the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
A decade-plus of Antarctic sea ice thickness and volume estimates from CryoSat-2 using a physical model and waveform-fitting
Steven Fons
Nathan Kurtz
Marco Bagnardi
Abstract. We utilize a physical waveform model and a waveform-fitting method to estimate the snow depth and snow freeboard of Antarctic sea ice from CryoSat-2, and use these estimates to calculate the sea ice thickness and volume over an 11+ year time series. We compare our snow depth and thickness estimates to other altimetry- and ship-based observations, and find good agreement overall with some discrepancies in certain regions and seasons. The time series is used to calculate trends in the data, and we find small but statistically significant negative trends in the Ross Sea autumn (-0.3 cm yr-1), the Eastern Weddell winter (-0.8 cm yr-1), and the Western Weddell autumn and annual-average (-2.6 and -1.6 cm yr-1, respectively). Significant positive trends are found in the pan-Antarctic summer (0.4 cm yr-1) and Amundsen-Bellingshausen winter and annual-average (2.3 and 0.9 cm yr-1, respectively). Though pan-Antarctic trends in sea ice thickness and volume are small between 2010–2021, we find larger-magnitude trends regionally and since 2014. We place these thickness estimates in the context of a longer-term, snow-freeboard-derived, laser-radar sea ice thickness time series that began with ICESat and continues with ICESat-2. Reconciling and validating this longer-term, multi-sensor time series will be important in better understanding changes in the Antarctic sea ice cover.
Steven Fons et al.
Status: closed
-
RC1: 'Comment on egusphere-2022-1287', Anonymous Referee #1, 09 Jan 2023
Review of
A decade-plus of Antarctic sea ice thickness and volume estimates from CryoSat-2 using a physical model and waveform-fitting
by
Fons, S., et al.
Summary:
This paper exploits an innovative method to use CryoSat-2 radar altimeter observations to retrieve first freeboard and snow thickness on sea ice before using both to estimate the sea ice thickness and volume - including a credible estimation of the uncertainty. The method has been developed and published in different publications and is therefore not described in-depth in this contribution. Here the focus lies in the illustration and discussion of pan-Antarctic and regional distributions of sea ice thickness and sea ice volume as based on 11 years of CryoSat-2 data - including a trend analysis.Certainly this is an interesting and also important piece of work which broadens our knowledge about the thickness and volume distribution of Antarctic sea ice.
General Comments:
GC1: You dedicate quite some part of your paper to a trend analysis. The relevance of this trend analysis is not sufficiently well motivated and not put into a credible context with the overall variability of both the Antarctic sea ice cover and its influencing factors. The added value of this analysis is not fully convincing. See my specific comments in this regard.GC2: Your paper contains only few elements of inter-comparing your product(s) with other, independent results. Here I feel your paper has substantial potential for improvement. On the one hand, the discussion included so far in the paper based on the comparisons carried out would strongly benefit from a more critical view of i) the limitations of the intercomparison data sets used and ii) a more careful investigation and discrimination of level versus deformed sea ice and/or mean versus modal sea ice thickness values. On the other hand, key intercomparison data sets are left out, kind of limiting the credibility of the results presented - especially when keeping in mind that the authors' estimation of freeboard, snow thickness and sea ice thickness are not independent and therefore require an even more careful evaluation. See my specific comments for more information.
Specific Comments:
Abstract:
To my opinion, the abstract should contain a bit more information about the method, the product and its evaluation and less detailed information about the trend analysis results - simply because this is a short time series in a highly variable environment, possibly requiring 30+ years to derive any reliable trend information. See also GC1.L18: All three references given relate to the Arctic. I have i) difficulties to understand your choice to not directly focus on Antarctic conditions - as this is the focus of your paper and it does not read well to introduce / motivate an Antarctic focus paper with exclusively Arctic focus referenes - and ii) even for the Arctic the selection of the references given seems rather arbitrary, missing out several of the more recent literature that is available. I recommend to revise the references.
Figure 1: I know, this is just a schematic figure. However, it wrongly implies that the part of the sea ice underneath the water surface is as thick as the part of the sea ice above the water surface. In addition, the thickness of the snow load almost certainly would lead to flooding of the ice-snow interface. Therefore, for the sake of displaying a more realistic schematic figure - that even lecturers might want to take from your paper - I recommend to replace this figure by one which has more realistic dimensions.
In the caption, h_fs is not mentioned yet.L71: "... these estimates are both assumed to be biased high ..." --> There has been a Cryosphere Discussion paper around for a while (tc-2021-227 by Wang et al.); if I recall correctly they took an independent look (not from the producer's side) at the ESA sea ice thickness product. It might be worth a look.
L121-125: ICESat had only several dedicated measurement periods while ICESat-2 has been operated continuously. I therefore assume the climatology maps have a different number of months as their baseline, i.e. for August or December it is possibly mostly ICESat-2 - aka data from 1 or 2 years, respectively, while for March it is data from one ICESat-2 year and five ICESat years. I recommend to include a short table detailing this difference in representativity of the climatology freeboard maps of the different months.
Not clear as well is how the different coverage of ICESat measurements over different months is taken into account in the respective monthly mean. Often these maps are bi-monthly maps derived e.g. half from February and half from March. How is this realized in your climatology? Did you use a Feb/Mar ICESat map for both February and March?
L131: "area of the grid cell" --> please provide the information where you obtained the grid cell areas from. Since this data is on a polar-stereographic projection the grid cell area varies with latitude and you possibly downloaded and used the respective file from NSIDC (?)
L137-139: Undoubtly the Worby et al. (2008) data set is a benchmark in this direction. I note, however, that it terminates in March 2005. Have you considered to take a look at the extension of this data set available here: https://www.cen.uni-hamburg.de/en/icdc/data/cryosphere/seaiceparameter-shipobs.html ?
