The work aims to study the sea ice freeboard in areas around the Antarctic peninsula. Improved sea ice surface topography is a useful product and can be implemented in other studies, such as ice drift product development and climate studies. The manuscript is reasonably well written and mostly easy to follow. Some of the terminology is at times somewhat confusing. There are a great many figures in the manuscript, could some perhaps be moved to supplementary information.

Major comments

n the abstract the terms sea ice DEM, i.e. snow freeboard is introduced. How does this relate to the sea ice topography? Why is the sea ice DEM not = sea ice and snow freeboard? How is the air-ocean-ice system related to the sea ice topography? The statement as it stands right now is a bit challenging to interpret.

R21. If we assume that the DEM is snow freeboard, should it then be assumed that no penetration if the snow is possible?

R22-23. Please elaborate how this product is essential for assessing the impact of climate change on sea ice.

R58-59. How can the DEM help separate the different ice types?

R67. What is meant with Ice Chart here? An operational ice charts such as those provided by the ice services. What is meant is explained on R108. This one of the terminology words introduced it the introduction that gets explained later in the manuscript. This terminology should either be removed from the introduction or needs to be explained here.

R91. Perhaps state how the denoising is done then why it’s useful/essential to do so here.

Table 1. One of the datasets (R5) has a higher HoA. Does this affect the results presented here?

R111. The spatial resolution of the Ice Charts is 10 x 10 km. How wide are the SAR images used? Will more than a few pixels be comparable between the Ice Charts and the SAR images?

Figure 3, 4 and 7. The schematic in Figure 3 in itself is good but it’s challenging to understand if perhaps Figure 4 is step 1, and if so why this isn’t stated in Figure 4. Please indicate how these 3 flow charts are interconnected. It appears as if Step 1 is in part explained in Figure 4 but it’s unclear as more information than the TanDEM-X SAR images are used as input data? And the classification map at the end of Figure 4 appears to perhaps be the first box in Step 1. Figure 7 appears to be an explanation of the top right box in Step 2 in Figure 3. Please clarify these flow charts.

R180-184. Are some parameters more important for one of specific ice types? Or is the importance level presented in Fig 5 universal?

Figure 9. Some of the leads appear to have a light blue color, not the same as for the YI. Why is that? Which ice type do they represent? They appear to in 1, 2 and 3 have the highest E. What is the unit E? Does a low SNR perhaps get mistaken as a thick sea ice? Perhaps could a noise analysis remove erroneous values?

Figure 11. In the top, upper middle and bottom figures, it appears as if the SAR estimates are underestimating the high and low peaks. Is this a resolution issue? Or is there some other explanation behind this?

Figure 12. This figure could perhaps be moved to supplementary information as it doesn’t add much to the understanding of the results. It’s very challenging to see the elevations, if kept perhaps make the SAR images a lot larger?

Figure 13, 14, 15. Consider coloring the y-axis and the color used in the plot the same color for easier interpretation of the information contained within the figures. Add a legend to the two rightmost columns, to explain what the blue and the orange represents.  

Minor comments

R2 “… a digital …” or “digital elevation models”

R2-3 should it be drifting sea ice instead of drift sea ice?

R60-61. “sea ice elevation” has already been defined earlier in the manuscript.

R76. With sequence is it meant orbit?

R143. Wakabayashi et al 2004 used L-band SAR, how will this compare to the X-band SAR used here? Can we derive sea ice thickness using X-band SAR?

R198. The reference can be shortened to (Meier, Markus and Comiso, 2018)

R248. “… in the Ice Charts”

R283. Sea ice doesn’t evolve from MYI to TI. TI can evolve to MYI through surviving at least 2 seasonal cycles.

R363-367. It this information needed here?

In this interesting paper, a new method for the retrieval of total ice freeboard (ice freeboard plus snow thickness) from single-pass interferometric SAR is developed and applied to the Weddell and Ross Seas. The SAR-derived sea ice topography is validated by independently measured sea ice freeboard profiles and analyzed in comparison to several studies, which support the results. The paper should definitely be published, but I recommend modifications which concern the use of certain terms and the need for additional information. The latter is in particular important for the description of the method.

Abstract 

(1)    Line 3: “accurate sea ice DEMs (i.e., snow freeboard)” The term “snow freeboard” (see also line 21 in the introduction) is misleading. Better use “total freeboard” which is ice freeboard plus snow layer thickness

Introduction:

(2)    Lines 21-22: It is the mass of the ice above the water surface plus snow load (not snow freeboard) from which ice thickness can be estimated.

