A Residual Neural Network Integrated with a Hydrological Model for Global Flood Susceptibility Mapping Based on Remote Sensing Datasets

Round 1

Reviewer 1 Report

This manuscript shows a robust way to apply ML for flood mapping. It is interesting to analyze and evaluate the model performance thoroughly. Although some deficiencies regarding the flood application. The manuscript is prepared well in structural and English writing, so minor revision is necessary.

 

Minor comments:

Explaining every parameter/variable of equations in the method is fundamental. Therefore, please supplement them for formulas 1 and 2.

In Figure 2, why are the ROC curves almost overlapping regarding different return periods?

In Discussion 5.1, what is the “original training data”? Does the result show in any figure or table? For example, from which evidence pointed out that “All pre-trained models had better performance in terms of AUC and accuracy. ” compared to the original one?

Although 50 years of RP showed the most significant improvement, in hydrological and flood studies, 100 and 200 years of RP are highly related to flood occurrence. If so, regardless of the performance of the ML method, why not show an analysis of RP100 or RP200 to contribute better to flood?

The manuscript is prepared well in structural and English writing.

Author Response

Dear Reviewer,

Thank you for your valuable feedback on our manuscript. We appreciate your comments and suggestions, and we have made the following revisions to address your concerns:

1. We have added an explanation of every parameter/variable of Equations 1 and 2 in the methods section.

2. The ROC curves in Figure 2 appear to overlap because the model's performance is not highly sensitive to the return periods.

3. We have revised Discussion 5.1 to clarify the meaning of "original training data", and Table 3 shows the evidence.

4. We appreciate your comment regarding the RP100 or RP200. We have added an analysis of RP100 and RP200 in discussion 5.3 to contribute better to flood studies.

We hope that our revisions have addressed your concerns, and we thank you again for your helpful comments. Please let us know if you have any further feedback or suggestions.

Best regards,

Junfei Liu

Reviewer 2 Report

This is an interesting paper, quite well design and fine in terms of the contents, that shows how a ResNet integrated with hydrological model can represents a key tool for flood susceptibility prediction. However, I suggest some additions/modifications below (other modifications can be found in the file attached).

lines 28-32. There are several studies in the literature addressing different models for the flooding hazard assessments, see e.g., 10.1109/JSTARS.2013.2284607; https://doi.org/10.3390/ijgi9120720. It is worth add further details regarding the flood susceptibility mapping and on its value for disaster risk reduction/mitigation, and references to make the introductory framework more complete.

lines 47-52. There are recent studies demonstrating precisely the performance of different ML models for flood detection, e.g., https://doi.org/10.3390/s18010018; https://doi.org/10.3390/drones7020070 

lines 100-102. The sediment transport, together with the roughness coefficient, represents a key aspect that affects the flooding harzard. There are several studies addressing this point also through sensitivity analyses, it is worth add more references to validate this statement. see e.g., https://doi.org/10.1016/j.scitotenv.2022.156736

line 103. In Eq. 1 and Eq. 2 add the meaning of acronyms.

lines 363-364. Referring also to the detailed comparison reported in paragraph 5.1 concerning the different models performance, why the model pretrained with RPs of 50 years showed better accuracy with respect to the others? It is worth trying to investigate the causes of this result. 

 

Comments for author File: Comments.pdf

Author Response

Dear Reviewer,

Thank you for your thorough review of our manuscript. We appreciate your valuable feedback and suggestions, which we have carefully considered in our revisions. Here are our responses to your specific comments:

1. We agree with your comment that further details regarding the flood susceptibility mapping and its value for disaster risk reduction/mitigation should be added to the introduction. We have added more information and references to make the introductory framework more complete.

2. We appreciate your suggestion to add recent studies demonstrating the performance of different ML models for flood detection. We have added these references to the manuscript to enhance the discussion.

3. We agree that sediment transport, along with the roughness coefficient, represents a key aspect that affects the flooding hazard. We have added more references to validate this statement and provide more context.

4. We have revised Eq. 1 and Eq. 2 to include the meaning of the acronyms.

5. We appreciate your comment regarding the better performance of the model pretrained with RPs of 50 years. We have investigated the causes of this result and have included our findings in discussion 5.4.