Table 1: In the text you state "angular backscatter efficiency"; I suggest to use the same expression in the table.
I note that snow depth seems to be given in cm while the roughness has a different unit. You could consider harmonizing this.
I am a bit puzzled about the bounds. For sigma you state bounds 0-1m; I assume this means that sigma is allowed to range between 0 and 1 m. However, for snow depth you specify a plus/minus range around the values suggested by the climatology rather than a range such as specified for sigma - otherwise the snow depth would need to range between -30 cm and +30 cm. Even when it is the range around the climatology values (which I assume) I am wondering what happens at a snow depth of 5 cm.
What is "std"?
I note that the static parameters don't have bounds even though you apply the retrieval year-round and backscatter / extinction characteristics of snow and ice may change throughout the year. Would it therefore make sense to introduce bounds here as well?
L175: "or until ... is reached" --> What are the output parameters taken in this case? The very last one? Also, I asssume that finding a minimum residual results in a quantitatively better fit. How often does the retrieval needs to reach 100 evaluations compared to finding an adequate minimum?
L183: "the nominal tracking bins" ... what is their difference and hence an approximation of the vertical resolution of the approach?
L207: "if at least three lead-type points exists within" --> I am sorry for asking this question but can successive points overlap or are they truly independent, i.e. adjacent footprints do not overlap because the along-track distance between their centers is larger than the along-track dimension of the footprint?
L237: I could have asked this question earlier in the context of the SSH approximation: For that approximation you need a minimum of 3 valid points within a 10-km segment. And then you first compute the parameters mentioned along track, i.e. for each valid floe-type point along-track, and then perform the gridding?! For the latter, does that have to be a minimum of valid floe-type points from which the parameters mentioned are computed? I can imagine that there are seasons and regions where you may have quite a number of valid lead-type points and a lot of mixed-type points but only few floe-type points.
Finally, the SSH derived is representative for 10-km segments, i.e. in the worst case a step-function in SSH, or is this derived using a running 10-km segment possibly providing a smoother representation of the SSH?
L278/279: While following the approach of Spreen et al. (2009) is at first place good, I am wondering whether it would make sense to back this value up by looking into sea-ice concentration uncertainty information that is provided with the OSI-450/OSI-430-b CDR/iCDR sea ice concentration data set (which includes smearing uncertainty contributions) or with the NOAA/NSIDC SIC CDR (even though this is basically a modified standard deviation)? Another source you could look into in this regard is this one:
https://egusphere.copernicus.org/preprints/2022/egusphere-2022-1189/L310: "more validation data" --> I would call your comparison to the KK20 product an inter-comparison. It is not a validation as the KK20 data is just one possible result of combining two satellite data sets to obtain a snow thickness product. For an evaluation or validation one would need ground-truth data which the KK20 data certainly is not. Hence, my suggestion is to stress that we need "validation data" (hence delete "more" as this implies that the KK20 data are already validation data) of the type ground-based measurements to really be in the position to perform a validation.
Figure 3: I am wondering whether it would make sense to provide an estimate of the actual number of values per month as I would assume that the number grid cells contributing to a February value in the Indian Ocean sector differs considerably from the region W-Weddell.
I note that the name for sector "Amundsen-Bellingshausen Sea" has different flavors. Consider using one.L334-338: "In Fig. 4 ... in Fig. 3)." --> How do these results compare to the snow thickness retrievals based on ICESat data in Kern and Ozsoy-Cicek (2016)? Didn't they also show an increase in the average snow thickness from autumn to winter to spring - in contrast to the snow thickness values derived using a modified version of the Markus and Cavalieri (1998, online 2013) approach?
Figure 4: The dashed line in the inset histogram denotes what?
The legend of the maps as well as the histograms is in meters. I suggest to then also set the binsize to 0.02 m.L355 / Figure 6: I am wondering about the real information content of these probability distribution values given the fact that the number of observations per region / season varies so much. Did you consider normalizing the histograms to 1?
Figure 5: Same comments as I had for Figure 4.
In addition: please swap "Spring" and "Autumn" below the bottommost row of panels.L363-368: "Despite ..." --> I have repeatedly used the Worby et al. (2008) data set (and its extension mentioned further up in my comments) for inter-comparison purposes and am well aware of its value. I am wondering, however, whether some additional information needs to be given here to underline how vague that information can be. These data have an observational negative bias because ships tend to avoid thicker sea ice. In summer, floes break different under the action of the ship's hull reducing integrity of level ice made of rafted ice floes. In summer, pancake ice which makes a substantial fraction of the observed sea ice cover, is essentially lacking. Leads, often followed by the ships, are covered by thinner ice types in summer than the freezing season. Also, in addition to these observational biases there could be biases by the regions traversed during the different seasons. For instance in the Ross sector, cruises hardly reached to the thicker sea ice parts in the eastern Ross Sea during winter and spring, simply because these areas are not accessible, but rather crossed the thinner sea ice in the sea ice export area of the Ross Ice Shelf polynya. Therefore, especially the high sea ice thickness reported in Worby et al. (2008) for summer could very well be caused by preferably entering areas with thicker sea ice compared to winter and spring. See GC2.
Figure 6: Are the values from Worby et al. those of the level ice or do these include the estimated contribution from ridged sea ice?
Figure 7: While I was trying to understand why I have the impression that the individual mean sea ice thickness values do not add up to the pan-Antarctic mean sea ice thickness value I figured out that the scales are not the same. How important is it (for your message) to show the pan-Antarctic sea ice thickness occupying more vertical space in the figure than the individual sea ice thickness time series? Would it make sense to try to show the time series for each region with the same vertical scale?
