Surface Properties Linked to Retrieval Uncertainty of Satellite Sea-Ice Thickness with Upward-Looking Sonar Measurements
Sine Munk Hvidegaard
Round 1
Reviewer 1 Report
Authors presenting a study on estimation satellite sea ice thickness data uncertainties by using sonar measurements. In my opinion the topic is very important and the significance of the content is high, since satellite collected data are more and more used by scientists. Therefore systematic analysis of the quality of the data with regards to sea ice thickness is required/. The introduction is quite short, authors only referring to the literature without discussing the kay sources of uncertainties. I wonder if a bit more details could be added here. I specifically wonder about local phenomena like ice ridges or hummocks – see for instance:
Girjatowicz, J.P., 2019. Characterization of Grounded Ice Hummocks Found in Coastal Lagoons of the Southern Baltic Sea. Journal of Coastal Research, 35(6), pp.1250-1259.
Kolerski, T., Zima, P., Szydłowski, M., 2019. Mathematical Modeling of Ice Thrusting on the Shore of the Vistula Lagoon (Baltic Sea) and the Proposed Artificial Island. Water.. doi:10.3390/w11112297
The Materials and Methods section is well describing both Radar and Sonar Data.
Figure 2 which presents comparison of the ULS measurements with three satellite-derived sea ice draft estimates in my opinion is too small, and it may be difficult to see what is described in the text. Please consider to enlarge it in a final version of the paper. I don’t have any comments to the rest of the figures in “Results and Discussion” chapter.
In line 302 authors present equation for hice calculation form hydrostatic equilibrium. And then discussing the largest uncertainties in the SIT. I wonder if snow density may not be also the reason for uncertainty? It may vary from about 20 to nearly 550 kg/m3 (Roebber, P.J., Bruening, S.L., Schultz, D.M. and Cortinas Jr, J.V., 2003. Improving snowfall forecasting by diagnosing snow density. Weather and forecasting, 18(2), pp.264-287.)
In general I found this paper very interesting and valuable , suggesting editors to be publish in Remote Sensing after minor revision
Author Response
Please see the attachment.
Dear Reviewer
We would like to thank you for your comments that helped to improve the manuscript. In the following, please, find the answers to your comments. Please also note that the changes in the revised manuscript are highlighted with the “Track Changes” function.
Point 1: The introduction is quite short, authors only referring to the literature without discussing the kay sources of uncertainties. I wonder if a bit more details could be added here. I specifically wonder about local phenomena like ice ridges or hummocks – see for instance:
Girjatowicz, J.P., 2019. Characterization of Grounded Ice Hummocks Found in Coastal Lagoons of the Southern Baltic Sea. Journal of Coastal Research, 35(6), pp.1250-1259.
Kolerski, T., Zima, P., Szydłowski, M., 2019. Mathematical Modeling of Ice Thrusting on the Shore of the Vistula Lagoon (Baltic Sea) and the Proposed Artificial Island. Water.. doi:10.3390/w11112297
Response 1: We agree that introduction would benefit of more detailed description and expanded description of the uncertainties. Also in response to the comment from another Reviewer we define the terms freeboard, lead and draft.
Point 2: Figure 2 which presents comparison of the ULS measurements with three satellite-derived sea ice draft estimates in my opinion is too small, and it may be difficult to see what is described in the text. Please consider to enlarge it in a final version of the paper.
Response 2: We enlarged the Figure 2 and slightly changed the scales for better visualization of the curves.
Point 3: In line 302 authors present equation for hice calculation form hydrostatic equilibrium. And then discussing the largest uncertainties in the SIT. I wonder if snow density may not be also the reason for uncertainty? It may vary from about 20 to nearly 550 kg/m3 (Roebber, P.J., Bruening, S.L., Schultz, D.M. and Cortinas Jr, J.V., 2003. Improving snowfall forecasting by diagnosing snow density. Weather and forecasting, 18(2), pp.264-287.)
