First, I apologize to the authors and editors for being so tardy in my review.
Review, Howell et al, Generating large-scale sea ice motion from Sentinel-1 and the RADARSAT Constellation Mission using the Environment and Climate Change Canada automated sea ice tracking system, Cryosphere tc-2021-223
This is a good summary of the SAR-derived ice motion product using two different C-band SAR sensors on 5 different platforms. I reviewed the revision made after comments provided by Reviewers 1 and 2 but I did not read the reviewers comments or response until after I read the revision. I certainly agree with many of these reviewers’ comments and revisions, specifically the inclusion of the 6.25 km product and the buoy comparisons. I also appreciate the revisions made by the authors, to emphasize the primary purpose of these two products which is for marine stakeholders. I believe it is very important to have papers such as this on describing approaches and algorithms for new products, especially those derived from satellite, particularly in journals where the papers will be read by the sea ice community in this case. I assume the editors of TC believe this paper and such related papers on products/algorithms are appropriate, once approved for submission of course through the review process.
I follow now with my review. In general, I found the paper well-written and worthy of publication after my relative minor suggestions are evaluated and commented on. I appreciate the scale of this effort to provide SAR ice motion products. I believe the paper could use some more discussion on some figures which I will mention. I am a little surprised that the buoy comparisons seem quite large especially on the basis of the selected grid size used for 6.25 km product and suggest that more description be included on the multiple sources of the errors (Section 5). The offsets are actually pretty large compared to previous publications using coarser resolution SAR imagery. Also, in comparing the SAR 7-day product with PM results (Section 6), it's understandable that the SAR product is better. I think this section could benefit from more discussion especially on why stakeholders might be more interested in using the SAR product vs. the PM, aside from the improved accuracy.
Detailed comments
1. Lines 61-62. Please rephrase this sentence to include the thought that the SAR data limitations prior to S1.
2. Line 65. Please add a sentence that mentions the rationale for near-real time SIM processing and derived products. The processing scenario would be more seamless without the need for near-real time processing. Helps explain need for Figure 7 discussion too.
3. Table 1. I suggest adding the PM sea ice products to this table or making an equivalent one, including pixel size and grid size, temporal spacing (not image count).
4. Sampling. From Figure 3, it seems pretty clear that the RCM mapping strategy in general is different from S1 for the Arctic. RCM has a more spatially/temporally distributed consistent sampling approach for the entire Arctic, while S1 has more intensive coverage in certain areas. Each mission is set up to meet its science/stakeholder requirements. Given the overlapping coverage shown in Figure 3C, which shows a major central portion of the Arctic with over 9 images per week, I am surprised at the sparse results shown later in Figure 10, since 5 sensors are being utilized. The enlargements shown in a-d suggest more extensive results than the large figure but I don’t find these that useful. I wonder if an additional large figure or two showing sequential results, say from March 15-17, then March 18-20, would be beneficial to show. Such additions might highlight some of the issues with uneven spatial sampling that both acquisition plans may have, perhaps, and would also indicate how a regional time series could be developed for further study.
Continuing on about sampling, as Reviewer 1 mentioned, and has mentioned in Lines 193-194 and paragraph starting with line 195, I too would be interested in hearing whether they tried mixing RCM and S1image pairs to derive a motion product and what the results were.
Also, regarding use of stacks, perhaps I missed this but for high coverage areas, say within the central Arctic, how do they handle more than one image pair within a 3-day period- are multiple pairs averaged together to make a single 3-day product?
5. Section 4
I assume ‘resolvable’ means grid cells with derived vectors that passed quality check. Please clarify.
The availability of 5 sensors collecting Arctic sea ice imagery has enabled the most complete picture of Pan-Arctic SAR-derived ice motion. Please clarify sentence starting on Line 214.
Its long been desired.
Neither Table 2 or the text really addresses the point about the most complete pan-Arctic SAR-derived maps, at least in terms of numbers.
6. Section 5
In the first paragraph, the authors should include the reference Lindsay and Stern, 2003 on RGPS offsets, where a mean displacement of 323 m was found for Radarsat-1
Paragraph starting line 309. As mentioned errors of 2.78 km seem quite large to me, basically since its about half of 6.25 km grid cell for winter, also in comparison to the above Lindsay/Stern results. I strongly suggest the authors add a paragraph that describes in more detail the sources of error- buoy, interpolated time to SAR image, spacecraft and SAR image location accuracy, identifying matchups between image pairs and so forth. This would at least enable readers and authors to point to major error sources which may be improved upon.
Figure 15A- the date on figure and in caption are mislabelled. For Figure 15B I found it interesting that a couple of locations within the CAA showed higher velocities (yellow) than surrounding areas. This figure warrants additional discussion.
7. Section 6
I too am a little puzzled by comparing SAR with PM, including down-sampling the SAR results.
But the PM products do at least have similar spatial scales to the weekly product. I am uncertain as to the meaning of the colors in Fig 16 and 17. Both appear to be heat maps. Both figures warrant additional discussion beyond lines 371-2.
Details:
Line 33- Torres is misspelled
Line 41- Kwok 2015 paper is not listed in the references
Lines 122-123 not a sentence
Line 122. How much computer time does it take to analyze 160 images per day?
Lines 153-155 shouldn’t the overlap be greater than 30%, for retention?
Line 156- you might mention that 32,000 km/2 is about 178 km2 or a little less than 1/2 a frame.
Line 191 – spatially complete
Line 198-199 – insert: …less representative results, often due to higher wind speeds. |