Skip to Content
You are currently on the new version of our website. Access the old version .
MDPIMDPI

Advanced

by
  • Qingjie Zhou 1,*,
  • Xishuang Li 1 and
  • Baohua Liu 2
  • et al.

Reviewer 1: Youkou Dong Reviewer 2: Hongshuai Qi Reviewer 3: Dmitrii G. Borisov Reviewer 4: Anonymous

Round 1

Reviewer 1 Report (Previous Reviewer 1)

The determination and classification of seafloor sediment types are crucial for the exploitation of marine resources, construction of marine engineering, and maintenance of marine ecological environments. Automatic classification of seafloor sediment based on acoustic data is an important method to quickly understand the type of a large range of sediment. This manuscript presents an approach for classifying seabed sediment using an enhanced U-Net and K-means clustering algorithm, which has certain innovation and novelty. In the manuscript, reasonable modifications have been made to the previously mentioned questions. The structure of this paper is basically reasonable, the logical thinking is clear, the point of view is accurate, the language is smooth, and the argumentation method is reasonable.This is a carefully done study and the findings are of considerable interest.

A few minor revisions are list below.

1. Abstract:

The abstract needs to be condensed again. It is necessary to fully express the advantages and disadvantages of the method and the reliability of the results

2. Relevant research background needs to be supplemented in INTRODUCTION. The background of U-Net and K-means cluster analysis needs to be supplemented.

3. Please check the text carefully for lexical accuracy and consistency. For example, in Line 60, Line 66, Line 69, the ‘sub-bottom profiling data’ the Line 69, ‘sub-bottom profile data’, and the Line 70 the ‘sub-bottom profiler data’.

4. Line 442, notice the format of formula (9).

5. Revising the conclusion. Conclusion is not to summarize what you have down in the study. It should straightly bring forward the logical knowledge and fruit by your research.

I suggest that the manuscript be published in Remote Sensing after minor revision.

I wish the authors all the best in the further development of their manuscript.

Author Response

Thanks for the reviewer’s kind suggestion. We have responded to relevant opinions point by point, please see the attachment for details.

Author Response File: Author Response.docx

Reviewer 2 Report (New Reviewer)

The authors provided an interesting case study of classification of seabed sediment based on improved U-Net. It is a meaningful try for classification of seabed sediment. The authors have revised to improve the manuscript greatly with the suggestions. However, the flaws that have mentioned by former reviews is still unchanged. The manuscript should provide more discussion on the effectiveness and accuracy with more validation and comparing analysis. Otherwise, The logic of the manuscript should be improved, (1)The manuscript should research on innovative method with case application rather than research on the result of sediment classification  in Northern Slope of South China Sea  (2)the Discussion only describe the four sedimentary environment zones in research area, but not discuss with previous research. It is not deep enough. (3) So, the second conclusion is not the highlight and key point of this manuscript. 

Author Response

Thanks for the reviewer’s kind suggestion. We have responded to relevant opinions point by point, please see the attachment for details.

Author Response File: Author Response.docx

Reviewer 3 Report (New Reviewer)

 

The work describes an interesting approach for estimating sediment grain-size based on the analysis of multibeam and sub-bottom profiling data using machine learning techniques. Such approaches are undoubtedly the future of marine science. However, the manuscript has several issues that prevent its publication in its current form.

 

MAJOR COMMENTS:

1. The manuscript was submitted as a file that included all previously made corrections, which made the reviewer’s job challenging. Some sections were impossible to read.

2. The authors did not provide sufficient information on the data used for the study. Specifically:

- What equipment was used for multibeam and sub-bottom profiling?

- When were the data acquired (cruise, ship, year)?

- How were sediment samples collected (grab, gravity corer, box corer)?

- How was the grain size of sediment samples analyzed (wet sieving, laser diffraction)?

A small part of the needed information is presented in the “Regional setting” section, but it should be moved to the “Data and Methods” section.