L390/391: "while Maksym ... thickness" --> This sounds like they used satellite microwave radiometry to estimate sea ice thickness but what they did is first of all not that simple and secondly their main statement refers to level, undeformed sea ice which is not 1-to-1 comparable to your work. I therefore invite you to check the reference one more time and to rephrase your sentence accordingly. It is important to check out which part of the sea-ice thickness distribution the respective publications refer to to be able to make appropriate statements here. In this context it might be a good idea to, in addition, introduce a discussion of modal sea ice thickness values representative of the level sea ice.
L398/399: "Williams et al. ..." --> You might find it enlightening to again take a look into Kern et al. (2016). Even though their 1-layer method results are possibly biased and rather refer to the total (sea ice plus snow) than the "true" sea ice thickness, the intercomparison of the other methods (including the ZIF) seems to provide a possible range (at least for the ICESat measurement period) of sea-ice thickness values obtained using different methods.
Section 4.4: In light of the substantially larger (and known) variation of the Antarctic sea ice cover - compared to the Arctic - I have a conceptual problem with dedicating a full sub-section to a trend analysis of an eleven years long time series. This looks like somebody wants to investigate an eleven years long precipitation time series of the U.K. in light of trends. But it is of course your decision to keep or delete this part of the manuscript. In case you keep it I strongly recommend to - beyond statistical significance estimates - state clearly that any trend found for these eleven years can simply be the part of a multi-decadal variation that cannot be resolved yet with the existing record of CrysoSat-2 sea ice thickness and volume observations. This would be a good motivation to i) discuss your results even more in the context of the work of other studies; ii) to advertize more work needs to be done to include Envisat and ERS1/2 RA altimeter data analysis to extent the time-series; iii) to advertize your own sub-section about expanding the CS-2 time series back in time to the ICESat periods.
L420: "contains at least four years of data" --> Is there any constraint as to when these four years need to contain data? Is it possible that all data are from the first 4 years?
L432: "Holland (2014)" --> How many years of your 11-year period overlap with the data used by Holland (2014)? Are those results therefore compatible with your results?
L432/434: Both, Garnier et al., (2022) and Xu et al. (2021) used data from a longer time series, didn't they? What is then the added value of performing such a trend analysis over a shorter time period? This is not entirely clear to me.
L441-443: How many years of CryoSat-2 data did Kwok and Cunningham (2015) use in their analysis? I checked it out: It is four winters. You investigate 11 years. I don't think your current writing (and citing that paper) does support further discussing the impact of an analysis of 11-years worth of sea ice thickness and volume in the Southern Ocean.
Below in this sub-section you will find more comments going into this direction. All I wish to trigger with these is to encourage you to one more time critically think whether the message you provide here is compelling, sustainable and worth the effort. Does it send out the right signal in view of already existing work and in view of what we know about the length time series of geophysical parameters should have in order to provide a meaningful statement about climatological features such as trends? See GC1.L455-457: "However, the same ... since 2014" --> Certainly. And if you shorten the time period even further, e.g. to a 4-years like Kwok and Cunningham did, then you will find an even larger decrease in sea ice thickness or volume for 2014-2017 while you may find an increase in sea ice thickness or volume for 2011-2014 and 2017-2020. Fine. And?
I find it kind of dangerous to refine the temporal granularity of such trend analysis in an area such as the Southern Ocean being influenced by at least three multi-decadal oscillations plus El Nino/La Nina events. I agree, Kwok and Cunningham (2015) did it with an even shorther time series, Kurtz and Markus (2012) as well ... but what did we learn from these?L463/464: "modeled studies into ... scenarios" --> Certainly. But this is not a surprizing finding and, in addition, it requires first some more work still to be done improving those models - see Roach et al., 2020, Geophys. Res. Lett., 47, who for good reason first looked at the Antarctic sea ice area in CMIP6 models finding it not well represented.
L472/473: "A longer-term time series ... implications." --> Exactly. Two other studies exist (almost certainly there are more in the meantime) that already looked into longer time series which complicates to see the immediate added value of your investigation in comparison to their studies.
L497-501: "Likely ... estimate Antarctic sea ice thickness." --> I have two comments here. The first one is related to whether you also looked into the work of Ricker et al., 2015, Impact of snow accumulation on CryoSat-2 range retrievals over Arctic sea ice: An observational approach with buoy data, Geophys. Res. Lett., 42. While being for Arctic conditions that work might be further enlightening with respect to your observations.
The second comment is about the observation that in the time series of h_i_70 the primary maximum mean sea ice thickness is not occurring in February anymore but occurs in late winter / spring in all but one year. What does this tell us in light of the fact that the primary maximum now occurs close to the maximum sea ice coverage - involving a large fraction of seasonal sea ice with different surface properties than encountered in February?
L505/506: "could come from ... its lifetime" --> I am aware of these changes but at the same time I am wondering i) which release of the ICESat GLAS data you used for your re-processing of the ZIF sea ice thickness values and ii) whether you did not correct for the different gain values that are reported along with the ICESat data?
L525 / Section 5:
I absolutely agree with you that it would be really nice to have ground-based observations that cover all three sensors' observation period. But we know that this is not possible. The only data sets I am aware of that covers all three sensors contain only estimates of the sea ice thickness: the ASPEcT data set and its extension mentioned further up. Arctic studies often tend to look into PIOMAS to see whether there is long-term consistency in the estimates. I am not deep enough involved into such studies to know whether GIOMAS data would be a viable alternative for the Southern Ocean.However, apart from these considerations, I am missing a more thoughtful evaluation of your sea ice thickness data / product for the CryoSat-2 period used. I have several concerns. One is the apparent lack of adequately discriminating between modal (level) and mean (level + deformed) sea ice thickness values in those parts of your intercomparisons where such a discrimination would be possible (e.g. the Worby et al., 2008 data). In that context I note again that you could have used the extended version of these data noted earlier in addition - even though these do not contain this discrimination into level and level+deformed ice.
