the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Seasonal evolution and parameterization of Arctic sea ice bulk density: results from the MOSAiC expedition and ICESat-2/ATLAS
Abstract. Satellite retrievals of Arctic sea ice thickness typically assume a constant sea ice bulk density (IBD), overlooking its seasonal variations influenced by ice internal texture and contaminants. This study unveils the initial insights into the seasonal evolution and parameterization of IBD during the Arctic freezing season from October to April. To retrieve IBD, we combined in situ observations obtained from ice mass balance buoys, snow pits, and snow transects during the Multidisciplinary drifting Observatory for the Study of Arctic Climate (MOSAiC) expedition, as well as laser freeboard data derived from the Ice, Cloud, and Land Elevation Satellite-2 (ICESat-2). Assuming hydrostatic equilibrium, local-scale IBDs for the level ice component of the MOSAiC ice floes, predominantly consisting of second-year ice, were obtained at a spatial scale of approximately 50 km. The results indicated a statistically significant seasonal decreasing trend in IBD at a rate of ~16 kg m−3 per month (P < 0.001) from mid-October to mid-January, likely attributable to increased internal porosity as the sea ice aged. This was followed by a relatively stable period from mid-January to mid-April, with an average IBD of ~897 ± 11 kg m−3. Core-based IBDs from eight MOSAiC sites showed a similar seasonal pattern, but with a narrower range of variation and an earlier onset of the relatively stable period, possibly owing to the spatial heterogeneity of the MOSAiC ice floes. Based on regression analyses, we developed updated parameterizations for IBD that are anticipated to be applicable throughout the freezing season, encompassing both first- and second-year ice. In particular, the ice draft-to-thickness ratio emerged as the most efficient parameter for determining IBD (R2 = 0.99, RMSE = 1.62 kg m−3), with potential application to multi-year ice and deformed ice as well. Our updated parameterizations have the potential to optimize basin-scale satellite-derived sea ice thickness, thereby contributing to more accurate monitoring of changes in sea ice volume.
- Preprint
(6207 KB) - Metadata XML
-
Supplement
(1122 KB) - BibTeX
- EndNote
Status: closed
-
CC1: 'Comment on egusphere-2024-1240', Arttu Jutila, 15 May 2024
Community comment on the manuscript by Zhou, Y., Wang, X., Lei, R., von Albedyll, L., Perovich, D. K., Zhang, Y., and Haas, C.: Seasonal evolution and parameterization of Arctic sea ice bulk density: results from the MOSAiC expedition and ICESat-2/ATLAS, EGUsphere [preprint], https://doi.org/10.5194/egusphere-2024-1240, 2024.
Dear Zhou et al.,
Dear handling editor,The competing interests policy of The Cryosphere prohibits me from acting as an official referee for this manuscript due to recent collaborations with some of the coauthors. Therefore, I am posting this comment as a member of the scientific community to discuss some matters related to it.
In the manuscript by Zhou et al., now under peer-review process and public discussion, sea-ice bulk density is derived using the hydrostatic equilibrium equation and values of modal total freeboard from the satellite laser altimeter ICESat-2, mean snow depth and sea-ice thickness from 15 autonomous ice mass-balance buoys (IMB) deployed within the MOSAiC Distributed Network (DN), and mean snow density from snow pit measurements conducted in the MOSAiC Central Observatory (CO) from October 2019 to April 2020.
In 2022, I have authored a paper in The Cryosphere on the same general topic, sea-ice bulk density, using a similar approach but simultaneous airborne multi-sensor measurements from the AWI IceBird program:
Jutila, A., Hendricks, S., Ricker, R., von Albedyll, L., Krumpen, T., and Haas, C.: Retrieval and parameterisation of sea-ice bulk density from airborne multi-sensor measurements, The Cryosphere, 16, 259–275, https://doi.org/10.5194/tc-16-259-2022, 2022.
This work is referenced many times and data originating from this study are used in the manuscript by Zhou et al.
First, I want to inform that there is a recently published new version of the AWI IceBird airborne sea-ice parameter dataset. In the new version, the quality flag identifying level and deformed ice has been rectified.
Jutila, A., Hendricks, S., Ricker, R., von Albedyll, L., and Haas, C.: Airborne sea ice parameters during the IceBird Winter 2019 campaign in the Arctic Ocean, Version 2, https://doi.org/10.1594/PANGAEA.966057, 2024.
Jutila, A., Hendricks, S., Ricker, R., von Albedyll, L., and Haas, C.: Airborne sea ice parameters during the PAMARCMIP2017 campaign in the Arctic Ocean, Version 2, https://doi.org/10.1594/PANGAEA.966009, 2024.
Regarding the study of Zhou et al., I would like to raise general concerns and perhaps some misunderstandings of my paper. The general points are the following:
- Spatial scales. While the presented study broadens the knowledge with the aspect of seasonal evolution of remotely sensed sea-ice bulk density, I am concerned about the different magnitudes of spatial scales utilized in the derivation. More specifically, you use total freeboard from the ICESat-2 satellite laser altimeter orbits extracted within a circle around the CO that has a diameter of 100 km; snow depth and sea-ice thickness derived from 15 autonomous IMBs in the DN within circle around the CO that has a diameter of 70 km (in the beginning of the drift, but how about later after being affected by sea-ice dynamics for months?) while the data are inherently point measurements; and snow density derived from snow pit measurements within the CO that extends over an area with a diameter of only few hundred meters while the data are inherently point measurements. None of these data sources have real spatial overlap with each other. This effectively diminishes the study to use ice-type-averaged values (not far from Alexandrov et al.’s (2010) multi-year ice density derivation with climatological values from literature) as the measurements are not from the same piece of ice – not even remotely.
In addition, using the term “local-scale” with data originating extending 100 km, when local is generally understood as <~1 km, definitely raised my eyebrows throughout the manuscript.
Why was ICESat-2 ATL10 rev5 used when rev6 is available? Actually, why not use the publicly available, more local MOSAiC helicopterborne laser scanner data by Hutter et al. (2023), which you also cite in the manuscript? I think that could be a feasible option to explore and it would back up better the local-scale aspect of this study. At this point, however, I must point out that I was involved in collecting and processing said data, too. - Level ice. You state that the chosen IMBs were deployed on level ice. I agree that this is a correct approach, to consider level ice only. However, the publicly available deployment documents for the buoys T63, T65, T70, and I1 indicate that ridged ice was already in close vicinity during deployment.
How was it ensured that ICESat-2 data was considered over level ice only? How long are the data segments, did they include only level data? While the modal value of the log-normal fitted freeboard is an estimate of the thermodynamically grown sea ice, it does not strictly exclude e.g. thin sea ice that has deformed and gained the same freeboard as thermodynamically grown undeformed sea ice.
How about snow pits, have you considered that pressure ridge sites were sampled on MOSAiC, too? Level ice tends to have thinner snowpack with larger temperature gradients that lead to snow metamorphism affecting the snow density profile.
When comparing your data to the AWI IceBird dataset, did you choose measurements on level ice only quality flag? I would suggest doing so, and in that case also using the updated version of the dataset.
More specific comments:
L62ff: Alexandrov et al. (2010) did not use airborne multi-sensor data. They used ground-based drill-hole measurements achieved through landing airplanes on the sea ice in the 1980s (Soviet Sever expeditions). So far, I am not aware of any other study utilizing airborne multi-sensor measurements to derive sea-ice bulk density than Jutila et al. (2022).
L79ff: While Shi et al. (2023) have more recently argued the point, it was mentioned earlier in Jutila et al. (2022), to which also Shi et al. (2023) refer.
L179ff: Sea-ice freeboard and thickness are found to follow log-normal or exponential distribution, but does total freeboard behave the same? And how about on the 100 km scale?
L232ff: Both your snow depth and sea-ice thickness measurements come from the IMBs. Therefore, are their uncertainties not independent and the assumption thus wrong?
L244ff: Were any other formulations than first and second order polynomials investigated?
Figure 5 & L313ff: Which values are you using for the three J22 data points? To my eyes, they do not match the values from Table 3 in Jutila et al. (2022) that list the average bulk densities on 800 m spatial scale. Or did you perhaps derive those values from the nominal resolution datasets? Did you use all values or only the level ice ones? Furthermore, I recommend using the same marker shape for the same ice type, adding citations also to the main text, and explaining the acronym “J22” (now only on L361).
L480ff: The data consists of several profiles covering a total distance of more than 3000 km (3410 km). Surveyed sea ice was primarily first-year ice (100 % in 2017) and multi-year ice, with only very little second-year ice. While you mention the spatial resolution of the data, I also think it’s important to distinguish between the nominal measurement spacing (5-6 m) and the footprint size (40 m) of the measurement.
L493ff & Figure 10: Jutila et al. (2022) applied inverse-uncertainty weighted mean, not inverse distance. Are all ice types included in this analysis, also level ice and first-year ice, even though you’re targeting to analyze rough and older ice? The AWI IceBird airborne sea-ice parameter datasets can easily distinguish different ice types using the provided quality flags.
L523ff: The “new approach proposed in this study to determine [ice bulk density] at the basin scale using satellite altimetry data” is not new as this capability has been previously demonstrated in Jutila et al. (2022). If you mean using satellite altimetry data in your approach to determine sea-ice bulk density (together with ground-based point measurements), you need to present and discuss the effect of different scales for the reasons brought up earlier. The study also seems to highlight the parameterization applying the ice draft-to-thickness ratio, but there is no current or planned satellite mission that can directly observe sea-ice draft, thickness, nor their ratio.
Citation: https://doi.org/10.5194/egusphere-2024-1240-CC1 -
AC1: 'Reply on CC1', Yi Zhou, 18 May 2024
Dear Dr. Arttu Jutila,
Thank you for your constructive comments. We will address your insights with comprehensive clarifications and revisions throughout our manuscript. We have outlined the original comments in black with our planned responses highlighted in blue. Kindly refer to the attached document.
Best regards,
Yi Zhou and other co-authors.
-
CC2: 'Reply on AC1', Arttu Jutila, 23 May 2024
Reply to author comment https://doi.org/10.5194/egusphere-2024-1240-AC1 by Zhou et al. for the manuscript by Zhou, Y., Wang, X., Lei, R., von Albedyll, L., Perovich, D. K., Zhang, Y., and Haas, C.: Seasonal evolution and parameterization of Arctic sea ice bulk density: results from the MOSAiC expedition and ICESat-2/ATLAS, EGUsphere [preprint], https://doi.org/10.5194/egusphere-2024-1240, 2024.