Response 3: Indeed lack of knowledge about snow density also contributes to the uncertainty of the freeboard to thickness conversion. However its effect is supposed to be comparatively small as snow density largely varies seasonally ranging on average from 250 kg/m3 in September to 320 kg/m3 in May. Although snow density may locally deviate significantly from these values, the variability of the mean values over hundreds of kilometer scale used in our analysis for comparison with ULS data is considered to be not so important comparing to snow depth and sea ice density.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The manuscript presents a very technical study of possible correspondences between various parameters related to retrieval of sea ice radar altimetry products. The outcome of the study provides important results for algorithm improvements for present and future radar altimetry mission to observe sea ice thickness from space.
To follow the argumentation in the paper, a very high level of prior knowledge about sea ice radar altimetry is needed. Therefore, I recommend that the authors consider whether the manuscript can be simplified and still reach the same conclusions – which are important to share with the scientific and space mission communities.
The language of the manuscript can in general be improved and need to be thoroughly checked. Specific sentences need to be re-written and many prepositions need to be corrected.
I recommend acceptance with major revisions of the manuscript.
Further detailed suggestions for improvements are listed in the attached document.
Comments for author File: Comments.pdf
Author Response
Please see the attachment.
Dear Reviewer
We would like to thank you for detailed review that helped to improve the manuscript In the following, please, find the answers to your comments. Please also note that the changes in the revised manuscript are highlighted with the “Track Changes” function.
General comments: To follow the argumentation in the paper, a very high level of prior knowledge about sea ice radar altimetry is needed. Therefore, I recommend that the authors consider whether the manuscript can be simplified and still reach the same conclusions – which are important to share with the scientific and space mission communities.
The language of the manuscript can in general be improved and need to be thoroughly checked. Specific sentences need to be re-written and many prepositions need to be corrected.
Response to general comments: We agree that information given in the paper require a certain level of the expertise in sea ice radar altimetry. Therefore, we highly appreciate your detailed review of the manuscript with the comments and suggestions that highlighted the points that could be improved or clarified. We hope that changes that we made in the revised manuscript would help to make the perception of the paper easier for the reader.
We also thank you for your detailed suggestions in language editing.
Specific comments
Point 1: L. 20-21 Please elaborate more on the reasons for the expectation for physical interpretations
Response 1: Here we refer to the theoretical basis of the SAR-mode radar altimeter measurements discussed in details by Wingham et.al. [27]. The multi-looking approach applied in the SAR-mode operation of Cryosat-2 radar altimeter over sea ice provides measurements of surface elevation within the beam-limited footprint. Therefore, in contrast to using data from conventional pulse-limited radar altimeters, when the half-power point of the waveform leading edge corresponds to the surface, the 80% retracker threshold located closer to the maximum-power point better represents the mean surface for the SAR-mode data. Thus, stating that the 80% threshold is expected to better represent the surface elevation is related to theoretical basis, but not to the findings reported in the paper. In the manuscript, where this feature of the SAR-mode data is discussed, we provide the reference to the paper by Wingham et.al. [27], but we suggest that it is not reasonable to discuss these details in the abstract. However we add the note in the abstract that comparison of sea ice draft time series derived using 50% and 80% retracker thresholds is related to Cryosat-2 data.
Point 2: L. 36-46 Help the reader not very familiar with sea ice altimetry by defining/explaining the terms “draft”, “freeboard”, and “lead”
Response 2: We define the terms freeboard, lead and draft in the Introduction of the revised manuscript. Please, note that in response to the comment of another Reviewer we added some details about sources of uncertainties in sea ice thickness retrieval.
Point 3: L. 44 Rewrite to e.g. “…allow validation of temporal…”
Response 3: Corrected.
Point 4: L. 44 Rewrite to e.g. “…Since draft observations represent the main…”
Response 4: Corrected.
Point 5: L. 46 and -> but
Response 5: Corrected.
Point 6: L. 76 Please argue why a 200 km radius is chosen. What are the considerations regarding areal average versus local drift speeds of the sea ice in the region?
