the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 28 Mar 2025
Mapping Sea Ice Concentration in the Canadian Arctic with CryoSat-2
Abstract. Sea ice concentration (SIC) is an essential parameter for understanding environmental change in the polar regions. Historically, SIC has been determined using satellite passive microwave (PMV) radiometry, and this has revealed a progressive decline in the extent of the ice cover in the Arctic since records began in 1979. At regional and local scale, classifications based on satellite radar and optical imagery are practical. Here, we use CryoSat-2 to derive a new SIC product in the Canadian Arctic (CA), a region that is vital for shipping, freshwater production, and multi-year ice transport but is frequently excluded from pan-Arctic sea ice satellite observations. The 300 m along-track sampling of CryoSat-2 allows the fine-scale distribution of sea ice to be resolved, and an empirical correction for the overestimation of leads and misclassification of floes allows SIC to be determined. In general, spatial and temporal variations in SIC determined from CryoSat-2 are in close agreement with those determined from PMV and synthetic aperture radar (SAR) imagery in ice charts. Across the CA region, the root mean square difference (RMSD) between SIC determined monthly from CryoSat-2 and PMV and weekly from ice charts are 8.4 and 10 %, respectively. A local comparison to SIC determined from 82 cloud-free Landsat 8 scenes acquired in the central CA shows an RMSD of 3.3 %. Our findings highlight the complementarity of SIC records determined from CryoSat-2 and their potential to expand our knowledge of ice conditions in the CA.
This preprint has been withdrawn.
-
Withdrawal notice
This preprint has been withdrawn.
-
Preprint
(13435 KB)
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2025-693', Anonymous Referee #1, 28 Apr 2025
The paper presents monthly sea ice concentration retrievals at 25 km resolution derived from CryoSAT-2 over a portion of the Canadian Arctic. The authors found a reasonable agreement between the CryoSAT-2 ice concentrations and ice concentrations provided by passive microwave, Canadian Ice Service ice charts, and Landsat optical data. The authors need to address the following comments before the paper may be considered for publication again.
Major comments:
My major concern is that the scientific value of this paper is not high enough for publication. The authors use an existing algorithm for deriving ice/lead/water from CryoSAT-2 and then aggregate these retrievals to come up with the ice concentration estimate in 25 km × 25 km grid cells. The algorithm was applied for winter only and for a relatively limited area in the Arctic (a portion of the Canadian Arctic) with a very low (monthly) temporal resolution that is typical to altimetry and low spatial resolution (25 km) that is close to the passive microwave resolution. The value of such CryoSAT-2 ice concentration dataset is not clear. For example, it is unclear how CryoSAT-2 monthly ice concentration retrievals are more useful or complementary to ice concentration retrievals from imaging SAR. Ice concentration retrieval from SAR (e.g., RCM) is not really computationally expensive (once trained) even at the pan-Arctic scale, and SAR provides much better spatiotemporal coverage than altimetry during both summer and winter.
One possible way to improve the paper is to expand the area to the pan-Arctic scale and to extend the time period to both summer and winter. I understand that, to this end, the algorithm for ice concentration retrieval from CryoSAT-2 will need to be substantially improved. For example, for summer time there exist an ice/lead detection approach in https://doi.org/10.1016/j.rse.2021.112744 that might be potentially adapted for ice concentration retrieval in summer.
Based on the above, I feel that the paper needs to be rejected at this moment, but could be potentially resubmitted again after implementing the suggested improvements.
I also suggest conducting verification against Interactive Multisensor Snow and Ice Mapping System (IMS) pan-Arctic ice/water data.
Minor comments:
Line 39-40. “hazardous” and “hazard” in the same sentence does not read well.
Line 75. “SAR” was already defined earlier.
Line 355. “calculated from” is repeated twice
Line 360. Please rephrase: “at approximately 0.3 km 1.5 km along- and across-track”.
Citation: https://doi.org/10.5194/egusphere-2025-693-RC1 -
AC1: 'Reply on RC1', Amy Swiggs, 30 Apr 2025
As the reviewer notes, we apply an existing algorithm (developed by us) for detecting ice, lead, and water using CryoSat-2 radar altimetry and we aggregate these classifications to produce gridded estimates of sea ice concentration across a 2 x 106 km2 sector of the Canadian Arctic. Our study is the first application of satellite radar altimetry to determine sea ice concentration and is therefore a novel addition to the literature.
The reviewer states that the value of our new sea ice concentration product is not clear. However, we do also present a detailed comparison of our product to three established estimates of sea ice concentration derived using three alternative approaches, and so we believe that its value is clearly articulated.
The reviewer suggests expanding our study to be pan-Arctic. However, this would deviate from our main objectives, which are to present and evaluate a novel retrieval of sea ice concentration. We focus on the Canadian Arctic because it is a region known to be challenging for passive microwave radiometry and satellite radar altimetry owing to the presence of islands. The method we present could be expanded to other regions, but that is not core to our objectives.
The reviewer states that our algorithm is applied to winter only and suggests extending the dataset to include summer months. In fact, we apply our algorithm to autumn, winter, and spring. We do not apply it to summer because treatment of sea ice surface melting in satellite radar altimetry is still in its infancy, and because in the Arctic Ocean the vast majority of all sea ice develops during the eight calendar months we include in our survey.
The reviewer states that the temporal resolution of satellite altimetry is very low. Quantifying and assessing this is in fact an objective of our study, and so we believe that it is a useful contribution to the literature. Moreover, fine temporal resolution is not always essential, for example in studies of long-term trends in the sea ice pack, which are typically evaluated using monthly records.
To avoid confusion, we could adjust the title of our paper to ‘Evaluating Sea Ice Concentration from CryoSat-2 in the Canadian Arctic’.
Citation: https://doi.org/10.5194/egusphere-2025-693-AC1
-
AC1: 'Reply on RC1', Amy Swiggs, 30 Apr 2025
-
RC2: 'Comment on egusphere-2025-693', Anonymous Referee #2, 20 May 2025
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-693/egusphere-2025-693-RC2-supplement.pdf
-
RC3: 'Comment on egusphere-2025-693', Anonymous Referee #3, 16 Jun 2025
Review of
Mapping Sea Ice Concentration in the Canadian Arctic with CryoSat-2
by
A. E. Swiggs et al.
Summary:
This manuscript tries to provide a sea ice concentration data product complementary to existing ones. The product is derived from CryoSat-2 waveforms which enable to discriminate between ice (floes), leads and open ocean. The product comes a monthly temporal resolution and is presented here at the commonly used grid resolution of other (passive-microwave satellite sensor based) sea ice concentration products of 25 km x 25 km. The region of focus is the Canadian Arctic Archipelago and adjacent seas, i.e. Baffin Bay and Beaufort Sea. Derived CryoSat-2 sea ice concentration data are compared to one commonly used passive microwave product, to Canadian Ice Charts and to a set of high-resolution Landsat-8 sea ice concentration data estimated from thermal infrared observations.