Dear Zhou et al.,
Thank you for your response to my comment and in particular for continuing the public discussion in the spirit of The Cryosphere’s interactive review process.
I am pleased to see that you have carefully considered my comments. Nevertheless, I still see some loose ends which I would like to discuss further.
Scales. Thank you for the very helpful additional Figure A1. While you clarify that the extraction radii for IS2 and IMBs were 50 km and 30-40 km, respectively, you have unfortunately misread my comment. I did not use the term radius but diameter, i.e. two times the radius, for a very specific reason. Let’s consider a rather standard satellite data product that has data in a regular grid format where the grid cells are square-shaped with each side measuring 25 km in length. This results in a spatial scale of 25 km – not 12.5 km that would be the radius of the largest possible circle drawn within the grid cell. In your case, for example looking at Figure 3c where the IS2 ground track passes nearly directly above the center of the circle or the MOSAiC CO, you end up extracting along-track data for a length of 100 km, not 50 km. That is the largest length scale of your input data. The same principle applies to the 40-m footprint size of the AWI IceBird measurements: it is the diameter, not the radius, of the EM-Bird footprint, within which snow depth and freeboard measurements are averaged and sea-ice bulk density is eventually derived.
In your consideration of the helicopter-borne laser scanner data in Hutter et al. (2023), you state having found it temporally insufficient with only about ten days of valid data for your study period. Did you include also the approximately weekly transect flights that extend beyond the MOSAiC CO floe for a few tens of kilometers into the DN, in most cases reaching the three L-sites? They cover several buoys, including seven IMBs of your study.
Furthermore, I would like to pose you a question: which of the two has more severe effects, lack of spatial or temporal overlap? Earlier you stated that during the winter season, which is the focus of your study, there was little sea-ice dynamics influencing the study area. Temporal interpolation should not cause too many problems then. After all, you already apply it for snow density. Figure 4d even shows that the IS2 modal freeboard values are very close to the trend line.
IS2 modal freeboard. I may have been too quick previously agreeing that modal freeboard is an estimate of level ice. Coming back to it now after a while, I would like you to carefully consider the log-normal distribution and the modal value with respect to different variables. For sea-ice thickness measurements, it is indeed well-known and supported by studies that the modal value represents the most common value and the thickness of thermodynamically-grown level ice. For freeboard variables, whether it is sea-ice freeboard or snow/total freeboard, I think it is not so straight forward. In your answer to my comment, you give a long list of references very much like Koo et al. (2021), who write:
“Since the modal thickness represents the thickness of the most frequently observed ice or level ice (Farrell et al., 2012; Hansen et al., 2013; Petty et al., 2016; Rack et al., 2021; Tian et al., 2020), we estimate the thermodynamic ice growth around the buoys by using the variations in the modal thickness.”
They, like many if not all the studies you referenced, first transform snow or sea-ice freeboard (depending on the sensor in question) into sea-ice thickness with auxiliary data using the hydrostatic equilibrium assumption before deriving the thickness distribution and its modal value. I suppose that looking only at freeboard distribution you are (1) mixing the terms snow freeboard and sea-ice freeboard and (2) “cutting corners” in your study. Are sea-ice and snow freeboards correlated with each other or with sea-ice thickness, do they always behave the same? Could a seemingly level snow freeboard conceal sea ice that is not level? A single snow freeboard value can correspond to a wide range of sea-ice thickness values depending on the snow load that cannot be deduced from laser altimetry alone. Let’s consider a fixed total freeboard value of 0.3 m, approximately the derived modal value in your Figure 3c. Additionally, let’s assign representative and fixed values for the densities of sea water, ice, and snow, but let snow depth vary from 0 to 0.3 m (it could be even larger for cases of negative freeboard). The sea-ice thickness values resulting from this single total freeboard value, but varying snow depth information, can vary by more than 1.5 m according to the hydrostatic equilibrium equation.
Snow pits. Regarding the snow pit data, I do recommend being more explicit and more detailed in the methods section. While the data descriptor paper by Macfarlane et al. (2023) states a total of 576 snow pits, you have used only a small part of that data in your analysis. In fact, from the MOSAiC snow pit snow density cutter dataset (https://doi.org/10.1594/PANGAEA.940214), I can count only 85 snow pits between 25 October 2019 and 30 April 2020. The large number of “snow pits” stems from the fact how snow pits of different complexity were defined during the expedition: even a single profile with the SnowMicroPen (SMP) instrument could count as a snow pit. In the context of your manuscript, however, this is misleading since you did not use the SMP density profiles in your analysis. Furthermore, I trust you have indeed dug deep into the MOSAiC jargon of locations: snow pit locations “FR” in January-February and “DR” in April refer to “Fort Ridge” and “David’s Ridge” sites, respectively, that do not represent level ice conditions.
AWI IceBird average densities. In your response you claim to have used the mean values of sea-ice bulk density from Table 3 of Jutila et al. (2022). I have now double-checked this, and while the values for SYI and MYI seem fine, I don’t think this is true for FYI. Table 3 of Jutila et al. (2022) states 929.3 +/- 16.0 kg m-3 for 2017 and 925.4 +/- 17.7 kg m-3 for 2019, in addition the main text states 928.5 +/- 16.4 kg m-3 as the overall average density for FYI. However, your code snippet “Sea_ice_bulk_density.m” in Zenodo (https://doi.org/10.5281/zenodo.11055727) shows that a value of 921.4222 +/- 18.5586 kg m-3, smaller than any other FYI average value given in Jutila et al. (2022), was used for plotting Figure 5.
Citation: https://doi.org/10.5194/egusphere-2024-1240-CC2 -
AC2: 'Reply on CC2', Yi Zhou, 12 Jun 2024
Dear Arttu Jutila,
Thank you for your valuable suggestions and the further discussions needed. We apologize for any previous misunderstandings concerning your remarks. We have made several significant revisions, as detailed in the attached document. The original comments are in black, and our replies are written in blue.
Best regards,
Yi Zhou and other co-authors.
-
AC2: 'Reply on CC2', Yi Zhou, 12 Jun 2024
-
CC2: 'Reply on AC1', Arttu Jutila, 23 May 2024
- Spatial scales. While the presented study broadens the knowledge with the aspect of seasonal evolution of remotely sensed sea-ice bulk density, I am concerned about the different magnitudes of spatial scales utilized in the derivation. More specifically, you use total freeboard from the ICESat-2 satellite laser altimeter orbits extracted within a circle around the CO that has a diameter of 100 km; snow depth and sea-ice thickness derived from 15 autonomous IMBs in the DN within circle around the CO that has a diameter of 70 km (in the beginning of the drift, but how about later after being affected by sea-ice dynamics for months?) while the data are inherently point measurements; and snow density derived from snow pit measurements within the CO that extends over an area with a diameter of only few hundred meters while the data are inherently point measurements. None of these data sources have real spatial overlap with each other. This effectively diminishes the study to use ice-type-averaged values (not far from Alexandrov et al.’s (2010) multi-year ice density derivation with climatological values from literature) as the measurements are not from the same piece of ice – not even remotely.
-
CC3: 'Comment on egusphere-2024-1240', Hoyeon Shi, 08 Jun 2024
Dear Zhou et al.,
Thank you for sharing the interesting results with the Cryosphere community. I enjoyed reading the manuscript, but I have two concerns, so I'm leaving them here. I would appreciate it if the authors could consider the following comments.
The first is about using the modal IS2 freeboard, which has already been raised in detail by “Reply on AC1” by Arttu Jutila. Although the authors provide two arguments in L179, the question that must be answered regarding this issue is “How could the modal freeboard be more physically compatible with the hydrostatic balance equation (used for IBD estimation) than the mean freeboard?” I would like to emphasize that my question is not related to the authors’ assumption of the log-normal distribution of freeboard. This is because both “Mean” and “Mode” are available for any kind of statistical surface height distribution. What would be the physical meaning of modal height? How much do the parameterizations obtained from the mean and modal freeboards differ? The hydrostatic balance equation describes the balance of snow and ice mass and buoyancy, which are the quantities proportional to physical volume. Considering that the volume is area times height, it sounds more natural to me to multiply mean height than the modal height to area. I would like to hear the authors’ opinions on this.
Second, the manuscript is missing some context about using bivariate parameters for the IBD parameterization. The authors wrote they examined bivariate parameters, unlike previous studies that focused solely on the univariate parameters. This might mislead readers into thinking that there has been no study that used bivariate parameters to parameterize IBD, which is not true. The two papers already cited in the manuscript, Alexandrov et al. (2010) and Shi et al. (2023), have used the ice freeboard-to-thickness ratio to parameterize the sea ice bulk density. It should be noted that the ‘ice draft-to-thickness ratio’ is actually exactly the same as 1 – ice freeboard-to-thickness ratio. Furthermore, since Eq. (8) of the manuscript is a first-order linear equation, the form of the parameterization suggested by this study is the same as the one previously suggested by the two papers, i.e.:
IBD = a1 * draft-to-thickness-ratio + a2 = a1 * (1 – freeboard-to-thickness-ratio) + a2 = -a1 * freeboard-to-thickness-ratio + (a1 + a2)
Previous studies interpreted “-a1” as the density difference between floating and submerged parts of sea ice and “a1 + a2” as the density of submerged part of sea ice. Moreover, this parameterization has been implemented in the simultaneous estimation method that consistently estimates snow depth, sea ice thickness, ice freeboard, and IBD (Shi et al., 2023). Although the authors wrote in L358, “The strong linear relationship between the two bivariate parameters and IBD …, supporting previous suggestions (Alexandrov et al., 2010; Shi et al., 2023)”, I would recommend authors to provide a more complete context of using bivariate parameters in section 2.2.4.
Nevertheless, I think this study's novelty comes from determining coefficients of parameterization based on the multisource observations rather than using representative values available from the literature. Accordingly, it would be very valuable to our community if the authors could include the following things.
(1) Parameterization of IBD using the ice freeboard-to-thickness ratio, i.e.:
IBD = a1 * freeboard-to-thickness-ratio + a2
(2) Comparison of the determined parameterization equation above and the equations used in the previous studies with physical interpretation of the difference between them.
