Review Taelman et al., (2026) “Empirical Evidence of overestimated Ku-Band sea ice radar freeboards in satellite altimetry” 

This manuscript makes an important and timely contribution to the ongoing debate surrounding the validity of the snow-penetration assumption in Ku-band satellite radar altimetry. By analyzing the temporal evolution of spaceborne laser (ICESat-2) and radar (CryoSat-2, Sentinel-3) freeboards over immobile landfast first-year ice (FYI) and multi-year ice (MYI) off the Greenland coast, Taelman et al. provide one of the first empirical, multi-sensor, in situ–supported assessments of when and under what conditions Ku-band freeboards become positively biased. The paper is carefully structured, the literature review is comprehensive, and the combination of field measurements with satellite time series represents a genuinely novel methodological approach. The finding that radar freeboard over FYI generally tracks the snow–ice interface during cold-season conditions; while exhibiting positive biases under saline snow or snowfall events - is both substantive and well-supported by the data. The implications for the Copernicus CRISTAL mission are clearly articulated and add strong relevance to the work.

That said, several methodological choices require additional justification, and a number of analytical gaps need to be addressed before the manuscript is ready for publication in The Cryosphere. The major concerns center on: (1) the treatment of multi-sensor sampling differences in the time series analysis; (2) the basis for the 100 km ATL10 lead tie-point accumulation distance; (3) the representation of snow depth uncertainty from LaKu; and (4) several insufficiently substantiated concluding statements. These are detailed below: 

Specific Comments: 

• Line 45: It would be very useful to link the citations directly to the specific snow dielectric properties they refer to. Consider placing the relevant citations immediately next to each property listed. 
• Line 50: Can you cite these studies in this line? 
• Introduction: The authors did a fantastic job with a very comprehensive literature review and stated the objectives very clearly. I would suggest re-emphasizing the methodological uniqueness of this study in a bit more detail towards the end of the introduction. 
• Line 161: Citing the study that mentioned the ~17 m diameter could be helpful since there have been other studies that claim even finer footprints for ICESat-2. 
• Line 188: Is there an approximate distance between the in-situ measurements and the ROIs that is used to quantitatively define "proximity"? 
• Line 190: It would be worthwhile to provide the uncertainty estimates for these instruments over sea ice. 
• Line 195: Please clarify the spatial distribution of the snow pits with respect to the FYI and MYI ROIs.
• Line 295: Is it possible to provide a map illustrating examples of these false leads? Additionally, please provide statistics on the extent of these false leads compared to each dataset, and quantify the impact that excluding them had on the overall measurements. 
• Section 4.5: The approach used to fill gaps in the ATL10 freeboard is interesting, but the rationale needs further justification. What was the basis for selecting a 100 km radius instead of, for example, 10 km for accumulating lead tie points? Were other distances tested to determine the optimal threshold? Furthermore, the inverse distance squared weighted (IDW) interpolation method might be too simplistic given the complex ice dynamics typical of near-shore environments; this limitation should be discussed. 
• Section 4.6: While constructing this time series for laser and radar freeboards provides a more comprehensive picture, it is difficult to confidently assume that all tracks intersecting the ROIs within the study period are similar enough to reconstruct a single time series. Given the fine spatial resolution of IS2, one can expect varying snow distributions between tracks and differences among beams. Similarly, the spatial distance between the CS2 and S3 tracks could result in observations of different snow distributions. I recommend providing further analysis to account for these spatial and temporal variabilities and to ensure consistency between the instruments. 
• Section 4.7.2: Using snow depth from LaKu is a positive step away from relying on climatology or models. However, did the authors consider the uncertainties inherent in these approaches, as noted in recent validation studies (e.g., Fredensborg Hansen et al., 2024; Saha et al., 2025)? Given that these uncertainties are often quite significant even over landfast ice (Saha et al., 2025), I suggest discussing how they might impact the subsequent freeboard and sea ice thickness derivations. 
• Figure 4 (Salinity): One snow pit appears to have a significantly different salinity profile (median ~4 PSU). Is there a physical explanation for this discrepancy, and what does it suggest about the snowpack on that specific FYI ROI? 
• Figure 4 (Hardness): The IQR for one of the FYI snow pits ranges widely between F and K (with a median of 1F). Is there a particular reason why the hardness values varied so significantly in this instance?
• Figure 5: This plot could potentially be simplified and made more readable by separating the ice and snow depth data into two distinct panels.
• Line 475: How were the uncertainty values for the snow depth and ice thickness measurements from the MagnaProbe and GEM determined? 
• Lines 490–495: The dichotomy presented here regarding the laser freeboard time series is interesting and warrants further discussion regarding the physical mechanisms that could potentially explain this phenomenon. 
• Lines 515–524: When comparing the time series between laser and radar freeboards, the inherent difference in footprint size must be carefully considered. I encourage the authors to include a discussion on how these differing footprint sizes impact the conclusions of the comparative analysis. Specifically, how much of the observed difference is likely due to the footprint size discrepancy rather than differences in snow/ice interaction?
• Figure 7: Excellent figure; the alignment of CS2 and S3 is impressive. A few suggestions: (i) What is the cause of the bi-modal distribution of the IS2 freeboard in the June plot? (ii) Please include the number of tracks ($n$) for each sensor within each sub-plot. (ii) For the 15-day periods with limited laser and radar tracks, please discuss how representative those time series are for that period. (iv) Consider using slightly more distinct colors for CS2 and S3 to improve contrast.