3. It is unclear what grain-size classification system was used in this work. I suggest that it was Folk's classification system, but it does not include the "Less muddy silt" type. Can you clarify on this point?

4. The authors did not describe the verification of the results obtained. Specifically, there is no discussion regarding the correlation between the grain-size analysis of sediment samples and the results obtained using the neural network and clustering techniques.

5. The authors should provide the limitations of the developed approach.

6. The geological terminology used in the manuscript is at times unclear or inaccurate.

7. I am not a native speaker but I strongly recommend English language editing. 

 

The geological terminology used in the manuscript is at times unclear or inaccurate.

SOME EXAMPLES:

I would recommend usage of the terms "submarine sediments", "seabed sediment" etc. The term "sediments"  imply its formation and location on the sea-floor. If the authors would like to emphasize that the study focuses on the sediment from the sea, the term "marine sediments should be preferred".

"Substrate" is also not the best choice. I would recommend using "surface sediments" or "deposits" instead.

sediment particle components (sediment composition?),

spatial distribution of the research area (location of the study area?),

shelf slope break line,

seabed shallow profile (?),

sedimentary environment zones (?),

and many others

Please look for other papers on sedimentology in the top level scientific journals (preferably with native English speakers as authors)

Author Response

Thanks for the reviewer’s kind suggestion. We have responded to relevant opinions point by point, please see the attachment for details.

Author Response File: Author Response.docx

Reviewer 4 Report (New Reviewer)

Please see the attachment

Comments for author File: Comments.pdf

Text is understandable and easy to read. Quality of English Language is fine.

Author Response

Thanks for the reviewer’s kind suggestion. We have responded to relevant opinions point by point, please see the attachment for details.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report (New Reviewer)

Authors has revised the manuscript carefully with suggestions. The result and discussion has improved greatly. its logic is more clear. More validation and comparing analysis were done to support the new classification method for seafloor sediment. Despite the method need practical examination with more cases, the manuscript is encouraged to publication.  

Author Response

Thanks for the reviewer’s kind suggestion. Please see the attachment for comments reply.

Author Response File: Author Response.docx

Reviewer 3 Report (New Reviewer)

The revised manuscript still suffers from numerous incorrect geological terms, typos, repeated figures, and fragments of text. The structure of the manuscript still needs to be changed. The captions for the figures, in most cases, do not sufficiently describe the contents of the figures.

 

MANUSCRIPT STRUCTURE

Please move rows 126-137 to the introduction or add to a new section "Regional settings"

 

2. Data and methods

2.1. Data collection

               2.1.1. Multibeam survey

               2.1.2. Sub-bottom profiling

               2.1.3. Sediment sampling

2.2. Improved U-Net

2.3.Optimized K-means clustering algorithm

 

MAJOR COMMENTS

The seafloor reflection intensity also depends on the calcium carbonate content in sediments. Why this factor was not taken into account? What is the composition of the studied sediments (siliclastic, calcareous)?

It states (182-183) that the reflection coefficient and acoustic impedance are closely linked to the physical properties of the sediment, such as mean grain size density, and porosity.  Why the porosity was not taken into account?

In conclusions in states that authors performed the comprehensive analysis and research on the classification and characteristics of seafloor sediment types. In fact, only two characteristics were studied: grain-size and seafloor reflection intensity. It is not a comprehensive study.

If possible, please put the averaged grain size obtained during the grain-size analysis of sediment samples on Fig.12 (just Φ units near the dots marking the location of sampling sites).

 

TERMINOLOGY

multi-beam depth -> multibeam bathymetry, multibeam survey data

topographic slope - what does it mean? slope angles?

sampling tests -> sampling sites?

seafloor reflection intensity of the sub-bottom profile -> seafloor reflection intensity (based on sub-bottom profiling data)

seafloor sediments is not the best term, I would recommend "marine sediments", "surface sediments" or "sediments"

surface sediments is a widely used term. In many papers this term implies the upper 2 centimeters of sediments. Please indicate which depth interval is meant by the term "surface sediments"

438: low amount of clayey - what does it mean, clayey? it's an adjective, where's the noun?