For instance, in Alexandrov et al. (2010), for multiyear sea ice, a1 is -370 kg m-3 (= 550 – 920) and a2 is 920 kg m-3. In Shi et al. (2023), a2 is the same, while a1 is -105 kg m-3 (= 815 – 920) for multiyear sea ice and -45 kg m-3 (= 875 – 920) for first-year sea ice.
Citation: https://doi.org/10.5194/egusphere-2024-1240-CC3 -
AC3: 'Reply on CC3', Yi Zhou, 12 Jun 2024
Dear Hoyeon Shi,
Thank you for your interest in our work and for providing valuable suggestions. We have provided clarification and discussion regarding your questions, and we have included several significant revisions in the revised manuscript, as detailed in the attached document. The original comments are in black, and our replies are written in blue.
Best regards,
Yi Zhou and other co-authors.
-
AC3: 'Reply on CC3', Yi Zhou, 12 Jun 2024
-
RC1: 'Comment on egusphere-2024-1240', Anonymous Referee #1, 09 Jun 2024
This manuscript, titled “Seasonal evolution and parameterization of Arctic sea ice bulk density: results from the MOSAiC expedition and ICESat-2/ATLAS”, deals with the sea ice density, which is a key issue in estimating the sea ice thickness in satellite altimetry. I consider it novel to use MOSAiC data for informing us of the seasonal evolution of sea ice density. And the majority of the analysis and result is sound. However, several major issues have to be made, which are listed below.
First, the spatial representation issue is central to the analysis, and has to be dealt with in a more systematic way. The two particular sources of uncertainty in Eqs. 5 are that of hf (total freeboard) and rho_s (snow density). For example, for hf and the analysis with buoy-measured hi and hs in Fig. 3, the uncertainty is actually two fold, under formal definitions. First, the uncertainty between the mode of the log-normal fitted IS2 hf and the areal mean level-ice hf, and second, that between the areal mean level-ice hf and that measured at the buoy (in order to compare with buoy hi & hs). The total uncertainty addressed here is only the difference between fitted mode and the maximum probability bin of hf. This uncertainty falls into part of the first uncertainty I mentioned, and hence the second uncertainty is not accounted for. The representation error in snow depth is potentially large as well, but could diminish more quickly at larger scales. The representation issue is also raised by two community comments.
Second, and consequently from the first point I raised, the apparent better fitting with the bivariate formulation (Sec. 3.3 and Fig. 7). One should be very careful in claiming that any fitting is better for parameterizing the ice density. Since if one look closely at Eqs. 5 and the first point I raised, it is immediate that the representation uncertainty will be a major source of correlation (R2>0.9 for both cases) with bivariate formulation, since: rho_i = rho_w*(hi+hs-hf)/hi + …, where (hi+hs-hf) is the derived sea ice draft, and contains a large representation error. This error gets carried to rho_i, in proportions, so that the values on both sides correlates really well (R2=0.9). In a sense rho_i and draft/thickness ratio are NOT independent due to the way rho_i is derived. And more importantly, the representation uncertainty dominates over the variability in the snow-related, second term in Eqs. 5. Let me be very clear here: I think there should exist significant correlation between the rho_i and draft/thickness ratio, which could arises from physical reasons (see also IceBird results in Sec. 4.4). I just don’t consider the argument here to be strict enough for comparing the parameterization schemes for ice density, and especially, whether the bivariates are better. Better quantification of uncertainty due to limited representation, is potentially needed before such claims.
Third, I think better sea ice topography data collected during MOSAiC campaign serve as a very good source of information for this study, which is a point already raised by Arttu (in first CC). However, maybe the dataset does not fully support the study of the whole winter, but it is definitely worth to look into and discussed in the paper.
Fourth, I consider the use of IceBird data could be improved. The scale dependency analysis is nice, but one has to be clear of two aspects. First, the uncertainty of SnowRadar (hence hs) and EM (hence hi+hs) over rough ice, which could compromise the analysis for this particularly important ice type. Second, the potential of apparent but superficial statistical correlation since the derived rho_i carries the measurement and representation error of the original measurements. Therefore, I suggest to change the analysis to level ice only with IceBird data, since such type info is available.
A minor comment: on line 452: the increase of hf may well be due to thermodynamic growth of ice, but purely/largely due to snow accumulation. So be more strict, as follows: These findings indicate that the magnitude of sea ice elevation changes exhibits significant spatial variability, possibly related to initial ice thickness, sea ice growth, and snow accumulation.
Citation: https://doi.org/10.5194/egusphere-2024-1240-RC1 -
AC4: 'Reply on RC1', Yi Zhou, 13 Jun 2024
Dear reviewer,
We sincerely appreciate your constructive comments, which have significantly contributed to the improvement of our manuscript. We have made thorough and detailed revisions according to your suggestions. Please refer to the attached document for a detailed review.
Best regards,
Yi Zhou and other co-authors.
-
AC4: 'Reply on RC1', Yi Zhou, 13 Jun 2024
-
CC4: 'Comment on egusphere-2024-1240', Mats Granskog, 10 Jun 2024
I also share some concerns already noted valuably by Arttu Jutila in an another comment to this manuscript. But I like want to here highlight that the "core-based density" IBD data used is not the most appropriate one from the MOSAiC Campaign, and I have grave concerns of the validity of some of the bold statements made would be (and are) based on flawed density data.
To my understanding the density used here is not measured from ice cores (as could perhaps be understood when reading the text), but calculated from other properties as was done in the paper by Angelopoulos et al. (this is the only ice density related data that is cited in the paper). It clearly needs to be stated how the density was derived.
But, what is more important is that the seasonal evolution of the IBD from the density measured directly from ice cores using the hydrostatic method (see Oggier et al., 2023ab, links at bottom comment), have been published and are available for use. This data shows a different story of the temporal density evolution, and contradicts what is shown here, and suggest that the calculated density used here is not correct.
Sea ice densities are up to 960 kg/m3 (Figs. 5,6) but with some simple logic, this would require no gas inclusions and very high brine volumes. This is not a feasible value for any but very new ice at very early stages of growth, not aged FYI or SYI. To me this is a fundamental flaw in the current manuscript, that the density values do not seem realistic and bold statements of the seasonal evolution are made based on calculated values that differ significantly from those measured in the field. And the description of the exact way the density data that is used has been derived can be misinterpreted.
I am also curious how data points shown in Figure 5 early in the time-series, do not appear in Figure 6. How is there a data point with a mean (FYI+SYI) density of 960 in Fig 5, but in Fig 6, the max of either FYI or SYI are than 935. How is the mean then of both combined up to 960?
The last comment is more generic to proper scientific conduct, and MOSAiC policies of using data derived by others. Inclusion of people who derived the data as co-authors would make the use of the data more robust, and what is done here does not really follow the practices that MOSAiC participants agreed to.
Data sets from both level FYI and SYI are available where the density was from rather high-accuracy hydrostatic weighing method. FYI: Oggier et al - https://doi.pangaea.de/10.1594/PANGAEA.956732 and SYI: Oggier et al - https://doi.org/10.1594/PANGAEA.959830 - in addition there is some ice density data from sea ice ridges (but from Leg 4 only) - https://doi.org/10.1594/PANGAEA.953865)yours truly,
Mats Granskog
with the hat as the MOSAiC task leader of Physical ice coring
Citation: https://doi.org/10.5194/egusphere-2024-1240-CC4 -
AC5: 'Reply on CC4', Yi Zhou, 15 Jun 2024
Dear Dr. Mats Granskog,
Thank you for your detailed review and valuable comments on our manuscript. We believe that your suggestions and comments will effectively improve the rigour of the relevant content of our manuscript. First of all, we would like to express our gratitude.
a) We must acknowledge a significant oversight in the use of ice core data. Yes, we used the sea ice density data provided by the BGC group, which was not directly measured but estimated based on some assumptions. As you mentioned, the ice core density data based on the relative rule sampling strategy and rigorous measurement process have also been released from the ice-core working group, so we will update the ice core data in the revision. During the revision, we will discuss with the responsible person of ice-core working group for describing the data more rigorously and potentially better explain the seasonal evolution mechanism of ice density. We will further acknowledge their contributions and invite them to join our author list if they are willing.
b) In addition, we believe that the sea ice bulk density obtained from very localized ice-core sampling also has limitations, especially for the representativeness of the MOSAiC DN scale (~50 km).
First, we must emphasize that we will carefully consider the IBD retrieval method and the issues of spatial variability raised by the community and reviewer RC1, and will make some revisions on this issue. Through a series of spatially adjusted methods, we can extract the mean IBD of the level ice at the DN scale, based on the assumption of hydrostatic equilibrium.
Given the significant spatial heterogeneity of sea ice at the MOSAiC DN scale, it is crucial to recognize that direct comparisons between IBD measured from ice cores from two sites and the mean IBD at the DN scale are not straightforward. The reason why we have gone to considerable effort to derive DN-scale IBD results from different observations during the MOSAiC freezing season is also due to the influence of spatial representativeness, in order to provide reliable references at the grid scales ( a few kilometers to tens of kilometers) of satellite remote sensing or numerical models.
Through our analysis of IBD at different buoy sites, and the recent introduction of IBD results at the L-site scale (see details in our responses to others), we have identified significant variation in IBD. These results suggest that there is strong spatial heterogeneity in IBD, caused primarily by ice age or thickness. Furthermore, according to the multi-sensor airborne IBD derived by Jutila et al. (2022), which is sufficiently spatially overlap, we observe that even along a single measurement trajectory (as shown in their Fig. 5), IBD can vary between approximately 800 and 1000 kg m-3. This variability further illustrates the significant spatial heterogeneity inherent in sea ice properties. Therefore, a significant difference in both the absolute value and its seasonal variation is expected between the DN-scale results and those from ice core sampling from very limited sites. In the revised version, we will also combine the SYI and FYI ice core density time series provided by the ice core working group to provide a more rigorous discussion of the different mechanisms affecting the seasonal evolution of sea ice density.
c) In response to your comments regarding Figures 5 and 6.
The description in our original manuscript may not be clear enough. We clarify that, Figure 5 shows the average IBD at the DN scale for level ice, while Figure 6 shows IBD results derived from ice cores provided by the BGC group (will be updated). Therefore, the results shown in these two figures are independent of each other.
Furthermore, we acknowledge your concerns about the rationality of the relatively high density values in autumn. However, from the perspective of DN-scale averaging, the mean sea ice parameters we obtain would include some very thin and young ice over the leads, which was an event younger than the FYI site, and thus would increase the spatial mean ice density to some extent. When the ice is still relatively thin in autumn, there may indeed be some uncertainty in obtaining the bulk density of sea ice based on the Archimedean principle. Therefore, based on the initial satellite altimetry data in autumn and combined with updated ice core observations, we can consider and test some constraint mechanisms to obtain more reasonable data or data interpretation.