• Line 573-575- This uniqueness may potentially be included in the introduction. Just a suggestion. 
• Line 590- I would recommend providing a statistics comparing the impact of including the negative freeboard vs excluding. What percentage of the readings were negative freeboards? 
• Line 594- I’m curious the extent of impact using the interpolation scheme had on the overall freeboard time series? Is this leading to any over-generalization in the freeboard values? Also, are the interpolated values being using for the freeboard comparisons e.g. in Figure 7 and 8? 
• Figure 10: For the SIT over MYI, there appears to be better correspondence between the modeled results and the in-situ data compared to the altimetry. What factors are driving this difference?
• Line 694- I’m not convinced that the time series trends challenging the common understanding that snow loading weighs down the ice down resulting in decrease in radar freeboard are well substantiated. Further analysis needs to be conducted before conclusively making this claim. 
• Line 669- Is there a reference to this assumption made? 
• Line 710: Year missing in the in-text citation. 
• Figure 11: Excellent Figure! 
• Line 719: Was this bias reported in the results section? If not, I would suggest providing a detailed analysis of this bias as high as 400% for FYI. 
• Line 738: While observations do support the conclusions drawn here,  I believe the authors need to state more explicitly that the exact percentage of impact of these factors during warming events could not be quantified. 
• Line 805: Some quantitative comparisons with these studies would be helpful here. 
• Code Availability: According to The Cryosphere (TC) guidelines, the data and code must be published open-source. Please double-check the journal's requirements and ensure that valid links to the available data/code are provided prior to publication.

[Synopsis and recommendation]

This manuscript presents an empirical analysis of Ku-band radar and laser altimeter freeboards over immobile landfast first-year ice and multi-year ice off northeast Greenland. By exploiting the fast-ice condition, the authors are able to follow approximately the same sea-ice areas from ice formation through winter and into the melt season. This is a valuable and novel observational framework, and the topic is highly relevant for The Cryosphere.

Overall, I find the study interesting and potentially important. However, I think several key interpretations need to be better qualified. In particular, the manuscript often interprets variations in retracked Ku-band freeboard as evidence for changes in the radar scattering horizon, but the retrieved freeboards are also affected by retracker behavior and processing choices. The manuscript would be strengthened by more clearly separating these two effects, or at least by explicitly acknowledging where they cannot be separated. I therefore recommend publication after major revision.

[General comments]

This is a very good and novel idea: obtaining time series of altimeter observations over the same ice floes by taking advantage of non-drifting fast ice. Our community needs more studies of this kind, and the approach is well suited to investigating the temporal behavior of laser and radar freeboards.

My main concern is that the interpretation currently mixes two different sources of variability: first, deviations between the assumed Ku-band radiative process and the actual scattering process in the snow–ice system; and second, limitations or biases of the retracking algorithm itself. Much of the analysis is based on retracked radar freeboard, which is affected by both. A positive bias in retrieved freeboard can arise not only because the relative contributions of air/snow, snow-volume, and snow/ice scattering change, but also because of how the LARM retracker handles fixed or constrained model parameters and the optimization procedure. Please discuss this coupled issue more explicitly throughout the manuscript.

