SMOS-derived Antarctic thin sea-ice thickness: data description and validation in the Weddell Sea

Kaleschke, Lars; Tian-Kunze, Xiangshan; Hendricks, Stefan; Ricker, Robert

doi:https://doi.org/10.5194/essd-2023-326

Preprints

https://doi.org/10.5194/essd-2023-326

Preprints

11 Sep 2023

| 11 Sep 2023

Status: a revised version of this preprint is currently under review for the journal ESSD.

SMOS-derived Antarctic thin sea-ice thickness: data description and validation in the Weddell Sea

Lars Kaleschke, Xiangshan Tian-Kunze, Stefan Hendricks, and Robert Ricker

Abstract. Accurate satellite measurements of the thickness of Antarctic sea ice are urgently needed but pose a particular challenge. The Antarctic data presented here were produced using a method to derive the sea-ice thickness from 1.4 GHz brightness temperatures previously developed for the Arctic, with only modified auxiliary data. The ability to detect thin sea- ice thicknesses using this method is limited to cold conditions, meaning it is only possible during the freezing period, typically March to October. The SMOS level 3 sea-ice thickness product contains estimates of the sea-ice thickness and its uncertainty up to a thickness of about 1 m. The sea-ice thickness is provided as daily average on a polar stereographic projection grid with a sample resolution of 12.5 km, while the SMOS brightness temperature data used has a footprint size of about 35–40 km in diameter. Data from SMOS have been available since 2010, and the mission’s operation has been extended to continue until at least the end of 2025.

Here we compare two versions of the SMOS Antarctic sea-ice thickness product which are based on different level 1 input data (v32 based on SMOS L1C v620, and v33 based on SMOS L1C 724). A validation is performed to have a first baseline reference for future improvements of the retrieval algorithm and synergies with other sensors.

Sea-ice thickness measurements to validate the SMOS product are particularly rare in Antarctica, especially during the winter season and for the valid range of thicknesses. From the available validation measurements, we selected datasets from the Weddell Sea that have varying degrees of representativeness: Helicopter-based EM Bird (HEM), Surface and Under-Ice Trawl (SUIT), and stationary Upward-Looking Sonars (ULS). While the helicopter can measure hundreds of kilometers, the SUIT’s use is limited to distances of a few kilometers and thus only captures a small fraction of an SMOS footprint. Compared to SMOS, the ULS are point measurements and multi-year time series are necessary to enable a statistically representative comparison. Only 4 of the ULS moorings have a temporal overlap with SMOS in the year 2010.

Based on selected averaged HEM flights and monthly ULS climatologies we find a small mean difference (bias) of less than 10 cm and a root-mean-square deviation of about 20 cm with a correlation coefficient R>0.9 for the valid sea-ice thickness range between zero and about one meter. The SMOS sea-ice thickness showed an underestimate of about 40 cm with respect to the less representative SUIT validation data in the marginal ice zone. Compared with sea-ice thickness outside the valid range we find that SMOS strongly underestimates the real values which underlines the need for combination with other sensors such as altimeters.

In summary, the overall validity of the SMOS sea-ice thickness for thin sea-ice up to a thickness of about 1 m has been demonstrated through validation with multiple datasets. To ensure the quality of the SMOS product, an independent regional sea-ice extent index was used for control. We found that the new version v3.3 is slightly improved in terms of completeness, indicating less missing data. However, it is worth noting that the general characteristics of both datasets are very similar, also with the same limitations. Archived data are available on the PANGAEA repository at https://doi.org/10.1594/PANGAEA.934732, (Tian-Kunze and Kaleschke, 2021) and operationally via https://spaces.awi.de/display/CS2SMOS.

Received: 11 Aug 2023 – Discussion started: 11 Sep 2023

Download & links

Lars Kaleschke, Xiangshan Tian-Kunze, Stefan Hendricks, and Robert Ricker

Status: final response (author comments only)