In this context I would like to remind you to adequately discuss the limitations of the data you used for your intercomparisons presented in this manuscript - as voiced further up in the context of the Worby et al. (2008) data.
What I am missing is consideration of Operation Ice Bridge data in your evaluation and discussion of the quality in this manuscript. There is a substantial amount of data available and even though flights mostly cover the Weddell and Bellingshausen Seas these are nevertheless a very valuable source for the evaluation of your product. Other air-borne data exist, such as helicopter-borne electromagnetic sounding but I am in fact not sure how many of these would be available within the CryoSat-2 period. For sure researchers organized from New Zealand obtained data in the southern Ross Sea.
In short, in view of recommendations I conveyed to other authors with a similar manuscript profile my main recommendation for you and your section 5 is to put more emphasis on more critically discussing the reliability of your results rather than discussing trends.
Editoral Comments / Typos:
L25: "snow freeboard" --> You could add that here the assumption is that the dominant scattering comes from the snow surface.Equation 2: I recommend to add the information that the second term actually results in a reduction of the sea ice thickness computed by the first term alone - which is opposite to Equation 1 - and which particularly in the Antarctic - the focus of your paper - is important to consider as snow freeboard might equal the snow thickness or may even be smaller than that in case of flooding.
"Kurtz and Markus, 2012" and "Kwok, 2011" are references in which one can find these two equations - however, I am wondering whether it wouldn't make more sense to go back to those publications where these equations were developed / introduced first ... which might be the Laxon et al. paper from 2003 in case of Equation 1 and one of the earlier Kwok (et al.) papers for Equation 2.
L36: ICESat facilitated "snow freeboard" measurements. Please correct.
L39-42: "In most of these ... Kurtz et al., 2009)" --> I suggest to place the Warren et al. reference behind "1954-1991"; otherwise it reads as if Warren et al. (1999) have used that climatology to convert freeboard to thickness.
I further suggest to not highlight that Kurtz et al. (2009) used snow thickness data from passive microwave sensors (which by the way do not provide "lower resolution" snow thickness data compared to the Warren et al climatology being based on interpolation using a polynomial function anyways) - simply because this is just one of the alternatives used by the various other groups already cited. How important it is for the Antarctic focus of your paper to introduce the reader to potential alternatives to the Warren et al. climatology which is not existing in the Antarctic?
L43: ICESat-2 --> Did you overlook the contributions by Kwok (et al.) - who also combined Cryosat-2 and ICESat-2 - on purpose here?
L45: "have found success in estimating sea ice freeboard over Arctic sea ice" --> In light of the fact that most of the studies you cited had freeboard-to-thickness conversion as their ultimate aim, I am wondering whether you might want to rephrase this along the lines: "were succesful in retrieving sea ice thickness from sea ice freeboard estimates over Arctic sea ice" ... or the like.
L57: "Markus and Cavalieri, 2013" --> This is the electronic version of the original book chapter from 1998, right? Has the content changed? If not, please check with EGUSphere how to cite to avoid the impression that this is a more recent work.
L74: "through the use of key snow depth assumptions" --> I might be wrong but it is only the Kurtz and Markus (2012) work which does this assumption. I therefore suggest to add something like "partly" or "for example" to make clear that assuming zero freeboad is ONE possible solution - with limited applicability though as one can figure out in the subsequently cited by you literature.
L78: "Zero ice freeboard ..." --> In addition to citing Willatt et al (2010) you could also include Ozsoy-Cicek et al. 2013, JGR-Oceans.
L79: Regarding this underestimation you could have cited the earlier study by Kwok and Maksym from 2014 (JGR-Oceans) using OIB data; also Kern et al. (2016) performed intercomparisons between different retrieval approches, Kurtz and Markus being on of these.
L95 "utilize CryoSat-2" --> It would not hurt to also mention Envisat here because with that one would have an uninterrupted time series produced using an independent sensor from 2003 through today (see also Paul et al., 2018).
L127: "is based off of" --> I would have written "is based on" ... but I am not a native English speaker ...
L160: You might want to change the font of P, I and p so that it matches "v" and equation (4).
L187: "R_n" needs to be "R_0" ?
L197: I am not sure I would throw the Schwegmann et al. paper into one pot with the Paul et al one because the latter used a considerably modified methodology. Hence citing Paul et al might be sufficient here.
L239-241: You could consider to delete the information about which algorithm you used and how you compute the sea ice area because you described this earlier.
L247: You might want to add that the demarcation in longitude is given in the unit "degrees East".
L262: The "s" in h_fs needs to be put in sub-script mode.
L267: "h_f" at the end of the line needs to be "h_fs"?
L293: "sections" --> "section"
L301: What is "IQR"?
L308/309: "Despite ... nevertheless" ... I guess one of these is enough; I'd discard the "nevertheless".
L325/326: "This could cause ... anomalous snow depths" --> I am wondering whether you could narrow this down towards that the snow-ice interface will most likely be located higher in the snow pack and then also state that this will lead to anomalously low snow thickness values?
L379/380: You are refering to "basal growth" here. For Southern Ocean sea ice a substantial portion of the sea ice volume (up to 1/3 in some places) is actually made of snow ice, i.e. snow that was first flooded at the ice-snow interface and then re-froze. This is not a basal growth. One solution could be to simply write "growth".
L384: "in ice thickness" --> in pan-Antarctic sea ice thickness"
"more that" --> "more than"L394: A reader would be happy to be reminded what this correction factor does and when it is applied.
L395: "was estimates" --> "was estimated"
L459: Sometime you use pan-Antarctic with a capital "P" sometimes not. You might decide for one version of how to write it. I don't know actually what would be correct grammatically.
L484: "about" --> "around"
L510: "h_i-total" is what?
L550/551: "It is clear that ... sea ice thickness," --> I encourage you to also include "snow thickness" here.
