Review Reports - Artificial Intelligence-Based Regional Flood Frequency Analysis Methods: A Scoping Review

Round 1

Reviewer 1 Report

The manuscript presents a scoping review on AI-based RFFA methods. The topic is very interesting and in the scope of Water, however the manuscript needs significant changes prior to considered for publication. Following is the summary of my comments (For detailed comments -- refer to the annotated pdf file).

1. The overall language of the manuscript is below the academic standards and needs editing. I would suggest this article to be proofread by a native language editor towards improving its readability.

2. Abstract needs to be reformulated. (specific comments in the pdf)

3. Methodology section is poorly written and structured. More comprehensive details of the methodology needs to be provided. Suggestions provided in the annotated pdf.

4. Overall structure of the manuscript needs to be improved. Suggestions provided in the annotated file.

5. Authors need to define some subjective criteria to evaluate the literature. Since each study was performed over different dataset, they can not be compared using the accuracy reported in each. An approach trained using quality data from real world floods reporting 80% accurate is far better in practice in comparison to another approach trained over very few samples of lab simulated data reporting 99% accuracy. Some generalized comparative measures are must to criticise the literature.

6. Additional compact section of Discussions of the bibliometric analysis is needed.

7. Additional detailed section on the challenges and future research directions is needed.

Comments for author File: Comments.pdf

Author Response

The overall language of the manuscript is below the academic standards and needs editing. I would suggest this article to be proofread by a native language editor towards improving its readability.

Authors’ response: The manuscript is thoroughly edited by a native English Speaker.

Abstract needs to be reformulated. (specific comments in the pdf)

Annotation 1: Correct the first sentence of the abstract:

Authors’ response: The first sentence is rewritten as below: “Flood is one of the most destructive natural disasters, which causes significant economic damage and loss of lives.” The motivation of the study has been incorporated. The abstract has been upgraded significantly. Writing has been improved.

Annotation 2: Always define the abbreviation for the first instance.

Authors’ response: All the abbreviated words are fully spelt out at their first appearances. Here, ANN is fully defined as artificial neural networks.

Annotation 3: Reformulate keywords.

Authors’ response: The keywords have been reformulated as suggested by you. The new keywords are: Regional flood frequency analysis;, Artificial neural networks, Flood; Artificial intelligence

In the Introduction: Paragraph 1, last line, add a reference.

Authors’ response: The following reference is added: Rajkhowa S, Sarma J. Climate change and flood risk, global climate change. InGlobal Climate Change 2021 Jan 1 (pp. 321-339). Elsevier.

Annotation 4: Provide reference and examples of data driven models.

Authors’ response: The following two references are added and examples of the data-driven methods are also added.

Sahu, R. T., Verma, M. K., & Ahmad, I. (2021). Regional frequency analysis using L-moment methodology—a review. Recent Trends in Civil Engineering, 811-832.

Rahman, A., Haddad, K., Zaman, M., Kuczera, G. and Weinmann, P.E. (2011). Design flood estimation in ungauged catchments: A comparison between the Probabilistic Rational Method and Quantile Regression Technique for NSW. Australian Journal of Water Resources, 14, 2, 127-137.

“On the other hand, data-driven models have been quite popular for flood estimation in recent years [Sahu et al., 2021C). Examples include quantile regression technique and probabilistic rational method (Rahman et al., 2011)”

Annotation 5: Remove one of SVM and SVR.

Authors’ response: SVM is removed.

Annotation 6: Cross validation issue

Authors’ response: The sentences are revised and a new reference is added.

“If adequate data exits, it is often possible to build, test, and evaluate an AI based model (similar to many other RFFA models) by dividing the data into training, test, and evaluation data sub-sets [67, 68]. Cross-validation is also often used in building RFFA models when less data samples are available [Jung et al., 2021].”

Annotation 7: Data used in modelling

Authors’ response: The sentence is revised as below. “However, it should be noted that data quality is of significant importance in developing and testing accurate models.”

Annotation 8: Why this scoping review is needed?

Authors’ response: A full justification is provided in the revised paper (last paragraph of the Introduction section).

Methodology section is poorly written and structured. More comprehensive details of the methodology needs to be provided. Suggestions provided in the annotated pdf.

Authors’ response: Many thanks for your suggestion. The methodology section is upgraded significantly as suggested by you.

Annotation 10: Why WOS and IEEE not considered?

Authors’ response: It was believed use of Scopus and Google Scholar would capture relevant articles on AI-based RFFA studies.

Annotation 11: Redraw figure 1 using PRISMA.

Authors’ response: Figure 1 has been updated.

