|Review of: |
“Assessment of Arctic Sea Ice Thickness Retrieval Ability of the Chinese HY-2B Ku-band Radar Altimeter”
By Dong et al.
# 2nd review
I’d like to thank the authors for the extensive edits, added work, and detailed response made to all the reviews. This is my second time reviewing this manuscript, so my comments will primarily focus on the edits made since the last review as well as the author’s response to the reviews, but I will also include some additional comments that I realized when reading it a second time. Unfortunately, even with the extensive edits and comments provided by the authors, I still have concerns that should be considered before publication.
The authors have since the original submission changed the processing slightly to include not-a-number values when sea surface height anomalies (SSHA) were not available (instead of providing a value of 0), as well as used the 15 lowest (instead of 9) points within a 25-km segment as a measure of the sea level within leads (SSHA). This has improved their statistics when comparing with reference data (Operation IceBridge data) for both sea ice freeboard and sea ice thickness estimates. They have produced additional studies in response to reviews, where they have assessed waveform parameters/features to use for discriminating between leads/floes instead of their current methodology. However, this did not yield better results than their current methodology (and have therefore not been included in the manuscript). The authors have amended their uncertainty estimation, which now presents more feasible estimations, and provided a more extensive discussion on the limitations of their methodology and impacts hereof.
I am happy to see the positive impact that the change in processing has shown in your comparison with OIB, however the statistics are still not a positive for HY-2B freeboard nor thickness estimates as the CryoSat-2 estimates. This, again, tells me that the processing chain of AWI (lead/floe discrimination, TFMRA50 re-tracker) is likely the way forward even with the different footprint and altimeter specifications of HY-2B compared with CryoSat-2. I do recommend the authors to investigate using waveform parameters for lead/floe discrimination as well as applying a re-tracker commonly used for sea ice (such as TFMRA50), however I understand that it will require quite the effort. If this is not possible, I think it is crucial that the manuscript clearly reflects (and mentions) the limitations of the methods and data used in this study. Below I have mentioned some suggestions for implementing this.
I think it is important to highlight that this study is a feasibility study. What I mean, is that you are in fact not truly assessing the retrieval ability of HY-2B for Arctic sea ice thickness, since you are not using re-trackers commonly used, or designed, for sea ice – in fact, you use observations provided by an ocean-preferred re-tracker. As such, your results are also limited by this. I therefore propose you change the title of your manuscript to reflect this. Suggestions could be: “Feasibility of retrieving Arctic sea ice thickness from the Chinese HY-2B Ku-band radar altimeter” or “Assessment of Arctic sea ice thickness retrieval ability of the Chinese HY-2B Ku-band radar altimeter: a feasibility study”. Also, I think it is crucial that you state the purpose of this study in the end of the introduction, highlighting that you are aiming to investigate whether HY-2B can be used for sea ice, but that you are limited to already provided higher-level products, and that it is not within the scope of the study to derive a freeboard product using your own re-tracker from the HY-2B product.
I’m happy to see a study using the waveforms and classifying the waveforms using pulse peakiness (PP), however slightly confused to see that it was in fact not satisfactory. Based on your response, I see some aspects that could potentially explain this: (1) PP can be calculated in different ways (that is to say, sometimes they are tweaked in different studies). I cannot see how you have calculated PP, but I do find it exceptionally odd that HY-2B is not able to use it to classify by – whereas all other radar altimeters have in fact successfully used this. Therefore, it may be due to how you have calculated PP; (2) For HY-2B we do not yet know which PP thresholds to use. You have tried several PP thresholds (not sure what the selection of thresholds were based on) but choosing high PP’s (which would normally yield a classification as lead) resulted in almost no lead observations. This might be a result of the already low data coverage, which you also suggest, but I do propose that for the next time you look at this, that instead of pre-defined PP’s, you choose a selection of waveforms and make a statistical analysis to derive the PP thresholds. I am aware that you state this is work for future studies, however it will unfortunately be those studies that have the true value then.
There is very little data available by HY-2B. Why is that? It is based on the selected re-tracker where some restrictions/flags are applied? There must be applied some post-processing steps that remove the data. Could you provide paragraph explaining when in the processing this data is removed, and why it was removed? Surely, this should not be the case when a more appropriate re-tracker is used. Could you also provide a measure of how much coverage HY-2B has compared with CryoSat-2? (e.g., how many points within each grid cell when gridding).