L611: You might want to replace this reference by the paper published in Earth and Space Science, 8(7), 2021 to have the link to the peer-reviewed version of your work.
Citation: https://doi.org/10.5194/egusphere-2022-1287-RC1 - AC1: 'Reply on RC1', Steven Fons, 09 Mar 2023
-
RC2: 'Comment on egusphere-2022-1287', Anonymous Referee #2, 20 Jan 2023
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2022/egusphere-2022-1287/egusphere-2022-1287-RC2-supplement.pdf
- AC2: 'Reply on RC2', Steven Fons, 09 Mar 2023
Status: closed
-
RC1: 'Comment on egusphere-2022-1287', Anonymous Referee #1, 09 Jan 2023
Review of
A decade-plus of Antarctic sea ice thickness and volume estimates from CryoSat-2 using a physical model and waveform-fitting
by
Fons, S., et al.
Summary:
This paper exploits an innovative method to use CryoSat-2 radar altimeter observations to retrieve first freeboard and snow thickness on sea ice before using both to estimate the sea ice thickness and volume - including a credible estimation of the uncertainty. The method has been developed and published in different publications and is therefore not described in-depth in this contribution. Here the focus lies in the illustration and discussion of pan-Antarctic and regional distributions of sea ice thickness and sea ice volume as based on 11 years of CryoSat-2 data - including a trend analysis.Certainly this is an interesting and also important piece of work which broadens our knowledge about the thickness and volume distribution of Antarctic sea ice.
General Comments:
GC1: You dedicate quite some part of your paper to a trend analysis. The relevance of this trend analysis is not sufficiently well motivated and not put into a credible context with the overall variability of both the Antarctic sea ice cover and its influencing factors. The added value of this analysis is not fully convincing. See my specific comments in this regard.GC2: Your paper contains only few elements of inter-comparing your product(s) with other, independent results. Here I feel your paper has substantial potential for improvement. On the one hand, the discussion included so far in the paper based on the comparisons carried out would strongly benefit from a more critical view of i) the limitations of the intercomparison data sets used and ii) a more careful investigation and discrimination of level versus deformed sea ice and/or mean versus modal sea ice thickness values. On the other hand, key intercomparison data sets are left out, kind of limiting the credibility of the results presented - especially when keeping in mind that the authors' estimation of freeboard, snow thickness and sea ice thickness are not independent and therefore require an even more careful evaluation. See my specific comments for more information.
Specific Comments:
Abstract:
To my opinion, the abstract should contain a bit more information about the method, the product and its evaluation and less detailed information about the trend analysis results - simply because this is a short time series in a highly variable environment, possibly requiring 30+ years to derive any reliable trend information. See also GC1.L18: All three references given relate to the Arctic. I have i) difficulties to understand your choice to not directly focus on Antarctic conditions - as this is the focus of your paper and it does not read well to introduce / motivate an Antarctic focus paper with exclusively Arctic focus referenes - and ii) even for the Arctic the selection of the references given seems rather arbitrary, missing out several of the more recent literature that is available. I recommend to revise the references.
Figure 1: I know, this is just a schematic figure. However, it wrongly implies that the part of the sea ice underneath the water surface is as thick as the part of the sea ice above the water surface. In addition, the thickness of the snow load almost certainly would lead to flooding of the ice-snow interface. Therefore, for the sake of displaying a more realistic schematic figure - that even lecturers might want to take from your paper - I recommend to replace this figure by one which has more realistic dimensions.
In the caption, h_fs is not mentioned yet.L71: "... these estimates are both assumed to be biased high ..." --> There has been a Cryosphere Discussion paper around for a while (tc-2021-227 by Wang et al.); if I recall correctly they took an independent look (not from the producer's side) at the ESA sea ice thickness product. It might be worth a look.
L121-125: ICESat had only several dedicated measurement periods while ICESat-2 has been operated continuously. I therefore assume the climatology maps have a different number of months as their baseline, i.e. for August or December it is possibly mostly ICESat-2 - aka data from 1 or 2 years, respectively, while for March it is data from one ICESat-2 year and five ICESat years. I recommend to include a short table detailing this difference in representativity of the climatology freeboard maps of the different months.
Not clear as well is how the different coverage of ICESat measurements over different months is taken into account in the respective monthly mean. Often these maps are bi-monthly maps derived e.g. half from February and half from March. How is this realized in your climatology? Did you use a Feb/Mar ICESat map for both February and March?
L131: "area of the grid cell" --> please provide the information where you obtained the grid cell areas from. Since this data is on a polar-stereographic projection the grid cell area varies with latitude and you possibly downloaded and used the respective file from NSIDC (?)
L137-139: Undoubtly the Worby et al. (2008) data set is a benchmark in this direction. I note, however, that it terminates in March 2005. Have you considered to take a look at the extension of this data set available here: https://www.cen.uni-hamburg.de/en/icdc/data/cryosphere/seaiceparameter-shipobs.html ?
Table 1: In the text you state "angular backscatter efficiency"; I suggest to use the same expression in the table.
I note that snow depth seems to be given in cm while the roughness has a different unit. You could consider harmonizing this.
I am a bit puzzled about the bounds. For sigma you state bounds 0-1m; I assume this means that sigma is allowed to range between 0 and 1 m. However, for snow depth you specify a plus/minus range around the values suggested by the climatology rather than a range such as specified for sigma - otherwise the snow depth would need to range between -30 cm and +30 cm. Even when it is the range around the climatology values (which I assume) I am wondering what happens at a snow depth of 5 cm.
What is "std"?
I note that the static parameters don't have bounds even though you apply the retrieval year-round and backscatter / extinction characteristics of snow and ice may change throughout the year. Would it therefore make sense to introduce bounds here as well?