Annotation 12: It seems like authors are taking only the regression methods as AI.

Authors’ response: We have considered all the AI-based RFFA studies (including both classification and estimation).

Overall structure of the manuscript needs to be improved. Suggestions provided in the annotated file.

Authors’ response: The structure of the manuscript is improved as suggested.

Authors need to define some subjective criteria to evaluate the literature. Since each study was performed over different dataset, they can not be compared using the accuracy reported in each. An approach trained using quality data from real world floods reporting 80% accurate is far better in practice in comparison to another approach trained over very few samples of lab simulated data reporting 99% accuracy. Some generalized comparative measures are must to criticise the literature.

Authors’ response: A comment is made on this aspect (in the last paragraph of methodology) as below: It should be noted that the relative accuracy of a RFFA study/method cannot be compared with another study/method directly using the reported error statistics, since in most cases they use different dataset to develop and test the methods. For example, an approach trained using quality data from real world floods reporting 80% accurate is far better in practice in comparison to another approach trained over very few samples of lab simulated data reporting 99% accuracy. Hence, the comparison of AI-based RFFA methods made in the following section is taken to be as a guide only. In real world ap-plications, several methods should be applied and compared to select a preferred method for design flood estimation. One method found to be better in one geographic region does not guarantee its superiority at another location.

Additional compact section of Discussions of the bibliometric analysis is needed.

Authors’ response: The discussion of the bibliometric analysis has been expanded.

Additional detailed section on the challenges and future research directions is needed.

Authors’ response: A new section on this has been added.

Annotation 13: Need a major structural change

Authors’ response: The manuscript has been restructured as suggested by you.

Annotation 14: Dataset, sample size etc to be added in Table 1

Authors’ response: Predictor variables used are in Column 4, Number of catchments used and years of streamflow data are added in Column 6.

Annotation 15: This information to be moved to new subsections 3.1, 3.2, …

Authors’ response: This is done.

Annotation 16: Move bibliometric analysis after methodology

Authors’ response: We prefer this to be here (as it is).

Annotation 17: Discuss more on the images

Authors’ response: The figures are discussed as suggested.

Annotation 18: Add a section on bibliometric analysis outcome.

Authors’ response: We have enhanced bibliometric analysis section. We believe a separate section is not needed.

Annotation 19: Add a section on Challenges and Future Direction

Authors’ response: Thanks. This new section is added.

Many thanks. Your comments have assisted a lot to improve the manuscript.

Reviewer 2 Report

This is well done and will be useful to anyone in the water control field. The bibliography is extensive and will be useful. The authors should be aware that the USEPA has done a similar program to do surveys of the literature on particular projects.

Author Response

Authors’ response: Many thanks for your positive comments.

We could not find the USEPA paper on AI-RFFA review. However, we found a paper by ASCE, which we have added.

“In 2000, American Society of Civil Engineers (ASCE) published a review paper on the applicability of ANN to hydrology, which however did not focus on RFFA [ASCE, 2000).”

The following reference is added in the list:

ASCE Task Committee on Application of Artificial Neural Networks in Hydrology. (2000). Artificial neural networks in hydrology. II: Hydrologic applications. Journal of Hydrologic Engineering, 5(2), 124-137.

Reviewer 3 Report

A good and comprehensive paper with only some minor suggestions for improvement.

Table 2 is incomplete, not all variables are listed.

The reference Haddad and Rahman (2012) is not in the comment reference format.

I think that more explanation of figure 5 is required. It is not clear how countries are positioned, or what the relative sizing means.

Author Response

Reviewer’s comment:

A good and comprehensive paper with only some minor suggestions for improvement.

Authors’ response: Many thanks for your positive comment.

Reviewer’s comment: Table 2 is incomplete, not all variables are listed.

Authors’ response: Table 2 is now upgraded.

Reviewer’s comment: The reference Haddad and Rahman (2012) is not in the comment reference format.

Authors’ response: The reference of this paper is corrected using Water Format:

Haddad, K.; Rahman, A. Regional Flood Frequency Analysis in Eastern Australia: Bayesian GLS Regression-Based Methods within Fixed Region and ROI Framework–Quantile Regression vs. Parameter Regression Technique. J. Hydrol. 2012, 430, 142–161.

Reviewer’s comment:

I think that more explanation of figure 5 is required. It is not clear how countries are positioned, or what the relative sizing means.

Authors’ response: Thanks for the comment. The relevant part (Page 6, paragraph 1) of the revised paper is significantly enhanced.

Round 2

Reviewer 1 Report

Authors have considered most of my comments from round 1 review. The manuscript looks far better and improved now.