We have also reviewed the core-based IBD results provided by the ice core working group, measured using the hydrostatic weighing method, and have tentatively identified some unusual phenomena: (1) IBD results for FYI are generally lower than those for SYI; (2) FYI cores show a significant increase in IBD from October to December. These results seem to contradict the conclusions of many previous studies. However, we continue to argue that sampling size plays a crucial role in these discrepancies. We advocate a detailed comparison of IBD results at different scales to fully understand these variations and their implications. Of course, ice core measurements help us to understand the seasonal evolution mechanism of a particular sea ice type. Therefore, we will also combine the updated ice core data to improve our discussion and make it more rigorous.
Overall, we are grateful for your reminder and are taking significant steps to improve the rigour and robustness of our study.
Best regards,
Yi Zhou and other co-authors.
Reference
Jutila, A., Hendricks, S., Ricker, R., von Albedyll, L., Krumpen, T., and Haas, C.: Retrieval and parameterisation of sea-ice bulk density from airborne multi-sensor measurements, The Cryosphere, 16, 259-275, 2022.
Oggier, M., Salganik, E., Whitmore, L., Fong, A. A., Hoppe, C. J. M., Rember, R., Høyland, K. V., Divine, D. V., Gradinger, R., Fons, S. W., Abrahamsson, K., Aguilar-Islas, A. M., Angelopoulos, M., Arndt, S., Balmonte, J. P., Bozzato, D., Bowman, J. S., Castellani, G., Chamberlain, E., Creamean, J., D'Angelo, A., Damm, E., Dumitrascu, A., Eggers, S. L., Gardner, J., Grosfeld, L., Haapala, J., Immerz, A., Kolabutin, N., Lange, B. A., Lei, R., Marsay, C. M., Maus, S., Müller, O., Olsen, L. M., Nuibom, A., Ren, J., Rinke, A., Sheikin, I., Shimanchuk, E., Snoeijs-Leijonmalm, P., Spahic, S., Stefels, J., Torres-Valdés, S., Torstensson, A., Ulfsbo, A., Verdugo, J., Vortkamp, M., Wang, L., Webster, M., Wischnewski, L., and Granskog, M. A.: First-year sea-ice salinity, temperature, density, oxygen and hydrogen isotope composition from the main coring site (MCS-FYI) during MOSAiC legs 1 to 4 in 2019/2020. PANGAEA, 2023a.
Oggier, M., Salganik, E., Whitmore, L., Fong, A. A., Hoppe, C. J. M., Rember, R., Høyland, K. V., Gradinger, R., Divine, D. V., Fons, S. W., Abrahamsson, K., Aguilar-Islas, A. M., Angelopoulos, M., Arndt, S., Balmonte, J. P., Bozzato, D., Bowman, J. S., Castellani, G., Chamberlain, E., Creamean, J., D'Angelo, A., Damm, E., Dumitrascu, A., Eggers, L., Gardner, J., Grosfeld, L., Haapala, J., Immerz, A., Kolabutin, N., Lange, B. A., Lei, R., Marsay, C. M., Maus, S., Olsen, L. M., Müller, O., Nuibom, A., Ren, J., Rinke, A., Sheikin, I., Shimanchuk, E., Snoeijs-Leijonmalm, P., Spahic, S., Stefels, J., Torres-Valdés, S., Torstensson, A., Ulfsbo, A., Verdugo, J., Vortkamp, M., Wang, L., Webster, M., Wischnewski, L., and Granskog, M. A.: Second-year sea-ice salinity, temperature, density, oxygen and hydrogen isotope composition from the main coring site (MCS-SYI) during MOSAiC legs 1 to 4 in 2019/2020. PANGAEA, 2023b.
Citation: https://doi.org/10.5194/egusphere-2024-1240-AC5
-
AC5: 'Reply on CC4', Yi Zhou, 15 Jun 2024
-
RC2: 'Comment on egusphere-2024-1240', Anonymous Referee #2, 14 Jun 2024
This review focuses on the manuscript titled “Seasonal evolution and parameterization of Arctic sea ice bulk density: results from the MOSAiC expedition and ICESat-2/ATLAS”. The research presented in this manuscript addresses a critical and highly relevant topic concerning the seasonal changes in the bulk density of Arctic sea ice and its implications for altimetry estimates. The central Arctic winter presents unique challenges and variability in ice density, which can significantly impact the accuracy of altimetry measurements. The study's integration of data from the MOSAiC expedition provides a comprehensive dataset, enriching our understanding of the physical properties of sea ice throughout the winter season. The study’s improved parameterization methods have the potential to significantly enhance the accuracy of satellite altimetry estimates. Addressing the online discussions about spatial variability will likely result in important revisions, further strengthening the manuscript's contributions to the scientific community. Overall, a well-constructed and well-written paper which I enjoyed reading.
General comments
I would be interested to know the implications of including a varying bulk density in satellite retrievals, you touched on this in the abstract. Would it change thickness estimates by a lot if changing the ice bulk density? How much would the 30-35% uncertainty (line 50) be reduced by including a varying IBD?
In altimetry models, FYI and SYI are treated very differently. Is it possible to make more of a separation between these ice types? You mention 4 buoys deployed in the FYI and 11 in the SYI, can you color these differently in Figure 1b and figure 5 can you split these into a FYI and SYI, as it appears to be quite a difference between the two in figure 6. I acknowledge you discuss the limitations in the discussion but I am curious how different the parameterizations would be for each ice type.
If the community is to take your methodology and apply it to satellite data products, we need to know how representative MOSAiC is to the general Arctic basin and how representative it is this year. Were the ocean, and atmospheric conditions typical for a season? Rinke (2021) explains if this year is representative atmospherically. Oceanographically, it might be worth looking through this Schultz (2023) accepted preprint to add a sentence or two about this.
Line 130: Snow-ice formation was suggested for site T72 (Figure 4), in Figure 8 T 72 appears to have increasingly positive freeboard over the season, implying that this is a very local effect on the SIMBA data. I believe there is a local process occurring, likely due to dynamics (which is suggested in Lei (2022)), which causes an ice surface depression and an increase in sea ice thickness. One reason could be a nearby ridge formation. I think it’s important to clarify this and ensure the reader does not assume flooding is a result of increased snow weight.
Additionally, regarding the platelet ice, wouldn’t we see this increase in ice thickness in multiple buoys? I don’t think there is enough evidence to confirm these processes, so I would recommend the author exclude this label.
I agree with your approach for snow bulk density due to the lower relative contribution to total uncertainty is around 1.7%; Macfarlane (2023) Table 3 gives the average density on FYI, SYI refrozen leads and ridges, and their Figure 7 seems to agree with your snow density trends too.
Please ensure all figures are color-blind friendly, for example, the color bar used in Figure 1b, and trend lines in Figure 5. A quick check can be made here (https://www.color-blindness.com/coblis-color-blindness-simulator/).
Minor revisions and figure comments
Choosing more appropriate acronyms, currently SI is used for local-scale IBD. At initial glance I was unsure what this meant, could this be changed to a clearer acronym, for example IBDSI? If it is standard in the literature then please ignore this comment.
Also, ICESat-2 is already an acronym, I don’t see the need to shorten it further to IS. This would help the readability.
Line 36: Include manual in situ measurements of ice thickness
L60: “mass/volume, submersion, and specific gravity methods, which require sampling, ice block preparation, and measurement.” As some of these methods are used in this manuscript it would be good to expand on each of these and the benefits of each.
L119 and L240: σℎi and σℎs were both set to 0.02m following Lei et al. (2022)- is this from the vertical resolution of the IMB buoys of 2 cm? how well can they identify the interface? Sledd (2024) addresses the uncertainties of the interface detection and found some “transition layers”, possibly worth including details on how you have handled this.
Figure 1: the drift track given under 1b should be for 1a, and please explain what the colours are for each buoy in 1b. “Snow density: Average for each depth” should this be “averaged for each snowpit”?
Figure 2: I appreciated this figure, it was very clear and informative.
Figure 4a: What about the increase in sea ice thickness in I3? I think the labels are speculative, and without observations about what was causing the ice increase, it is hard to be certain what the increase in ice thickness was a result of.
Figure 4b: The label “strong horizontal blowing snow” only appears to span February and March; there were also blowing snow events in November and December (Gong 2023), so I believe this arrow is a bit misleading.
Figure 5: “based on” in situ observations, if I understand correctly this should just be “from” in situ observations
Figure 7: Figure captions should be easy to interpret without the body of the text. Consider including the measurements that have been used to obtain a) sea ice thickness, (b) total freeboard, (c) sea ice draft… etc. Is this the hydrostatic equilibrium approach (x-axis) and the ice core obtained bulk density (y-axis)?
Figure 9: Could the y-axes be changed to black? Unsure why the red was chosen and what it represents.
Additional references
Rinke, A., Cassano, J.J., Cassano, E.N., Jaiser, R. and Handorf, D., 2021. Meteorological conditions during the MOSAiC expedition: Normal or anomalous?. Elem Sci Anth, 9(1), p.00023. https://online.ucpress.edu/elementa/article/9/1/00023/118092/Meteorological-conditions-during-the-MOSAiC
Sledd, A., Shupe, M.D., Solomon, A., Cox, C.J., Perovich, D. and Lei, R., 2024. Snow thermal conductivity and conductive flux in the Central Arctic: Estimates from observations and implications for models. Elementa: Science of the Anthropocene, 12(1).https://doi.org/10.1525/elementa.2023.00086
Gong, X., Zhang, J., Croft, B., Yang, X., Frey, M.M., Bergner, N., Chang, R.Y.W., Creamean, J.M., Kuang, C., Martin, R.V. and Ranjithkumar, A., 2023. Arctic warming by abundant fine sea salt aerosols from blowing snow. Nature Geoscience, 16(9), pp.768-774. https://www.nature.com/articles/s41561-023-01254-8
Schulz, K., Koenig, Z., Muilwijk, M., Bauch, D., Hoppe, C.J., Droste, E., Hoppmann, M., Chamberlain, E.J., Laukert, G., Stanton, T. and Zurita, A.Q., 2023. The Eurasian Arctic Ocean along the MOSAiC drift (2019-2020): An interdisciplinary perspective on properties and processes. https://eartharxiv.org/repository/view/5902/
Macfarlane, A.R., Löwe, H., Gimenes, L., Wagner, D.N., Dadic, R., Ottersberg, R., Hämmerle, S. and Schneebeli, M., 2023. Temporospatial variability of snow's thermal conductivity on Arctic sea ice. The Cryosphere, 17(12), pp.5417-5434. https://tc.copernicus.org/articles/17/5417/2023/
Citation: https://doi.org/10.5194/egusphere-2024-1240-RC2
Status: closed
-
CC1: 'Comment on egusphere-2024-1240', Arttu Jutila, 15 May 2024
Community comment on the manuscript by Zhou, Y., Wang, X., Lei, R., von Albedyll, L., Perovich, D. K., Zhang, Y., and Haas, C.: Seasonal evolution and parameterization of Arctic sea ice bulk density: results from the MOSAiC expedition and ICESat-2/ATLAS, EGUsphere [preprint], https://doi.org/10.5194/egusphere-2024-1240, 2024.