L175: "or until ... is reached" --> What are the output parameters taken in this case? The very last one? Also, I asssume that finding a minimum residual results in a quantitatively better fit. How often does the retrieval needs to reach 100 evaluations compared to finding an adequate minimum?
L183: "the nominal tracking bins" ... what is their difference and hence an approximation of the vertical resolution of the approach?
L207: "if at least three lead-type points exists within" --> I am sorry for asking this question but can successive points overlap or are they truly independent, i.e. adjacent footprints do not overlap because the along-track distance between their centers is larger than the along-track dimension of the footprint?
L237: I could have asked this question earlier in the context of the SSH approximation: For that approximation you need a minimum of 3 valid points within a 10-km segment. And then you first compute the parameters mentioned along track, i.e. for each valid floe-type point along-track, and then perform the gridding?! For the latter, does that have to be a minimum of valid floe-type points from which the parameters mentioned are computed? I can imagine that there are seasons and regions where you may have quite a number of valid lead-type points and a lot of mixed-type points but only few floe-type points.
Finally, the SSH derived is representative for 10-km segments, i.e. in the worst case a step-function in SSH, or is this derived using a running 10-km segment possibly providing a smoother representation of the SSH?
L278/279: While following the approach of Spreen et al. (2009) is at first place good, I am wondering whether it would make sense to back this value up by looking into sea-ice concentration uncertainty information that is provided with the OSI-450/OSI-430-b CDR/iCDR sea ice concentration data set (which includes smearing uncertainty contributions) or with the NOAA/NSIDC SIC CDR (even though this is basically a modified standard deviation)? Another source you could look into in this regard is this one:
https://egusphere.copernicus.org/preprints/2022/egusphere-2022-1189/L310: "more validation data" --> I would call your comparison to the KK20 product an inter-comparison. It is not a validation as the KK20 data is just one possible result of combining two satellite data sets to obtain a snow thickness product. For an evaluation or validation one would need ground-truth data which the KK20 data certainly is not. Hence, my suggestion is to stress that we need "validation data" (hence delete "more" as this implies that the KK20 data are already validation data) of the type ground-based measurements to really be in the position to perform a validation.
Figure 3: I am wondering whether it would make sense to provide an estimate of the actual number of values per month as I would assume that the number grid cells contributing to a February value in the Indian Ocean sector differs considerably from the region W-Weddell.
I note that the name for sector "Amundsen-Bellingshausen Sea" has different flavors. Consider using one.L334-338: "In Fig. 4 ... in Fig. 3)." --> How do these results compare to the snow thickness retrievals based on ICESat data in Kern and Ozsoy-Cicek (2016)? Didn't they also show an increase in the average snow thickness from autumn to winter to spring - in contrast to the snow thickness values derived using a modified version of the Markus and Cavalieri (1998, online 2013) approach?
Figure 4: The dashed line in the inset histogram denotes what?
The legend of the maps as well as the histograms is in meters. I suggest to then also set the binsize to 0.02 m.L355 / Figure 6: I am wondering about the real information content of these probability distribution values given the fact that the number of observations per region / season varies so much. Did you consider normalizing the histograms to 1?
Figure 5: Same comments as I had for Figure 4.
In addition: please swap "Spring" and "Autumn" below the bottommost row of panels.L363-368: "Despite ..." --> I have repeatedly used the Worby et al. (2008) data set (and its extension mentioned further up in my comments) for inter-comparison purposes and am well aware of its value. I am wondering, however, whether some additional information needs to be given here to underline how vague that information can be. These data have an observational negative bias because ships tend to avoid thicker sea ice. In summer, floes break different under the action of the ship's hull reducing integrity of level ice made of rafted ice floes. In summer, pancake ice which makes a substantial fraction of the observed sea ice cover, is essentially lacking. Leads, often followed by the ships, are covered by thinner ice types in summer than the freezing season. Also, in addition to these observational biases there could be biases by the regions traversed during the different seasons. For instance in the Ross sector, cruises hardly reached to the thicker sea ice parts in the eastern Ross Sea during winter and spring, simply because these areas are not accessible, but rather crossed the thinner sea ice in the sea ice export area of the Ross Ice Shelf polynya. Therefore, especially the high sea ice thickness reported in Worby et al. (2008) for summer could very well be caused by preferably entering areas with thicker sea ice compared to winter and spring. See GC2.
Figure 6: Are the values from Worby et al. those of the level ice or do these include the estimated contribution from ridged sea ice?
Figure 7: While I was trying to understand why I have the impression that the individual mean sea ice thickness values do not add up to the pan-Antarctic mean sea ice thickness value I figured out that the scales are not the same. How important is it (for your message) to show the pan-Antarctic sea ice thickness occupying more vertical space in the figure than the individual sea ice thickness time series? Would it make sense to try to show the time series for each region with the same vertical scale?
L390/391: "while Maksym ... thickness" --> This sounds like they used satellite microwave radiometry to estimate sea ice thickness but what they did is first of all not that simple and secondly their main statement refers to level, undeformed sea ice which is not 1-to-1 comparable to your work. I therefore invite you to check the reference one more time and to rephrase your sentence accordingly. It is important to check out which part of the sea-ice thickness distribution the respective publications refer to to be able to make appropriate statements here. In this context it might be a good idea to, in addition, introduce a discussion of modal sea ice thickness values representative of the level sea ice.
L398/399: "Williams et al. ..." --> You might find it enlightening to again take a look into Kern et al. (2016). Even though their 1-layer method results are possibly biased and rather refer to the total (sea ice plus snow) than the "true" sea ice thickness, the intercomparison of the other methods (including the ZIF) seems to provide a possible range (at least for the ICESat measurement period) of sea-ice thickness values obtained using different methods.