Related to this point, many previous studies used different retracking algorithms, which makes it difficult to synthesize earlier findings into a single physical interpretation. If possible, I recommend that the authors provide information on the retracker used in the studies cited in the Introduction, for example TFMRA50, TFMRA70, LARM, or other approaches. This would help readers understand whether apparent differences among studies reflect physical snow/radar interactions, retracker dependence, or both.

I also wonder whether any time evolution of ice freeboard from in situ observations is available. It would be very helpful to verify whether the suspicious increases in radar freeboard are actually unphysical, rather than real changes in the ice/snow system. A reliable reference is critical when using terms such as “overestimation” and “underestimation.” For example, if SIMBA-derived snow depth and ice thickness are available, they might be converted into freeboard using assumed snow, ice, and seawater densities. Even if such a comparison is limited to MYI or only part of the season, it would provide useful context.

Finally, regarding objective 3, I found it difficult to see how the stated hypotheses were directly tested. The analysis is persuasive in places, but it is mostly observational and interpretive rather than a controlled test. Since LARM is based on a physics-based waveform simulator, the authors could consider adding a controlled numerical experiment. For example, they could define a reference snow–ice scene representing the physical conditions that they argue cause positive radar-freeboard bias, simulate the corresponding waveforms, and then retrack those waveforms under the usual cold, non-saline, snow/ice-interface-dominated assumption. Comparing the retracked radar freeboard with the known ideal radar freeboard would provide a more quantitative test of the proposed mechanism.

[Specific comments]

Line 66: “Landy et al.” appears to be missing the year.

Lines 68–70: The manuscript states that only a few studies report satellite-based empirical observations of biases between retracked radar freeboards and the expected snow/ice interface. Armitage and Ridout (2015) and Shi et al. (2024, especially Sect. 5.1) also address closely related issues. Please consider citing these studies.

Line 381: Similar to Jutila et al. (2022), Shi et al. (2023) reanalyzed the SEVER expedition dataset, which was the basis of the widely used two-density approach of Alexandrov et al. (2010), and found that MYI density may be closer to FYI density than commonly assumed. Please consider citing this study if relevant.

Figure 6: It is somewhat difficult to understand the distinction between relative and absolute time. Please clarify this in the caption and/or main text.

Line 623: “For both ice types and during most of the winter, the time series suggests that the retracked heights at Ku-band frequency correspond to locations well below the air/snow interface, and likely close to the snow/ice interface.” What are the clearest pieces of evidence supporting the phrase “likely close to the snow/ice interface”? Please make this reasoning more explicit.

Line 647: “Zhu et al., 2917” should be “Zhu et al., 2024.”?

Lines 649–653: As far as I understand, LARM optimizes an ice-surface roughness parameter during retracking. Does this mean that the roughness effect associated with frost flowers is not fully represented by the LARM waveform model? Please clarify whether such frost-flower-related effects can be absorbed by the LARM roughness optimization, or whether they could still lead to a positive bias in the retracked radar freeboard.

Line 701: “Landy et al.” is missing the year.

Line 710: “Khvorostovsky et al.” is missing the year.

Lines 744–745: The manuscript states that the radar-freeboard increase is temporary. Which specific results support this? Please refer to the relevant time series or quantify the duration more clearly.

Lines 752–753: Is the relevant supporting figure Fig. B2? Please cite it explicitly in the text.

Line 807: “Landy et al.” is missing the year.

Conclusions: Where do the values of 85 % and 60 % come from, and how were they evaluated? Please explain this more clearly in the Results or Discussion before using these numbers in the Conclusions.

Line 833: The phrase “short-term periods with heightened radar freeboards” needs clearer definition. Short-term relative to what timescale, and heightened compared to which reference level?

Line 854: “Kern et al. 2235” should be corrected. 

[References]

Armitage, T. W. K. and Ridout, A. L. (2015), https://doi.org/10.1002/2015GL064823

Shi et al. (2023), https://doi.org/10.1109/TGRS.2023.3265274

Shi et al. (2024), https://doi.org/10.1029/2024EA003715