Dear Zhou et al.,
Dear handling editor,The competing interests policy of The Cryosphere prohibits me from acting as an official referee for this manuscript due to recent collaborations with some of the coauthors. Therefore, I am posting this comment as a member of the scientific community to discuss some matters related to it.
In the manuscript by Zhou et al., now under peer-review process and public discussion, sea-ice bulk density is derived using the hydrostatic equilibrium equation and values of modal total freeboard from the satellite laser altimeter ICESat-2, mean snow depth and sea-ice thickness from 15 autonomous ice mass-balance buoys (IMB) deployed within the MOSAiC Distributed Network (DN), and mean snow density from snow pit measurements conducted in the MOSAiC Central Observatory (CO) from October 2019 to April 2020.
In 2022, I have authored a paper in The Cryosphere on the same general topic, sea-ice bulk density, using a similar approach but simultaneous airborne multi-sensor measurements from the AWI IceBird program:
Jutila, A., Hendricks, S., Ricker, R., von Albedyll, L., Krumpen, T., and Haas, C.: Retrieval and parameterisation of sea-ice bulk density from airborne multi-sensor measurements, The Cryosphere, 16, 259–275, https://doi.org/10.5194/tc-16-259-2022, 2022.
This work is referenced many times and data originating from this study are used in the manuscript by Zhou et al.
First, I want to inform that there is a recently published new version of the AWI IceBird airborne sea-ice parameter dataset. In the new version, the quality flag identifying level and deformed ice has been rectified.
Jutila, A., Hendricks, S., Ricker, R., von Albedyll, L., and Haas, C.: Airborne sea ice parameters during the IceBird Winter 2019 campaign in the Arctic Ocean, Version 2, https://doi.org/10.1594/PANGAEA.966057, 2024.
Jutila, A., Hendricks, S., Ricker, R., von Albedyll, L., and Haas, C.: Airborne sea ice parameters during the PAMARCMIP2017 campaign in the Arctic Ocean, Version 2, https://doi.org/10.1594/PANGAEA.966009, 2024.
Regarding the study of Zhou et al., I would like to raise general concerns and perhaps some misunderstandings of my paper. The general points are the following:
- Spatial scales. While the presented study broadens the knowledge with the aspect of seasonal evolution of remotely sensed sea-ice bulk density, I am concerned about the different magnitudes of spatial scales utilized in the derivation. More specifically, you use total freeboard from the ICESat-2 satellite laser altimeter orbits extracted within a circle around the CO that has a diameter of 100 km; snow depth and sea-ice thickness derived from 15 autonomous IMBs in the DN within circle around the CO that has a diameter of 70 km (in the beginning of the drift, but how about later after being affected by sea-ice dynamics for months?) while the data are inherently point measurements; and snow density derived from snow pit measurements within the CO that extends over an area with a diameter of only few hundred meters while the data are inherently point measurements. None of these data sources have real spatial overlap with each other. This effectively diminishes the study to use ice-type-averaged values (not far from Alexandrov et al.’s (2010) multi-year ice density derivation with climatological values from literature) as the measurements are not from the same piece of ice – not even remotely.
In addition, using the term “local-scale” with data originating extending 100 km, when local is generally understood as <~1 km, definitely raised my eyebrows throughout the manuscript.
Why was ICESat-2 ATL10 rev5 used when rev6 is available? Actually, why not use the publicly available, more local MOSAiC helicopterborne laser scanner data by Hutter et al. (2023), which you also cite in the manuscript? I think that could be a feasible option to explore and it would back up better the local-scale aspect of this study. At this point, however, I must point out that I was involved in collecting and processing said data, too. - Level ice. You state that the chosen IMBs were deployed on level ice. I agree that this is a correct approach, to consider level ice only. However, the publicly available deployment documents for the buoys T63, T65, T70, and I1 indicate that ridged ice was already in close vicinity during deployment.
How was it ensured that ICESat-2 data was considered over level ice only? How long are the data segments, did they include only level data? While the modal value of the log-normal fitted freeboard is an estimate of the thermodynamically grown sea ice, it does not strictly exclude e.g. thin sea ice that has deformed and gained the same freeboard as thermodynamically grown undeformed sea ice.
How about snow pits, have you considered that pressure ridge sites were sampled on MOSAiC, too? Level ice tends to have thinner snowpack with larger temperature gradients that lead to snow metamorphism affecting the snow density profile.
When comparing your data to the AWI IceBird dataset, did you choose measurements on level ice only quality flag? I would suggest doing so, and in that case also using the updated version of the dataset.
More specific comments:
L62ff: Alexandrov et al. (2010) did not use airborne multi-sensor data. They used ground-based drill-hole measurements achieved through landing airplanes on the sea ice in the 1980s (Soviet Sever expeditions). So far, I am not aware of any other study utilizing airborne multi-sensor measurements to derive sea-ice bulk density than Jutila et al. (2022).
L79ff: While Shi et al. (2023) have more recently argued the point, it was mentioned earlier in Jutila et al. (2022), to which also Shi et al. (2023) refer.
L179ff: Sea-ice freeboard and thickness are found to follow log-normal or exponential distribution, but does total freeboard behave the same? And how about on the 100 km scale?
L232ff: Both your snow depth and sea-ice thickness measurements come from the IMBs. Therefore, are their uncertainties not independent and the assumption thus wrong?
L244ff: Were any other formulations than first and second order polynomials investigated?
Figure 5 & L313ff: Which values are you using for the three J22 data points? To my eyes, they do not match the values from Table 3 in Jutila et al. (2022) that list the average bulk densities on 800 m spatial scale. Or did you perhaps derive those values from the nominal resolution datasets? Did you use all values or only the level ice ones? Furthermore, I recommend using the same marker shape for the same ice type, adding citations also to the main text, and explaining the acronym “J22” (now only on L361).
L480ff: The data consists of several profiles covering a total distance of more than 3000 km (3410 km). Surveyed sea ice was primarily first-year ice (100 % in 2017) and multi-year ice, with only very little second-year ice. While you mention the spatial resolution of the data, I also think it’s important to distinguish between the nominal measurement spacing (5-6 m) and the footprint size (40 m) of the measurement.
L493ff & Figure 10: Jutila et al. (2022) applied inverse-uncertainty weighted mean, not inverse distance. Are all ice types included in this analysis, also level ice and first-year ice, even though you’re targeting to analyze rough and older ice? The AWI IceBird airborne sea-ice parameter datasets can easily distinguish different ice types using the provided quality flags.
L523ff: The “new approach proposed in this study to determine [ice bulk density] at the basin scale using satellite altimetry data” is not new as this capability has been previously demonstrated in Jutila et al. (2022). If you mean using satellite altimetry data in your approach to determine sea-ice bulk density (together with ground-based point measurements), you need to present and discuss the effect of different scales for the reasons brought up earlier. The study also seems to highlight the parameterization applying the ice draft-to-thickness ratio, but there is no current or planned satellite mission that can directly observe sea-ice draft, thickness, nor their ratio.
Citation: https://doi.org/10.5194/egusphere-2024-1240-CC1 -
AC1: 'Reply on CC1', Yi Zhou, 18 May 2024
Dear Dr. Arttu Jutila,
Thank you for your constructive comments. We will address your insights with comprehensive clarifications and revisions throughout our manuscript. We have outlined the original comments in black with our planned responses highlighted in blue. Kindly refer to the attached document.
Best regards,
Yi Zhou and other co-authors.
-
CC2: 'Reply on AC1', Arttu Jutila, 23 May 2024
Reply to author comment https://doi.org/10.5194/egusphere-2024-1240-AC1 by Zhou et al. for the manuscript by Zhou, Y., Wang, X., Lei, R., von Albedyll, L., Perovich, D. K., Zhang, Y., and Haas, C.: Seasonal evolution and parameterization of Arctic sea ice bulk density: results from the MOSAiC expedition and ICESat-2/ATLAS, EGUsphere [preprint], https://doi.org/10.5194/egusphere-2024-1240, 2024.
Dear Zhou et al.,
Thank you for your response to my comment and in particular for continuing the public discussion in the spirit of The Cryosphere’s interactive review process.
I am pleased to see that you have carefully considered my comments. Nevertheless, I still see some loose ends which I would like to discuss further.
Scales. Thank you for the very helpful additional Figure A1. While you clarify that the extraction radii for IS2 and IMBs were 50 km and 30-40 km, respectively, you have unfortunately misread my comment. I did not use the term radius but diameter, i.e. two times the radius, for a very specific reason. Let’s consider a rather standard satellite data product that has data in a regular grid format where the grid cells are square-shaped with each side measuring 25 km in length. This results in a spatial scale of 25 km – not 12.5 km that would be the radius of the largest possible circle drawn within the grid cell. In your case, for example looking at Figure 3c where the IS2 ground track passes nearly directly above the center of the circle or the MOSAiC CO, you end up extracting along-track data for a length of 100 km, not 50 km. That is the largest length scale of your input data. The same principle applies to the 40-m footprint size of the AWI IceBird measurements: it is the diameter, not the radius, of the EM-Bird footprint, within which snow depth and freeboard measurements are averaged and sea-ice bulk density is eventually derived.
In your consideration of the helicopter-borne laser scanner data in Hutter et al. (2023), you state having found it temporally insufficient with only about ten days of valid data for your study period. Did you include also the approximately weekly transect flights that extend beyond the MOSAiC CO floe for a few tens of kilometers into the DN, in most cases reaching the three L-sites? They cover several buoys, including seven IMBs of your study.