RC1: 'Comment on essd-2023-326', Anonymous Referee #1, 22 Oct 2023

Review report of “SMOS-derived Antarctic thin sea-ice thickness: data description and validation in the Weddell Sea” submitted to Earth System Science Data.
The goal of this paper is to construct a data record of Antarctic thin ice thickness from SMOS measurements and the authors try to validate the data record initially. This is an interesting goal and I sincerely thank for the effort of the authors to provide an important ice thickness data set over the Antarctic. However, I have critically wondered that the results of this paper do not fit the scope of the journal which is “the reuse of high-quality data of benefit to Earth system science”. The algorithm described in this paper for Antarctic thin ice is just from the existing algorithm applicable to Arctic sea ice. As the authors discussed, the characteristic differences between Arctic and Antarctic sea ice are significant. However, I can’t find any effort to consider or analyze the different characteristics between them that can affect the retrieved SMOS thin ice thickness at all throughout the paper. In the case that it will be done in the paper, I can’t accept the validation results which are not sufficient for the potential user to use this product having convincing. For instance, thin ice thickness is very important to be used in data assimilation systems in sea ice models. In order to use the ice thickness for this purpose, a much more relevant error analysis should be preceded to provide an observation error covariance matrix. The data produced here may be biased to the real state vectors over different regions. At least, error propagation analysis should be accompanied, however, there is no effort on this in the paper. I totally agree that few observation data compared to Arctic areas. In addition, several assumptions were used to estimate thin ice thickness, however, there is no evidence or error analysis to prove that the assumptions are valid. For instance, the algorithm assumes 100% ice condition which rarely exits over Antarctic thin ice distributions even in the middle of austral wintertime, however, the authors just discussed that less than 100% ice condition doesn’t matter because the product shows well growth of the seasonal ice. I can’t agree with this. I feel that the authors would compare SMOS sea ice extent to other data in order to show that the SMOS ice thickness data over SMOS is plausible. However, I don’t understand why this comparison can provide information on the general quality and completeness of the SMOS sea ice thickness product. As known, sea ice thickness is vertical information while sea ice extent gives horizontal spreads of sea ice. In addition, specific comments are listed below.

(1) Page 2, Line 39: ‘record’ can be ‘records’. There are numerous subject-verb agreement problems. I don’t want to review all the same problems in this paper. Please check carefully throughout the paper.
(2) Page 2, Line 41: add ‘last’ between ‘since’ and ‘five’ or specify the years you referred to.
(3) Page 2, Line 43: what is the ‘can be also be’? If it is a typo, correct this.
(4) Page 4, Line 106: does the ‘full polarization’ mean full Stokes’ component including third and fourth Stokes’ polarizations? If not, change it into ‘first two polarizations’.
(5) Page 4, Line 109: This sentence can be modified as “SMOS measures brightness temperatures with a spatial resolution of about 35 km at nadir on a daily basis in the polar regions”.
(6) Page 4, Line 116: perhaps, oN and oS can be switched.
(7) Page 4, Lines 119-120: Define both ‘JRA’ and ‘GMT’ when they first appear.
(8) Page 5, Lines 126-127: Again, please define ‘GECCO2’ and ‘MITgcm’
(9) Page 5, Line 128: ‘are’ can be ‘were’. Please use ‘past sentence’ when you did something for this work throughout the paper. There are tons of these kinds of issues in the paper. I don’t want to spend time to peak all issues.
(10) Page 5, Line 132: Define ‘HEM’ and ‘SUIT’
(11) Page 6, Line 141: ‘ULS’ was already defined in the introduction.
(12) Page 6, Lines 141-142: ‘have’ can be changed into ‘had’.
(13) Page 7, Line 164: how to prescribe snow depth, snow density, and water density in order to utilize the hydrostatic equilibrium equation.
(14) Page 7, Line 170: the authors mentioned that “in addition, ice thickness variation within the SMOS grid are considered”. why did you consider this for what?
(15) Page 7-8, Line 176-177: This sentence is imperfect. Please rewrite it. This sentence gives weird information that “inhomogeneities are much smaller than the wavelength of 21 cm.”
(16) Page 9, Line 193: Define ‘ISEA 4H9’.
(17) Page 14, Line 233: the authors should discuss the difference between v3.2 and v3.3 in detail in the paper.
(18) Page 15, Lines 261-262: ULS sees different ice floes as sea ice flows by ocean current. In order to make a comparison between ULS and SMOS ice thicknesses, the authors should track the ice floes even for monthly comparison, which process was done in many papers for Arctic sea ice studies. Perhaps the Lagrangian tracking method would be useful.
(19) Regarding the sea surface salinity (SSS) dataset: the SSS climatology used in this study was based on a model simulation over the years 1952-2001 which is far from the SMOS period. Is there any other SSS climatology covering the SMOS observation period? If so, recommend replacing the SSS dataset.
(20) Regarding the ULS dataset: the authors converted the ice draft into total thickness using an empirical linear fit equation, which is a very important relationship in the validation section. It is weird that there is no discussion of this fit equation in the paper.
(21) Table 1: why several ULS data were neglected in this paper? I don

Citation: https://doi.org/10.5194/essd-2023-326-RC1
RC2: 'Comment on essd-2023-326', Anonymous Referee #2, 09 Nov 2023

This article, titled “SMOS-derived Antarctic thin sea-ice thickness: data description and validation in the Weddell Sea”, introduces the sea ice thickness (SIT) product based on L-band radiometer of SMOS for the Weddell Sea. The results presented include promising results and sound validations with various independent observations, and the capability of SMOS for SIT retrieval in Southern Oceans is demonstrated. The scope of the work falls right into those of the ESSD journal. And furthermore, the data link works totally fine, with the data files formatted according to the regulations widely used by the climate science community. However, I do have the following concerns and comments which need further clarification and possibly revisions.