"The spatial distribution characteristics of water depth" sounds not good; it is just sea-floor topography

179-180: "seabfloor, which is the interface between water and sediment at a depth range of 1m" this definition is totally unclear

185: surface soil? There is no soil on the sea-floor

 

FIGURES

Figure 1

what is CNOOC? please provide the description?

Figure 3. a) time or TWT, be consistent. Why the time values in Fig.3a are over 1500 ms while in Fig.3b it is less than 20 ms? What units has horizontal axis in Fig.3b? Figure 3c was not discussed in the text. What is it for? What is SD, CD in the Figure 3c?

Figure 8 repeats the upper part of figure 12.

Figure 9 is not needed. It is a well-known scheme

Figure 10. Is this figure really nessesary? By the way: sandy silt not SAND silt. The figure is not clear. Why sand, silt and CLAYEY (not clay?)

Figure 11 rpeats the inset in Figure 13.

Figure 13. it is not clear from the caption that the scheme in the up-right is a result of this study.

 

MINOR COMMENTS

77: which velocity?

87-88: Berthold et al. used GoogLeNet to preliminarily classify gravel, mud, sand, and mixed sediments. Based on which data?

145: surface sediment sampling

438: low amount of clayey - what does it mean, clayey? it's an adjective, where's the noun?

438: Based on the figure... Which figure? Figure 10? The relative composition of grain-size fractions is not evident in the figure.

515-523: Repeat of the text in rows 126-137.

 

555-560 should be excluded from the conclusions.

 

-

Author Response

Thanks for the reviewer’s kind suggestion. Please see the attachment for comments reply.

Author Response File: Author Response.docx

Reviewer 4 Report (New Reviewer)

After reading the reviewed paper, I believe that the paper is significantly improved. However, I think one of my comments in “Major issues” was not properly addressed, so please find my comments below. I also have two minor comments as follows:

1. I believe that authors still didn't present the overall accuracy of their dataset. To do so, an additional set of samples not used for classification is needed, that is than used to independently test accuracy of the dataset used for training. It is a common practice to show it as a proof that training set is viable. I suggest two papers, but feel free to search for more recent, especially in Remote Sensing Journal.

Stephens, D.; Diesing, M. A Comparison of Supervised Classification Methods for the Prediction of Substrate Type Using Multibeam Acoustic and Legacy Grain-Size Data. PLoS One 2014, 9, doi:10.1371/journal.pone.0093950.

Diesing, M.; Green, S.L.; Stephens, D.; Lark, R.M.; Stewart, H.A.; Dove, D. Mapping Seabed Sediments: Comparison of Manual, Geostatistical, Object-Based Image Analysis and Machine Learning Approaches. Cont. Shelf Res. 2014, 84, 107–119, doi:10.1016/j.csr.2014.05.004.

2. As mentioned in my first comments for Line 132-133 (maybe it was not sufficiently clear in writing, so I apologize), I would like to repeat the comment that maps in Figures 1, 3, 8, 10 and 12 should have a scale bar.

 3. It is not necessary to put Folk sediment classification diagrams, as they are a standard. If you want, I suggest to plot grain-size data on them (as points). That is not too complicated to do, otherwise it is really not necessary to put the diagrams with no data in them into the paper.

No comments

Author Response

Thanks for the reviewer’s kind suggestion. Please see the attachment for comments reply.

Author Response File: Author Response.docx

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 manuscript presents an automated approach for classifying seabed sediment using an enhanced U-Net and K-means clustering algorithm, which has certain innovation and novelty. The proposed method employs multi-beam water depth, sub-bottom profile, and sample test data from the Northern Slope of the South China Sea to perform the classification analysis. Furthermore, the sediment classification outcomes are integrated with the geological context to summarize and examine the sedimentary environment of the study area. The results of this study offer a valuable addition to the existing classification techniques for seafloor sediments and provide significant insights into the sedimentary environment and processes of the Northern Slope of the South China Sea. This is a carefully done study and the findings are of considerable interest. A few minor revisions are list below.