Furthermore, I would like to pose you a question: which of the two has more severe effects, lack of spatial or temporal overlap? Earlier you stated that during the winter season, which is the focus of your study, there was little sea-ice dynamics influencing the study area. Temporal interpolation should not cause too many problems then. After all, you already apply it for snow density. Figure 4d even shows that the IS2 modal freeboard values are very close to the trend line.
IS2 modal freeboard. I may have been too quick previously agreeing that modal freeboard is an estimate of level ice. Coming back to it now after a while, I would like you to carefully consider the log-normal distribution and the modal value with respect to different variables. For sea-ice thickness measurements, it is indeed well-known and supported by studies that the modal value represents the most common value and the thickness of thermodynamically-grown level ice. For freeboard variables, whether it is sea-ice freeboard or snow/total freeboard, I think it is not so straight forward. In your answer to my comment, you give a long list of references very much like Koo et al. (2021), who write:
“Since the modal thickness represents the thickness of the most frequently observed ice or level ice (Farrell et al., 2012; Hansen et al., 2013; Petty et al., 2016; Rack et al., 2021; Tian et al., 2020), we estimate the thermodynamic ice growth around the buoys by using the variations in the modal thickness.”
They, like many if not all the studies you referenced, first transform snow or sea-ice freeboard (depending on the sensor in question) into sea-ice thickness with auxiliary data using the hydrostatic equilibrium assumption before deriving the thickness distribution and its modal value. I suppose that looking only at freeboard distribution you are (1) mixing the terms snow freeboard and sea-ice freeboard and (2) “cutting corners” in your study. Are sea-ice and snow freeboards correlated with each other or with sea-ice thickness, do they always behave the same? Could a seemingly level snow freeboard conceal sea ice that is not level? A single snow freeboard value can correspond to a wide range of sea-ice thickness values depending on the snow load that cannot be deduced from laser altimetry alone. Let’s consider a fixed total freeboard value of 0.3 m, approximately the derived modal value in your Figure 3c. Additionally, let’s assign representative and fixed values for the densities of sea water, ice, and snow, but let snow depth vary from 0 to 0.3 m (it could be even larger for cases of negative freeboard). The sea-ice thickness values resulting from this single total freeboard value, but varying snow depth information, can vary by more than 1.5 m according to the hydrostatic equilibrium equation.
Snow pits. Regarding the snow pit data, I do recommend being more explicit and more detailed in the methods section. While the data descriptor paper by Macfarlane et al. (2023) states a total of 576 snow pits, you have used only a small part of that data in your analysis. In fact, from the MOSAiC snow pit snow density cutter dataset (https://doi.org/10.1594/PANGAEA.940214), I can count only 85 snow pits between 25 October 2019 and 30 April 2020. The large number of “snow pits” stems from the fact how snow pits of different complexity were defined during the expedition: even a single profile with the SnowMicroPen (SMP) instrument could count as a snow pit. In the context of your manuscript, however, this is misleading since you did not use the SMP density profiles in your analysis. Furthermore, I trust you have indeed dug deep into the MOSAiC jargon of locations: snow pit locations “FR” in January-February and “DR” in April refer to “Fort Ridge” and “David’s Ridge” sites, respectively, that do not represent level ice conditions.
AWI IceBird average densities. In your response you claim to have used the mean values of sea-ice bulk density from Table 3 of Jutila et al. (2022). I have now double-checked this, and while the values for SYI and MYI seem fine, I don’t think this is true for FYI. Table 3 of Jutila et al. (2022) states 929.3 +/- 16.0 kg m-3 for 2017 and 925.4 +/- 17.7 kg m-3 for 2019, in addition the main text states 928.5 +/- 16.4 kg m-3 as the overall average density for FYI. However, your code snippet “Sea_ice_bulk_density.m” in Zenodo (https://doi.org/10.5281/zenodo.11055727) shows that a value of 921.4222 +/- 18.5586 kg m-3, smaller than any other FYI average value given in Jutila et al. (2022), was used for plotting Figure 5.
Citation: https://doi.org/10.5194/egusphere-2024-1240-CC2 -
AC2: 'Reply on CC2', Yi Zhou, 12 Jun 2024
Dear Arttu Jutila,
Thank you for your valuable suggestions and the further discussions needed. We apologize for any previous misunderstandings concerning your remarks. We have made several significant revisions, as detailed in the attached document. The original comments are in black, and our replies are written in blue.
Best regards,
Yi Zhou and other co-authors.
-
AC2: 'Reply on CC2', Yi Zhou, 12 Jun 2024
-
CC2: 'Reply on AC1', Arttu Jutila, 23 May 2024
- Spatial scales. While the presented study broadens the knowledge with the aspect of seasonal evolution of remotely sensed sea-ice bulk density, I am concerned about the different magnitudes of spatial scales utilized in the derivation. More specifically, you use total freeboard from the ICESat-2 satellite laser altimeter orbits extracted within a circle around the CO that has a diameter of 100 km; snow depth and sea-ice thickness derived from 15 autonomous IMBs in the DN within circle around the CO that has a diameter of 70 km (in the beginning of the drift, but how about later after being affected by sea-ice dynamics for months?) while the data are inherently point measurements; and snow density derived from snow pit measurements within the CO that extends over an area with a diameter of only few hundred meters while the data are inherently point measurements. None of these data sources have real spatial overlap with each other. This effectively diminishes the study to use ice-type-averaged values (not far from Alexandrov et al.’s (2010) multi-year ice density derivation with climatological values from literature) as the measurements are not from the same piece of ice – not even remotely.
-
CC3: 'Comment on egusphere-2024-1240', Hoyeon Shi, 08 Jun 2024
Dear Zhou et al.,
Thank you for sharing the interesting results with the Cryosphere community. I enjoyed reading the manuscript, but I have two concerns, so I'm leaving them here. I would appreciate it if the authors could consider the following comments.
The first is about using the modal IS2 freeboard, which has already been raised in detail by “Reply on AC1” by Arttu Jutila. Although the authors provide two arguments in L179, the question that must be answered regarding this issue is “How could the modal freeboard be more physically compatible with the hydrostatic balance equation (used for IBD estimation) than the mean freeboard?” I would like to emphasize that my question is not related to the authors’ assumption of the log-normal distribution of freeboard. This is because both “Mean” and “Mode” are available for any kind of statistical surface height distribution. What would be the physical meaning of modal height? How much do the parameterizations obtained from the mean and modal freeboards differ? The hydrostatic balance equation describes the balance of snow and ice mass and buoyancy, which are the quantities proportional to physical volume. Considering that the volume is area times height, it sounds more natural to me to multiply mean height than the modal height to area. I would like to hear the authors’ opinions on this.
Second, the manuscript is missing some context about using bivariate parameters for the IBD parameterization. The authors wrote they examined bivariate parameters, unlike previous studies that focused solely on the univariate parameters. This might mislead readers into thinking that there has been no study that used bivariate parameters to parameterize IBD, which is not true. The two papers already cited in the manuscript, Alexandrov et al. (2010) and Shi et al. (2023), have used the ice freeboard-to-thickness ratio to parameterize the sea ice bulk density. It should be noted that the ‘ice draft-to-thickness ratio’ is actually exactly the same as 1 – ice freeboard-to-thickness ratio. Furthermore, since Eq. (8) of the manuscript is a first-order linear equation, the form of the parameterization suggested by this study is the same as the one previously suggested by the two papers, i.e.:
IBD = a1 * draft-to-thickness-ratio + a2 = a1 * (1 – freeboard-to-thickness-ratio) + a2 = -a1 * freeboard-to-thickness-ratio + (a1 + a2)
Previous studies interpreted “-a1” as the density difference between floating and submerged parts of sea ice and “a1 + a2” as the density of submerged part of sea ice. Moreover, this parameterization has been implemented in the simultaneous estimation method that consistently estimates snow depth, sea ice thickness, ice freeboard, and IBD (Shi et al., 2023). Although the authors wrote in L358, “The strong linear relationship between the two bivariate parameters and IBD …, supporting previous suggestions (Alexandrov et al., 2010; Shi et al., 2023)”, I would recommend authors to provide a more complete context of using bivariate parameters in section 2.2.4.
Nevertheless, I think this study's novelty comes from determining coefficients of parameterization based on the multisource observations rather than using representative values available from the literature. Accordingly, it would be very valuable to our community if the authors could include the following things.
(1) Parameterization of IBD using the ice freeboard-to-thickness ratio, i.e.:
IBD = a1 * freeboard-to-thickness-ratio + a2
(2) Comparison of the determined parameterization equation above and the equations used in the previous studies with physical interpretation of the difference between them.
For instance, in Alexandrov et al. (2010), for multiyear sea ice, a1 is -370 kg m-3 (= 550 – 920) and a2 is 920 kg m-3. In Shi et al. (2023), a2 is the same, while a1 is -105 kg m-3 (= 815 – 920) for multiyear sea ice and -45 kg m-3 (= 875 – 920) for first-year sea ice.
Citation: https://doi.org/10.5194/egusphere-2024-1240-CC3 -
AC3: 'Reply on CC3', Yi Zhou, 12 Jun 2024
Dear Hoyeon Shi,
Thank you for your interest in our work and for providing valuable suggestions. We have provided clarification and discussion regarding your questions, and we have included several significant revisions in the revised manuscript, as detailed in the attached document. The original comments are in black, and our replies are written in blue.
Best regards,
Yi Zhou and other co-authors.
-
AC3: 'Reply on CC3', Yi Zhou, 12 Jun 2024
-
RC1: 'Comment on egusphere-2024-1240', Anonymous Referee #1, 09 Jun 2024
This manuscript, titled “Seasonal evolution and parameterization of Arctic sea ice bulk density: results from the MOSAiC expedition and ICESat-2/ATLAS”, deals with the sea ice density, which is a key issue in estimating the sea ice thickness in satellite altimetry. I consider it novel to use MOSAiC data for informing us of the seasonal evolution of sea ice density. And the majority of the analysis and result is sound. However, several major issues have to be made, which are listed below.