(3)    Line 35: As far as I remember does the Dierking paper discuss problems and requirements for retrieving the sea ice surface topography of drifting ice but demonstrates it only for landfast ice.

(4)    Line 43 and lines 54-55: “Antarctic old ice” – what precisely is “old ice”? The separation between “young ice” and “old ice” based on the criterion of penetration depth (the difference between DMS and SAR elevation) is not suitable, since salinity (as the major factor influencing the µ-wave penetration) is not only linked to ice age but also to other factors (e.g. saline snow crusts at the ice surface, effects of ice flooding). This is also visible in your data, Figs. 13-15. I propose that you instead use the categories “low-penetration condition” and “large-penetration condition”.

(5)    Lines 59-61: Sentences: “A root-mean-square error (RMSE) of 0.26m between the derived DEM and reference data signifies a precise elevation mapping for both YI and OI. Throughout the paper, “sea ice elevation" is the entire vertical height (including snow depth) above the local sea surface.” Actually, 0.26 m (for averages over areas of several meters side length) can locally be a rather high (but mostly acceptable) uncertainty, considering that a large fraction of Antarctic sea ice is first-year with a thickness of around one meter (https://www.climate.gov/news-features/understanding-climate/understanding-climate-antarctic-sea-ice-extent) and correspondingly much less elevation above the water surface. “Precise” means that repeated measurements are close to one another – here the term “accurate” may be more appropriate. 

Data processing

(6)    Line 87: Here it is ground-range? Is the pixel size of 10 x 10 m used for both the classification process and for elevation retrieval? Should be stated.

(7)    Line 97: The vertical accuracy of the DMS data (line 232) should also be mentioned here. Which reference surface was used for the height values? The local water surface or a reference ellipsoid? In the User Guide by Dotson and Arvesen I found “The IceBridge DMS L3 Photogrammetric DEMs are GeoTIFF imagery, in meters and above the WGS-84 ellipsoid.” (page 5). The WGS-84 ellipsoid is usually not at the same level as the local water surface.

(8)    Lines 115-117: I checked the types of ice charts at the US National Ice Center (https://usicecenter.gov/Products/AntarcHome). I recommend that you provide a link for your reference (“U.S. National Ice Center., 2020”) and show the ice chart which you actually used. E.g., I did not find any hint that the ice charts are averages over 7 days, but are produced once a week (https://nsidc.org/sites/default/files/documents/user-guide/g10013-v001-userguide.pdf). 

(9)    Section 2.4: The definition of an “average ice type” does not make sense. With regard to the notation, it is more an “ice condition index”. A meaningful comparison between ice type and topography is achieved when the concentration of the respective ice type in your window is sufficiently large, e.g. > 80% or even larger. One has to consider that the topography for one ice type can be highly variable (determined by zones of deformation and their areal fractions relative to the smooth level ice areas). Hence one better concentrates on windows for which one ice type is clearly dominant. In your case that should not be a problem.

(10)    Line 148: what are “Pauli-1” and “Pauli-2”-polarizations? Do you refer to the Pauli-representation? Then Pauli-1 is the first and Pauli-2 the second component of the Pauli decomposition which relates surface and volume scattering?

(11)    Lines 166 – 172: see comment (4) above. There is also a misprint on line 172, it should read “… h_pene ≥ 0.3 m are OI”. Was there a special criterion for selecting a threshold of 0.3 m for separating YI and OI?

(12)    Lines 185 – 186: The question is how your “YI” class is related to the ice types listed in Table 2. The WMO-category defines young ice as ice between 10 cm and 30 cm in thickness, as correctly listed in your table. It can be assumed that a penetration depth up to 0.3 m (your “YI”) covers the categories “Thin Ice” and “First-Year Ice”. The “old ice” with a penetration depth ≥ 0.3 m cover the thicker FY ice and MY ice. Conclusion is that you should not use the notation YI for penetration depths < 0.3m, see comment 4 above. Check also notations used in section “Conclusions”.

(13)    Figure 6 and Section 2.5:  How do you relate h_InSAR to the local water surface? Or in other words: which reference surface do you actually use when calculating h_InSAR? I suppose that in the initial InSAR processing it is not the local water surface but also the WGS84-elliposid?

