the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 14 Jan 2025
Sea Ice Concentration Estimates from ICESat-2 Linear Ice Fraction. Part 2: Gridded Data Comparison and Bias Estimation
Abstract. Sea ice coverage is a key indicator of changes in the global climate. Estimates of sea ice area and extent are primarily derived from satellite measurements of surface microwave emissions, from which local sea ice concentration (SIC) is derived. Passive microwave (PM) satellite sensors remain the sole global product for understanding SIC variability, but may be sensitive to consistent biases. In part I we explored these in a multi-sensor intercomparison of optical, passive microwave, and lidar data, showing that a new, independent SIC product, the linear ice fraction (LIF), derived from ICESat-2 (IS2) laser altimetry, could be used to quantify and understand PM SIC biases. Here in part II, we develop and assess the reliability of larger-scale estimates of SIC from IS2 LIF. We develop an LIF emulator that samples optical imagery using the distribution of possible orientation angles for IS2 to understand the limitations of this one-dimensional product. We find that the error qualities of the LIF product are improved when combining multiple IS2 tracks, and discuss intrinsic but correctable biases that emerge in the combination of multiple IS2 measurements. We use these to develop a monthly LIF product, covering up to 54 % of the Arctic sea ice cover, with has similar-or-better error qualities compared to PM data. We then discuss pathways to enhancing PM-SIC data with IS2 LIF in the future.
- Preprint
(1690 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2024-3864', Anonymous Referee #1, 10 Feb 2025
This paper presents a method to estimate gridded sea ice concentrations (SIC) from a novel Linear Ice Fraction (LIF) product derived using an emulator of ICESat-2 tracks. The paper is a companion to part 1 which compares the LIF product to passive microwave SIC products. Here in part 2, the authors examine and quantify the potential sources of bias when using a high-resolution linear product with gridded, low-resolution passive microwave SIC observations. The results look promising and the paper highlights the potential for these linear observations to enhance current passive microwave SIC products. I think this paper needs some revision before it is ready for publication, specifically to improve clarity of Section 2.1. There are also quite a few subtleties about passive microwave sea ice data that are not quite captured in the current paper. My comments below describe these issues in more detail.
Comments
L5: ATL07 relies on a passive microwave SIC product and a SIC threshold, thus, LIF is not really an independent SIC estimate.
L14: Although they both measure the amount of area covered by ice, sea ice concentration (expressed as a percentage) and sea ice fraction (expressed from 0 to 1) are different. The two are used interchangeably and imprecisely throughout the paper. Here (L14), I would change “fraction” to “percentage” since you are defining concentration but consider if the language throughout the paper should be revised to use only sea ice fraction to simplify the discussion since this is how you defined your LIF product.
L55: This applies in other parts of the paper as well, but the SIC products used in this paper need to be defined (e.g., acronyms spelled out) and cited appropriately. I see this is included in part 1, but it needs to be in part 2 as well.
L63: Similar comment as above, there is risk of passive microwave biases in your product because the data are not independent.
L73: I disagree here. 1) A PM satellite covers the entire Arctic approximately twice per day. 2) PM observations are influenced by cloud cover, especially liquid clouds or clouds with heavy precipitation. I think what you are meaning to point out here is that PM SIC observations can be obtained daily even in cloudy conditions.
L76-77: This is also true for most PM SIC products as well. They are usually produced from drop-in-the-bucket daily- or twice-daily- averaged brightness temperatures.
L80: Is ATL07 version 7 available somewhere or is there a reference to this information? I see release 6 is currently available at NSIDC.
Fig 2c: Related to my comment above about concentration versus fraction, the unit for ice fraction is not “%” as labeled on the y-axis. You also have legend labels for LIF and SIC for data that is plotted only as a 0-1 range.
Figure 3 and L136-176: I understand what you are accomplishing with the image and sampling uncertainty estimation and think it is a reasonable approach; however, there are many problems with the labeling on Figure 3 that make it extremely hard to follow this section of the text. I’m having a hard time parsing which parts of the figure the text is referring to because of the errors on the figure, in the caption, and in the text. This section and figure need a careful edit before they make any sense.
L202: “4 or more IS2 crossings” – Does this mean overflights or beams? L174 mentions 2 or 3 IS2 overflights. Does this mesh with L202? Please clarify.
L228-229: I don’t see any discussion of why the LIF and NASA Team are both bimodal. Why might that be? Or conversely, why are the other products not bimodal?
Table 1: The months noted for “summer” and “winter” periods do not match with the text (L214 and 244). Which is correct?
L255: As above, LIF is not quite independent.
Technical/typographical comments
Operation IceBridge appears in a variety of different permutations throughout the paper. Please use consistent capitalization and/or just use OIB after defining it at the first use.
L11: “with has similar” to “which has similar”
L20: “thus hence”, delete one.
L26: typo – concentration
L52: Which version of ATL07 was used?
L70: Where is Supporting Figure S1?
L81: XX% - fix
L97: Should the in text figure reference here be to Fig. 2?
L137: “(c, black line)” which figure?
L155: Is this referring to panel a in Figure 2?
Figure 4 and L195: These SIC products need to be defined in this paper.
Figure 5: In the caption and legend – should it be non-summer instead of winter?
L225-226: Add an in text reference to Table 1.
L234: Use “IS2” to be consistent throughout the paper.
L251: Define SSMI/S.
Citation: https://doi.org/10.5194/egusphere-2024-3864-RC1 -
RC2: 'Comment on egusphere-2024-3864', Anonymous Referee #2, 14 Feb 2025
This paper introduces a new gridded dataset for linear lead fraction (LIF) derived from NASA’s ICESat-2 laser altimeter. The study is a continuation of an earlier submission to The Cryosphere, now split into two parts based on prior feedback. The first part focused on OIB imagery classifications and the introduction of the LIF variable, along with initial comparisons with visual imagery. This second part presents a novel satellite emulator to analyze the impact of altimetry crossings on LIF estimates, followed by monthly gridded Arctic LIF estimates and their comparison to other sea ice concentration (SIC) products. I have reviewed both parts.