First, I find it worthy a second consideration for the paper’s focus on Weddell Sea. The dataset provided covers the whole Antarctic realm. And arguably, Weddell Sea is possibly not the best place for thin ice retrieval, given the dominant thick, perennial ice in the region. I understand that most validations are carried out in Weddell Sea, but what limits the validation in other part of the Southern Ocean? In the Indian/Pacific sector, there are also many available data from ASPECT. And many MIZ campaigns provide SIT over the thin ice, which could be invaluable source of validation especially for SMOS’s retrievable range.

Second, I consider that the retrieval result needs more in-depth analysis. For example, with some inspection of the daily SIT fields, one can easily discover large temporal (i.e., day-to-day) variability of SIT across the whole field. The (unrealistic) fluctuation of the retrieved SIT is evidently of atmospheric origin, possibly due to passing cyclones and the ensuing thermodynamic signals in the sea ice and snow. The effect of synoptic systems is more pronounced on thick ice (SIT>1m), but potentially present on thin ice as well. This observed phenomenon is directly linked to the thermodynamic quasi-equilibrium assumption (or the lack of) for the retrieval algorithm, as well as the timing of SMOS’s passes. I think an immediately available analysis is to explore what is the optimal retrieval interval by SMOS, which is definitely above (coarser than) 1-day. Also, collaterally, how the monthly mean SIT should be computed: mean of SIT or SIT based on mean TB?

Third, a related issue to my second comment is, a better quantification of uncertainty and a clear indication of the usability of the SIT product are needed. For SIT over 1m, the SIT product contains no information. Besides, very large uncertainty is present for SIT over 50cm. Whether the cyclones affects the uncertainty (see above) and whether this uncertainty is accounted for is in unknown. Furthermore, with a better quantified uncertainty, a simple field of confidence of SIT can be provided, for example: the relative uncertainty lower than 30% (high confidence), lower than 60% (low confidence), or higher than 60% (very low confidence). This would facilitate downstream users who are not experts of the data retrieval. The specific threshold values and the terms could be chosen more carefully, but I do suggest adding such information to avoid potential misuses of the data.

Fourth, an outstanding issue of the retrieval method is that it is originally developed for the Arctic. With potential difference in snowfall rate and snow stratigraphy between the two poles, some assumptions may not hold and need justification. At least, the uncertainty caused by them should be estimated. They include: 10% snow depth of sea ice depth, salinity in snow, as well as the potential of snow-ice.

Besides, I find that Section 6 is not an integral part of the paper which focuses on SIT. I suggest moving it to a more proper place, for example, as an appendix.

There also exist some minor issues, with examples below:
l122: “days temperature” should be “days’ temperature”
l293: missing “.”

Citation: https://doi.org/10.5194/essd-2023-326-RC2
AC1: 'Comment on essd-2023-326', Lars Kaleschke, 15 Jan 2024

We thank the two reviewers and the editor for their valuable comments and suggestions.

Enclosed you will find our answers, introduced by a general comment.

Citation: https://doi.org/10.5194/essd-2023-326-AC1

Lars Kaleschke, Xiangshan Tian-Kunze, Stefan Hendricks, and Robert Ricker

Data sets

SMOS-derived sea ice thickness in the Antarctic from 2010 to 2020 Xiangshan Tian-Kunze and Lars Kaleschke https://doi.pangaea.de/10.1594/PANGAEA.934732

Lars Kaleschke, Xiangshan Tian-Kunze, Stefan Hendricks, and Robert Ricker

Viewed

Total article views: 967 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
738	199	30	967	34	34

HTML: 738
PDF: 199
XML: 30
Total: 967
BibTeX: 34
EndNote: 34

Views and downloads (calculated since 11 Sep 2023)

Month	HTML	PDF	XML	Total
Sep 2023	233	55	8	296
Oct 2023	120	30	3	153
Nov 2023	60	11	1	72
Dec 2023	93	23	3	119
Jan 2024	83	29	2	114
Feb 2024	47	21	3	71
Mar 2024	66	20	5	91
Apr 2024	36	10	5	51

Cumulative views and downloads (calculated since 11 Sep 2023)

Month	HTML	PDF	XML	Total
Sep 2023	233	55	8	296
Oct 2023	120	30	3	153
Nov 2023	60	11	1	72
Dec 2023	93	23	3	119
Jan 2024	83	29	2	114
Feb 2024	47	21	3	71
Mar 2024	66	20	5	91
Apr 2024	36	10	5	51

Viewed (geographical distribution)

Total article views: 962 (including HTML, PDF, and XML) Thereof 962 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 27 Apr 2024

Short summary

We describe a sea ice thickness dataset based on SMOS satellite measurements, initially designed for the Arctic but adapted for Antarctica. We validated it using limited Antarctic measurements. Our findings show promising results, with a small difference in thickness estimation and a strong correlation with validation data within the valid thickness range. However, improvements and synergies with other sensors are needed, especially for sea ice thicker than one meter.


Total:	0
HTML:	0
PDF:	0
XML:	0