(14)    Section 3.2. Here, many things are unclear to me. In summary, I recommend to rewrite this section for the sake of clarity. Single issues: (a) what is the exact definition of m, does it refer to thickness of layer 1 and layer 2? (b) Is h_mod the surface elevation above the water surface (or reference ellipsoid)? (c) Which AMSR Level 3 data did you use for retrieving the snow depth over your test sites? How large is the uncertainty of those snow depth values? (d) line 202: “h_DMS can be transformed into φ_DMS by Eq. (3)” => equation 3 describes the relation between h_InSAR and φ_γ. I wonder whether this equation can be simply applied using h_DMS to derive φ_DMS because for the DMS data there is no height of ambiguity. Is  φ_DMS assumed to be the phase at the snow-air interface, hence φ_DMS = φ₀? (e) From equation 8, you obtain m and h_v, according to the given definition and Fig. 6 h_v is ice thickness which could be mentioned. (f) Why is the second step needed, namely to combine the SAR features with m and h_v to obtain modified values m’ and h_v’ ? With the classified images (your maps of ice types), one can directly link the m and h_v values from the first step in Fig. 7 with the corresponding ice classification map. The second step is not needed. 

(15)    Fig 9: the right-most color bar is “E”, is this the DEM given in meters?

(16)    Figure 11: What is the explanation for the data gaps in the second profile from the top?

(17)    Results shown in Figs. 9, 10, 11: h_{mod_SAR} is elevation = total freeboard relative to the water surface (or reference ellipsoid), retrieved from pixels of 10 x 10 m in size? 

(18)    Again Figs. 9, 10, 11: For “old ice”, height values = total freeboard of even larger than 3 m are measured. Intuitively, this seems to be not very realistic, although DMS and SAR values match. Is there any information available about the area regarding ice and snow conditions which seems to be special when comparing to the results of the other profiles shown in Figs. 13-15 which reveal smaller heights? If you used the WGS84 ellipsoid as reference: Is the difference between reference ellipsoid and mean sea level larger close to the Antarctic Peninsula? Perhaps also icebergs biased the measurements?

(19)    Figure 12: The color bars show values of h_{mod_SAR} down-sampled to a resolution of 500 m?

(20)    Figures 13,14,15, left column: The percentage of “OI” refers to the large-penetration-depth category. It would be much easier to discuss the relationship between the OI and real ice types when concentrations of MY, FY, and TI are shown instead of an index related to the “average ice type”. Lines 241-243: In Fig. 13, the percentage of OI is only 58% for distances between 0 and 160 m, but the ice type in the ice chart seems to be almost 100% MYI ice since the “average ice type” value is close to 3. In the range between approximately 300 and 600 km, MYI is also close to 100% concentration, but the percentage of OI is down to 20-25%. (Note that according to section 2.4, the indices for TI, FYI, and MYI should be 0,1,2, respectively, and not 1,2,3) This demonstrates that the penetration depth is not directly linked to the ice types shown in the ice chart.

(21)    Line 253-261: Referring to the statement: “In general, the region with thicker ice (e.g., MYI) is anticipated to display higher elevation or larger roughness compared to the area with thinner ice, such as FYI and TI”: Locally, rough FYI may reveal a larger roughness than smooth level MYI, and it may reveal a higher elevation when covered by a very thick snow cover. This may also explain discrepancies. 

(22)    Line 262: “sea ice …exhibits…highest elevation”. You should again clearly state that the values of elevation are total freeboard, i.e. also include the snow layer.

(23)    Line 265: Figs. 6g-l in the paper by Wang et al (2020) are indeed well suited for comparing with your results. The values they show (Fig. 6l for 2017), however, have only a narrow peak at 1.5 to 2.5 m, otherwise values are lower. Figs 6g-l should also be mentioned with regard to your Figs 13-15, considering window sizes for averaging the elevation.

(24)    Line 283: ice type “MTI”? I think it should be MYI.

(25)    Line 294-295: Sentence: “The variation of the roughness along the R1 segment also highlights the importance of combining topographic mapping with ice category mapping to comprehensively characterize sea ice features.” Since there is no direct relationship between ice type and topography data, it is not per se “important” to combine both, but can be useful in certain cases, e.g. for operational ice charting. Since the edges of ice floes with open water between the floes contribute to the ice roughness, ice concentration may also be a useful parameter to be combined with ice topography in some cases.

A point of my interest: For radar applications, it would also be useful to show how large deviations between DMS and SAR values (= your penetration depth) can be, and where this occurs (spatial distribution along your tracks)..