Response 6: The radius of 200 km was chosen following the approach used by Laxon et al. [18]. In addition, we formed draft time series using the 25 km radius that corresponds to grid size of the CCI product. Though with this smaller radius the results are, in general, similar, the formed time series are notably noisier. This may be caused by the lower number of available satellite altimeter measurements as well as due to the effect of the sea ice drift. The sea ice drift associated with the Beaufort gyre in the study area is about hundred kilometers per month. Although good consistency of spatial patterns between the means and change rates of altimeter parameters is indicated on Figures 4 and 5, using the 200 km radius allows to reduce the effect of the sea ice drift when comparing altimeter retrievals and ULS measurements.
In order to provide the basis for the choice of the radius in the revised manuscript we added the reference to paper by Laxon et al.
Point 7: L. 77-> Add uncertainty estimates of the data products used
Response 7: We specified the accuracy of the ULS measurements in Section 2.2 of the revised manuscript
Point 8: L. 96-> Please comment on the dependency between draft and freeboard and the implications for accuracy of the draft estimates
Response 8: In the Introduction section we describe that estimation of sea ice thickness using satellite altimetry is based on freeboard retrieval and on applying hydrostatic equilibrium of floating ice, and note the main sources of the uncertainties for this approach. We also note that the draft is the main portion of the sea ice thickness. In response to your comment we also added explanation of the terms ‘freeboard’ and ‘draft’. Therefore we presume that description given in the Introduction provide information about dependency between draft and freeboard with the sources of their uncertainties.
Point 9: L. 109 Subscript error for “draft ULS”
Response 9: Corrected.
Point 10: L. 114 Figure 2. The blue curve is difficult to distinguish from the black curve. Perhaps choose cyan or similar.
Response 10: In response to the comment of another reviewer we enlarged the Figure 2 and also slightly changed the scales for better visualization of the curves. We believe that the curves on the new version of the figure can be easier distinguished.
Point 11: L. 151 Figure 3. Include in the description of the figure that the plotted data is an average of all values within one month and 200 km radius. This could also be added to the figure caption in line 154.
Response 11: We added the phrase that monthly time series are formed in the vicinity of the moorings in the first paragraph of section 3.2 of the revised manuscript and edited the sentence appropriately. In the Materials and Methods section we note that all altimeter measurements are averaged within the radius 200 km from the moorings. But we are not sure that it is necessary to repeat this detail in the description of each figure with time series.
Point 12: L. 156 Rewrite to e.g. “…For most of the winter seasons an increase in the LeW…
Response 12: Corrected.
Point 13: L. 158 Rewrite to e.g. “…increase with increase in the fraction of deformed ice…” (Is this what is meant?)
Response 13: Yes, this is what we meant. Corrected.
Point 14: L 160 Help the reader by explaining why the negative difference in freeboard is to be expected
Response 14: The effect of lower freeboard estimates using 80% retracker threshold is seen also from the draft time series that is mentioned in the manuscript above – in the second paragraph of section 3.1. Therefore in the revised manuscript we add this explanation to the description of Figure 2 in section 3.1.
Point 15: L. 170 Rewrite to e.g. “…LeW (PP), which suggests a relationship between the inter-annual variability of these parameters”
Response 15: Corrected.
Point 16: L. 172-184 This section in particular need to be clarified.
Response 16: This paragraph notes the finding that is the basis of one of the paper conclusion that the difference between satellite-derived and ULS-measured sea ice draft seasonal growth depends on the surface properties preconditioned by the melt intensity during preceding summer. Probably the clarification of what we mean by the term ‘initial conditions’ is required. Therefore we corrected one sentence in this paragraph appropriately.
Point 17: L. 174 Rewrite to e.g. “…the mean and the change rate…”
Response 17: Corrected.
Point 18: L. 176 Rewrite to e.g. “…CRS data (marked with bold in Table 2).”
Response 18: Corrected.
Point 19: L. 183 Rewrite to e.g. “…winter show no significant change...”
Response 19: Corrected.
Point 20: L. 185 Table 2. Consider to inter-change columns “All moorings” and “Mooring A” for consistency with figure 2 and 3 (the same applies for Tables 3 and 4)
Response 20: Corrected.