While the manuscript's topic is not un-interesting - for the sake of doing research and testing what is possible and/or useful - the results that are obtained do not sufficiently well map to the goal to provide a useful complementary sea ice concentration data set for the sea ice community / for shipping / for off-shore activities. The added value is marginal and the quality of the obtained sea ice concentration data product is questionable outside the Canadian Arctic Archipelago.I have several concerns with the current version of the manuscript
General Comments:
GC1: The manuscript would benefit from a more thorough scan of the sea ice concentration data products that are available from PMV. In that context it would be very helpful to distill the pitfalls and deficiencies PMV products have better to more clearly know where actually the knowledge gap is this manuscript wants to fill.GC2: The manuscript would benefit a lot from a better mapping between the motivation and the actually received results. While the motivation is very much directed towards an improvement of conditions for shipping and off-shore activities in the complex ice conditions of the Canadian Arctic Archipelago the results do not sufficiently well respond to the motivation - simply because the work presented in this manuscript is scientifically interesting but does not / cannot fulfil the voiced "requirements".
GC3: The way the Landsat-8 sea ice concentration data set is presented in this manuscript is not sufficient to i) understand how the authors end up with their classification into what is open water and what is sea ice, and to ii) trust in the results that are presented. The examples shown in the manuscript are not convincing with respect to the maturity of this data set that is used here as a form of ground truth - which I suspect it is not.
GC4: The manuscript would benefit from a better description of how the sea ice concentration is derived from the classified CS-2 waveforms with respect to how (and why) monthly mean sea ice concentration estimates are derived. Did the authors try to come up with an along-track version of their sea ice concentration estimates? How many waveforms are discarded? How are mixed waveforms treated? Are all leads taken as open water (and why)? While the authors try to explain why they observe such a considerable underestimation of the sea ice concentration than the other two products used in the Beaufort Sea and the Baffin Bay, the results of the intercomparison could have been discussed more thoroughly and with a better view on the actual conditions (key word: weather influence, land-spillover effect, wet snow - see my specific comments here). Here the manuscript appears to be overly hypothetically. Into the same topic falls the issue of the bias correction carried out which seems to be required mainly because the Landsat-8 classification into floes and "no-floes" might not be as correct and/or useful as the authors have thought.
Specific Comments:
L14: "is frequently excluded ..." --> I don't understand this statement. All passive microwave based algorithms also estimate sea ice concentration in the CA region - as you define it, but also in the Canadian Arctic Archipelago, i.e. excluding Baffin Bay and Beaufort Sea - where possible. And there are more and more activities towards using SAR imagery which in fact has been the backbone for the Canadian Ice Service for decades and is now, that coverage is more regular and handling of large data amounts has become more easy, the 2nd most source to derive information about the sea ice cover in the CA region.L15/16: This empirical correction reads as if it is not easy to use CS-2 for this purpose - which makes it even less interesting - compared to PMV and SAR.
L18-20: You compare the CS-2 sea ice concentration data at monthly and weekly temporal scale with other datasets. What it the native temporal resolution of the CS-2 sea ice concentration product?
L21: How complementary are CS-2 sea ice concentrations with respect to SAR? Is there any proven benefit / advantage over SAR?
L35/36: The three papers cited here might have excluded the CA region from their analysis but not necessarily because it is so "super" difficult to retrieve the sea ice concentration in this region. Tschudi et al is about a sea-ice motion product which natively requires a larger area than the 25 km grid cell size to retrieve the sea ice motion vector. The other two papers deal with sea-ice thickness retrieval based on satellite altimetry. These approaches require leads or cracks to refer the elevation measurements over sea ice to those of open water or very thin ice in leads. The authors might want to correct me but isn't most of the sea ice cover in the Canadian Arctic Archipelago (CAA) during winter landfast sea ice with little to no or at least substantially less anchor points to approximate the local sea surface height required to infer the freeboard from the altimeter's elevation measurement? I am therefore not content with the three references given as a (the?) motivation why the CA region has been left out in pan-Arctic studies. If I am not mistaken there are several papers wherein one can find regional trends in sea ice area and extent in the CA region -- another hint that this region is very well observed with respect to its sea ice cover - also using PMV, if not primarily.
L44-46: Well, yes, the "standard" sea ice concentration algorithms only offer 25 km grid resolution. However, with the advent of AMSR-E / AMSR2 there have been 12.5 km products since 2002.
The ARTIST Algorithm (Kaleschke et al., 2001; Spreen et al. 2008) offers substantially finer spatial resolution with a 6.25 km product available for AMSR-E and with 3.125 km product available for AMSR2, e.g. from Meereisportal at AWI or from the University of Bremen. Hence there ARE finer resolved products based on PMV. Having said that, even the NOAA/NSIDC CDRv5 now offers 12.5 km grid resolution.
One issue the authors have not commented on, which is potentially far more important - particularly for the shoulder seasons and for summer, is the so-called land-spill over effect which can create spurious sea ice concentration values along the coast line if not corrected for properly (see e.g., Maass et al., 2010, in Tellus).
In summary: While your argument that the sea ice conditions may have become more hazarduous for shipping during recent years, which in undoubtly true, the remainder of your motivation why it is recommendable to invest time developing sea ice concentration algorithms allowing for finer grid resolutions is not convincing the way written.L47-53: While all what you write here is true, the most recent version of the product cited with Wulf et al. does have one-percent granularity and is a fully gridded sea ice concentration product with uncertainty estimates. There has been quite some progress made.
L58-60: I recommend to stress that an altimeter - in contrast to all other sensors that are used to generate sea ice concentration maps or ice charts - is NOT an imaging system. Its spatial coverage is therefore tremenduously more sparse. Given your motivation to support shipping, this is a serious drawback in light of the timeliness with which captians need an update about the sea ice conditions.
L61-63: I can understand that scientifically it is kind of interesting to try to retrieve sea ice concentration from CS-2 data ... but on the other hand, I am really wondering how a "concentration" will be derived from ONE altimeter echo for one footprint. In contrast to PMV sea ice concentration retrieval where it can be assumed that the measured brightness temperature can be seen as a linear combination of the contributions from the surface types within the footprint, this is not the case for an altimeter signal. One can classify between leads and ice floes and then one has mixed signatures which, however, do not provide information about the FRACTION of either of the surface types sensed. Hence, overall, the retrieval must be much less straightforward than for PMV and SAR, and I assume quite a number of assumptions have to be made - in addition to the mentioned correction of misclassified waveforms.
