The review is attached as a PDF.

Review of the MS: “Assessment of Arctic sea ice thickness retrieval ability of the Chinese HY-2B Radar altimeter” by Dong et al.

The HY-2B satellite is carrying a dual frequency Ku- and C- band radar altimeter. The data are used for deriving the sea ice thickness and it is compared to similar retrievals from CryoSat2 and airborne campaign data (Operation Ice Bridge). This is an urgent and welcome topic. However, important details are missing from the MS about the instrument and data and discussion of the processing steps and choices made.  I did not see it stated clearly, but this study is not using the dual-frequency capability of HY-2B when deriving the freeboard? Only Ku-band? I think that the answers to these questions are important for evaluating the MS and the novelty of it.

The snow data used in the processing are the same as the AWI snow depth (line 224). This is logical when HY-2B data are compared to AWI CS2 data. Are all the other processing steps and auxiliary data used in the processing identical to the AWI processing so that a proper comparison can be made? In line 24 it is mentioned that some of the differences were due to the applied sea surface height anomaly extraction method. Is it due to differences in the methodology or sensor specific differences? Please clarify.

A number of studies focused on Ku- and Ka-band radar altimeter applications from AltiKa and CryoSat and for preparation of the planned CRISTAL mission. This MS does not continue that discussion using C- and Ku-band radar altimetry for sea ice applications. Anyway, I think that some justification of the choices made in this paper are needed, for example, it is implicitly understood that radar scattering is at the snow-ice interface. Is that a good assumption? And how would that differ for Ku- and C-band. Would the different foot-print sizes at the two frequencies have an impact for the derived ice thickness?

The snow depth dataset is a strange combination of retrieved snow depth over first-year ice and snow depth climatology over multiyear ice and some smoothing and filtering. What does the different snow depths and ice densities in first-year ice and multiyear ice areas mean for the retrieved ice thickness? How does this compare to the variation in derived freeboards for the ice thickness variability? What is the impact of the fixed ice densities and the ice type classification with its different snow depths etc.? Is the snow depth needed at all? To me it seems that the snow data are massaged until the “right” ice thickness is achieved.

Some specific comments:

Line 13: “… it is of great significance…” this is a subjective statement without real justification. Please reformulate.

Line 14: delete “values”

Line 15: “… ice growing cycles” replace by “winters”.

Line 17: delete “recorded”.

Line 18: replace “verified” with “compared”.

Line 40: delete “effect”.

Around line 45: It is confusing and inaccurate what is written about AMOC. Please reformulate or delete.

Line 52: add “and extent” after “density”.

Line 57: use “derive” instead of “estimate”, also line 61.

Line 75: “few reports” please list these few reports as references.

Line 77: delete “as a supplementary means”

Line 79: delete sentence starting with “Therefore…”.

Line 90: Add some details about the HY-2B altimeter.

Line 91: delete “successfully”.

Line 168: “Snow depth” it is unclear which snow depth dataset is actually used. Please clarify this and explain if you are using AWI snow depths or IS-2/CS2 snow depths.

Line 224: delete “depth”.

Line 226: Eq. 3 units in meters? Please add units, sometimes [m] sometimes [cm]… use the same units throughout.

Line 230: Eq. 4, what is the sea water density? Sometimes units are [kgm-3] sometimes [kg/m3], stick to one form. Multiplication is sometimes a dot and sometimes x?

Line 322: I think that I know, but this is confusing: what is a “deviation”? standard deviation? or bias? Throughout the MS.

Line 324: replace ”independent” with “OIB”

This manuscript introduces the retrieval of Arctic sea ice freeboard and thickness with the radar altimeter onboard HY-2B. The inclination angle of HY-2B allows the coverage of up to 82N, and the satellite potentially constitutes an important source of information for sea ice of both polar regions. Specifically in the manuscript, the authors focus on the processing of converting existing range (or elevation) product of HY-2B into radar/ice freeboard and thickness. I consider the data from HY2B and this submission a good contribution to the community, but I do have the following comments that in my opinion that should be addressed first. 

