This is a review of Gupta et al (2024).

The work uses ICESat-2 altimetry to understand properties of the shape and size of sea ice floes in the Weddell Sea. Generally the paper is well-written and interesting - as a case study in how laser altimetry can be used to explore the properties of the sea ice surface. I think there are some methodological questions I am hoping to have addressed in a revised manuscript.

The authors do discuss this, but it is worth emphasizing. The use of radiometrically-defined “dark lead”s, remains a challenge with ICESat-2, because of a lack of confidence in their meaning. I would like to see your Fig. A3 as an anomaly rather than side-by-side plots to show the differences. Perhaps in addition, showing the histograms next to one another of all lead spacings and chord lengths. Chord/spacing measurements with IS-2 have often been confusing, as with the high along-track resolution it is very easy to chop a floe in half with a misclassification. It might be in the Weddell this is not an issue, so that’s exciting! But more details here might be helpful.

More details are needed on the power law slopes. For example, how are you fitting alpha to the FCD? A treatment of this mathematically is in several places, including Virkar and Clauset (2014), but has to be done carefully. The use of binning can introduce spurious errors in the slope of such a distribution - see Stern et al 2018. The VC method is simple to apply and doesn’t rely on the binning. The statistical tests examined there should be applied before discussing power-law fits (if they are not already used). 

I would suggest renaming your alphas (one is a power law, others are not) because they correspond to different distributions. 

It is also important to discuss that there is fITD uncertainty, much like FCD uncertainty, because of the unknown misalignment of the laser with sea ice structure (in this case, ridges, say).

The extremely high correlation (R^2 = 0.98) between mean freeboard heights and mean chord length is striking, but somewhat concerning - do you think this is because thicker ice is easier to separate from open water? In that case, thicker ice means fewer missed classifications means wider FCD. Have you examined this possibility? Can you explain why this correlation is so strong?

For the floe roundness elements - an important question here is what is the intended usage of floe freeboard roundness information? I assume this is to infer something about the dynamical behavior of the floes contacting on another? This could be better justified because this metric, especially with a laser altimeter that has uneven sampling and only measures freeboard, is somewhat unclear. It would be useful to see exactly how ICESat-2 “sees” an individual floe rather than the composite normalized image of Figure 7.  - because floe surfaces are very non-round (see your Figure 1!) Freeboard variability can be from changes in sea ice density, and “floe roundness” is realistically more closely compared to the surface roundness alone. How round the floe is depends on factors under the ocean surface that ICESat-2 can’t examine here.   

It would be helpful to discuss averaging here - I presume in the paragraph beginning on L264 you mean you average over all floes, not average along the floe. How much variance is there in the resulting plots shown in Figure 7? This could be a good visualization - showing the variability associated with your compositing.