Section 4.4: In light of the substantially larger (and known) variation of the Antarctic sea ice cover - compared to the Arctic - I have a conceptual problem with dedicating a full sub-section to a trend analysis of an eleven years long time series. This looks like somebody wants to investigate an eleven years long precipitation time series of the U.K. in light of trends. But it is of course your decision to keep or delete this part of the manuscript. In case you keep it I strongly recommend to - beyond statistical significance estimates - state clearly that any trend found for these eleven years can simply be the part of a multi-decadal variation that cannot be resolved yet with the existing record of CrysoSat-2 sea ice thickness and volume observations. This would be a good motivation to i) discuss your results even more in the context of the work of other studies; ii) to advertize more work needs to be done to include Envisat and ERS1/2 RA altimeter data analysis to extent the time-series; iii) to advertize your own sub-section about expanding the CS-2 time series back in time to the ICESat periods.
L420: "contains at least four years of data" --> Is there any constraint as to when these four years need to contain data? Is it possible that all data are from the first 4 years?
L432: "Holland (2014)" --> How many years of your 11-year period overlap with the data used by Holland (2014)? Are those results therefore compatible with your results?
L432/434: Both, Garnier et al., (2022) and Xu et al. (2021) used data from a longer time series, didn't they? What is then the added value of performing such a trend analysis over a shorter time period? This is not entirely clear to me.
L441-443: How many years of CryoSat-2 data did Kwok and Cunningham (2015) use in their analysis? I checked it out: It is four winters. You investigate 11 years. I don't think your current writing (and citing that paper) does support further discussing the impact of an analysis of 11-years worth of sea ice thickness and volume in the Southern Ocean.
Below in this sub-section you will find more comments going into this direction. All I wish to trigger with these is to encourage you to one more time critically think whether the message you provide here is compelling, sustainable and worth the effort. Does it send out the right signal in view of already existing work and in view of what we know about the length time series of geophysical parameters should have in order to provide a meaningful statement about climatological features such as trends? See GC1.L455-457: "However, the same ... since 2014" --> Certainly. And if you shorten the time period even further, e.g. to a 4-years like Kwok and Cunningham did, then you will find an even larger decrease in sea ice thickness or volume for 2014-2017 while you may find an increase in sea ice thickness or volume for 2011-2014 and 2017-2020. Fine. And?
I find it kind of dangerous to refine the temporal granularity of such trend analysis in an area such as the Southern Ocean being influenced by at least three multi-decadal oscillations plus El Nino/La Nina events. I agree, Kwok and Cunningham (2015) did it with an even shorther time series, Kurtz and Markus (2012) as well ... but what did we learn from these?L463/464: "modeled studies into ... scenarios" --> Certainly. But this is not a surprizing finding and, in addition, it requires first some more work still to be done improving those models - see Roach et al., 2020, Geophys. Res. Lett., 47, who for good reason first looked at the Antarctic sea ice area in CMIP6 models finding it not well represented.
L472/473: "A longer-term time series ... implications." --> Exactly. Two other studies exist (almost certainly there are more in the meantime) that already looked into longer time series which complicates to see the immediate added value of your investigation in comparison to their studies.
L497-501: "Likely ... estimate Antarctic sea ice thickness." --> I have two comments here. The first one is related to whether you also looked into the work of Ricker et al., 2015, Impact of snow accumulation on CryoSat-2 range retrievals over Arctic sea ice: An observational approach with buoy data, Geophys. Res. Lett., 42. While being for Arctic conditions that work might be further enlightening with respect to your observations.
The second comment is about the observation that in the time series of h_i_70 the primary maximum mean sea ice thickness is not occurring in February anymore but occurs in late winter / spring in all but one year. What does this tell us in light of the fact that the primary maximum now occurs close to the maximum sea ice coverage - involving a large fraction of seasonal sea ice with different surface properties than encountered in February?
L505/506: "could come from ... its lifetime" --> I am aware of these changes but at the same time I am wondering i) which release of the ICESat GLAS data you used for your re-processing of the ZIF sea ice thickness values and ii) whether you did not correct for the different gain values that are reported along with the ICESat data?
L525 / Section 5:
I absolutely agree with you that it would be really nice to have ground-based observations that cover all three sensors' observation period. But we know that this is not possible. The only data sets I am aware of that covers all three sensors contain only estimates of the sea ice thickness: the ASPEcT data set and its extension mentioned further up. Arctic studies often tend to look into PIOMAS to see whether there is long-term consistency in the estimates. I am not deep enough involved into such studies to know whether GIOMAS data would be a viable alternative for the Southern Ocean.However, apart from these considerations, I am missing a more thoughtful evaluation of your sea ice thickness data / product for the CryoSat-2 period used. I have several concerns. One is the apparent lack of adequately discriminating between modal (level) and mean (level + deformed) sea ice thickness values in those parts of your intercomparisons where such a discrimination would be possible (e.g. the Worby et al., 2008 data). In that context I note again that you could have used the extended version of these data noted earlier in addition - even though these do not contain this discrimination into level and level+deformed ice.
In this context I would like to remind you to adequately discuss the limitations of the data you used for your intercomparisons presented in this manuscript - as voiced further up in the context of the Worby et al. (2008) data.
What I am missing is consideration of Operation Ice Bridge data in your evaluation and discussion of the quality in this manuscript. There is a substantial amount of data available and even though flights mostly cover the Weddell and Bellingshausen Seas these are nevertheless a very valuable source for the evaluation of your product. Other air-borne data exist, such as helicopter-borne electromagnetic sounding but I am in fact not sure how many of these would be available within the CryoSat-2 period. For sure researchers organized from New Zealand obtained data in the southern Ross Sea.
In short, in view of recommendations I conveyed to other authors with a similar manuscript profile my main recommendation for you and your section 5 is to put more emphasis on more critically discussing the reliability of your results rather than discussing trends.