First, the spatial representation issue is central to the analysis, and has to be dealt with in a more systematic way. The two particular sources of uncertainty in Eqs. 5 are that of hf (total freeboard) and rho_s (snow density). For example, for hf and the analysis with buoy-measured hi and hs in Fig. 3, the uncertainty is actually two fold, under formal definitions. First, the uncertainty between the mode of the log-normal fitted IS2 hf and the areal mean level-ice hf, and second, that between the areal mean level-ice hf and that measured at the buoy (in order to compare with buoy hi & hs). The total uncertainty addressed here is only the difference between fitted mode and the maximum probability bin of hf. This uncertainty falls into part of the first uncertainty I mentioned, and hence the second uncertainty is not accounted for. The representation error in snow depth is potentially large as well, but could diminish more quickly at larger scales. The representation issue is also raised by two community comments.
Second, and consequently from the first point I raised, the apparent better fitting with the bivariate formulation (Sec. 3.3 and Fig. 7). One should be very careful in claiming that any fitting is better for parameterizing the ice density. Since if one look closely at Eqs. 5 and the first point I raised, it is immediate that the representation uncertainty will be a major source of correlation (R2>0.9 for both cases) with bivariate formulation, since: rho_i = rho_w*(hi+hs-hf)/hi + …, where (hi+hs-hf) is the derived sea ice draft, and contains a large representation error. This error gets carried to rho_i, in proportions, so that the values on both sides correlates really well (R2=0.9). In a sense rho_i and draft/thickness ratio are NOT independent due to the way rho_i is derived. And more importantly, the representation uncertainty dominates over the variability in the snow-related, second term in Eqs. 5. Let me be very clear here: I think there should exist significant correlation between the rho_i and draft/thickness ratio, which could arises from physical reasons (see also IceBird results in Sec. 4.4). I just don’t consider the argument here to be strict enough for comparing the parameterization schemes for ice density, and especially, whether the bivariates are better. Better quantification of uncertainty due to limited representation, is potentially needed before such claims.
Third, I think better sea ice topography data collected during MOSAiC campaign serve as a very good source of information for this study, which is a point already raised by Arttu (in first CC). However, maybe the dataset does not fully support the study of the whole winter, but it is definitely worth to look into and discussed in the paper.
Fourth, I consider the use of IceBird data could be improved. The scale dependency analysis is nice, but one has to be clear of two aspects. First, the uncertainty of SnowRadar (hence hs) and EM (hence hi+hs) over rough ice, which could compromise the analysis for this particularly important ice type. Second, the potential of apparent but superficial statistical correlation since the derived rho_i carries the measurement and representation error of the original measurements. Therefore, I suggest to change the analysis to level ice only with IceBird data, since such type info is available.
A minor comment: on line 452: the increase of hf may well be due to thermodynamic growth of ice, but purely/largely due to snow accumulation. So be more strict, as follows: These findings indicate that the magnitude of sea ice elevation changes exhibits significant spatial variability, possibly related to initial ice thickness, sea ice growth, and snow accumulation.
Citation: https://doi.org/10.5194/egusphere-2024-1240-RC1 -
AC4: 'Reply on RC1', Yi Zhou, 13 Jun 2024
Dear reviewer,
We sincerely appreciate your constructive comments, which have significantly contributed to the improvement of our manuscript. We have made thorough and detailed revisions according to your suggestions. Please refer to the attached document for a detailed review.
Best regards,
Yi Zhou and other co-authors.
-
AC4: 'Reply on RC1', Yi Zhou, 13 Jun 2024
-
CC4: 'Comment on egusphere-2024-1240', Mats Granskog, 10 Jun 2024
I also share some concerns already noted valuably by Arttu Jutila in an another comment to this manuscript. But I like want to here highlight that the "core-based density" IBD data used is not the most appropriate one from the MOSAiC Campaign, and I have grave concerns of the validity of some of the bold statements made would be (and are) based on flawed density data.
To my understanding the density used here is not measured from ice cores (as could perhaps be understood when reading the text), but calculated from other properties as was done in the paper by Angelopoulos et al. (this is the only ice density related data that is cited in the paper). It clearly needs to be stated how the density was derived.
But, what is more important is that the seasonal evolution of the IBD from the density measured directly from ice cores using the hydrostatic method (see Oggier et al., 2023ab, links at bottom comment), have been published and are available for use. This data shows a different story of the temporal density evolution, and contradicts what is shown here, and suggest that the calculated density used here is not correct.
Sea ice densities are up to 960 kg/m3 (Figs. 5,6) but with some simple logic, this would require no gas inclusions and very high brine volumes. This is not a feasible value for any but very new ice at very early stages of growth, not aged FYI or SYI. To me this is a fundamental flaw in the current manuscript, that the density values do not seem realistic and bold statements of the seasonal evolution are made based on calculated values that differ significantly from those measured in the field. And the description of the exact way the density data that is used has been derived can be misinterpreted.
I am also curious how data points shown in Figure 5 early in the time-series, do not appear in Figure 6. How is there a data point with a mean (FYI+SYI) density of 960 in Fig 5, but in Fig 6, the max of either FYI or SYI are than 935. How is the mean then of both combined up to 960?
The last comment is more generic to proper scientific conduct, and MOSAiC policies of using data derived by others. Inclusion of people who derived the data as co-authors would make the use of the data more robust, and what is done here does not really follow the practices that MOSAiC participants agreed to.
Data sets from both level FYI and SYI are available where the density was from rather high-accuracy hydrostatic weighing method. FYI: Oggier et al - https://doi.pangaea.de/10.1594/PANGAEA.956732 and SYI: Oggier et al - https://doi.org/10.1594/PANGAEA.959830 - in addition there is some ice density data from sea ice ridges (but from Leg 4 only) - https://doi.org/10.1594/PANGAEA.953865)yours truly,
Mats Granskog
with the hat as the MOSAiC task leader of Physical ice coring
Citation: https://doi.org/10.5194/egusphere-2024-1240-CC4 -
AC5: 'Reply on CC4', Yi Zhou, 15 Jun 2024
Dear Dr. Mats Granskog,
Thank you for your detailed review and valuable comments on our manuscript. We believe that your suggestions and comments will effectively improve the rigour of the relevant content of our manuscript. First of all, we would like to express our gratitude.
a) We must acknowledge a significant oversight in the use of ice core data. Yes, we used the sea ice density data provided by the BGC group, which was not directly measured but estimated based on some assumptions. As you mentioned, the ice core density data based on the relative rule sampling strategy and rigorous measurement process have also been released from the ice-core working group, so we will update the ice core data in the revision. During the revision, we will discuss with the responsible person of ice-core working group for describing the data more rigorously and potentially better explain the seasonal evolution mechanism of ice density. We will further acknowledge their contributions and invite them to join our author list if they are willing.
b) In addition, we believe that the sea ice bulk density obtained from very localized ice-core sampling also has limitations, especially for the representativeness of the MOSAiC DN scale (~50 km).
First, we must emphasize that we will carefully consider the IBD retrieval method and the issues of spatial variability raised by the community and reviewer RC1, and will make some revisions on this issue. Through a series of spatially adjusted methods, we can extract the mean IBD of the level ice at the DN scale, based on the assumption of hydrostatic equilibrium.
Given the significant spatial heterogeneity of sea ice at the MOSAiC DN scale, it is crucial to recognize that direct comparisons between IBD measured from ice cores from two sites and the mean IBD at the DN scale are not straightforward. The reason why we have gone to considerable effort to derive DN-scale IBD results from different observations during the MOSAiC freezing season is also due to the influence of spatial representativeness, in order to provide reliable references at the grid scales ( a few kilometers to tens of kilometers) of satellite remote sensing or numerical models.
Through our analysis of IBD at different buoy sites, and the recent introduction of IBD results at the L-site scale (see details in our responses to others), we have identified significant variation in IBD. These results suggest that there is strong spatial heterogeneity in IBD, caused primarily by ice age or thickness. Furthermore, according to the multi-sensor airborne IBD derived by Jutila et al. (2022), which is sufficiently spatially overlap, we observe that even along a single measurement trajectory (as shown in their Fig. 5), IBD can vary between approximately 800 and 1000 kg m-3. This variability further illustrates the significant spatial heterogeneity inherent in sea ice properties. Therefore, a significant difference in both the absolute value and its seasonal variation is expected between the DN-scale results and those from ice core sampling from very limited sites. In the revised version, we will also combine the SYI and FYI ice core density time series provided by the ice core working group to provide a more rigorous discussion of the different mechanisms affecting the seasonal evolution of sea ice density.
c) In response to your comments regarding Figures 5 and 6.
The description in our original manuscript may not be clear enough. We clarify that, Figure 5 shows the average IBD at the DN scale for level ice, while Figure 6 shows IBD results derived from ice cores provided by the BGC group (will be updated). Therefore, the results shown in these two figures are independent of each other.
Furthermore, we acknowledge your concerns about the rationality of the relatively high density values in autumn. However, from the perspective of DN-scale averaging, the mean sea ice parameters we obtain would include some very thin and young ice over the leads, which was an event younger than the FYI site, and thus would increase the spatial mean ice density to some extent. When the ice is still relatively thin in autumn, there may indeed be some uncertainty in obtaining the bulk density of sea ice based on the Archimedean principle. Therefore, based on the initial satellite altimetry data in autumn and combined with updated ice core observations, we can consider and test some constraint mechanisms to obtain more reasonable data or data interpretation.
We have also reviewed the core-based IBD results provided by the ice core working group, measured using the hydrostatic weighing method, and have tentatively identified some unusual phenomena: (1) IBD results for FYI are generally lower than those for SYI; (2) FYI cores show a significant increase in IBD from October to December. These results seem to contradict the conclusions of many previous studies. However, we continue to argue that sampling size plays a crucial role in these discrepancies. We advocate a detailed comparison of IBD results at different scales to fully understand these variations and their implications. Of course, ice core measurements help us to understand the seasonal evolution mechanism of a particular sea ice type. Therefore, we will also combine the updated ice core data to improve our discussion and make it more rigorous.
Overall, we are grateful for your reminder and are taking significant steps to improve the rigour and robustness of our study.
Best regards,
Yi Zhou and other co-authors.
Reference
Jutila, A., Hendricks, S., Ricker, R., von Albedyll, L., Krumpen, T., and Haas, C.: Retrieval and parameterisation of sea-ice bulk density from airborne multi-sensor measurements, The Cryosphere, 16, 259-275, 2022.