L64: "October to May" ... This makes me to assume that using CS-2 it is not possible to retrieve sea ice concentrations during the summer melt period. This is particularly bad in light of the motivation of this study, to come up with an alternative sea ice concentration product for the CA region, because especially during summer PMV sea ice concentration products would be less reliable because of the effect of snow and sea ice melt and because of the above-mentioned land-spill over effect.
L122/123: Please provide more details of how you "map" the sea ice concentration from LS8 thermal infrared observations. Your description is too short to fully understand what you did. What about those areas classified as leads? These might be ice covered and hence contribute to the sea ice concentration. Hence, in my opinion, only using the floes identified might lead to a considerable, unknown, underestimation of the true sea ice concentration.
What is the resolution of the LS8 sea ice concentration product you use for the comparison? How did you deal with clouds? Did you limit the LS8 data to a certain incidence angle range? Did you evaluate your LS8 sea ice concentration retrieval? What is the uncertainty / bias of that product?
I am aware that you began to describe the product in the lines below the line I used a reference. There you write that you use true color images derived from bands 1-7 ... but these bands partly cover the near-infrared range as well so how can these be true-color images?
When you write "brightness temperature thresholds" you mean "infrared brightness temperature thresholds"?L127-128 / Figure C2: I looked at Figure C2 and there are things that limit the credibility of the analysis. At least for two of the examples shown it is clear that what is classified as open water (lead) is not open water in the true-color image. This could be an explanation for the surprisingly low overall LS8 sea ice concentration value you are reporting upon later in Section 4. Also, I note that there are clouds in the scenes shown, which to me implies that clouds have not been masked out sufficiently well in your LS8 data set (unlike in Kern et al., 2022), which is a considerable drawback of the credibility of the data set you use for evaluation.
L140: "still" --> "calm". What happens with wider leads where wind roughens the water surface? I recognize that further down you talk about the open-ocean surface and mention wind there. But it is not only the open ocean where waves are generated that could have an influence on the altimeter return.
L158: Do you have an idea about the fraction of discarded waveforms?
L159: I am not sure I fully understand what you mean by "lead and floe densities" ... is the lead density the number of CS-2 returns classified as lead divided by the number of all valid CS-2 returns within the 25 km grid cell?
L160/161: How did you compute the matching LS8 lead and floe densities? Did you overlay the CS-2 footprints onto the respective classified LS8 image and did the counting as for CS-2? This is not sufficiently clear.
L169/170: How did you select this threshold of 50? I am sure it has to do with the minimum and maximum number of CS-2 footprints within a 25 km grid cell but I don't understand how you ended up with 50.
Figure 2: What is the time period on which this mean count is based? Is it one particular year or is it the entire time period?
Where do the stripes in the maps showing the mean count for all surface types come from?
What about the fraction of the ambiguous surface type? How large is this and how far away are we from adding up to 100% when we sum over the three maps shown for one month?L192-202: Frankly speaking, I find little value in the intercomparison of the multi-annual overall sea ice concentration (mean and maximum) from these three different methodologies. The only thing we learn is that CIS and NT are quite close together while CS-2 shows substantially smaller sea ice concentration values.
Neither the sentence about the heavy ridging nor the one about the largely immobile CAA icepack is sufficiently well linked to the CS-2 results, I find.Figure 3 b: While there is a seasonal cycle in the standard deviation for CIS and NT sea ice concentration estimates, which resembles well the seasonal cycle of sea ice development, there is much less of such a cycle in the CS2 sea ice concentration estimates. Hypothetically, one could say that the noise that is generated by the retrieval itself is so large that the seasonal cycle is not visible while the retrieval noise in the other two products is much smaller. The standard deviation given by CS2 for most of the winter season in the Baffin Bay and the Beaufort Sea is clearly too high.
L227-229: Another point you should mention here is that ice charts are made for shipping and in case of doubt rather report more sea ice than less; there is a tendency for a positive (conservative) bias.
L230: I was pointing out the need to get numbers of the excluded data earlier - the overall fraction of discarded waveforms (per month) would be a useful information to give.
L235-238: I agree to the statement that PMV algorithms tend to underestimate sea ice concentration in case of thin ice. You find a figure in the Natalia et al. (2015) paper you cited from which you could even quantify the amount of underestimation for the NT algorithm.
I am not so sure I am content with the second statement. To me this overestimation you are referring to is less an effect of little thin ice in open water but rather in general the fact that with the coarse spatial resolution of PMV sensors one has a smearing effect near the ice edge and the ice edge is not as clearly mapped as it should be - hence the "apparent" overestimation of the sea ice concentration near the ice edge.L241-244: No, thin ice cannot be the issue here. I am convinced that the low correlation simply stems from the fact that most sea ice concentration values are close to 100% - especially for NT and CIS - while those for CS-2 are not. Hence you correlate a bunch of values near 100% with a bunch of values near 90% ... the result is a low correlation if the values to not vary in phase which they don't. I was wondering what the bias between CS2 and NT and CIS sea ice concentrations is for every month. This might be a much better estimate for the mis-estimation of the sea ice concentration using CS2 and could give a handle to where the retrieval could and should be improved.
L244-248: Since you are only considering months October to May there is very little influence from melting surfaces or even wet snow. If at all this might be relevant on a monthly scale only in May. It is, however, unlikely that there is already that much wet snow that the C-Band backscatter drops to values of open water. And even if so I would be pretty sure that the CIS analysts are very much aware of that and do not misclassify sea ice as open water at that time of the year. I recommend to delete this entire hypothetical discussion about the impact of wet snow. It is not adding value here.
L253-255: "Atmospheric ... 2007)." --> Atmospheric effects can impact PMV sea ice concentration retrieval, this is true, but this rarely occurs in the middle of winter and not in the CAA which is far away from open water.
It is true that ice variability near 100% is not always well represented by PMV sea ice concentration retrieval algorithms; often they saturate near 100%. This has been taken up by research groups to provide (Lavergne et al., 2019) and evaluate (Kern et al., 2019) the full, naturally retrieved sea ice concentration distribution around 100%. However, having said that, the NT sea ice concentration is a bit less prone to this saturation effect (see Kern et al., 2019) and the likelihood that the sea ice concentration variability you found is "real" is comparably high.L258-266: "Ambiguity ... other data sets" --> This whole explanation is quite hypothetical. I was wondering whether the authors could show examples of what they attempt to explain here.