Editoral Comments / Typos:
L25: "snow freeboard" --> You could add that here the assumption is that the dominant scattering comes from the snow surface.Equation 2: I recommend to add the information that the second term actually results in a reduction of the sea ice thickness computed by the first term alone - which is opposite to Equation 1 - and which particularly in the Antarctic - the focus of your paper - is important to consider as snow freeboard might equal the snow thickness or may even be smaller than that in case of flooding.
"Kurtz and Markus, 2012" and "Kwok, 2011" are references in which one can find these two equations - however, I am wondering whether it wouldn't make more sense to go back to those publications where these equations were developed / introduced first ... which might be the Laxon et al. paper from 2003 in case of Equation 1 and one of the earlier Kwok (et al.) papers for Equation 2.
L36: ICESat facilitated "snow freeboard" measurements. Please correct.
L39-42: "In most of these ... Kurtz et al., 2009)" --> I suggest to place the Warren et al. reference behind "1954-1991"; otherwise it reads as if Warren et al. (1999) have used that climatology to convert freeboard to thickness.
I further suggest to not highlight that Kurtz et al. (2009) used snow thickness data from passive microwave sensors (which by the way do not provide "lower resolution" snow thickness data compared to the Warren et al climatology being based on interpolation using a polynomial function anyways) - simply because this is just one of the alternatives used by the various other groups already cited. How important it is for the Antarctic focus of your paper to introduce the reader to potential alternatives to the Warren et al. climatology which is not existing in the Antarctic?
L43: ICESat-2 --> Did you overlook the contributions by Kwok (et al.) - who also combined Cryosat-2 and ICESat-2 - on purpose here?
L45: "have found success in estimating sea ice freeboard over Arctic sea ice" --> In light of the fact that most of the studies you cited had freeboard-to-thickness conversion as their ultimate aim, I am wondering whether you might want to rephrase this along the lines: "were succesful in retrieving sea ice thickness from sea ice freeboard estimates over Arctic sea ice" ... or the like.
L57: "Markus and Cavalieri, 2013" --> This is the electronic version of the original book chapter from 1998, right? Has the content changed? If not, please check with EGUSphere how to cite to avoid the impression that this is a more recent work.
L74: "through the use of key snow depth assumptions" --> I might be wrong but it is only the Kurtz and Markus (2012) work which does this assumption. I therefore suggest to add something like "partly" or "for example" to make clear that assuming zero freeboad is ONE possible solution - with limited applicability though as one can figure out in the subsequently cited by you literature.
L78: "Zero ice freeboard ..." --> In addition to citing Willatt et al (2010) you could also include Ozsoy-Cicek et al. 2013, JGR-Oceans.
L79: Regarding this underestimation you could have cited the earlier study by Kwok and Maksym from 2014 (JGR-Oceans) using OIB data; also Kern et al. (2016) performed intercomparisons between different retrieval approches, Kurtz and Markus being on of these.
L95 "utilize CryoSat-2" --> It would not hurt to also mention Envisat here because with that one would have an uninterrupted time series produced using an independent sensor from 2003 through today (see also Paul et al., 2018).
L127: "is based off of" --> I would have written "is based on" ... but I am not a native English speaker ...
L160: You might want to change the font of P, I and p so that it matches "v" and equation (4).
L187: "R_n" needs to be "R_0" ?
L197: I am not sure I would throw the Schwegmann et al. paper into one pot with the Paul et al one because the latter used a considerably modified methodology. Hence citing Paul et al might be sufficient here.
L239-241: You could consider to delete the information about which algorithm you used and how you compute the sea ice area because you described this earlier.
L247: You might want to add that the demarcation in longitude is given in the unit "degrees East".
L262: The "s" in h_fs needs to be put in sub-script mode.
L267: "h_f" at the end of the line needs to be "h_fs"?
L293: "sections" --> "section"
L301: What is "IQR"?
L308/309: "Despite ... nevertheless" ... I guess one of these is enough; I'd discard the "nevertheless".
L325/326: "This could cause ... anomalous snow depths" --> I am wondering whether you could narrow this down towards that the snow-ice interface will most likely be located higher in the snow pack and then also state that this will lead to anomalously low snow thickness values?
L379/380: You are refering to "basal growth" here. For Southern Ocean sea ice a substantial portion of the sea ice volume (up to 1/3 in some places) is actually made of snow ice, i.e. snow that was first flooded at the ice-snow interface and then re-froze. This is not a basal growth. One solution could be to simply write "growth".
L384: "in ice thickness" --> in pan-Antarctic sea ice thickness"
"more that" --> "more than"L394: A reader would be happy to be reminded what this correction factor does and when it is applied.
L395: "was estimates" --> "was estimated"
L459: Sometime you use pan-Antarctic with a capital "P" sometimes not. You might decide for one version of how to write it. I don't know actually what would be correct grammatically.
L484: "about" --> "around"
L510: "h_i-total" is what?
L550/551: "It is clear that ... sea ice thickness," --> I encourage you to also include "snow thickness" here.
L611: You might want to replace this reference by the paper published in Earth and Space Science, 8(7), 2021 to have the link to the peer-reviewed version of your work.
Citation: https://doi.org/10.5194/egusphere-2022-1287-RC1 - AC1: 'Reply on RC1', Steven Fons, 09 Mar 2023
-
RC2: 'Comment on egusphere-2022-1287', Anonymous Referee #2, 20 Jan 2023
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2022/egusphere-2022-1287/egusphere-2022-1287-RC2-supplement.pdf
- AC2: 'Reply on RC2', Steven Fons, 09 Mar 2023
Steven Fons et al.
Steven Fons et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 309 | 173 | 14 | 496 | 5 | 7 |
- HTML: 309
- PDF: 173
- XML: 14
- Total: 496
- BibTeX: 5
- EndNote: 7
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1