Oggier, M., Salganik, E., Whitmore, L., Fong, A. A., Hoppe, C. J. M., Rember, R., Høyland, K. V., Divine, D. V., Gradinger, R., Fons, S. W., Abrahamsson, K., Aguilar-Islas, A. M., Angelopoulos, M., Arndt, S., Balmonte, J. P., Bozzato, D., Bowman, J. S., Castellani, G., Chamberlain, E., Creamean, J., D'Angelo, A., Damm, E., Dumitrascu, A., Eggers, S. L., Gardner, J., Grosfeld, L., Haapala, J., Immerz, A., Kolabutin, N., Lange, B. A., Lei, R., Marsay, C. M., Maus, S., Müller, O., Olsen, L. M., Nuibom, A., Ren, J., Rinke, A., Sheikin, I., Shimanchuk, E., Snoeijs-Leijonmalm, P., Spahic, S., Stefels, J., Torres-Valdés, S., Torstensson, A., Ulfsbo, A., Verdugo, J., Vortkamp, M., Wang, L., Webster, M., Wischnewski, L., and Granskog, M. A.: First-year sea-ice salinity, temperature, density, oxygen and hydrogen isotope composition from the main coring site (MCS-FYI) during MOSAiC legs 1 to 4 in 2019/2020. PANGAEA, 2023a.
Oggier, M., Salganik, E., Whitmore, L., Fong, A. A., Hoppe, C. J. M., Rember, R., Høyland, K. V., Gradinger, R., Divine, D. V., Fons, S. W., Abrahamsson, K., Aguilar-Islas, A. M., Angelopoulos, M., Arndt, S., Balmonte, J. P., Bozzato, D., Bowman, J. S., Castellani, G., Chamberlain, E., Creamean, J., D'Angelo, A., Damm, E., Dumitrascu, A., Eggers, L., Gardner, J., Grosfeld, L., Haapala, J., Immerz, A., Kolabutin, N., Lange, B. A., Lei, R., Marsay, C. M., Maus, S., Olsen, L. M., Müller, O., Nuibom, A., Ren, J., Rinke, A., Sheikin, I., Shimanchuk, E., Snoeijs-Leijonmalm, P., Spahic, S., Stefels, J., Torres-Valdés, S., Torstensson, A., Ulfsbo, A., Verdugo, J., Vortkamp, M., Wang, L., Webster, M., Wischnewski, L., and Granskog, M. A.: Second-year sea-ice salinity, temperature, density, oxygen and hydrogen isotope composition from the main coring site (MCS-SYI) during MOSAiC legs 1 to 4 in 2019/2020. PANGAEA, 2023b.
Citation: https://doi.org/10.5194/egusphere-2024-1240-AC5
-
AC5: 'Reply on CC4', Yi Zhou, 15 Jun 2024
-
RC2: 'Comment on egusphere-2024-1240', Anonymous Referee #2, 14 Jun 2024
This review focuses on the manuscript titled “Seasonal evolution and parameterization of Arctic sea ice bulk density: results from the MOSAiC expedition and ICESat-2/ATLAS”. The research presented in this manuscript addresses a critical and highly relevant topic concerning the seasonal changes in the bulk density of Arctic sea ice and its implications for altimetry estimates. The central Arctic winter presents unique challenges and variability in ice density, which can significantly impact the accuracy of altimetry measurements. The study's integration of data from the MOSAiC expedition provides a comprehensive dataset, enriching our understanding of the physical properties of sea ice throughout the winter season. The study’s improved parameterization methods have the potential to significantly enhance the accuracy of satellite altimetry estimates. Addressing the online discussions about spatial variability will likely result in important revisions, further strengthening the manuscript's contributions to the scientific community. Overall, a well-constructed and well-written paper which I enjoyed reading.
General comments
I would be interested to know the implications of including a varying bulk density in satellite retrievals, you touched on this in the abstract. Would it change thickness estimates by a lot if changing the ice bulk density? How much would the 30-35% uncertainty (line 50) be reduced by including a varying IBD?
In altimetry models, FYI and SYI are treated very differently. Is it possible to make more of a separation between these ice types? You mention 4 buoys deployed in the FYI and 11 in the SYI, can you color these differently in Figure 1b and figure 5 can you split these into a FYI and SYI, as it appears to be quite a difference between the two in figure 6. I acknowledge you discuss the limitations in the discussion but I am curious how different the parameterizations would be for each ice type.
If the community is to take your methodology and apply it to satellite data products, we need to know how representative MOSAiC is to the general Arctic basin and how representative it is this year. Were the ocean, and atmospheric conditions typical for a season? Rinke (2021) explains if this year is representative atmospherically. Oceanographically, it might be worth looking through this Schultz (2023) accepted preprint to add a sentence or two about this.
Line 130: Snow-ice formation was suggested for site T72 (Figure 4), in Figure 8 T 72 appears to have increasingly positive freeboard over the season, implying that this is a very local effect on the SIMBA data. I believe there is a local process occurring, likely due to dynamics (which is suggested in Lei (2022)), which causes an ice surface depression and an increase in sea ice thickness. One reason could be a nearby ridge formation. I think it’s important to clarify this and ensure the reader does not assume flooding is a result of increased snow weight.
Additionally, regarding the platelet ice, wouldn’t we see this increase in ice thickness in multiple buoys? I don’t think there is enough evidence to confirm these processes, so I would recommend the author exclude this label.
I agree with your approach for snow bulk density due to the lower relative contribution to total uncertainty is around 1.7%; Macfarlane (2023) Table 3 gives the average density on FYI, SYI refrozen leads and ridges, and their Figure 7 seems to agree with your snow density trends too.
Please ensure all figures are color-blind friendly, for example, the color bar used in Figure 1b, and trend lines in Figure 5. A quick check can be made here (https://www.color-blindness.com/coblis-color-blindness-simulator/).
Minor revisions and figure comments
Choosing more appropriate acronyms, currently SI is used for local-scale IBD. At initial glance I was unsure what this meant, could this be changed to a clearer acronym, for example IBDSI? If it is standard in the literature then please ignore this comment.
Also, ICESat-2 is already an acronym, I don’t see the need to shorten it further to IS. This would help the readability.
Line 36: Include manual in situ measurements of ice thickness
L60: “mass/volume, submersion, and specific gravity methods, which require sampling, ice block preparation, and measurement.” As some of these methods are used in this manuscript it would be good to expand on each of these and the benefits of each.
L119 and L240: σℎi and σℎs were both set to 0.02m following Lei et al. (2022)- is this from the vertical resolution of the IMB buoys of 2 cm? how well can they identify the interface? Sledd (2024) addresses the uncertainties of the interface detection and found some “transition layers”, possibly worth including details on how you have handled this.
Figure 1: the drift track given under 1b should be for 1a, and please explain what the colours are for each buoy in 1b. “Snow density: Average for each depth” should this be “averaged for each snowpit”?
Figure 2: I appreciated this figure, it was very clear and informative.
Figure 4a: What about the increase in sea ice thickness in I3? I think the labels are speculative, and without observations about what was causing the ice increase, it is hard to be certain what the increase in ice thickness was a result of.
Figure 4b: The label “strong horizontal blowing snow” only appears to span February and March; there were also blowing snow events in November and December (Gong 2023), so I believe this arrow is a bit misleading.
Figure 5: “based on” in situ observations, if I understand correctly this should just be “from” in situ observations
Figure 7: Figure captions should be easy to interpret without the body of the text. Consider including the measurements that have been used to obtain a) sea ice thickness, (b) total freeboard, (c) sea ice draft… etc. Is this the hydrostatic equilibrium approach (x-axis) and the ice core obtained bulk density (y-axis)?
Figure 9: Could the y-axes be changed to black? Unsure why the red was chosen and what it represents.
Additional references
Rinke, A., Cassano, J.J., Cassano, E.N., Jaiser, R. and Handorf, D., 2021. Meteorological conditions during the MOSAiC expedition: Normal or anomalous?. Elem Sci Anth, 9(1), p.00023. https://online.ucpress.edu/elementa/article/9/1/00023/118092/Meteorological-conditions-during-the-MOSAiC
Sledd, A., Shupe, M.D., Solomon, A., Cox, C.J., Perovich, D. and Lei, R., 2024. Snow thermal conductivity and conductive flux in the Central Arctic: Estimates from observations and implications for models. Elementa: Science of the Anthropocene, 12(1).https://doi.org/10.1525/elementa.2023.00086
Gong, X., Zhang, J., Croft, B., Yang, X., Frey, M.M., Bergner, N., Chang, R.Y.W., Creamean, J.M., Kuang, C., Martin, R.V. and Ranjithkumar, A., 2023. Arctic warming by abundant fine sea salt aerosols from blowing snow. Nature Geoscience, 16(9), pp.768-774. https://www.nature.com/articles/s41561-023-01254-8
Schulz, K., Koenig, Z., Muilwijk, M., Bauch, D., Hoppe, C.J., Droste, E., Hoppmann, M., Chamberlain, E.J., Laukert, G., Stanton, T. and Zurita, A.Q., 2023. The Eurasian Arctic Ocean along the MOSAiC drift (2019-2020): An interdisciplinary perspective on properties and processes. https://eartharxiv.org/repository/view/5902/
Macfarlane, A.R., Löwe, H., Gimenes, L., Wagner, D.N., Dadic, R., Ottersberg, R., Hämmerle, S. and Schneebeli, M., 2023. Temporospatial variability of snow's thermal conductivity on Arctic sea ice. The Cryosphere, 17(12), pp.5417-5434. https://tc.copernicus.org/articles/17/5417/2023/
Citation: https://doi.org/10.5194/egusphere-2024-1240-RC2
Data sets
Sea ice bulk density during the MOSAiC expedtion Yi Zhou https://zenodo.org/doi/10.5281/zenodo.11055727
IS2 modal freeboard during the MOSAiC expedtion Yi Zhou https://zenodo.org/doi/10.5281/zenodo.11055727
Snow depth and sea ice thickness data derived from SIMBA buoy measurements Ruibo Lei https://doi.org/10.1594/PANGAEA.938244
Snow depth and sea ice thickness data derived from SIMB buoy measurements Donald K. Perovich https://doi.org/10.18739/A20Z70Z01
Model code and software
IS2 modal freeboard extraction and sea ice bulk density retrieval Yi Zhou https://zenodo.org/doi/10.5281/zenodo.11055727
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
600 | 160 | 54 | 814 | 64 | 26 | 22 |
- HTML: 600
- PDF: 160
- XML: 54
- Total: 814
- Supplement: 64
- BibTeX: 26
- EndNote: 22
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1