1. Errors of language：

Line 20, ‘U-convolutional neural network’ would be ‘U-Net’.

Line 37, ‘acoustic imaging’ would be ‘acoustic data’.

Line 109, Line 110, Line 119, Line 249, ‘shallow profile’ would be ‘sub-bottom profile’, the noun should be unified.

Line 330, ‘northeast-southwest’ would be ‘NE-SW’.

2. Relevant research background needs to be supplemented in INTRODUCTION. The background of U-Net and K-means cluster analysis needs to be supplemented.

3. CONCLUSIONS needs more in it, as it's more of an afterthought. The technical advantages of the substrate classification method based on the combination of improved U-Net network and k-means cluster analysis are not summarized.

I suggest that the manuscript be published in Remote Sensing after minor revision.

I wish the authors all the best in the further development of their manuscript.

Author Response

Thanks for the reviewer’s kind suggestion.

  1. The language in this paper has been repolished and the language errors have been corrected.
  2. The research background is supplemented in the introduction, including U-Net and K-means clustering analysis.
  3. The conclusion is supplemented. The technical advantages of sediment classification method based on improved U-Net network and k-means cluster analysis are summarized.

Reviewer 2 Report

This study is very interesting, but unfortunately, I don't think it can be published. The entire manuscript has significant flaws. Please carefully revise your manuscript. Wish you all the best.

 

1. In the abstract, do the words 'seafloor', 'submarine', and 'seabed' mean the same thing?

2. In the abstract and introduction, do the words 'categories', 'types', and 'classes' mean the same thing?

3. I am not clear about the content expressed in the abstract. What is the relationship between a single data type and the drawbacks of U-Net, and why is the word 'however' used to connect these two sentences. Should 'however' be changed to 'besides'?

4. Please provide evidence that supports the following points in the manuscript: "The optimization algorithm has several advantages, including easy programming, few iteration times, and stable clustering results. It has been found to be effective in the classification of seafloor substrate."

5. The references discussed by the authors are outdated, and more references from the past three years should be discussed.

6. I did not see the encoder on the left side of Figure 2, nor did I see the decoder on the right side of Figure 2. Line 182, line 190.

7. What is pooled index upsampling technique? The author did not introduce it in this manuscript, only cited a reference, which is very irresponsible.

8. The methods proposed in this manuscript are not novel enough, they only cite others' methods, and even do not bother to introduce the details of so-called innovative methods in the manuscript. Section 3.1.

9. Lines 209-210, "It has been found to be effective in the classification of seafloor substrate." Please cite references that support this conclusion.

10. Lines 225-234 should not appear in the current position. This only introduces that distinguishing between seafloor sediments with high similarity is difficult, and it should be placed in the Section 6.

11. Section 4 is named "Classification and result analysis of floor seating", but I did not see any experimental results or analysis in this section. In this section, only the process of classification is stated.

12. In Section 5, only the results of the so-called improved method are listed, while the results of the original method are not listed. The author needs to supplement experiments for comparative analysis.

13. In the Section 6, the reason for the effectiveness of the proposed method was not discussed, but rather something else was mentioned, which made me very confused.

14. The manuscript should be no less than 18 pages.

15. The pictures are too blurry.

The language of the entire manuscript needs to be carefully reorganized.

Author Response

Thanks for the reviewer’s advice.

