Semi-automated tracking of iceberg B43 using  Sentinel-1 SAR images via Google Earth Engine

Koo, YoungHyun; Xie, Hongjie; Ackley, Stephen F.; Mestas-Nuñez, Alberto M.; Macdonald, Grant J.; Hyun, Chang-Uk

doi:https://doi.org/10.5194/tc-15-4727-2021

Articles | Volume 15, issue 10

https://doi.org/10.5194/tc-15-4727-2021

© Author(s) 2021. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/tc-15-4727-2021

© Author(s) 2021. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 15, issue 10

Research article

|

07 Oct 2021

Research article |

| 07 Oct 2021

Semi-automated tracking of iceberg B43 using Sentinel-1 SAR images via Google Earth Engine

YoungHyun Koo, Hongjie Xie, Stephen F. Ackley, Alberto M. Mestas-Nuñez, Grant J. Macdonald, and Chang-Uk Hyun

Download

Final revised paper (published on 07 Oct 2021)
Preprint (discussion started on 17 May 2021)

Interactive discussion

Status: closed

CC1:
'Comment on tc-2021-131', David G. Long, 27 May 2021

This paper describes an approach to detect, locate, and measure Antarctic tabular icebergs. With its finer spatial resolution, the data can supplement optical tracking by the U.S. National Ice Center and scatterometer tracking by the Scatterometer Climate Record Pathfinder (www.scp.byu.edu). The latter maintains an extensive database of daily iceberg positions going back to the late 1970's (www.scp.byu.edu/iceberg; Budge and Long, 2017; Stuart and Long, 2011; Long et al. 2002).
The authors should be sure to note in the second sentence of the introduction that the paper exclusively considers (large) Antarctic tabular icebergs to avoid confusion with smaller icebergs that occur in both polar regions. While the paper mainly consider SAR observations, radar scatterometers have also been extensively used for detecting and tracking icebergs. Though the scatterometers observations have much lower resolution and less precision in measuring iceberg area, the wide swath of the scatterometers facilitate daily position observations (Budge and Long, 2011; Stuart and Long, 2011). This fine temporal resolution is helpful during period of rapid iceberg motion, for example when icebergs such as A68 move from the Weddell Sea into the South Atlantic. The rapid motion and spinning of this icebeg and others such as B10A can be revealed in animations of the daily scatterometer images, e.g., www.scp.byu.edu/iceberg/A68tracking.html The authors are encouraged to include references to scatterometry as another tool for iceberg tracking.
It has been previously noted (Budge and Long, 2017; Stuart and Long, 2011), that the disintegration of icebergs is slowed when the iceberg is encapulated with sea ice, suggesting that the encapulating sea ice may protects the iceberg from exposure to wave stress. This scatterometer-derived resul supports the authors' postulation.

References
J.S. Budge and D.G. Long, "A Comprehensive Database for Antarctic Iceberg Tracking Using Scatterometer Data," IEEE Journal of Selected Topics in Applied Earth Observations, Vol. 11, No. 2, pp. 434-442, doi:10.1109/JSTARS.2017.2784186, 2017.
K.M. Stuart and D.G. Long, "Iceberg Size and Orientation Estimation using SeaWinds", Cold Regions Science and Technology, doi:10.1016/j.coldregions.2011.07.006, Vol. 69, pp. 39-51, 2011.
K.M. Stuart and D.G. Long, "Tracking large tabular icebergs using the SeaWinds Ku-band microwave scatterometer", Deep-Sea Research Part II, doi:10.1016/j.dsr2.2010.11.004, Vol. 58, pp. 1285-1300, 2011.
D.G. Long, J. Ballantyne, and C. Bertoia, "Is the Number of Icebergs Really Increasing?" EOS, Transactions of the American Geophysical Union, Vol. 83, No. 42, pp 469 & 474, 15 Oct. 2002.

