Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Brief summary

The paper shares the results of analysis of flood risk in an area of Brazil which relies on rail for the movement of goods but due to the terrain is at risk of flooding. To better understand this risk the authors have employed a MCA and ML approaches. Whilst I expect the analysis that has been conducted is valuable it has not been sufficiently well presented to warrant publication without a major review. There needs to be a much greater line of sight to existing literature, the article requires rewording in many places and the results need to be more robustly critiqued in reference to the existing literature.

General concept comments

1.       Whilst there is inclusion of literature in the first section of the paper, there are areas where this is sparse (see detailed feedback) and I would expect to see a dedicated review of the methodologies that the authors employ such that they can justify more robustly their choices.

2.       I would suggest your aim comes before the hypothesis an research questions. This research aims to….. We will achieve this aim by proving/disproving the hypothesis by answering these research questions….. I also think you’d be better showing the research questions just as a numbered list. I’m not sure they need to be in a table.

3.       Section one would benefit from a summary paragraph at the end which outlines the structure of the paper.

4.       In figure three the precipitation database does not appear to feed into nay other steps of the method. Further the iterative nature of the review/monitor drainage system could be better presented if the intention is to apply this method once further new climate models are available.

5.       I would like to see a more clear explanation and justification for how you’ve developed your methodology. Where are these techniques or similar used in the literature. How did you decide that this would be the best/most appropriate method? This point would be best addressed through also addressing point 1 above.

6.       Where do the outputs of figure 4 feed into the stages of the research?  These links can be more clearly made. Think about how figure 2, 3 and 4 link together. How can you more clearly communicate this?

7.       The Otto Pfafstetter method needs to be critically evaluated and noted as a limitation. This has been developed using meteorological recordings from the 1950’s things have changed drastically since then and weather patterns now are not the same as then. I wouldn’t expect you to alter this method necessarily but I don’t think you have appropriately critiqued this approach nor reflected on what it means for your analysis.

8.       Noting you are only looking at surface drainage. What values for R are you using? What cross-sectional shape of ditches are you considering? Or where are you getting these values from? Presumably pluvio? But I think you need to explain how you know the shape of each of the ditches in the drainage system. Or if you don’t what assumptions are you making and why?

9.       Section 2.7 is useful context. Why isn’t this in the introduction? This evidences the case for looking at railways in this location.

10.   In section 2.8 you describe different types of ML. This is fine for justification of the ML approach you decide to you use however this would be better framed in the context of a literature review of methods to employ to undertake your analysis. Instead of textbooks for references I would expect reference to papers that have employed these methods to carry out similar analysis with success. This section can be shortened and reframed.

11.   Table three needs to consider elevation as you have applied this to the MCA as shown in the results.

12.   It is not clear how the layers of risk are combined. Are they weighted? If so why and how and if not why not?

Detailed Feedback

1.       Whilst there is extensive and good quality references, there are sections of the paper where you have not included references where I would expect to see them. This list is not exhaustive but please review:

a.       Section 1 lines 35-40: I would expect some references about climate change. Perhaps from the IPCC

b.       Section 1 para 2: needs references.

c.       Figure 1: You should provide references for the sources of the maps.

d.       Section 2.1 para 5: Distributed in text references throughout the paragraph in order to highlight which information is from which source.

e.       Section 2.2 para 6 the “rational method” should be referenced. Further you should signpost to the section where you discuss this further I think in section 2.6

f.        A reference on line 221 would be good.

g.       Section 2.5 para 1-4 should include references.

h.       Section 2.6 para 2 you need to support these arguments with references.

i.         Section 2.6 Para 3 you need to provide the reference to the Brazilian regulations.

j.         Figure 5 the referencing needs correcting.

k.       Additional reference to consider: Ferranti, E.J.S., Fontana Oberling, D. and David Quinn, A., 2022. Transport resilience to weather and climate: an interdisciplinary view from Rio de Janeiro. Proceedings of the Institution of Civil Engineers-Urban Design and Planning, 175(3), pp.103-121.

2.       The paper needs a thorough proofread as there are some sections which are quite difficult to read. In particular please review:

a.       Section 1 Para 1 – clumsy wording and some translation issues. For example it is not compulsory that changes are made, but it is critical, or they are changes that are urgently needed.

3.       There are some minor language changes to be made. I’ve made a note of some things that can be changed but this list is not exhaustive.

a.       Line 53: full stop.

b.       103: MCA should be defined here. Check all of your abbreviations are correctly introduced at their earliest mention. See line 495

c.       Line 180 phase should be stage for consistency

d.       Line 216: reword representative representation

e.       316, 323 and others. I’m not sure project is the word that you mean to use. Drainage system might be better? Or assets?

f.        Line 329 and 336 I don’t think you mean rail pavement. You might mean track and track bed. This is likely a translation issue.

g.       Line 376 repetition rational method.

h.       Line 398: is platform the word you mean? Do you mean track or railway/railroad?

i.         437: rain not rains may be better to use the word precipitation anyway.

j.         Be consistent, railway or railroad? Pick one and stick with it.

k.       Drainage assets not devices e.g. line 449

l.         Line 555 Eerrovia = Ferrovia

4.       You do not need to list all 38 municipalities. In section 2.1

5.       Figure 4 could be better quality. It is slightly blurred.

6.       Table 3 doesn’t need to be followed by so much text this can be reduced. Add references to understand where these criteria have come from. Both in the text and possibly in the table as well.