L278-282: "where the sea ice is thickest" --> I don't subscribe to this statement of that mean differences between the investigated products is particularly small where the sea ice is thickest; you also don't show where the sea ice is particularly thick next to a map of the bias or the RMSD.
Yes, the CAA is a region of complex ice conditions but so is the Beaufort Sea and basically all regions where a mixture of multiyear ice, first-year ice and young ice is present. Landfast sea ice has no influence on the retrieval accuracy of PMV sea ice concentrations and the land-spillover effect takes it bill during periods with open water along the coast - hence mainly in summer and, in your case, during October and potentially a bit of November. Once everything is frozen the land-spillover has no effect.L300-304: While I agree that sea ice concentration near the ice edge is resulting in more noise and sligthly less realiable sea ice concentration values than within the ice pack, I don't see that these are particularly challenging conditions. All what you state here has nothing to do with the bias you observed in the CS2 sea ice concentrations in the Beaufort Sea and also in the Baffin Bay. It has also nothing to do with the different spatio-temporal sampling. I think the CS2 retrieval simply has a problem with i) classifying all floes and with ii) assigning open water to all footprints that are classified as a lead. In winter these are mostly ice covered and contribute to the sea ice concentration. There has been a paper by Kwok in the 1990s or early 2000s demonstrating that in the high Arctic during winter sea ice concentrations are very close to 100%, actually 99.5% or so. The fact that you have a smaller bias between CS2 and the other two products in the CAA supports this idea - simply because the concentration of leads is substantially smaller in the CAA than in the two other sub-regions of your CA region.
L306-308: "There are differences ... 2024)." --> What you write here is not correct as written. There are two processes one needs to discriminate. A) the Nares Strait Arc has formed. Then the NOW forms and can be seen as an ice factory, producing a considerable amount of new ice advected southwards into the Baffin Bay. B) The Nares Strait Arc has NOT formed. Then indeed sea ice is exported out of the Arctic Ocean into the Baffin Bay and the NOW does not form or is at least not as persistent as in winters with an ice arc. You can find studies about this ice arc and also about the efficiency of the Nares Strait ice export out of the Arctic Ocean in the literature in case you really think these considerations are relevant for your intercomparison study.
L321: The very low sea ice concentration value for LS8 for October, paired with my observation that the classification shown in Figure C2 for the one example from October clearly shows an over-estimation of the area classified as lead and hence ocean makes me to doubt the maturity of your LS8 data set used for the evaluation. I asked this before (and I did not check the Swiggs et al., 2024 paper) but what means did you use to evaluate your sea ice concentration estimates based on the LS8 images? My impression is that your method considerably underestimates the sea ice concentration contribution from young ice in form of thin ice sheets - namely dark and light nilas and potentially also grey ice.
L323: At this point I recommend to not go further because it does not make sense to try to evaluate a monthly mean sea ice concentration product which high-resolution daily LS8 sea ice concentration estimates the way you attempt ... this is all far too hypothetical and involves too many assumptions. You set off with this work to provide sea ice concentration data at a finer spatial resolution than is offered by the daily available PMV sea ice concentration products to assist shipping and other activities focusing on the CAA. Now you come up with a monthly mean sea ice concentration at the same grid resolution and try to evaluate it with daily, high-resolution sea ice concentration estimates. I cannot see the benefit here and don't think that the rest of this sub-section is fit for purpose in the sense that it would demonstrate the usefulness and/or added value of the CS2 product.
L415-417: Regarding what you write about SAR. Yes, it requires extensive computational power but we are there. More and more applications are using SAR, Deep-Learning and similar statistical methods are improving constantly, and the availability of SAR - which was a major hindrance in the past - has substantially increased thanks to the Canadian and European efforts. I am therefore convinced that SAR clearly outweighs the CS-2 SIC product that is presented in this work. The CS-2 SIC product can neither convince with its spatial resolution (at least not the way written in this manuscript) and it cannot convince with its temporal resolution which is clearly inferior to ALL other products available and clearly jeopardizes intentions to provide a SIC product that would aid shipping and off-shore activities in a timely manner.
L429-431: "Whilst ... of the year" --> The CS-2 product would be a useful contribution if it would be possible to retrieve SIC during the melt period. This is where basically all other products have their deficiencies. It is a pity that CS-2 cannot fill this gap.
In addition, writing about the "dynamics" in relation to a SIC data set that comes at monthly temporal resolution when there are products with daily temporal resolution that allow, for instance, the assessment of break-up and freeze-up more precisely in time. So, also here, the CS-2 SIC product is not convincing.L434/435: It is true that most of the currently available pan-hemispheric sea ice thickness data sets are produced on monthly timescales - at least when one looks at the Level-3 and Level-4 products. However, the derivation process itself relies on daily information about the sea-ice concentration and it is not sufficient to use monthly SIC data. In addition, more and more users of such sea ice thickness data sets actually ask for along-track information - which clearly requires SIC data at daily temporal resolution.
L436-439: "However ..." I am not sure the first sentence is complete in its wording ... please check.
In addition, I was wondering whether the authors considered looking into publications that are related to detecting and mapping leads in CS-2 observations, e.g. this one: Wernecke, A., and L. Kaleschke, Lead detection in Arctic sea ice from CryoSat-2: quality assessment, lead area fraction and width distribution, The Cryosphere, 9, 1955-1968, 2015, doi: 10.5194/tc-9-1955-2015
Typos / Editoral Comments:
L74: What is the CS-2 sampling distance along track? Please add this information.L92: "thermal brightness data" would be infrared for me. Perhaps you should change it to "microwave brightness data"?
L97: Could you please specify a bit better what your "common grid" is?! Is this an EASE grid or is this a polar-stereographic grid?
L115: If you only used LS8 images and LS8 was launched in 2013 then I recommend to write 2013 instead of 2010 here.
Figure 1 a): So what you show here must be your derived monthly ice chart and not the original weekly ice chart, right? Please clarify, because you state in the sentence before: "prior to processing".
c): Can I assume that the coverage of the NASA Team sea ice concentration is from one day only? Which day?L174: This sentence about the error of "our CS2 data" is quite vague. What are exactly "our CS-2 data"?
L184: Why "extremely"? It gets to 100%, this is it. Where is the extreme?
L240: "these sensor ..." --> something seems to be missing in this half sentence.