Point 21: L. 191-192 Rewrite to e.g. “…that a strong relationship between the mean and the change of the surface characteristic over mooring sites is indicated despite the fact that we…”
Response 21: Corrected.
Point 22: L. 194 on -> in
Response 22: Corrected.
Point 23: L. 201 delete “are”
Response 23: Corrected.
Point 24: L. 203 “of the corresponding”
Response 24: Corrected.
Point 25: L. 205 Figure 4. The moorings are difficult to see – a bit larger, please
Response 25: The points depicting moorings are enlarged.
Point 26: L. 212 Rewrite to e.g. “…can also be noted, …”
Response 26: Corrected.
Point 27: L. 214-215 Rewrite to e.g. “This suggests that sea ice freeboard retrieval uncertainties to a large extent are able to explain the seasonal and…”
Response 27: Corrected.
Point 28: L. 241 uncertainty
Response 28: Corrected.
Point 29: L. 282-292 This section need to be clarified or perhaps skipped
Response 29: The first paragraph of this section continues the discussion presented in the previous paragraph about the difference between freeboard estimates derived using 50% and 80% retracker thresholds depending on the sea ice type. Here the effect of this difference on the sea ice draft estimates and their deviation from the ULS measurements are discussed.
The second paragraph of this section discusses one more possible reason of the underestimation of the SIT seasonal growth related to the methods of accounting for the lower speed of radar waves propagation in snow. We edited both paragraphs for clarification.
Point 30: L. 284 Rewrite to e.g. “…The time series show, as expected, that MYI fraction.
Response 30: We incorporated this phrase in the edited text.
Point 31: L. 293 The MYI fraction plotted is difficult to read – thicker dotted line please
Response 31: Corrected.
Point 32: L. 306 “In principle, …”
Response 32: Corrected.
Point 33: L. 312-313 Rewrite sentence
Response 33: We rewrote the sentence for clarification
Point 34: L. 314-315 “…could not result from…”
Response 34: Corrected.
Point 35: L. 316 Please explain why this is unlikely (because other literature shows the opposite might be the case)
Response 35: To our knowledge there are publications showing that Warren climatology overestimates snow depth due to decrease of the MYI fraction over the last decades (e.g. [24] or by Rostosky et al., [JGR Oceans, Vol. 123, 2018]). However we are not sure that there is evidence that Warren climatology systematically underestimates the seasonal snow depth growth.
Point 36: L. 317 Rewrite to e.g. “…over the considered time-periods of…”
Response 36: Corrected.
Point 37: L. 319 “…ice density result from …”
Response 37: Corrected.
Point 38: L. 322 “accumulated convergence considered as a measure of” could be deleted
Response 38: Corrected.
Point 39: L. 323 “blocky sea ice structures, one needs to…”
Response 39: Corrected.
Point 40: L. 326 “Therefore, as sea ice …”
Response 40: We corrected the phrase with “Therefore, since sea ice …” instead in order to avoid using word ‘as’ twice.
Point 41: L. 327 “…parameters, this may lead…”
Response 41: Corrected.
Point 42: L. 328 “…could also explain…”
Response 42: Corrected.
Point 43: L. 339 “…conditions of the preceding winter…”
Response 43: Here we mean “…conditions that precede winter seasons …”. We edited this phrase appropriately.
Point 44: L. 344 “When the ocean is covered by sea ice, the…”
Response 44: Corrected.
Point 45: L. 346 characteristic -> measure
Response 45: Corrected.
Point 46: L. 349 indicate -> show
Response 46: Corrected.
Point 47: L. 353 The sentence is unclear
Response 47: In this sentence we refer to MYI fraction time series on Figure 6. We edited this sentence for clarification
Point 48: L. 357 Rewrite to e.g. “…for the three considered satellite-derived sea ice draft estimates, Table 4 presents…”
Response 48: Corrected.
Point 49: L. 258-359 Why are the timeseries combined? Will that add more information or only additional uncertainty? Please explain the benefit of this choice.