Citation: https://doi.org/10.5194/tc-2021-131-CC1
- RC1:
  'CC1 again as RC', David G. Long, 03 Jun 2021
  
  This paper describes an approach to detect, locate, and measure Antarctic tabular icebergs. With its finer spatial resolution, the data can supplement optical tracking by the U.S. National Ice Center and scatterometer tracking by the Scatterometer Climate Record Pathfinder (www.scp.byu.edu). The latter maintains an extensive database of daily iceberg positions going back to the late 1970's (www.scp.byu.edu/iceberg; Budge and Long, 2017; Stuart and Long, 2011; Long et al. 2002).
  
  The authors should be sure to note in the second sentence of the introduction that the paper exclusively considers (large) Antarctic tabular icebergs to avoid confusion with smaller icebergs that occur in both polar regions. While the paper mainly consider SAR observations, radar scatterometers have also been extensively used for detecting and tracking icebergs. Though the scatterometers observations have much lower resolution and less precision in measuring iceberg area, the wide swath of the scatterometers facilitate daily position observations (Budge and Long, 2011; Stuart and Long, 2011). This fine temporal resolution is helpful during period of rapid iceberg motion, for example when icebergs such as A68 move from the Weddell Sea into the South Atlantic. The rapid motion and spinning of this icebeg and others such as B10A can be revealed in animations of the daily scatterometer images, e.g., www.scp.byu.edu/iceberg/A68tracking.html The authors are encouraged to include references to scatterometry as another tool for iceberg tracking.
  
  It has been previously noted (Budge and Long, 2017; Stuart and Long, 2011), that the disintegration of icebergs is slowed when the iceberg is encapulated with sea ice, suggesting that the encapulating sea ice may protects the iceberg from exposure to wave stress. This scatterometer-derived resul supports the authors' postulation.
  
  References
  
  J.S. Budge and D.G. Long, "A Comprehensive Database for Antarctic Iceberg Tracking Using Scatterometer Data," IEEE Journal of Selected Topics in Applied Earth Observations, Vol. 11, No. 2, pp. 434-442, doi:10.1109/JSTARS.2017.2784186, 2017.
  
  K.M. Stuart and D.G. Long, "Iceberg Size and Orientation Estimation using SeaWinds", Cold Regions Science and Technology, doi:10.1016/j.coldregions.2011.07.006, Vol. 69, pp. 39-51, 2011.
  
  K.M. Stuart and D.G. Long, "Tracking large tabular icebergs using the SeaWinds Ku-band microwave scatterometer", Deep-Sea Research Part II, doi:10.1016/j.dsr2.2010.11.004, Vol. 58, pp. 1285-1300, 2011.
  
  D.G. Long, J. Ballantyne, and C. Bertoia, "Is the Number of Icebergs Really Increasing?" EOS, Transactions of the American Geophysical Union, Vol. 83, No. 42, pp 469 & 474, 15 Oct. 2002.
  
  Citation: https://doi.org/10.5194/tc-2021-131-RC1
  - AC1: 'Reply on RC1', YoungHyun Koo, 03 Jun 2021
    
    We appreciate your great comments and suggestions. We agree that scatterometer is another good tool to track large icebergs. We will add more descriptions and references about scatterometers in the second paragraph of introduction as following:
    
    Radar remote sensing, such as scatterometer or synthetic aperture radar (SAR), is an efficient tool for monitoring both movements and area changes of icebergs. While multispectral images can be useful for observing icebergs, they cannot be used during polar night or under cloudy conditions. In contrast, radar remote sensing can be used for analysis regardless of the weather conditions or time of year (Han et al., 2019; Mazur et al., 2017; Wesche and Dierking, 2012). In particular, although scatterometers facilitate daily position and motion observations of large icebergs (Budge and Long, 2011; Stuart and Long, 2011a, Stuart and Long, 2011b), SAR instruments have a more significant advantage in precise observations of iceberg area changes thanks to their relatively fine spatial resolutions. Indeed, various SAR instruments have been used for detecting or tracking icebergs in the polar oceans including ERS-1 (Young et al., 1998; Willis et al., 1996), ENVISAT (Li et al., 2018; Howell et al., 2004; Mazur et al., 2017), Radarsat-1 (Wesche and Dierking, 2015; Lane et al., 2002; Power et al., 2001), Radarsat-2 (Scheuchl et al., 2004; Denbina and Collins, 2014), TerraSAR-X (Frost et al., 2016), and Sentinel-1 (Lopez-Lopez et al., 2020; Moctezuma-Flores and Parmiggiani, 2017; Heiselberg, 2020; Han et al., 2019).
    