L255-257: "However ... an underestimate." --> I don't understand the connection between the fact that both, NT and CS-2 can underestimate the true sea ice concentration. What do we learn here?
L274/274: "For NASA Team ... 2022)" --> Did DiGirolamo carry out an evaluation study like those published by Kern et al., 2020 and 2022? How fit for purpose is this reference?
Figure 5: I suggest to change the annotation of the legend to "sea ice concentration difference (%)"
L298-300: "Increasing melt ... satellite retrieval." --> no, the NT algorithm has tie points for first-year ice and multiyear ice and is therefore not overly sensitive to changes in these ice types. On the contrary, one could say that this algorithm is particularly "fail-proof" when it comes to different ice types.
Citation: https://doi.org/10.5194/egusphere-2025-693-RC3
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2025-693', Anonymous Referee #1, 28 Apr 2025
The paper presents monthly sea ice concentration retrievals at 25 km resolution derived from CryoSAT-2 over a portion of the Canadian Arctic. The authors found a reasonable agreement between the CryoSAT-2 ice concentrations and ice concentrations provided by passive microwave, Canadian Ice Service ice charts, and Landsat optical data. The authors need to address the following comments before the paper may be considered for publication again.
Major comments:
My major concern is that the scientific value of this paper is not high enough for publication. The authors use an existing algorithm for deriving ice/lead/water from CryoSAT-2 and then aggregate these retrievals to come up with the ice concentration estimate in 25 km × 25 km grid cells. The algorithm was applied for winter only and for a relatively limited area in the Arctic (a portion of the Canadian Arctic) with a very low (monthly) temporal resolution that is typical to altimetry and low spatial resolution (25 km) that is close to the passive microwave resolution. The value of such CryoSAT-2 ice concentration dataset is not clear. For example, it is unclear how CryoSAT-2 monthly ice concentration retrievals are more useful or complementary to ice concentration retrievals from imaging SAR. Ice concentration retrieval from SAR (e.g., RCM) is not really computationally expensive (once trained) even at the pan-Arctic scale, and SAR provides much better spatiotemporal coverage than altimetry during both summer and winter.
One possible way to improve the paper is to expand the area to the pan-Arctic scale and to extend the time period to both summer and winter. I understand that, to this end, the algorithm for ice concentration retrieval from CryoSAT-2 will need to be substantially improved. For example, for summer time there exist an ice/lead detection approach in https://doi.org/10.1016/j.rse.2021.112744 that might be potentially adapted for ice concentration retrieval in summer.
Based on the above, I feel that the paper needs to be rejected at this moment, but could be potentially resubmitted again after implementing the suggested improvements.
I also suggest conducting verification against Interactive Multisensor Snow and Ice Mapping System (IMS) pan-Arctic ice/water data.
Minor comments:
Line 39-40. “hazardous” and “hazard” in the same sentence does not read well.
Line 75. “SAR” was already defined earlier.
Line 355. “calculated from” is repeated twice
Line 360. Please rephrase: “at approximately 0.3 km 1.5 km along- and across-track”.
Citation: https://doi.org/10.5194/egusphere-2025-693-RC1 -
AC1: 'Reply on RC1', Amy Swiggs, 30 Apr 2025
As the reviewer notes, we apply an existing algorithm (developed by us) for detecting ice, lead, and water using CryoSat-2 radar altimetry and we aggregate these classifications to produce gridded estimates of sea ice concentration across a 2 x 106 km2 sector of the Canadian Arctic. Our study is the first application of satellite radar altimetry to determine sea ice concentration and is therefore a novel addition to the literature.
The reviewer states that the value of our new sea ice concentration product is not clear. However, we do also present a detailed comparison of our product to three established estimates of sea ice concentration derived using three alternative approaches, and so we believe that its value is clearly articulated.
The reviewer suggests expanding our study to be pan-Arctic. However, this would deviate from our main objectives, which are to present and evaluate a novel retrieval of sea ice concentration. We focus on the Canadian Arctic because it is a region known to be challenging for passive microwave radiometry and satellite radar altimetry owing to the presence of islands. The method we present could be expanded to other regions, but that is not core to our objectives.
The reviewer states that our algorithm is applied to winter only and suggests extending the dataset to include summer months. In fact, we apply our algorithm to autumn, winter, and spring. We do not apply it to summer because treatment of sea ice surface melting in satellite radar altimetry is still in its infancy, and because in the Arctic Ocean the vast majority of all sea ice develops during the eight calendar months we include in our survey.
The reviewer states that the temporal resolution of satellite altimetry is very low. Quantifying and assessing this is in fact an objective of our study, and so we believe that it is a useful contribution to the literature. Moreover, fine temporal resolution is not always essential, for example in studies of long-term trends in the sea ice pack, which are typically evaluated using monthly records.
To avoid confusion, we could adjust the title of our paper to ‘Evaluating Sea Ice Concentration from CryoSat-2 in the Canadian Arctic’.
Citation: https://doi.org/10.5194/egusphere-2025-693-AC1
-
AC1: 'Reply on RC1', Amy Swiggs, 30 Apr 2025
-
RC2: 'Comment on egusphere-2025-693', Anonymous Referee #2, 20 May 2025
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-693/egusphere-2025-693-RC2-supplement.pdf
-
RC3: 'Comment on egusphere-2025-693', Anonymous Referee #3, 16 Jun 2025
Review of
Mapping Sea Ice Concentration in the Canadian Arctic with CryoSat-2
by
A. E. Swiggs et al.
Summary:
This manuscript tries to provide a sea ice concentration data product complementary to existing ones. The product is derived from CryoSat-2 waveforms which enable to discriminate between ice (floes), leads and open ocean. The product comes a monthly temporal resolution and is presented here at the commonly used grid resolution of other (passive-microwave satellite sensor based) sea ice concentration products of 25 km x 25 km. The region of focus is the Canadian Arctic Archipelago and adjacent seas, i.e. Baffin Bay and Beaufort Sea. Derived CryoSat-2 sea ice concentration data are compared to one commonly used passive microwave product, to Canadian Ice Charts and to a set of high-resolution Landsat-8 sea ice concentration data estimated from thermal infrared observations.
