Improving Grain Size Analysis to Characterize Sedimentary Processes in a Low-Energy River: A Case Study of the Charente River (Southwest France)
Round 1
Reviewer 1 Report
-Figure 1 is not clear, pls improve it, pls.
Author Response
Response to Reviewer 1 Comments
Point 1 : Figure 1 is not clear, pls improve it, pls.
Response 1 : Thank you for reviewing our publication. I re-downloaded tne figure 1 in the resolution of 300 dpi to make it more readable. However, I can't change the colors or the map legend.
According to comments from other reviewers, the article has been modified to improve the quality of the English and to make the text easier to read (especially in the Results section).
Reviewer 2 Report
Excellent work. However, we can regret that the figures are not sufficiently resolved, which sometimes makes the text difficult to read.
It might also be a good idea for the EMMAgo R package to appear in the title, but I leave it up to the authors whether or not to do so.
Author Response
Response to Reviewer 2 Comments
Point 1 : Excellent work. However, we can regret that the figures are not sufficiently resolved, which sometimes makes the text difficult to read.
Response 1 : Thank you for reviewing our publication. The article has been modified to improve the quality of the English and to make the text easier to read (additional paragraphs, numbers rounded or removed, especially in the Results section).
Point 2 : It might also be a good idea for the EMMAgo R package to appear in the title, but I leave it up to the authors whether or not to do so.
Response 2 : We prefer to maintain the original title and not add the R package EMMAgeo in the title because we are using 5 statistical methods, 2 of which are showing more promising results (PCA+CAH on grain size data and the EMMA approach). On the other hand, EMMA appears in the abstract and keywords.
Reviewer 3 Report
This paper compares the effectiveness of five different methods for processing grain-size data to characterize low energy alluvial plain deposits. The statistical methods were applied to a 9m-long core from the fluvial island of la Baine, located in France. The results indicate that elementary statistical parameters have limited usefulness in describing and interpreting fine fluvial deposits. Textural analysis is more informative, but heavily dependent on the classification scheme. Multivariate statistics and end-member modelling analysis are the most effective methods for identifying sub-units, but with some limitations. Overall, the paper suggests that further developments are needed to improve the ability to connect end-member classes to sedimentary processes.
In general, the paper is well-written. The topic is interesting. It is important to see how statistical learning and data mining can be applied to real-life situations. In support of this paper, the authors should consider the flowing points to improve the paper's quality further.
- The Introduction is quite long, I suggest authors move some text from lines 63 to 111 to a new section called "Related work" or "Literature Review".
- In the Introduction, the authors should spend a few lines discussing the role of data-driven approaches in this field.
- Figures 2, 3, 5, 6, 7, 9, 10, 11, and 12 are not in good form; I suggest authors improve the resolution of these figures and make the text readable and zoomable. You can move some figures into the supplementary material.
- A table of main notations used in the paper should be included in section 3 or the Appendix.
- Section 4 contains abundant text in some subsections, making it harder for readers to follow. Thus, I suggest authors explain in detail the method that gives the best results and briefly explain other methods.
- In section 5, since sedimentary data contains categorical features, the authors may discuss some possible methods besides the five methods used in the paper. The authors can refer to a partitional clustering framework introduced in [Estimating the Optimal Number of Clusters in Categorical Data Clustering by Silhouette Coefficient] in the discussion.
Author Response
Response to Reviewer 3 Comments
Thank you for reviewing our article.
Point 1 : The Introduction is quite long, I suggest authors move some text from lines 63 to 111 to a new section called "Related work" or "Literature Review".
Response 1 : The initial introduction has been divided into 2 sections to make it shorter: 1. Introduction and 2. Literature Review.
Point 2 : In the Introduction, the authors should spend a few lines discussing the role of data-driven approaches in this field.
Response 2 : The introduction has been re-modelled. All the approaches developed here are data-based and do not necessarily require further detail (they are both presented in section 2. Literature Review and in section 4. Materials and Methods).
Point 3 : Figures 2, 3, 5, 6, 7, 9, 10, 11, and 12 are not in good form; I suggest authors improve the resolution of these figures and make the text readable and zoomable. You can move some figures into the supplementary material.
Response 3 : The figures were re-registered at 300 dpi resolution. The figures are clear and the text legible with or without zoom. No figures have been moved in the supplementary material.
Point 4 : A table of main notations used in the paper should be included in section 3 or the Appendix.
Response 4 : A table of main notations used in this paper has been created in the Appendix. It's the Table 9. Main notations used in this paper
Point 5 : Section 4 contains abundant text in some subsections, making it harder for readers to follow. Thus, I suggest authors explain in detail the method that gives the best results and briefly explain other methods.
Response 5 : The article has been modified to improve the quality of the English and to make the text easier to read especially in the Results section (additional paragraphs, numbers rounded or removed, shorter phrases).
Point 6 : In section 5, since sedimentary data contains categorical features, the authors may discuss some possible methods besides the five methods used in the paper. The authors can refer to a partitional clustering framework introduced in [Estimating the Optimal Number of Clusters in Categorical Data Clustering by Silhouette Coefficient] in the discussion.
Response 6 : We've included a paragraph in the Interpretation section on the possible use of alternative partitional clustering methods such as k-means or k-medoids.
Reviewer 4 Report
Very high quality research
Author Response
Response to Reviewer 4 Comments
Thank you for reviewing our article and for your comment. According to comments from other reviewers, the article has been modified to improve the quality of the English and to make the text easier to read (especially in the Results section).
Reviewer 5 Report
Paper can be accepted in present form.
Author Response
Response to Reviewer 5 Comments
Thank you for reviewing our article and for your comment. According to comments from other reviewers, the article has been modified to improve the quality of the English and to make the text easier to read (especially in the Introduction and Results section).
Round 2
Reviewer 3 Report
The paper reaches the acceptance level.
