products?

General: I suggest each author has another readthrough and checks for clarity and accuracy in the text. I noticed some issues with grammar/typos/formatting (citations and symbols).

This paper provides an evaluation of passive microwave (PM) sea ice concentration (SIC) estimates using classified airborne visual imageryfrom NASA’s Operation IceBridge (OIB) and introduces a new ICESat-2-based Linear Ice Fraction (LIF) dataset. The comparison includes classified imagery from four satellite imagery scenes (Sentinel-2 and WorldView). The study is a follow-up to a previous submission, which I also reviewed. In response to that feedback, the authors have now split the original manuscript into two parts to allow for a more focused discussion. This is a good approach, but I am somewhat uncertain about how well it has worked in practice.

In this Part 1 study, much of the paper is dedicated to presenting classified airborne imagery from Operation IceBridge, which is used to highlight biases in PM SIC data. However, we already know from past research (e.g., Kern et al.) that PM SIC products contain biases, and the comparisons here mostly serve to confirm those findings. While it is valuable to have more insight into these biases, the OIB imagery is not actually used to assess the new ICESat-2 LIF dataset, even though LIF appears to be the main focus of the study. More than half of the paper is therefore spent reaffirming known PM SIC biases, rather than contributing directly to the LIF validation. My impression from the earlier review was that the authors planned to expand the comparisons with coincident imagery, ideally incorporating OIB data that overlaps with ICESat-2 tracks. There were indeed OIB cal/val flights in 2019 and a summer calibration campaign in 2022 that could have been utilized for this.

The second half of the paper presents the LIF analysis, where the authors introduce four Sentinel-2 and WorldView scenes to evaluate ICESat-2-derived LIF estimates. However, they only show one example in detail, and then provide a summary table, which makes the evaluation of this new dataset feel quite limited. A major advantage of working with a small number of high-resolution scenes should be the ability to explore different surface types, environmental conditions, and classification performance in depth, but this aspect is underdeveloped. For example, it would be interesting to analyze how different ATL07 classification types influence LIF retrievals, how well the drift correction works, or how dark leads, which have known retrieval issues and are no longer included in the sea surface height retrievals (Kwok et al., 2021), affect the results. Similarly, while drift correction is applied, the fact that only four scenes are used means that manual adjustments could have been done instead. Recent studies such as Koo et al. (2023) and Liu et al. (2025) analyzed 17-18 coincident Sentinel-2 scenes and manually adjusted them, so this approach should be considered. Having at least one summer scene and scenes from other parts of the Arctic and the Southern Ocean would be highly beneficial.

Another aspect that could be explored in more detail is the definitional differences between PM SIC, optical imagery SIC, and ICESat-2 LIF. This is only briefly mentioned in L219, where the authors state that “new ice that appears gray in color is considered ice for SIC and LIF calculations.” However, this is a significant issue that deserves more discussion. How do different ice types (open water, leads, gray ice) compare across these datasets? Addressing this would improve the interpretation of the results.

Specific Comments:

• L142: The authors state that they examine images where MPF ≤ 50% in summer to avoid outliers and misclassified images in the unsupervised analysis. However, wouldn’t it be more informative to include scenes with high melt pond fractions? Why are these considered outliers?
• L151: The paper states that PM SIC products on average overestimate SIC, but there is significant spread.
• L160: It is mentioned that PM SIC products have a bias on average, but they are highly variable. However, OSI SAF and ASR do not appear to be biased on average—can this be clarified?
• L218: The authors mention “other pixels” in the text, but in Figure 4, the term “new ice” is used instead. This should be consistent.
• L225: How well does the drift correction work? This was not very clear in the methods section. Given that only four scenes are used, why wasn’t a manual adjustment tested as an alternative?
• L236: The statement that the May 7, 2022 image represents an area of highly fractured sea ice that four PM SIC products classify as completely ice-covered is interesting. It would be useful to include a visual example of this to illustrate the discrepancy.
• Figure 3: This figure is hard to interpret, and a better way to display this data should be considered.
• ASI Data in Figure 3: Why is ASI data only shown in the 0-5% SIC interval? Shouldn’t all products be included in every bin?
• All Sentinel-2/WorldView scenes: Since there is space, why not show all four Sentinel-2/WorldView comparisons with ICESat-2 LIF? The Figure 4 scene also looks quite small—is this the full scene, or just a zoomed-in version?
• Comparison with MODIS/Landsat PM Evaluations: The authors reference previous Kern et al. studies but do not clearly compare their results. How do the biases in this study compare with those found in MODIS and Landsat SIC evaluations? Are there any new insights gained from the OIB comparisons?
• Methods and Results Organization: The methods and results are somewhat mixed together, which makes it harder to follow. It may be better to fully separate them, even if this shortens the results section.

References

Kern, S., Lavergne, T., Notz, D., Pedersen, L. T., Tonboe, R. T., Saldo, R., and Soerensen, A. M.: Satellite Passive Microwave Sea-Ice Concentration Data Set Intercomparison: Closed Ice and Ship-Based Observations, The Cryosphere, pp. 1–55, https://doi.org/10.5194/tc- 2019-120, 2019.

Kern, S., Lavergne, T., Notz, D., Pedersen, L. T., and Tonboe, R.: Satellite passive microwave sea-ice concentration data set inter-comparison for Arctic summer conditions, The Cryosphere, 14, 2469–2493, https://doi.org/10.5194/tc-14-2469-2020, 2020.

Koo, Y., Xie, H., Kurtz, N. T., Ackley, S. F., and Wang, W.: Sea ice surface type classification of ICESat-2 ATL07 data by using data-driven machine learning model: Ross Sea, Antarctic as an example, Remote Sensing of Environment, 296, 113726, https://doi.org/10.1016/j.rse.2023.113726, 2023.

Kwok, R., Petty, A. A., Bagnardi, M., Kurtz, N. T., Cunningham, G. F., Ivanoff, A., and Kacimi, S.: Refining the sea surface identification approach for determining freeboards in the ICESat-2 sea ice products, The Cryosphere, 15, 821–833, https://doi.org/10.5194/tc-15-821-2021, 2021.

Liu, W., Tsamados, M., Petty, A., Jin, T., Chen, W., and Stroeve, J.: Enhanced sea ice classification for ICESat-2 using combined unsupervised and supervised machine learning, Remote Sensing of Environment, 318, 114607, https://doi.org/10.1016/j.rse.2025.114607, 2025.