While the manuscript's topic is not un-interesting - for the sake of doing research and testing what is possible and/or useful - the results that are obtained do not sufficiently well map to the goal to provide a useful complementary sea ice concentration data set for the sea ice community / for shipping / for off-shore activities. The added value is marginal and the quality of the obtained sea ice concentration data product is questionable outside the Canadian Arctic Archipelago.I have several concerns with the current version of the manuscript
General Comments:
GC1: The manuscript would benefit from a more thorough scan of the sea ice concentration data products that are available from PMV. In that context it would be very helpful to distill the pitfalls and deficiencies PMV products have better to more clearly know where actually the knowledge gap is this manuscript wants to fill.GC2: The manuscript would benefit a lot from a better mapping between the motivation and the actually received results. While the motivation is very much directed towards an improvement of conditions for shipping and off-shore activities in the complex ice conditions of the Canadian Arctic Archipelago the results do not sufficiently well respond to the motivation - simply because the work presented in this manuscript is scientifically interesting but does not / cannot fulfil the voiced "requirements".
GC3: The way the Landsat-8 sea ice concentration data set is presented in this manuscript is not sufficient to i) understand how the authors end up with their classification into what is open water and what is sea ice, and to ii) trust in the results that are presented. The examples shown in the manuscript are not convincing with respect to the maturity of this data set that is used here as a form of ground truth - which I suspect it is not.
GC4: The manuscript would benefit from a better description of how the sea ice concentration is derived from the classified CS-2 waveforms with respect to how (and why) monthly mean sea ice concentration estimates are derived. Did the authors try to come up with an along-track version of their sea ice concentration estimates? How many waveforms are discarded? How are mixed waveforms treated? Are all leads taken as open water (and why)? While the authors try to explain why they observe such a considerable underestimation of the sea ice concentration than the other two products used in the Beaufort Sea and the Baffin Bay, the results of the intercomparison could have been discussed more thoroughly and with a better view on the actual conditions (key word: weather influence, land-spillover effect, wet snow - see my specific comments here). Here the manuscript appears to be overly hypothetically. Into the same topic falls the issue of the bias correction carried out which seems to be required mainly because the Landsat-8 classification into floes and "no-floes" might not be as correct and/or useful as the authors have thought.
Specific Comments:
L14: "is frequently excluded ..." --> I don't understand this statement. All passive microwave based algorithms also estimate sea ice concentration in the CA region - as you define it, but also in the Canadian Arctic Archipelago, i.e. excluding Baffin Bay and Beaufort Sea - where possible. And there are more and more activities towards using SAR imagery which in fact has been the backbone for the Canadian Ice Service for decades and is now, that coverage is more regular and handling of large data amounts has become more easy, the 2nd most source to derive information about the sea ice cover in the CA region.L15/16: This empirical correction reads as if it is not easy to use CS-2 for this purpose - which makes it even less interesting - compared to PMV and SAR.
L18-20: You compare the CS-2 sea ice concentration data at monthly and weekly temporal scale with other datasets. What it the native temporal resolution of the CS-2 sea ice concentration product?
L21: How complementary are CS-2 sea ice concentrations with respect to SAR? Is there any proven benefit / advantage over SAR?
L35/36: The three papers cited here might have excluded the CA region from their analysis but not necessarily because it is so "super" difficult to retrieve the sea ice concentration in this region. Tschudi et al is about a sea-ice motion product which natively requires a larger area than the 25 km grid cell size to retrieve the sea ice motion vector. The other two papers deal with sea-ice thickness retrieval based on satellite altimetry. These approaches require leads or cracks to refer the elevation measurements over sea ice to those of open water or very thin ice in leads. The authors might want to correct me but isn't most of the sea ice cover in the Canadian Arctic Archipelago (CAA) during winter landfast sea ice with little to no or at least substantially less anchor points to approximate the local sea surface height required to infer the freeboard from the altimeter's elevation measurement? I am therefore not content with the three references given as a (the?) motivation why the CA region has been left out in pan-Arctic studies. If I am not mistaken there are several papers wherein one can find regional trends in sea ice area and extent in the CA region -- another hint that this region is very well observed with respect to its sea ice cover - also using PMV, if not primarily.
L44-46: Well, yes, the "standard" sea ice concentration algorithms only offer 25 km grid resolution. However, with the advent of AMSR-E / AMSR2 there have been 12.5 km products since 2002.
The ARTIST Algorithm (Kaleschke et al., 2001; Spreen et al. 2008) offers substantially finer spatial resolution with a 6.25 km product available for AMSR-E and with 3.125 km product available for AMSR2, e.g. from Meereisportal at AWI or from the University of Bremen. Hence there ARE finer resolved products based on PMV. Having said that, even the NOAA/NSIDC CDRv5 now offers 12.5 km grid resolution.
One issue the authors have not commented on, which is potentially far more important - particularly for the shoulder seasons and for summer, is the so-called land-spill over effect which can create spurious sea ice concentration values along the coast line if not corrected for properly (see e.g., Maass et al., 2010, in Tellus).
In summary: While your argument that the sea ice conditions may have become more hazarduous for shipping during recent years, which in undoubtly true, the remainder of your motivation why it is recommendable to invest time developing sea ice concentration algorithms allowing for finer grid resolutions is not convincing the way written.L47-53: While all what you write here is true, the most recent version of the product cited with Wulf et al. does have one-percent granularity and is a fully gridded sea ice concentration product with uncertainty estimates. There has been quite some progress made.
L58-60: I recommend to stress that an altimeter - in contrast to all other sensors that are used to generate sea ice concentration maps or ice charts - is NOT an imaging system. Its spatial coverage is therefore tremenduously more sparse. Given your motivation to support shipping, this is a serious drawback in light of the timeliness with which captians need an update about the sea ice conditions.
L61-63: I can understand that scientifically it is kind of interesting to try to retrieve sea ice concentration from CS-2 data ... but on the other hand, I am really wondering how a "concentration" will be derived from ONE altimeter echo for one footprint. In contrast to PMV sea ice concentration retrieval where it can be assumed that the measured brightness temperature can be seen as a linear combination of the contributions from the surface types within the footprint, this is not the case for an altimeter signal. One can classify between leads and ice floes and then one has mixed signatures which, however, do not provide information about the FRACTION of either of the surface types sensed. Hence, overall, the retrieval must be much less straightforward than for PMV and SAR, and I assume quite a number of assumptions have to be made - in addition to the mentioned correction of misclassified waveforms.
L64: "October to May" ... This makes me to assume that using CS-2 it is not possible to retrieve sea ice concentrations during the summer melt period. This is particularly bad in light of the motivation of this study, to come up with an alternative sea ice concentration product for the CA region, because especially during summer PMV sea ice concentration products would be less reliable because of the effect of snow and sea ice melt and because of the above-mentioned land-spill over effect.
L122/123: Please provide more details of how you "map" the sea ice concentration from LS8 thermal infrared observations. Your description is too short to fully understand what you did. What about those areas classified as leads? These might be ice covered and hence contribute to the sea ice concentration. Hence, in my opinion, only using the floes identified might lead to a considerable, unknown, underestimation of the true sea ice concentration.