Once again, we appreciate your comments and suggestions, and we hope that our revisions have addressed your concerns. Please let us know if you have any further feedback or suggestions.

Best regards,

Junfei Liu

Reviewer 3 Report

Dear authors,

A Residual Neural Network-Integrated Hydrological Model for Global Flood Susceptibility Mapping Based on Remote Sensing Datasets

This paper proposes a novel approach for flood susceptibility prediction. The authors integrated a ResNet-18 with a 2D hydrological model for global flood susceptibility mapping using remote sensing datasets. The authors showed which model performance is improved through physics-based initialization and, the pretrained model achieves better performance with incomplete training data. The paper deals with an important topic that can be used to improve public policies and providing valuable insights how to improve identification floods and flood susceptibility mapping. I suggest a minor revision.

 

 

1. The abstract needs to clarify the purpose of this paper.

 

2. The introduction needs more details of the important this study.

 

3. The introduction also needs to clarify the importance of this study, justifying the reason for this research

 

4. There is no connection between the paragraphs in introduction.

 

5. The objectives are not well described.

 

6. In general, the results are well presented and demonstrate the importance of the study. However, a better discussion with the factors that contribute these results.

 

7. You need to conclusion better your work. And it does not answer all the proposed objectives. It needs to be improved.

 

Author Response

Dear Reviewer,

Thank you for taking the time to review our manuscript. We appreciate your comments and suggestions, and we have taken them into account in our revisions. Here are our responses to your specific comments:

1. We agree with your comment that the abstract needs to clarify the purpose of this paper. We have revised the abstract to better reflect the purpose of the paper.

2. We appreciate your comment regarding the introduction needing more details about the importance of this study. We have added more information to the introduction to better explain the relevance and importance of the study.

3. We have revised the introduction to clarify the importance of the study and justify the reason for conducting this research.

4. We understand your comment regarding the lack of connection between the paragraphs in the introduction. We have revised the introduction to better connect the paragraphs and provide a more coherent flow of information.

5. We have carefully reviewed the objectives of the study and have rephrased and added more details to better describe them.

6. We appreciate your feedback on the presentation of the results. We have added more discussion to our manuscript to better explain the factors that contribute to the results.

7. We agree that the conclusion needs to be improved to better summarize the work and answer all the proposed objectives. We have revised the conclusion to better reflect the contributions of the study and to address all of the proposed objectives.

Once again, we appreciate your comments and suggestions, and we hope that our revisions address your concerns. Please let us know if you have any further feedback or suggestions.

Best regards,

Junfei Liu

Reviewer 4 Report

Dear Authors,

The presented research is quite interesting. The authors investigated the prediction of flood susceptibility by integrating ResNet-18 with a 2D hydrological model for global flood susceptibility mapping based on remote sensing datasets. Main contributions included the integration of hydrological simulation and deep learning, the pretrained model better performance achievement with incomplete training data, the improvement of model performance through physics-based initialization. Generally, the manuscript is well written and structured. I think it will be of interest to international readers.

I have only minor comments as follows:

-        Line, 7: Please change “1School” to “School”.

-        Line 97: Please describe the Table (Table 1) before its first appearance in the text.

-        Lines 101-104: the components of Equations 1 and two need to be described in the text

 

English language can be improved.

Author Response

Dear Reviewer,

Thank you very much for reviewing our manuscript. We appreciate your positive comments on the research and are glad to know that you found the manuscript well-written and structured.

We thank you for your valuable feedback, and we have addressed your minor comments as follows:

• We have corrected the error on line 7 by changing "1School" to "School".
• We have described Table 1 before its first appearance in the text as per your suggestion on line 97.
• We have also added descriptions of the components of Equations 1 and 2 in the text as per your comment on lines 101-104.

Once again, thank you for your time and efforts in reviewing our manuscript, and we hope that our revisions meet your expectations. Please let us know if there are any further suggestions or comments.

Best regards,

Junfei Liu

Round 2

Reviewer 2 Report

I appreciate the effort of the authors for addressing to my suggestions. I think that now the paper is more clear and complete, and it is ready to be published.