7.       Table 4 is difficult to understand without some kind o commentary for without having key information highlighted through formatting. I would suggest thus is moved to he appendix anyway.

8.       Suggest removal of the squares for data points in figure 9 for improved clarity. Further is it useful to overlay all of these to see the fit?

9.       The section on the Machine Learning models/Linear regression needs t be clear on what this means. This can be done without discussing the data.

10.   Generally your discussion needs to be more detailed. With in text references of the literature. What do you results mean? How does this compare to the findings of others reported in the literature? What are the limitations of the analysis? What does this mean for your results? How could this be improved in the future?

Comments on the Quality of English Language

Needs to be improved. There are the usual typos but also unclear phrases which I think are a translation issue. For example, rather than drainage devices drainage assets would probably be better in English. 

Author Response

We would like to sincerely thank the reviewer for their detailed and thoughtful analysis of our manuscript. The insightful comments and suggestions have significantly contributed to improving the quality and clarity of our work. We, the authors, deeply appreciate the time and dedicated effort to this review, and we are confident that the revisions made based on the feedback will enhance the overall strength of the article as we strive for its successful publication.

General Concept Comments:

1. Literature review: A comprehensive review of methodologies such as MCA and machine learning has been included. The literature review now provides a robust justification for the selected approaches, referencing relevant studies on risk assessment and infrastructure resilience.

2. Research aim and questions: The aim now precedes the hypothesis and research questions. The research questions are presented as a numbered list for clarity, as suggested.

3. Section structure: A summary paragraph outlining the paper’s structure has been added at the end of the introduction to improve flow and clarity.

4. Methodology clarification: The precipitation database and iterative nature of the review have been clarified. We also mentioned how future climate models can be integrated to improve accuracy.

5. Methodology development: The chosen methods are now more clearly justified with references to similar studies and literature. This aligns with point 1 by adding a stronger foundation for the selected approaches.

6. Link between figures: Figures 2, 3, and 4 have been revised to better explain how they relate to each other. Clearer links are made between the outputs of each stage of the research.

7. Otto Pfafstetter method: We acknowledge the limitations of the Otto Pfafstetter method and critique it in the context of modern meteorological patterns, explaining its relevance despite its age.

8. Drainage assumptions: Assumptions about ditch shapes and values for “R” are now explained in the methodology. The values used for the drainage system, sourced from Plúvio, are clearly documented.

9. Section 2.7 placement: This section has been moved to the introduction to better contextualize the importance of railways in Minas Gerais.

10. Machine Learning justification: Section 2.8 has been reframed to connect ML methods to literature rather than relying on textbooks, focusing on studies that have used similar techniques.

11. Elevation in Table 3: Elevation is now referenced in Table 3, and its relevance to the MCA is clarified in the results section.

12. Risk layers: The combination of risk layers has been explained more clearly in the methodology section, and the weights applied in the MCA are justified.

Detailed Feedback:

1. Missing references: The missing references for climate change, rational methods, and Brazilian regulations have been added in the specified sections. Figure 1 and Table 3 now have proper citations.

2. Proofreading: The entire manuscript has been thoroughly proofread to address clumsy wording and translation issues, particularly in Section 1. Phrasing like "drainage devices" has been replaced with "drainage assets."

3. Minor language issues:

   • MCA has been defined earlier in the text.
   • Words like "project" have been replaced with more appropriate terms like "assets" or "drainage system."
   • Inconsistencies like “railway” vs. “railroad” have been corrected for consistency.

Reviewer 2 Report

Comments and Suggestions for Authors

The article is of great relevance for flood prevention in infrastructure projects.

The case study, as well as fundamental concepts and hypotheses are explained in detail in the methodology.

The discussion of the results is clearly presented.

Figure 9 must be mentioned in the text.

Line 550: correct "Ferrovia Centro Atlantica"

Line 76: correct "rational method rational method"

Author Response

We would like to sincerely thank the reviewer for their detailed and thoughtful analysis of our manuscript. The insightful comments and suggestions have significantly contributed to improving the quality and clarity of our work. We deeply appreciate the time and effort to this review, and we are confident that the revisions made based on your feedback will enhance the overall strength of the article as we strive for its successful publication.

Figure 9 is now mentioned in the text.

Line 550 : "Ferrovia Centro Atlantica" is the name.

