On the Detection and Long-Term Path Visualisation of A-68 Iceberg
Round 1
Reviewer 1 Report
This paper proposed the iceberg detection algorithm using Sentinel-1 SAR image, which aims at analysis the dynamic of A-68 iceberg. The fuzzy logic approach based procedure is used to enhance the contrast of the SAR image, so as to achieve the quasi-automatic iceberg extraction. For the time series analysis of the A-68 iceberg, perimeter, area, rotation angle, major axis length as well as the positon of the iceberg is investigated. The manuscript is well written and the results is good. Some comments and suggestion as below:
- For iceberg detection using fuzzy approach, the fuzzier parameter q is set to 0.6. Is there any influence if q is set to different value? If so, please give some explanation for selection of q.
- For the iceberg detection result in Figure 2, there is a limite difference between the proposed method and the comparison approaches. In order to demonstrate the performance of the proposed method, the quantitative evaluation is suggested.
- For the time series analysis in Figure 4, it shows a different tendency between perimeter and area in some cases. For example, from October 2017 to December 2017, there is a significant increase of perimeter, while the area is slightly decrease. What is the main reason to cause the different patterns between perimeter and area, please give more description.
The topic addressed can be framed in the piece of research devoted to provide further insights on continuous iceberg observation. I agree that the work need collect to large amount of data from Sentinel-1 for inter comparison, and the current presentation of this manuscript need some revision so as to be published in Remote Sensing. I would like to give major revision.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The authors have in this resubmission adressed most of my previous questions and comments, although they have not answered several of them.
This work is suitable for a technical note.
Accept as is, if the other referees agree.
Author Response
Please see the attachment.
We appreciate the reviewer's opinion.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The author has addressed all the comments which listed in the first round. I have no more comments at the moment. In my view, the manuscript is satisfied to be published in Remote Sensing.
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
This paper provide the iceberg detection algorithm based on fuzzy enhancement and stochastic segmentation using Sentinel-1 SAR imagery. Then the time series of A-68 iceberg including displacement, area and rotation are discussed. The result indicate that the proposed method can obtain a satisfied iceberg detection result and small variations in area, perimeter and major axis length parameters can also be observed. The manuscript is well written and the results is good. Some comments and suggestion as below:
- The parameter a are defined twice in the manuscript, see equation 2 and equation 5. If they represent different parameters, it is better to use different symbol.
- To evaluate the performance of the iceberg detection algorithm, the Quantitative analysis of proposed method should be compared with state-of-the-art iceberg segmentation algorithm.
- The area and the perimeter shows different tendency on time series in figure 4, What is the main reason to cause different pattern on area and perimeter, please give more explanation.
The topic addressed can be framed in the piece of research devoted to provide further insights on iceberg extraction. I agree that the work need collect to large amount of Sentinel-1 SAR data, but the current presentation of this manuscript might is not suitable to be accepted in Remote Sensing. I would like to reject it.
Reviewer 2 Report
The paper proposes a long-term tracking approach which is based on a two-stage mechanism: (1) a fuzzy-logic based contrast enhancement, and (2) Ising model based segmentation and tracking. The target is a huge iceberg, the location and possible trajectory of which is highly important for marine security applications. Even though I like the idea and the presentation, I have various concerns which are based on the application part of the work.
Data and Method: There are 12 Sentinel-1 SAR images for a period of around 1.5 years and for each, a classical segmentation approach is applied.
1) There is no comparison with the ground truth data (location, size, etc.), which would basically express how good the proposed detection algorithm works. In absence of the ground truth data, the superiority of the proposed detection method can be shown by comparing it to some other kinds of detection methods. I can see that there is a comparison to binary K-means, the number of examples could be increased.
2) Actually scientifically, the method is not a tracking method since there is no state-space model description and a measurement model to apply a tracking algorithm. When reading the title, I was expecting a proper tracking method, maybe including a comparison study for the proposed methods and the classical state of the art, such as Kalman Filters. As it stands I am afraid naming this work as a "tracking" study is not fair. I would recommend using "long-term path visualisation of A-68 Iceberg".
3) In both equation (2) and (5) there are \alpha parameters. I think they are different parameters, so please replace one. If they are the same, please clarify the relationship.
4) Is Iceberg A68B plotted somewhere in the paper? I have not seen it, unfortunately.
5) It sounds like 40 iterations are quite less than normal for a Bayesian sampling methodology. Is there a specific reason for this? Please discuss these details more.
6) Related to (5), the non-convex optimisation solving step with the SA algorithm should be discussed in the paper. The authors just mentioned SA is utilised.
7) Also, the fitting results in Figures 4 and 6 are not informative. I mean they provide some information about the characteristics of the iceberg, but what is the point of this work? For example, can the proposed method be used in a prediction work to approximately know where the iceberg will be in 2022 January? According to me, those fitting results are highly logical if the work includes a simple prediction example as well.
9) I also suggest the authors add a block-diagram, flow diagram or an algorithm in order to visually summarise the proposed method from acquiring the SAR data to detect and segment the iceberg. It should include all the steps.
10) As I said at the beginning, I actually like the idea and presentation of the paper in general. However, from a scientific point of view, there is a way to go before the publication.
Reviewer 3 Report
This work tracks a very large Antarctic iceber A-68 long time. It is a simple task well described.
It would be nice with a better understanding and modeling of their results. For example:
- In Fig.4 the area and perimeter (circumference?) display a damped oscillation with season as could be expected a priory from simple melting at summer and freezing at winter. However, in the Antartic winter is around juli where area melting is largest. This is counter intuitive?
- The area and circumference should be related somehow. A simple model would be a square root, but the data seem to indicate some more?
- How is the major axis length shown in FIg. 6 related to area and circumference? Some simple understanding and relations are missing in this work.
- Why does the iceberg rotate 160 degrees in FIg. 6? Is it the currents, some Coriolis force or what?
- FIg. 6 caption text say temporal displacement, but is actually angle and major axis length that are plotted.
- Fig. 4 caption text, red and blue should be switched, as they do not show area and perimeter (respectively) as written but vica versa.
In summary, I suggest that more understanding and modelling is added, which can be used for other cases.