  1. The words 'seafloor', 'submarine', and 'seabed' mean the same thing. It has been revised in the article to use ‘seafloor’ as a unified expression.
  2. It has been revised in the article to use ‘type’ as a unified expression.
  3. The expression here is wrong. The disadvantages of single data type and U-Net are juxtaposition.
  4. The relevant contents are supplemented in the article. The algorithm optimization process is described in detail in this paper.
  5. This paper makes a supplement to the recent references.
  6. The expression is incorrect. The text has been corrected.
  7. The paper makes a supplementary explanation to the pooling technology.
  8. The improvement of the method is supplemented in this paper.
  9. This point has been modified and supplemented in the article.
  10. According to Shepard's classification of sediment type structure, we classified the sediment type in the study area, and we still think it is more appropriate to place it here.
  11. The corresponding position in the paper has been modified and improved.
  12. The experimental results of the original method are supplemented and compared.
  13. In Section 6, the effectiveness of the method is supplemented. Combined with the sedimentary environment of the study area, the validity of the results of sediment classification is proved.
  14. The article expands to 18 pages.
  15. The picture has been modified and perfected.

Reviewer 3 Report

This study uses multibeam bathymetry data and sub-bottom profile data to propose an automatic classification method for seafloor sediment based on improved U-Net and K-means clustering algorithms and delineates four sedimentary environment zones. Although this study is interesting, however, it suffers from some important flaws that do not encourage publication.

1. For the journal "Remote Sensing", the manuscript requirement is at least 18 pages, but the submitted manuscript is only 13 pages, which is far from enough.

2. In the Introduction, the survey of previous research is not sufficient, please add more discussion and analysis of the research in the field of seafloor sediment classification.

3. In lines 182 and 190, it should be Figure 4 and not Figure 2.

4. In recent years, many studies have used multi-beam bathymetric topography and sub-bottom profiles data for seafloor sediment classification, but the authors have not discussed these studies and consider the use of these data as a contribution to seafloor sediment classification, which is not a novel idea.

5. This study proposes an optimized K-means clustering algorithm. And, it is stated in the paper, "The optimization algorithm has several advantages, including easy programming, few iteration times, and stable clustering results. It has been found to be effective in the classification of seafloor substrate." Please validate the above advantages with experimental analysis or theory, rather than simply stating.

6. Please use vector graphics in your manuscript so that readers can clearly see your results.

7. Please explain in more detail what the pyramid pooling technique is, so that the reader can understand it more clearly.

8. Please discuss in detail the advantages of using data such as multibeam bathymetry and shallow profilers to classify seafloor sediments compared to using only backscattered intensity data.

9. Most of the references listed in this study are outdated, please cite some references within the last 3 years.

10. Please analyze in detail how the pooled index upsampling technique makes the image edges sharper, how it avoids the need to learn to upsample, and how it reduces the network training parameters.

11. Please select some of the most advanced methods and compare them with the proposed methods in this study to demonstrate their effectiveness.

12. In the Discussion and analysis, the four sedimentary environmental zoning maps of the study area are simply stated, and the effectiveness of the proposed method is not analyzed and discussed.

13. There are many different technical words in the manuscript but they all mean the same thing, for example: "seabed", "seafloor", and "submarine'', it is recommended to use only one of them.

The logic of the language in the manuscript is somewhat confusing and many words are used incorrectly, so it needs further editing.

Author Response

Thanks for the reviewer’s kind suggestion.

  1. The article expands to 18 pages.
  2. In the introduction part, the research in the field of submarine sediment classification is discussed and analyzed.
  3. The corresponding position in the paper has been modified.
  4. The discussion and analysis of multibeam substrate classification are supplemented in this paper.
  5. In this paper, the optimized K-means clustering algorithm is supplemented in detail.
  6. The picture has been modified and perfected.
  7. In this paper, the pyramid pool technology is supplemented
  8. Since there is no multi-beam backscattering intensity data in the study area, no comparative analysis has been made for this part.
  9. This paper makes a supplement to the recent references.
  10. The article makes a supplementary explanation to this part.
  11. The article makes a supplementary explanation to this part.
  12. In Section 6, the effectiveness of the method is supplemented. Combined with the sedimentary environment of the study area, the validity of the results of sediment classification is proved.
  13. It has been revised in the article to use ‘seafloor’ as a unified expression.