Experimental Results of Underwater Sound Speed Profile Inversion by Few-Shot Multi-Task Learning
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThe work may be published with minor modifications.
- (line 67) decipher the abbreviations CDT and SVP;
- (line 184) decipher the abbreviation XCDT; describe the difference between CDT and XCDT;
- (line 196-197) indicate what the numbers in brackets mean (6.774, 8.392, −18.8603);
- (line 198-211) What is the time interval between adjacent SSP measurements?;
- Figure 7. Additional panels for 7a and 7b, say 7c and 7d, will significantly improve the visibility of the graph
let's say additional graphs (panels, subplots) that will show the differences between Original SSP and the rest
(Original SSP - Reconstructed SSP,... and for everyone else)
- Figure 8. Similar changes as for Figure 7.
- Also, the work will be enriched with graphic information from USBL.
Graph in coordinates X axis - time, in coordinates Y axis - depth, in coordinates Z axis (color) - USBL output data
I suppose that the work can be published if the above recommendations are followed.
Author Response
We would like to thank you and all the reviewers for the precious time and efforts spent on reviewing our manuscript. The comments are valuable and helpful for improving the quality of our manuscript. We have carefully addressed all the comments raised by the reviewers in the revised manuscript and listed all the point-by-point response to the reviewer's comments. The amendments are highlighted in red color for clarity.
Lastly, we would like to express our greatest appreciation once more to you and the reviewers for the valuable comments.
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsExplain in details measurement procedure, which transducers are used and why are receivers in the Figure all on the same depth...
Wouldn't be better to use receivers when measuring speed of sound at different depths.
Explain main reason of error in ML algorithm for estimating the sound speed profile, what is the influence of ocean data (temperature, salinity) on the results...
Author Response
We would like to thank you and all the reviewers for the precious time and efforts spent on reviewing our manuscript. The comments are valuable and helpful for improving the quality of our manuscript. We have carefully addressed all the comments raised by the reviewers in the revised manuscript and listed all the point-by-point response to the reviewer's comments. The amendments are highlighted in red color for clarity.
Lastly, we would like to express our greatest appreciation once more to you and the reviewers for the valuable comments.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsThe manuscript describes a multi-task learning (MTL) model with partial parameter sharing among different training tasks, aimed at a fast and accurate estimation of underwater sound speed profiles. While the method is described in detail there is no clear description of the signal propagation data acquired in the experiment, which is crucial for the MTL method to be tested.
Comments on the Quality of English LanguagePlease retouch some sentences like, for instance, "can lead to great Snell effect."
Author Response
We would like to thank you and all the reviewers for the precious time and efforts spent on reviewing our manuscript. The comments are valuable and helpful for improving the quality of our manuscript. We have carefully addressed all the comments raised by the reviewers in the revised manuscript and listed all the point-by-point response to the reviewer's comments. The amendments are highlighted in red color for clarity.
Lastly, we would like to express our greatest appreciation once more to you and the reviewers for the valuable comments.
Author Response File: Author Response.pdf
Reviewer 4 Report
Comments and Suggestions for AuthorsThe paper is devoted to an actual topic. The authors solve the inverse problem of estimating sound speed profile out of the acoustic data. Machine learning methods can help to find the solution faster than conventional methods like deapth/bredth first search, Monte-Calro etc. However, there are some concerns about this manuscript.
1. The first two paragraphs of the introduction are too simple. They do not give a reader an idea of real challenging problems in this field.
2. The authors are encouraged to describe a possible practical scenario for application of this method. You see, an acoustic positioning system was used to carry out the experiment. It is a quite complicated equipment, much more complicated and expensive than a regular SSP (CTD) probe. Please, provide a case, when a high-quality acoustic data is available, but CTD cannot be applied. Additionally comment if you always need CTD to fine-tune the model in the place of the experiment.
3. The tested case was too simple for inversion. The actual SSP was too smooth and “too classical”. Perhaps, a range-depth ratio, when the range was 10 km and the depth was 3 km, lead to the fact that multipath propagation effects from the surface to the bottom were not so critical. So authors are welcome to provide some more examples (maybe simulated ones) in order to estimate the limitations of the proposed method.
4. The idea of experiment some have some defect inside it. First, you rely on acoustic positioning to get the actual coordinates of beacons. The results of the first step are obviously influenced by SSP. Second, you use acoustical data and coordinates to get the actual SSP. Is that so? Please comment on that.
Author Response
We would like to thank you and all the reviewers for the precious time and efforts spent on reviewing our manuscript. The comments are valuable and helpful for improving the quality of our manuscript. We have carefully addressed all the comments raised by the reviewers in the revised manuscript and listed all the point-by-point response to the reviewer's comments. The amendments are highlighted in red color for clarity.
Lastly, we would like to express our greatest appreciation once more to you and the reviewers for the valuable comments.
Author Response File: Author Response.pdf
Round 2
Reviewer 4 Report
Comments and Suggestions for AuthorsAll issues, raised by the reviewer were addressed. The text of the paper was improved. The authors did a good job regarding that. I think the paper can be published in the present form.