Flood Susceptibility Assessment in Urban Areas via Deep Neural Network Approach

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This manuscript (sustainability-3113937) aims to proposes a methodology for pluvial flood assessment in urban areas using a multiclass classification approach with a Deep Neural Network (DNN) optimized through hyperparameter tuning with Genetic Algorithms (GA) leveraging remote sensing data of comprehensive pluvial flood dataset for the Ibadan metropolis, Nigeria and Metro Manila, Philippines. Generally speaking, it is an interesting and promising research. I have some suggestions:

- . It is uncommon that some references have been cited in the Abstract, please move them to the main text.

- . In the Introduction Section, the authors should clearly and specifically state the research question and identify the objectives, hypotheses, and expected outcomes of the study.

- . What is the reason for the selected model (such as Deep Neural Network combined with Genetic Algorithms)? Is there a good reason to support this choice?

- . This manuscript did not describe in detail the specific sources, coverage, time span, spatial resolution and other key information of the remote sensing data used, which is essential for evaluating the model.

- . The manuscript lacks a detailed description of the data preprocessing steps, such as data cleaning, normalization, missing value processing.

- . The specific implementation details of the deep neural network (DNN) and genetic algorithm (GA) used in the study, such as network structure, parameter settings, data set selection and preprocessing, were not described in detail.

- . The manuscript did not clearly state the model validation methods, such as whether cross-validation, independent test set validation, were used, and what are the results of these validations.

- . Although deep learning models are highly accurate, it is often difficult to explain their inner workings. Does this affect the interpretation and application of the results?

- . The discussion section briefly explains the results, without analysis of the reasons behind the model's performance, and without mentioning possible limitations, such as how the model performs in a specific region or under specific conditions, or whether there is an overfitting problem.

- . In future studies, more relevant influencing factors behind urban floods need to be taken into account, such as the urban vertical patterns. Please refer to below.

Assessing the scale effect of urban vertical patterns on urban waterlogging: an empirical study in Shenzhen, 2024, 107486

The Resilience of the Built Environment to Flooding: The Case of Alappuzha District in the South Indian State of Kerala. Sustainability. 2024

- . Based on the results obtained, the authors need to put forward suggestions and policy implications for urban flood prevention and control.

- . The discussion section should analyze the experimental results in depth, explore the advantages and disadvantages of the model, as well as possible application scenarios and future research directions. The author should discuss the potential and limitations of the model in practical applications based on the experimental results.

Author Response

Dear Reviewer,

Thank you very much for your review of my work. Your comments and suggestions have been extremely helpful and will significantly improve the quality of the research. I greatly appreciate your time and contribution to the development of my work. Please find attached file with point-by-point ansewrs on your suggestions.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This study tried to apply DNN+GA to flood assessment in urban areas. But this approach is not innovative anymore. Experiment design is not appropriate. Derived conclusions do not have a strong basis. Detailed suggestions are listed below. 

1.        Figure 1: The colors for the risk levels do not match common perception. Blue and Green typically represent low risks.

2.        Figure 2: What is the source of flood susceptibility used in this figure?

3.        Figure 5: The caption is invisible.

4.        Two cases in this study cannot be compared. The flood map in Figure 1 exhibits a dendric pattern like river networks. It looks like the results of fluvial floods rather than pluvial floods. The second case in Manila has a more reasonable spatial pattern on historical flood events.

5.        Does the number of classes in the target variable impact the model performance? What if the flood height in the Manila dataset was reclassified into five classes?

6.        Conclusion: Which part of the analysis supports this point?

Author Response

Dear Reviewer,

I am sincerely grateful for your thorough review of my work. Your insightful comments and suggestions have been immensely valuable and will greatly enhance the quality of the research. I deeply appreciate the time and effort you invested in providing such detailed feedback, and I am grateful for your contribution to the development of my work. Please find attached file with point-by-point ansewrs on your suggestions.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Overall, the research presented in this paper well-composed. The authors applied deep neural network-based approach for assessment of pluvial floods in two case studies located in urban regions of Ibadan, Nigeria and Mania, Philippines. A major revision is recommended at this stage for improvement of paper.

Major Comments

 

1.      The authors have summarized previous studies one-by-one in line no. 31 – 114 and in Table 4, however, the authors have not critically analyzed previous studies in order to discuss research gaps that existed in previous studies, the significance of this research, and your main contribution/novelty for the advancement of research in this particular domain. The introduction section needs to be extensively revised to make it more coherent, logical and comprehensive.

2.      Line 182-185: The authors must discuss quantitative details of DL methods (their structure) and parameters adopted for optimization of hyperparameters. How was the genetic algorithm connected with the DNN algorithm? What was the range of various parameters used in the genetic algorithm, initial population, cross-over rate, mutation rate, iterations, etc., and their defining criteria? Figure 5 can be improved with a better representation of the architecture of the DNN model.

3.      Figure 6: The accuracy of the Manila Floods is far less than Ibadan Floods, which is a maximum of around 0.4. Are the DNN predictions for the Manila case reliable for further applications? Please discuss your understanding of the results with logical reasoning. Does the quantity and quality of data have implications on model training and further implications?

4.      Please discuss significance of DNN-based flood assessment in comparison to physics-based hydraulic modeling. Why researchers, policy-makers and other stakeholders should consider DNN-based flood assessment compared to physics-based hydraulic modeling, considering the latter case provide realistic flood results which can be calibrated and validated, and reliable for further applications?

5.      The current contribution/results are insufficient; therefore, the authors are suggested to do more analysis/evaluation of different scenarios and discuss their results in the paper.

 

 

Minor Comments

1.      Line no. 74: The full names of “RMSE, NSCE and MAPE” were not introduced in their first appearance, before the usage of these abbreviations, therefore, the authors are suggested to check the complete manuscript and define abbreviations in their first appearance before using them in short forms.

2.      Figure 1 and Figure 3: The authors are suggested to represent a map of the country including representation of study area on the map.

 

3.      Please discuss the computation time for each studied case of each DNN model.

Comments on the Quality of English Language

The English language of the manuscript needs improvement.

Author Response

Dear Reviewer,

I am sincerely grateful for your thoughtful review of my work. Your insightful comments and valuable suggestions have been instrumental in enhancing the quality of my research. I deeply appreciate the time and effort you invested in providing such a thorough evaluation. Please find attached file with point-by-point ansewrs on your suggestions.

Thank you once again for your contribution

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript is publishable in this journal.

Author Response

Dear Reviewer1,

Thank you for carefully considering our manuscript and providing valuable comments. We greatly appreciate the time and effort you have invested in evaluating our work.

best wishes, Ivan.

Reviewer 3 Report

Comments and Suggestions for Authors

No Comments

Author Response

Dear Reviewer3,

Thank you for carefully considering our manuscript and providing valuable comments. We greatly appreciate the time and effort you have invested in evaluating our work.

best wishes, Ivan.