    We will also add the references you suggested (Budge and Long, 2017; Stuart and Long, 2011) in the section 4.1, with the description that sea ice may protect icebergs from exposure to wave stress.
    
    Thank you again for your comment for our paper.
    Sincerely,
    
    Young
    
    Citation: https://doi.org/10.5194/tc-2021-131-AC1
RC2:
'Comment on tc-2021-131', Anonymous Referee #2, 03 Jun 2021
In this paper, the authors do present and apply a semi-automatic approach for iceberg motion tracking from SAR imagery using the Google Earth Engine (GEE) as a base environment for data acquisition and processing. Then, is discussed the main drift patterns and area decay observed along with the B-43 iceberg correlating it with the major forces involved in the iceberg lifetime.

A significant problem is addressed and, while the manuscript does not add that much new information to improve the search for reliable methods able to improve iceberg detection/tracking in the complex polar environment and also, to enhance our understanding of icebergs and their drift patterns, the paper contributes to icebergs studies bringing to light the potential of the application of the cloud-based GEE as a platform to improve massive data processing. However, the manuscript could be strengthened by putting more effort into a validation analysis about the robustness of the method (segmentation, classification e tracking scores) for different scenes in which icebergs appear with different (i) radiometric signatures, (ii) shapes, (iii) surrounded by sea-ice/ice mélange, (iv) proximity to clusters of small icebergs/ice shelves, and others issues that may be present in different SAR scenes. It could do assist the reader in the reproducibility of the method on a large scale. Once, automated approaches able to detect and tracking icebergs on large scale can be very useful to construct reliable iceberg-coupled models for climate studies.

Following some specific questions, that could help to make clear if the method is semi-automatic for iceberg detection or B-43 detection. As described in the manuscript, the method was applied only on the iceberg B-43 (target of the study), leading the reader to ponder about the specificity and limitations of the method that could do not allow its reproducibility for a bigger number of icebergs.

The chosen iceberg is very well defined with a shape very descriptive. In the conclusions, this issue is mentioned however would be useful to elaborate better on this kind of o issue once this can be a serious limitation. Also about the need for the polygon to be manually digitized to be used as a reference to track.

Don’t you think the superpixel approach used is sensitive to the iceberg surface complexity (i.e. texture)? Although a smooth process is performed which can help to search the target B-43. Aspects such as edges definition can be decreased for smaller bergs reducing segmentation performance, maybe demanding specific segmentation tunning for all different icebergs;

As discussed in previous papers, the classification approach based directly on the iceberg backscattering strength is more suitable for bright icebergs or icebergs in open water. This can reduce even more the automatic aspect demanding more human intervention to constantly manually find the references used to computed the centroid distance histograms (CDH’s).

As described, to perform motion tracking the method is sensitive to shape changes, which can increase the total number of human interventions for continuing the tracking (What is unfeasible for a large number of icebergs). In this sense, would be very useful an analyze how much shape change the tracking method is reliable to overpass in order to keep tracking without manual intervention.