The overall uncertainty analysis needs to be more clear and precise, especially in terms of the determination of freeboard uncertainty. First, the (claimed) uncertainty used for further analysis is sigma_SGDR, which is only 2cm and should be a lower bound of the actual range uncertainty. This uncertainty, in its true quantity, applies to both SSHA and individual freeboard. Since SGDR is the upstream dataset this work relies on, it is not necessary a quantity that needs much elaboration within this work. However, its value needs to be justified. For example, what is the gate resolution of HY-2B? Is the 2cm uncertainty due to the combination of several footprints (therefore smaller)? Furthermore, the authors state that Ricker et al. (2014) `believed’ the random uncertainty of radar freeboard to be determined by radar speckle noise. Is it possible to provide any other proof of how is the reference relevance to the treatment here? 

Second, the uncertainty of SSH is computed as the standard deviation (SD) of SSH points (of the along-track 25km segment). I think this is a crude guess, and unfortunately, probably an underestimation of SSH-induced uncertainty. Because the along-track points that are not used for determining SSH share the same uncertainty caused by the same set of SSH points, and therefore the uncertainty of the retrieved freeboard samples on the same track is systematic (rather than random). Therefore, in Eq. (11) it should be averaged out and diminished much more slowly.

Third, although the authors treat the snow-induced uncertainty as random error (possibly a typo on line 400), it is hardly the case, despite that the snow depth is based on both climatology and PMI retrieval. A simple counter argument is that: the PMI product is based on C-band PMI onboard AMSR-E/2, which is at about 60km in resolution. Not to mention the 8-grid smoothing that is carried out over the snow depth retrieval. Then there exists large local correlation of the snow depth uncertainty, given that 25km grid is used in this study. Arguably more importantly, the climatology of snow depth from W99 plays a very important role in the snow-depth composite of AWI, which is evident in the respective technical report. Given that W99 is halved for the combined product, its induced uncertainty is still very large and will dominate the portion caused by snow in the uncertainty of the final ice thickness product. Therefore, I will be very cautions to treat the snow depth related uncertainty as random errors.

Finally, the density of snow and ice are also treated as random errors. This should be also a very `optimistic’ estimation, especially the effect on the final ice thickness uncertainty. Overall I suggest that the authors reconsider the uncertainty quantification process and divide into the random part and the systematic part, and be cautious about which category each term belongs to. Differentiation of the uncertainty also ensures fairer comparison with the AWI product, and the author seemed to have used the random error of it, which is much lower in may part of the basin (e.g., Beaufort Sea).

Another major issue I noticed is the limited data availability of HY-2B even in the later months of the winter. As shown in Figure 6 and 8, there are large areas with no valid radar freeboard on the monthly scale: on the Eurasia continental shelf, Beaufort Sea, etc. What is the cause of the missing data? Invalid waveforms. I suppose that at this latitude HY-2B should have better coverage than CS2. 

Also, the authors mainly compared the HY-2B retrieval with CS2. However, Sentinel-3 and AltiKa have the same inclination angle as HY-2B, and especially Sentinel-3 satellites work on Ku-band (although they are of delay-Doppler type). Is it better to compare against Sentinel-3 retrievals? This is only a suggestion, out of my curiosity. The authors can decide whether this is an option or not. 

The comparison against CS2-IS2 data contains inconsistencies in the methodology. The comparison of radar freeboard is inherently between two CS2 retrievals (one from AWI and the other one carried out by Kwok). Therefore it’s not a fair comparison, since the two products based on CS2 arises from the same source of information. So many issues are not present, such as limited representation by altimetry. 

Minor issues:

Some figures contain maps that are too small to read, such as Fig. 6 and 8. Pls increase the font and maps accordingly. 

For the third-party and auxiliary datasets, I suggest that the authors just introduce them, without too much comments on the specific advantages. For each product there are potentially uncertainties that are fully addressed. I think just stating the basic status quo of the products is already enough. 

The reference list is not strictly ordered.

The language usage needs improvements. General suggestions include avoiding long sentences, and avoiding complex statements. Several typos are present. I suggest that the authors give an overhaul of the manuscript after the major issues are addressed.