What is the resolution of the LS8 sea ice concentration product you use for the comparison? How did you deal with clouds? Did you limit the LS8 data to a certain incidence angle range? Did you evaluate your LS8 sea ice concentration retrieval? What is the uncertainty / bias of that product?
I am aware that you began to describe the product in the lines below the line I used a reference. There you write that you use true color images derived from bands 1-7 ... but these bands partly cover the near-infrared range as well so how can these be true-color images?
When you write "brightness temperature thresholds" you mean "infrared brightness temperature thresholds"?L127-128 / Figure C2: I looked at Figure C2 and there are things that limit the credibility of the analysis. At least for two of the examples shown it is clear that what is classified as open water (lead) is not open water in the true-color image. This could be an explanation for the surprisingly low overall LS8 sea ice concentration value you are reporting upon later in Section 4. Also, I note that there are clouds in the scenes shown, which to me implies that clouds have not been masked out sufficiently well in your LS8 data set (unlike in Kern et al., 2022), which is a considerable drawback of the credibility of the data set you use for evaluation.
L140: "still" --> "calm". What happens with wider leads where wind roughens the water surface? I recognize that further down you talk about the open-ocean surface and mention wind there. But it is not only the open ocean where waves are generated that could have an influence on the altimeter return.
L158: Do you have an idea about the fraction of discarded waveforms?
L159: I am not sure I fully understand what you mean by "lead and floe densities" ... is the lead density the number of CS-2 returns classified as lead divided by the number of all valid CS-2 returns within the 25 km grid cell?
L160/161: How did you compute the matching LS8 lead and floe densities? Did you overlay the CS-2 footprints onto the respective classified LS8 image and did the counting as for CS-2? This is not sufficiently clear.
L169/170: How did you select this threshold of 50? I am sure it has to do with the minimum and maximum number of CS-2 footprints within a 25 km grid cell but I don't understand how you ended up with 50.
Figure 2: What is the time period on which this mean count is based? Is it one particular year or is it the entire time period?
Where do the stripes in the maps showing the mean count for all surface types come from?
What about the fraction of the ambiguous surface type? How large is this and how far away are we from adding up to 100% when we sum over the three maps shown for one month?L192-202: Frankly speaking, I find little value in the intercomparison of the multi-annual overall sea ice concentration (mean and maximum) from these three different methodologies. The only thing we learn is that CIS and NT are quite close together while CS-2 shows substantially smaller sea ice concentration values.
Neither the sentence about the heavy ridging nor the one about the largely immobile CAA icepack is sufficiently well linked to the CS-2 results, I find.Figure 3 b: While there is a seasonal cycle in the standard deviation for CIS and NT sea ice concentration estimates, which resembles well the seasonal cycle of sea ice development, there is much less of such a cycle in the CS2 sea ice concentration estimates. Hypothetically, one could say that the noise that is generated by the retrieval itself is so large that the seasonal cycle is not visible while the retrieval noise in the other two products is much smaller. The standard deviation given by CS2 for most of the winter season in the Baffin Bay and the Beaufort Sea is clearly too high.
L227-229: Another point you should mention here is that ice charts are made for shipping and in case of doubt rather report more sea ice than less; there is a tendency for a positive (conservative) bias.
L230: I was pointing out the need to get numbers of the excluded data earlier - the overall fraction of discarded waveforms (per month) would be a useful information to give.
L235-238: I agree to the statement that PMV algorithms tend to underestimate sea ice concentration in case of thin ice. You find a figure in the Natalia et al. (2015) paper you cited from which you could even quantify the amount of underestimation for the NT algorithm.
I am not so sure I am content with the second statement. To me this overestimation you are referring to is less an effect of little thin ice in open water but rather in general the fact that with the coarse spatial resolution of PMV sensors one has a smearing effect near the ice edge and the ice edge is not as clearly mapped as it should be - hence the "apparent" overestimation of the sea ice concentration near the ice edge.L241-244: No, thin ice cannot be the issue here. I am convinced that the low correlation simply stems from the fact that most sea ice concentration values are close to 100% - especially for NT and CIS - while those for CS-2 are not. Hence you correlate a bunch of values near 100% with a bunch of values near 90% ... the result is a low correlation if the values to not vary in phase which they don't. I was wondering what the bias between CS2 and NT and CIS sea ice concentrations is for every month. This might be a much better estimate for the mis-estimation of the sea ice concentration using CS2 and could give a handle to where the retrieval could and should be improved.
L244-248: Since you are only considering months October to May there is very little influence from melting surfaces or even wet snow. If at all this might be relevant on a monthly scale only in May. It is, however, unlikely that there is already that much wet snow that the C-Band backscatter drops to values of open water. And even if so I would be pretty sure that the CIS analysts are very much aware of that and do not misclassify sea ice as open water at that time of the year. I recommend to delete this entire hypothetical discussion about the impact of wet snow. It is not adding value here.
L253-255: "Atmospheric ... 2007)." --> Atmospheric effects can impact PMV sea ice concentration retrieval, this is true, but this rarely occurs in the middle of winter and not in the CAA which is far away from open water.
It is true that ice variability near 100% is not always well represented by PMV sea ice concentration retrieval algorithms; often they saturate near 100%. This has been taken up by research groups to provide (Lavergne et al., 2019) and evaluate (Kern et al., 2019) the full, naturally retrieved sea ice concentration distribution around 100%. However, having said that, the NT sea ice concentration is a bit less prone to this saturation effect (see Kern et al., 2019) and the likelihood that the sea ice concentration variability you found is "real" is comparably high.L258-266: "Ambiguity ... other data sets" --> This whole explanation is quite hypothetical. I was wondering whether the authors could show examples of what they attempt to explain here.
L278-282: "where the sea ice is thickest" --> I don't subscribe to this statement of that mean differences between the investigated products is particularly small where the sea ice is thickest; you also don't show where the sea ice is particularly thick next to a map of the bias or the RMSD.