Furthermore, the manuscript is well-written with good-quality figures. Just a few additional analyses about the method's robustness and some comparison/discussion with other state-of-art methods could strengthen the paper.
Citation: https://doi.org/10.5194/tc-2021-131-RC2
- AC2: 'Reply on RC2', YoungHyun Koo, 09 Aug 2021
  
  We appreciate your critical comments and suggestions. As you mentioned, although our method works successfully for our target iceberg B43, this method should be further tuned and enhanced to be applied to other small icebergs. We add more discussions about the performance of this method (similarity scores and some Sentinel-1 images as references; Appendix A) and how to improve our method in future. Here are our responses to your specific comments:
  - We will add more discussions about the issue of iceberg shape. In addition, we will add more details about the need of manual digitization.
  - Yes. The image segmentation method is sensitive to the iceberg surface or texture. Indeed, although this was not a critical issue in detecting our target iceberg B43, this will be a critical issue for detecting other small icebergs. We will add more details about this discussion for a further improvement of our method.
  - Yes. It is well-known that icebergs have bright backscatter compared to the surrounding open water and sea ice. Actually, we use this bright backscatter feature to identify icebergs. We will make this clearer in the updated manuscript.
  - Thank you for your suggestions. In Appendix, we will add the analysis of similarity scores and show a few Sentinel-1 images for low similarity scores.
  Thank you again for your comments for our paper.
  
  Sincerely,
  Young
  
  Citation: https://doi.org/10.5194/tc-2021-131-AC2

Peer review completion

AR: Author's response | RR: Referee report | ED: Editor decision | EF: Editorial file upload

ED: Reconsider after major revisions (further review by editor and referees) (16 Aug 2021) by Stef Lhermitte

Dear YoungHyun Koo and colleagues,

Your TCD manuscript "Semi-automated tracking of iceberg B43 using Sentinel-1 SAR images via Google Earth Engine" received two constructive reviews where both reviewers identified some issues and questions, which you already clarified in your author comments. Both reviewers scored your manuscript ranging from fair to good where both suggest a revision. In this revised version I suggest you also take into account my comments made during the access-review (repeated here below) .

After your author's response, I would like to ask you now to re-upload the:
1 revised version of the manuscript
2. author's response as you have already (partlt) done in the interactive discussion, but now make it more detailed.
3. track change version so the reviewers can easily check what has been changed.

Based on this response and your revised version, I plan to reconsult the reviewers before taking a final decision.

Best regards,
Stef Lhermitte

Editor comments:
- the introduction provides a good context (i.e. GEE opportunity for iceberg tracking), but fails to address any of the topics the paper actually discusses. I miss some background on the iceberg tracking methodologies and/or the scientific questions related to the role of environmental variables on iceberg trajectories. The introduction only sketches the GEE opportunity, but fails to provide the scientific background on the real scientific questions etc.
- L65-67 is a bit strange in the context of an introduction as it already contains the main result in the introduction
- When introducing a new methodology (in this case the GEE tracking algorithm), it is always important to benchmark it to current state-of-the-art methodologies. You compare it to NIC data, but I think this comparison should be extended (and quantified) to show better the validity of the method. I would also start the results with the comparison with existing tracking methods and not introduce it halfway the results.
- I helps readability if results and discussion sections are separated.
- I would recommend you to change the color palette of Fig 7 (right column). Currently, this figure uses rainbow palettes. There are many good reasons to avoid rainbow palettes (see https://www.nature.com/articles/s41467-020-19160-7?s=09), so I think it is important to communicate with more correct color palettes.
- As the method is applied on a small iceberg, I miss some discussion on how the method would work on larger icebergs (e.g. A68) that often are not captured within a single S1 scene

Hide

AR by YoungHyun Koo on behalf of the Authors (07 Sep 2021) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (07 Sep 2021) by Stef Lhermitte

RR by Anonymous Referee #2 (08 Sep 2021)

ED: Publish as is (17 Sep 2021) by Stef Lhermitte

AR by YoungHyun Koo on behalf of the Authors (21 Sep 2021) Author's response Manuscript

Short summary

This study demonstrates for the first time the potential of Google Earth Engine (GEE) cloud-computing platform and Sentinel-1 synthetic aperture radar (SAR) images for semi-automated tracking of area changes and movements of iceberg B43. Our novel GEE-based iceberg tracking can be used to construct a large iceberg database for a better understanding of the behavior of icebergs and their interactions with surrounding environments.