From a scientific perspective, the main difference between this study and the previous submission is the use of more realistic crossing angles (azimuths) in the crossings/emulation analysis, as well as a more detailed discussion of resulting biases. The methodology for generating monthly gridded LIF estimates and comparing them with SIC products remains largely unchanged, with only minor differences in the results and how they are presented. A few additional caveats have been introduced.
Overall, I find the study to be scientifically valuable, particularly regarding the emulation methodology and the way crossing angles and profiling differences are analyzed. However, I still have quite a few concerns that I believe should be addressed before publication.
- Figure 3 and Its Interpretation
This figure, which is central to the study, was difficult to follow. The panel labels were incorrect, and axis labels were missing, making interpretation challenging. Additionally, it was not always clear whether the panels referred to a single image (Fig. 2) or all images. Would it be possible to first introduce these concepts using a single-image example, before presenting the full dataset? I read the description multiple times and still found parts of the explanation unclear. Overall, this section felt a bit rushed.
- Validation of LIF at Low Concentrations
If LIF is to be presented as a product rather than a proof-of-concept, there needs to be a better validation at lower SIC levels. In Part 1 (and the original paper), LIF was only validated on high concentration winter scenes, but here it is applied to multiple months, including SIC down to 15%. The ability of ICESat-2 to profile large regions of open water or low-concentration ice reliably remains uncertain, yet this dataset includes these regions. This discrepancy should be addressed, as it is essential if LIF is intended to be a comprehensive gridded dataset. Ideally, this enhanced validation should have been integrated into Part 1.
- Definitional Differences Between SIC and LIF
In the response to the original submission, the authors stated that:
“The interpretation of lead type from IS-2 is not the goal of this manuscript, but it is certainly a heavily focused-on problem in the IS-2 community.”
This seems like a weak argument, given that lead classification is crucial for LIF estimation and the construction of a gridded LIF dataset. The paper acknowledges that the agreement between LIF and passive microwave SIC is much better when dark leads are excluded. To what extent, then, is this discrepancy simply an issue of definition? Perhaps it would be beneficial to focus on high concentration regions and analyze only specular leads to explore this further.
- Azimuthal Crossings and Beam Configuration
The introduction of more realistic azimuth-dependent crossings is a good improvement. However, in reality, the ICESat-2 beams are spaced ~3 km apart, and weak beams profile the same ice as strong beams with a 90 m separation across track but 2.5 km along-track spacing (ICESat-2 Specs). On L74, the text states the beams are 25 km wide, which seems like a typo? Additionally, the analysis does not appear to account for the fact that three beams have the same orientation—should these crossings be considered differently? On L202, the text mentions requiring 4 crossings, but it is unclear whether this refers to 4 out of the 6 beams or something else.
- Using ICESat-2 for Internal Cross-Validation
Given that ICESat-2’s multiple beams profile the same ice, this could be used to cross-validate the results internally. Specifically, the authors could compare weak beam results to strong beams and assess whether there is a consistent bias between the two. This would help strengthen the validity of the dataset.
- Use of Passive Microwave SIC at High Concentrations
Perhaps passive microwave SIC should not be used at the highest SIC levels? The analysis could have provided more insight into at which SIC ranges the discrepancies appear and why. As mentioned earlier, using ICESat-2 in low concentration regimes may not be the best approach and I would probably still trust passive microwave more. I do agree the high concentration results are compelling.
Code/Data Comments
- Providing the MATLAB Code is great, but how usable and well-documented is it? If this is meant to be a software package, ideally at least one reviewer should test it.
- I could not find the LIF data at the provided Zenodo link—is this available?
Specific Comments
- L21 – Is that statement fully accurate?
- L22 – Reanalyses also provide an alternative method, so perhaps clarify this point.
- L31 – Important to emphasize that previous studies (e.g., Kern et al., 2020) already identified these biases.
- L34 – Magruder’s study states the footprint as 11 m.
- L65 – Should also mention dark leads.
- L74 – Possible typo? ICESat-2 has a 6.6 km along-track swath in total.
- L82 – Now that the papers are split, Part 1 should better address SIC biases, as there is more space. Could the 2.4% bias be improved using different classification subsets or better aligning IS-2 with imagery?
- Figure 2 – Could you include the latitude of this example, along with the azimuths being applied?
- L114 – Minor point: the term “reference” in “synthetic reference ground track” seems unnecessary, as it is just a line.
- L136 – “The progression from LIF0 to Bi is path-dependent”—this phrasing is quite difficult to understand.
- L151 – What exactly does P represent here?
- Figure 3 – The panel labels are incorrect, making this figure confusing. Please revise.
- L177 – Should just be called Arctic, if that’s all that is shown.
- L202 – In Section 2, you mention requiring 8 total beam intersections, but here it states 4—please clarify.
- L218 – Would be good to explicitly reference Part 1 here, as well as previous studies (Kern et al.), which found similar PM SIC biases.
- L220 – The impact of dark lead removal seems more significant in summer than in winter. The phrasing here should be adjusted.
- L221 – Again, this highlights why Part 1 needs a more detailed analysis of dark leads across different seasons.
Citation: https://doi.org/10.5194/egusphere-2024-3864-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 116 | 26 | 5 | 147 | 3 | 8 |
- HTML: 116
- PDF: 26
- XML: 5
- Total: 147
- BibTeX: 3
- EndNote: 8
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1