Yes, the CAA is a region of complex ice conditions but so is the Beaufort Sea and basically all regions where a mixture of multiyear ice, first-year ice and young ice is present. Landfast sea ice has no influence on the retrieval accuracy of PMV sea ice concentrations and the land-spillover effect takes it bill during periods with open water along the coast - hence mainly in summer and, in your case, during October and potentially a bit of November. Once everything is frozen the land-spillover has no effect.L300-304: While I agree that sea ice concentration near the ice edge is resulting in more noise and sligthly less realiable sea ice concentration values than within the ice pack, I don't see that these are particularly challenging conditions. All what you state here has nothing to do with the bias you observed in the CS2 sea ice concentrations in the Beaufort Sea and also in the Baffin Bay. It has also nothing to do with the different spatio-temporal sampling. I think the CS2 retrieval simply has a problem with i) classifying all floes and with ii) assigning open water to all footprints that are classified as a lead. In winter these are mostly ice covered and contribute to the sea ice concentration. There has been a paper by Kwok in the 1990s or early 2000s demonstrating that in the high Arctic during winter sea ice concentrations are very close to 100%, actually 99.5% or so. The fact that you have a smaller bias between CS2 and the other two products in the CAA supports this idea - simply because the concentration of leads is substantially smaller in the CAA than in the two other sub-regions of your CA region.
L306-308: "There are differences ... 2024)." --> What you write here is not correct as written. There are two processes one needs to discriminate. A) the Nares Strait Arc has formed. Then the NOW forms and can be seen as an ice factory, producing a considerable amount of new ice advected southwards into the Baffin Bay. B) The Nares Strait Arc has NOT formed. Then indeed sea ice is exported out of the Arctic Ocean into the Baffin Bay and the NOW does not form or is at least not as persistent as in winters with an ice arc. You can find studies about this ice arc and also about the efficiency of the Nares Strait ice export out of the Arctic Ocean in the literature in case you really think these considerations are relevant for your intercomparison study.
L321: The very low sea ice concentration value for LS8 for October, paired with my observation that the classification shown in Figure C2 for the one example from October clearly shows an over-estimation of the area classified as lead and hence ocean makes me to doubt the maturity of your LS8 data set used for the evaluation. I asked this before (and I did not check the Swiggs et al., 2024 paper) but what means did you use to evaluate your sea ice concentration estimates based on the LS8 images? My impression is that your method considerably underestimates the sea ice concentration contribution from young ice in form of thin ice sheets - namely dark and light nilas and potentially also grey ice.
L323: At this point I recommend to not go further because it does not make sense to try to evaluate a monthly mean sea ice concentration product which high-resolution daily LS8 sea ice concentration estimates the way you attempt ... this is all far too hypothetical and involves too many assumptions. You set off with this work to provide sea ice concentration data at a finer spatial resolution than is offered by the daily available PMV sea ice concentration products to assist shipping and other activities focusing on the CAA. Now you come up with a monthly mean sea ice concentration at the same grid resolution and try to evaluate it with daily, high-resolution sea ice concentration estimates. I cannot see the benefit here and don't think that the rest of this sub-section is fit for purpose in the sense that it would demonstrate the usefulness and/or added value of the CS2 product.
L415-417: Regarding what you write about SAR. Yes, it requires extensive computational power but we are there. More and more applications are using SAR, Deep-Learning and similar statistical methods are improving constantly, and the availability of SAR - which was a major hindrance in the past - has substantially increased thanks to the Canadian and European efforts. I am therefore convinced that SAR clearly outweighs the CS-2 SIC product that is presented in this work. The CS-2 SIC product can neither convince with its spatial resolution (at least not the way written in this manuscript) and it cannot convince with its temporal resolution which is clearly inferior to ALL other products available and clearly jeopardizes intentions to provide a SIC product that would aid shipping and off-shore activities in a timely manner.
L429-431: "Whilst ... of the year" --> The CS-2 product would be a useful contribution if it would be possible to retrieve SIC during the melt period. This is where basically all other products have their deficiencies. It is a pity that CS-2 cannot fill this gap.
In addition, writing about the "dynamics" in relation to a SIC data set that comes at monthly temporal resolution when there are products with daily temporal resolution that allow, for instance, the assessment of break-up and freeze-up more precisely in time. So, also here, the CS-2 SIC product is not convincing.L434/435: It is true that most of the currently available pan-hemispheric sea ice thickness data sets are produced on monthly timescales - at least when one looks at the Level-3 and Level-4 products. However, the derivation process itself relies on daily information about the sea-ice concentration and it is not sufficient to use monthly SIC data. In addition, more and more users of such sea ice thickness data sets actually ask for along-track information - which clearly requires SIC data at daily temporal resolution.
L436-439: "However ..." I am not sure the first sentence is complete in its wording ... please check.
In addition, I was wondering whether the authors considered looking into publications that are related to detecting and mapping leads in CS-2 observations, e.g. this one: Wernecke, A., and L. Kaleschke, Lead detection in Arctic sea ice from CryoSat-2: quality assessment, lead area fraction and width distribution, The Cryosphere, 9, 1955-1968, 2015, doi: 10.5194/tc-9-1955-2015
Typos / Editoral Comments:
L74: What is the CS-2 sampling distance along track? Please add this information.L92: "thermal brightness data" would be infrared for me. Perhaps you should change it to "microwave brightness data"?
L97: Could you please specify a bit better what your "common grid" is?! Is this an EASE grid or is this a polar-stereographic grid?
L115: If you only used LS8 images and LS8 was launched in 2013 then I recommend to write 2013 instead of 2010 here.
Figure 1 a): So what you show here must be your derived monthly ice chart and not the original weekly ice chart, right? Please clarify, because you state in the sentence before: "prior to processing".
c): Can I assume that the coverage of the NASA Team sea ice concentration is from one day only? Which day?L174: This sentence about the error of "our CS2 data" is quite vague. What are exactly "our CS-2 data"?
L184: Why "extremely"? It gets to 100%, this is it. Where is the extreme?
L240: "these sensor ..." --> something seems to be missing in this half sentence.
L255-257: "However ... an underestimate." --> I don't understand the connection between the fact that both, NT and CS-2 can underestimate the true sea ice concentration. What do we learn here?
L274/274: "For NASA Team ... 2022)" --> Did DiGirolamo carry out an evaluation study like those published by Kern et al., 2020 and 2022? How fit for purpose is this reference?
Figure 5: I suggest to change the annotation of the legend to "sea ice concentration difference (%)"
L298-300: "Increasing melt ... satellite retrieval." --> no, the NT algorithm has tie points for first-year ice and multiyear ice and is therefore not overly sensitive to changes in these ice types. On the contrary, one could say that this algorithm is particularly "fail-proof" when it comes to different ice types.
Citation: https://doi.org/10.5194/egusphere-2025-693-RC3
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 357 | 49 | 16 | 422 | 20 | 31 |
- HTML: 357
- PDF: 49
- XML: 16
- Total: 422
- BibTeX: 20
- EndNote: 31
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1