Response 49: Combining time series allows evaluation of the relationships from the longer time series thereby increasing the confidence of the estimated correlations. Although combined time series are derived using data from different satellite missions the draft drEnv, as noted in Materials and Methods section, is estimated by the approach that minimizes inter-mission bias with respect to the drCS2_50 estimates [3].
Point 50: L. 397 “… and the difference between…”
Response 50: Corrected.
Point 51: L. 400 “…over the considered time…”
Response 51: Corrected.
Author Response File: Author Response.pdf
Reviewer 3 Report
The authors use upward looking sonar measurements from four moorings in the Beaufort Sea to examine relationships to satellite derived sea ice draft estimates, in order to elucidate the role that sea ice properties (and time series changes in properties) play on altimeter based thickness estimates by way of two commonly used retracker algorithms. The authors provide evidence of underestimation of seasonal sea ice draft during the winter sea ice growth season, and make sound linkages to the role that sea ice preconditioning and sea ice density plays in the underestimation.
Overall, the study is well laid-out, the methodology is mostly sound, and the results logically presented both visually and in the the text. The paper was well written and both informative and enjoyable to read. For example the statements on lines 178-184 were enjoyable to read. This study will be of interest to the remote sensing and sea ice communities, as efforts to improve sea ice thickness retrievals are ongoing, and essential for climate research.
Regarding the methodology, it is unclear about how it was decided to use the 200 km radius for comparison with the ULS measurements. The rationale for this choice should be given, as it begs the question as to what impact changing the size might have on the result. Also regarding the ULS measurements, what is the measurement error?
Additional comments:
- Where did the MYI fraction information come from?
- Can the authors make a recommendation in the conclusion for how systematic underestimating of sea ice density and its change over the winter season may be better understood and implemented into processing (if the latter is possible)?
Author Response
Please see the attachment.
Dear Reviewer
We would like to thank you for your comments that helped to improve the manuscript. In the following, please, find the answers to your comments. Please also note that the changes in the revised manuscript are highlighted with the “Track Changes” function.
Point 1: Regarding the methodology, it is unclear about how it was decided to use the 200 km radius for comparison with the ULS measurements. The rationale for this choice should be given, as it begs the question as to what impact changing the size might have on the result.
Response 1: The radius of 200 km was chosen following the approach used by Laxon et al. [18]. In addition, we formed draft time series using the 25 km radius that corresponds to grid size of the CCI product. Though with this smaller radius the results are, in general, similar, the formed time series are notably noisier. This may be caused by the lower number of available satellite altimeter measurements as well as due to effect of the sea ice drift. The sea ice drift associated with the Beaufort gyre in the study area is about hundred kilometers per month. Although good consistency of spatial patterns between the means and change rates of altimeter parameters is indicated (Figures 4 and 5), using the 200 km allows to reduce the effect of the sea ice drift when comparing altimeter retrievals and ULS measurements. In order to provide the basis for the choice of the radius we added the reference to paper by Laxon et al.
Point 2: Also regarding the ULS measurements, what is the measurement error?
Response 2: The ULS measurement errors are added to the data description in Materials and Methods section of the revised manuscript.
Point 3: Where did the MYI fraction information come from?
Response 3: We use data on MYI fraction provided by the Ocean and Sea Ice Satellite Application Facility and used for the SIT retrieval in the course of satellite altimeter data processing. This information is provided in Section 3.3 where data on MYI fraction are first mentioned in the paper.
Point 4: Can the authors make a recommendation in the conclusion for how systematic underestimating of sea ice density and its change over the winter season may be better understood and implemented into processing (if the latter is possible)?
Response 4: Indeed, information about underestimation of sea ice density and its change during winter could be valuable for the SIT retrieval. However, unfortunately, presently we can only speculate and make assumptions about it. Further thorough analysis like that presented by Belter et al. [15] is required to make conclusion and recommendation.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Thank you for the revised manuscript.
The manuscript has been sufficiently improved and I recommend it for publishing.
A few very minor corrections added to the attached document.
Comments for author File: Comments.pdf