Line 76 was corrected "rational method"

Reviewer 3 Report

Comments and Suggestions for Authors

This paper comprehensively assesses flood risks to the railway network in Minas Gerais, Brazil, under extreme rainfall events by integrating Multi-Criteria Analysis (MCA) with machine learning techniques. The research team has collected and analyzed vital data such as topography, drainage capacity, land use, and land cover, employing GIS technology to pinpoint areas susceptible to flooding. Hydrological models were calibrated and validated to develop machine learning models for predicting flow rates and monitoring drainage systems. The study revealed that the Linear Regression model was the most effective in forecasting precipitation and flood risks, offering a precise and dependable tool for assessment. This research deepens the comprehension of the vulnerability of railway infrastructure amidst climate change and lays a scientific foundation for developing adaptive risk management and mitigation strategies. After a thorough review of the paper, the following suggestions are offered:

1. Please review the document for consistency, such as in Table 1, where there are two question marks at the end of the "Question" in the third row under "Third."；

2. In section 2.1 regarding the study area, please refine the text to highlight the aspects that are most relevant to the subjects of this paper, thereby improving the narrative flow；

3. In section 2.5, where the authors present "Multicriteria classes for flood susceptibility assessment," it is asked whether these criteria are universally applicable to all soil types. For instance, are sandy soils and clays considered differently, and if so, how；

4. Section 2.5 requests clarification on how the weights for different risk variables are assigned during the multi-criteria analysis. Are these weights derived from expert opinions, historical data, or an objective benchmark? Additionally, is there a discussion on the justification for the weight distribution；

5. In section 2.8, the authors mention a 7:2:1 ratio for the training, testing, and validation datasets. Inquiry is made as to whether there are specific standards for these ratios or how these proportions were determined within the context of this paper；

6. In section 3.2, the authors are asked to elucidate the definition of "Overfitting" to facilitate understanding for the readers.

Comments on the Quality of English Language

Minor editing of English language required.

Author Response

We would like to sincerely thank the reviewer for their detailed and thoughtful analysis of our manuscript. The insightful comments and suggestions have significantly contributed to improving the quality and clarity of our work. We deeply appreciate the time and effort to this review, and we are confident that the revisions made based on your feedback will enhance the overall strength of the article as we strive for its successful publication.

• Table 1 formatting: Thank you for noticing the inconsistency in Table 1. The two question marks in the third row have been corrected, and the table has been reviewed for overall consistency.

• Section 2.1 refinement: We appreciate your suggestion to improve the narrative flow in Section 2.1. The text has been refined to focus more specifically on the geographical and climatic aspects of the study area that are most relevant to the paper’s objectives, emphasizing factors such as topography, rainfall patterns, and their impact on flood susceptibility.

• Criteria applicability to soil types (Section 2.5): In response to your comment, we have clarified that the multi-criteria classes are designed to be flexible and adaptable to different soil types. However, we acknowledge that certain soils, such as sandy or clay soils, behave differently in terms of water retention and permeability. We have expanded the discussion to explain how these differences are accounted for within the model, with sandy soils generally contributing to higher drainage capacity, and clays contributing to higher flood susceptibility.

• Weights in the Multi-Criteria Analysis (Section 2.5): Thank you for raising this point. The weights for different risk variables were derived through a combination of expert judgment and historical data specific to the region. We have expanded the discussion to provide a justification for the weight distribution, which was based on previous studies in similar geographical settings and the relative importance of each variable in determining flood risk.

• Training, testing, and validation ratios (Section 2.8): The 7:2:1 ratio for the division of training, testing, and validation datasets was based on common practice in machine learning literature. This proportion balances the need for sufficient data in the training set while preserving enough data for meaningful testing and validation. We have added a brief discussion justifying this choice within the context of the study and referenced relevant standards and practices in the field.

• Overfitting definition (Section 3.2): We have added a clear definition of “overfitting” in Section 3.2 to improve readability for non-expert audiences. This includes a brief explanation that overfitting occurs when a model learns the training data too well, capturing noise rather than general patterns, leading to poor performance on unseen data.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Where additional paragraphs have been added these are well written and do address some of the recommendations I made to improve the paper. However, they are not nearly extensive enough. The paper requires a significant review. I am not satisfied that there is an appropriate review of literature communicated in the paper in its present form, or that the methods developed and applied are sufficiently communicated. I would like the authors to return to my original comments and consider them again. Please also consult other papers of a similar nature and consider their structure and content.

Comments on the Quality of English Language

Can still be improved I think not all the detailed comments have been addressed.

Author Response

Dear Reviewer,

We are attaching a document with a thorough and detailed review of the observations made during your initial review. We are confident that we have adequately addressed and/or justified our actions within the article.

Best regards,

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This paper studies rainfall simulation based on machine learning combined with actual engineering data, adopts machine learning to analyze rainfall events, and studies the efficiency of machine learning, which has certain theoretical significance. In future research, it can be considered to increase the application of the current new neural network architecture and supplement more actual data to reveal more application value of machine learning.

Author Response

Dear Reviewer,

We appreciate your valuable recommendation and have incorporated it for future research between lines 793 and 801 of the article.

Kind Regards