ICESat-2 Bathymetric Signal Reconstruction Method Based on a Deep Learning Model with Active–Passive Data Fusion
Round 1
Reviewer 1 Report
Summary:
In this paper, an active-passive fusion ICESat-2 bathymetric signal reconstruction method is proposed. By proposing a new method, the author has improved the accuracy of inversion of nearshore bathymetry using ICESat-2 data and optical remote sensing image data when altimetry data are partially missing. The effectiveness of the method is verified by practical experiments, and the results are clear and reliable. The preliminary results show that the proposed method is able to reconstruct the missing bathymetric signal and the accuracy of the bathymetric inversion of the reconstructed ICESat-2 data is better than the results when the data are missing. At the same time, the author also compares and analyzes the accuracy performance of bathymetric inversion results under different bathymetric inversion models and different resolution remote sensing images. The above conclusions are gratifying, and further research will undoubtedly promote the accuracy of altimetry in nearshore bathymetry. Therefore, this is a well-written manuscript, and it provides a good reference for related research. However, there are some comments which, if addressed, could provide a more impactful manuscript. My comments are below.
Comments:
1) Lines 43-45: consider whether more examples of inverse bathymetry using fusion of ICESat-2 data and multispectral remote sensing data can be provided, especially relevant studies when altimetry data are locally missing.
2) Lines 86-95, it is possible to provide which version of ICESat-2 ATL03 data was used in the experiment, and whether the four ICESat-2 laser data chosen for the paper were strong or weak beams, and why.
3) Line 140, consider adding the meaning of PMS in “GF-2 PMS”.
4) Line 268, What is the matching method between the in-situ bathymetry values and the inverse bathymetry values, (e.g. interpolation or averaging), a more detailed description needs to be given here.
5) Line 274: Why were four bathymetric inversion models introduced and what are the differences between several models? Consider a more detailed description.
6) Lines 328-338, Signal reconstruction is the focus of this paper. Only qualitative descriptions are available here, such as poor results in study area A and better results in study areas B and C. It is suggested to add specific statistical indicators to quantitatively describe the validity and reliability of the reconstructed values at the three sites.
7) Lines 436-441: The (b)-(d) are shown twice in the caption of Figure 10. The second one is (e)-(g)?
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 2 Report
Your paper brings some originality in using deep learning algorithm in order to reconstruct IceSat-2 bathymetric profiles. However, it is the view of the reviewer that your paper suffers several major flaws:
1) Qualitative results where there should be more quantification. You should strongly avoid "other reaseons", "superior", too high"....
2) Likewise your result figures should describe in a more explicit form the difference between the results of the different models. You can use difference map, histograms of difference, violin distribution of the difference, etc.
3) The complexity of some of your sentances and/or the way they are formulated deserve the paper. (see below)
4) LSTM and DBN should be described in more details.
l36: Not entirely true. Physic based SDB estimation has proven to work without in-situ measurement.
l54-56 A figure of those issue would be welcomme. Actually this figure could be fig 13. It seems to me that, you could drop your last 55 section
l63 with -> which
l67. As this is first introduced here, you should introduce the concept of active-passive data fusion
l79 "compared.". The point should be suppressed or the next sentance modified. Most likely both
l81-82 reformulate
l97-98 reformulate
l99-102 reformulate
l104 the verb should not be in the future tense.
l123 large range -> large coverage
l126-127 reformulate
l147-148. You can further describe the quality of in-situ data using the International Hydrographic Organization's Special Publication number 44 (version 6)
l193-197: Long sentance. Difficult to understand
l197 LSTM is first introduced here. The acronym should be detailed. I would have introduced it earlier
l266: various is vague please be more specific
l298: "Its proposal". Proposal is not the appropriate word
l330: very ideal -> subjective. Be more objective.
l332 relative ... subjective. Be more objective.
Table 4 Unless I missunderstood, I feel this is a bit akward. To assess the performance of the reconstruction, I would have compared original vs reconstructed ( at the location where data haven been removed) and not all the profile
Figure 7: You should also compare with existing data when it is possible.
Plot x = depth from sooundings, y corresponding value of estimated depth + plot linear regression (should be close to one)
Figure 9 These figures are hard to compare. Difference maps may be a solution to highlight differences. There are other possible solutions.
l417-418 Not sure I understand the sentance
l420 "to some extent" too vague
l423 "relatively divergent" too vague
l429-430 rephrase, as it does not sounds english
l431 too low, vague, best would be to provide a range of resolution
l433 Capital after point. Please reformulate
l494-l497 too complex of a sentance to clearly understand what you mean
l532 "and other reasons" too vague
l558-559 poor english structure of the sentance
References. You might want to add DOIs
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report
Please clarify the following:
a. Describe in more detail the refraction correction procedure you undertaken ti reduce the ICESat-2 data.
b. Specify why the fact that the DBN model is generally superior than the other models used (line 495).
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 4 Report
Have no comments
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
The paper has greatly improved. Thanks for your work !