Your method of using 25 km segments and choosing leads based on the lowest 15 points within a segment still hasn’t fully convinced me. As mentioned, several times throughout your study, this will have different impacts: higher freeboards during freeze-up, lower freeboards during spring, and a difference across first-year ice (FYI) and multi-year ice (MYI). So, essentially, you have a known bias due to your choice of methodology that you are not accounting for. I think it would be necessary to provide a measure (average, modal, median – take your pick) of how often this is the case, e.g., by providing a statistical value of how many points you usually have within a 25 km segment, as well as the standard deviation of this (or min-max range, quantiles – again, take your pick) to understand the variability. In truth, we do not know how often it is the case that your method will be impacted using the 15 lowest points.
Furthermore, your method ensures a lead point every 25 km (if 15 points within a segment). Is this ‘enough’? Here, I am considering the fact that HY-2B’s footprint is 1.9 km across-track. In the study of Tilling et al. (2019), they compared CryoSat-2 and EnviSat, and saw that a lead-to-floe echo distance for EnviSat ranged from 0-20 km (average 11.3 km) – a satellite with a larger footprint, whereas for CryoSat-2 it ranged 0-4 km (mean of 1.0 km) – a satellite with comparable footprint. Somehow one lead observation every 25 km seems low.
Figure 1 in response to Reviewer#1 suggests that something odd is going on due to the limited lead observations identified using the selected PP threshold, since you are simply using a nearest neighbor interpolation – which makes me wonder; how come you use this interpolation? Many other studies use either linear or cubic interpolations, making the interpolation dependent on several points rather than just one. Furthermore, do you have a limit on how far away points are allowed to be from a lead observation, e.g., 200 km, which other studies have required.
In response to GC2 for Reviewer#1, you present a comparison with TFMRA50 product – which is not available in the SGDR product. How come you have not used this sea-ice specific re-tracker for the study instead of the SGDR? Please provide some justification/explanation for the choice of this re-tracker in the manuscript.
In response to GC2 from Reviewer#2, you conclude that the assumption of Ku-band penetrating to the snow-ice interface still holds based on former studies. However, more recent studies (Stroeve et al. 2020, 2022; Nab et al. 2022) have questioned this (with good reason). Since we do not currently have methods that can take into account this change of scattering horizon within the snowpack, I’d say the assumption is fair (and still widely used), however I do think that the topic warrants a discussion which is currently not given in the manuscript.
Also, based on several review comments, the uncertainty estimation procedure has been refined, and the results look more promising. However, when comparing with other studies (Ricker et al. 2014, Landy et al. 2020), the uncertainties estimated in this study (for both CS-2 and HY-2B) are in the lower range, which warrants a discussion. Also, consider separating your uncertainty estimates into ranges for MYI and FYI (like Ricker et al., 2014), since you also mention that you see different results depending on ice type.
Finally, I tried retrieving the data again. It was indeed necessary to create an account to access the FTP server, so I was not able to look at the data used for this study without creating a user (the user was not activated within deadline of this review). I suggest you write, in the data availability section, how to properly retrieve the data as you wrote in the response to reviews (especially since the website is not available with an English translation, thus limiting the potential user pool) to help other users get a hold of this data.
Please check again that all references are properly written. E.g., line 144 (new manuscript) should be ‘Tonboe et al. (2016)’ not ‘Rasmus et al. (2016)’.
Specific comments (references to line numbers in the new manuscript)
I highly encourage the authors to look over the text again. With the introduction of a several new paragraphs, there are several places where it could benefit from some proofreading.
Line 15-17. Your methodology has yet to be described, yet you mention specifics about how it is done. I suggest generalizing this sentence.
Line 17-20. Could you include a short sentence of how CryoSat-2 compare with OIB, for a perspective?
Line 22. I’m happy to see that the abstract has been shortened. Perhaps, you could include a short line on future work or current limitations for your work with HY-2B observations.
Line 59. “corrections” -> “processing steps”
Line 67. “The overall difference (…) AWI data is ” -> “They noted the average difference (…) AWI data to be ”
Line 69. “The radar freeboards are generally higher for HY-2B than CS-2” -> “They generally observed higher radar freeboards for HY-2B than CS-2”.
Line 70. “(…), the HY-2B satellite can be used to observe polar sea ice” -> “(…), the feasibility of using the HY-2B satellite to map the polar sea ice must be explored”. Please, also present limitations of your study (using already provided and re-tracked data rather than applying your own re-tracker etc.).
Line 84-85. Is “take into account the observations of sea ice” part of its main mission? I suggest rephrasing this sentence for clarity.
Line 88. Please provide a description of which stage the satellite is currently at and when the stages have started/ended.
Line 90. Waveforms have not been mentioned yet. Consider rephrasing this for clarity.
Line 96-97. Are the tracking modes named after the different re-trackers (OCOG, SMLE) or does HY-2B have to tracking modes based on surface, that tracks with different re-trackers? Consider rephrasing this, while also including a reference to what the “Brown” model is.
Line 114. Include which version (baseline) of ESA CryoSat-2 product you are using.
Line 196. MSS defined (by acronym) later than first used (used in MSS data section).
Line 199. You mention residual error – be careful with terms like error/uncertainties. Consider using a different term. Also, this is the first time you are mentioning anything about this residual “error”. Could you describe it more?
Line 207. Has SSHA been defined yet?
Line 209-210. “Since the (…)” -> consider rephrasing this or write the limitations in the introduction already, and then highlighting here, that you are simply using already provided elevations in the SGDR product.
Line 232 (Section 3.2). Aren’t these results? Consider moving them to the result section.
Line 241-242. “(…), which may have been caused by the fact that not all points within the 25 km segment are leads” -> “(…), which may have been caused by the fact that not all points used to estimate the SSHA within the 25 km segments originate from leads”.
Line 248. “totally overlapped” -> “fully coincident”
Line 249. Could you provide the exact time difference and overall spatial difference? If the difference is significant (in hours), perhaps a measure of drift might be provided as well, to imply how big of an impact drift may have in this comparison.
Line 308-310. Consider combining these two sentences (“Except (…)”) in the sentence starting in line 307: “The HY2-B sea ice thicknesses are thicker (…)” for clarity. Also, throughout the text, be aware of too many repetitions of practically the same text.
Line 331. What underestimation? It can be seen on the figures but has not been described in the text so far. Please include it.
Line 334. “The majority of the spread” -> “The majority of the spread (shown by RMSE or MAE)” – or however you see this spread, but link to it.
Line 339. The IS-2 snow freeboard is not subtracted from the AWI snow depths to obtain the sea ice freeboard. The AWI snow depths are subtracted from the IS-2 snow freeboards to obtain the sea ice freeboard. Please correct.
Line 342. Add which section this slower wave propagation correction has already been explained in, in the end of the sentence -> “(see Section …).
Line 348. “In addition (…)” – consider rephrasing this sentence for clarity.
Line 367. What is meant by a “larger” SSHA? I suggest rephrasing for clarity.
Line 434. Similar to the changes for the abstract, consider generalizing this sentence since you are talking specifics about a methodology that has not been described in your conclusion yet.
Landy, J. C., Petty, A. A., Tsamados, M., & Stroeve, J. C. (2020). Sea ice roughness overlooked as a key source of uncertainty in CryoSat-2 ice freeboard retrievals. Journal of Geophysical Research: Oceans, 125, e2019JC015820. https://doi.org/10.1029/2019JC015820
Nab, C., Mallett, R., Gregory, W., Landy, J., Lawrence, I., Willatt, R., et al. (2023). Synoptic variability in satellite altimeter-derived radar freeboard of Arctic sea ice. Geophysical Research Letters, 50, e2022GL100696. https://doi.org/10.1029/2022GL100696
Ricker, R., Hendricks, S., Helm, V., Skourup, H., and Davidson, M.: Sensitivity of CryoSat-2 Arctic sea-ice freeboard and thickness on radar-waveform interpretation, The Cryosphere, 8, 1607–1622, https://doi.org/10.5194/tc-8-1607-2014, 2014.
Stroeve, J., Nandan, V., Willatt, R., Dadic, R., Rostosky, P., Gallagher, M., Mallett, R., Barrett, A., Hendricks, S., Tonboe, R., McCrystall, M., Serreze, M., Thielke, L., Spreen, G., Newman, T., Yackel, J., Ricker, R., Tsamados, M., Macfarlane, A., Hannula, H.-R., and Schneebeli, M.: Rain on snow (ROS) understudied in sea ice remote sensing: a multi-sensor analysis of ROS during MOSAiC (Multidisciplinary drifting Observatory for the Study of Arctic Climate), The Cryosphere, 16, 4223–4250, https://doi.org/10.5194/tc-16-4223-2022, 2022.
Stroeve, J., Nandan, V., Willatt, R., Tonboe, R., Hendricks, S., Ricker, R., Mead, J., Mallett, R., Huntemann, M., Itkin, P., Schneebeli, M., Krampe, D., Spreen, G., Wilkinson, J., Matero, I., Hoppmann, M., and Tsamados, M.: Surface-based Ku- and Ka-band polarimetric radar for sea ice studies , The Cryosphere, 14, 4405–4426, https://doi.org/10.5194/tc-14-4405-2020, 2020.
Tilling, R., Ridout, A., & Shepherd, A. (2019). Assessing the impact of lead and floe sampling on Arctic sea ice thickness estimates from Envisat and CryoSat-2. Journal of Geophysical Research: Oceans, 124, 7473– 7485. https://doi.org/10.1029/2019JC015232