An Investigation into the Susceptibility to Landslides Using Integrated Learning and Bayesian Optimization: A Case Study of Xichang City
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThe authors proposed a paper titled “An investigation into the susceptibility to landslides using integrated learning and Bayesian optimization: a case study of Xichang city”, which revolves around the issues related to using integrated learning and Bayesian hyperparameter optimization to assess landslide susceptibility. The research topic is very important and the paper has an interesting findings that can help in order to solve the problem of assessing the susceptibility to landslides.
The following minor modifications are suggested to improve the quality of the paper:
- -To further enhance the interpretability of your modelsauthors are requested to use the SHAP (SHapley Additive exPlanations) method. SHAP provides detailed insights into the impact of each of the 13 influencing factors on the model's predictions.
- -A comprehensive review of related work on the same research topic should be conducted. By comparing the merits and demerits of these studies, the essential differences between this work and the existing ones should be highlighted to underscore its contributions.
Author Response
We have received your suggestions and thank you for your valuable comments on our manuscript.We have responded to each of them.Please see the attachment.
Author Response File: 
Author Response.docx
Reviewer 2 Report
Comments and Suggestions for AuthorsThe manuscript is useful and interesting on landslides susceptibility. Nevertheless, the paper requires to go through a careful revision before publication. I invite the authors to modify their manuscript according to the below comments.
The introduction effectively sets the context of the study area's geological and seismic history, but it could provide more background on the socio-economic impact of past landslides to emphasize the importance of the research. In addition, related landslide research are suggested (https://doi.org/10.1016/j.enggeo.2024.107563 ; https://doi.org/10.1016/j.gsf.2022.101493 ).
The section on data sources is detailed, but it could benefit from a discussion on the limitations of the data, such as potential inaccuracies or gaps in the landslide distribution data.
The research methods section provides a good overview of the integrated learning principles. However, it could be improved by comparing and contrasting the chosen algorithms (XGBoost, LightGBM, and RF) with other potential methods not selected, to justify the choice of these specific models.
The explanation of the Random Forest algorithm is clear. However, the text could benefit from a more detailed explanation of how the bagging integration concept specifically applies to landslide susceptibility assessment.
The description of the XGBoost algorithm mentions its higher computational efficiency. It would be helpful to include empirical evidence or references to support this claim.
Bayesian Optimization Algorithm:
The Bayesian optimization algorithm is well-explained, but the text could further clarify how this method addresses the limitations of traditional population optimization algorithms.
Performance Evaluation of the Models:
The performance evaluation section is thorough. However, it could discuss the potential biases introduced by the choice of performance metrics and how these were mitigated.
The results section is comprehensive, but it could benefit from a more detailed discussion on the implications of the findings for the local community and disaster management strategies.
The text should provide a clearer explanation of the significance of the AUC value and how it relates to the model's predictive power.
The correlation analysis is a strong aspect of the study. However, the text should discuss the potential for multicollinearity among the factors and how this was addressed in the analysis.
The susceptibility mapping is well-presented, but the text should include a discussion on how the model could be validated against actual landslide events that occurred after the study period.
Importance of Evaluation Factors:
The importance analysis is a valuable contribution, but the text should discuss the potential for overfitting when using the BO-XGBoost model and how this was addressed.
The discussion is insightful, but it could benefit from a broader perspective, including a comparison with landslide susceptibility studies in other regions with similar geological and seismic conditions.
The text should discuss the limitations of the study in more detail, including the potential for the model to be affected by unconsidered variables or changes in the environment over time.
The conclusion effectively summarizes the study's findings but should also include recommendations for future research directions, such as the integration of real-time data and the exploration of other machine learning techniques.
Author Response
We have received your suggestions and thank you for your valuable comments on our manuscript.We have responded to each of them.Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report
Comments and Suggestions for AuthorsThis manuscript explores the application of ensemble learning and Bayesian Optimization (BO) for landslide susceptibility assessment in the Xichang area. The study demonstrates that the BO-XGBoost model achieves an AUC value of 0.8677 in the study area, exhibiting better generalization ability and higher prediction accuracy compared to BO-LightGBM and BO-RF models. Overall, the research presents good scientific merit and practical value. However, I recommend the authors address the following points to further enhance the manuscript:
The concluding statements in the abstract, such as "The study demonstrates that the AUC value of the BO-XGBoost model in the study area is 0.8677, which has a better generalisation ability and higher prediction accuracy than the BO-LightGBM and BO-RF models," require more robust justification. Please elaborate on the reasons behind the superior performance of the BO-XGBoost model to highlight the novelty of the study.
The second paragraph of the introduction would benefit from a more comprehensive review of the applications of machine learning and Bayesian optimization in the field of geohazard assessment. Please discuss relevant studies and connect them to the present research.
Given the significant role of faults in landslide occurrences, I suggest providing a detailed explanation of the essential elements of faults within the geological setting section.
Please ensure the accurate usage of geographical terms. For instance, consider whether terms like "Luzhun Mountain," "Zhongshan Mountains," and "Yak Mountains" should be simply "Luzhun Mountain," "Zhongshan Mountains," and "Yak Mountains," respectively. Carefully check all specialized vocabulary, particularly those related to landslides.
Consider replacing "Data name" with "Data." Thoroughly review other data-related terminology for accuracy.
The manuscript appears to lack primary geological survey data, with the results primarily based on the analysis of geological maps. Please address this limitation and discuss its potential implications on the findings.
The discussion should include validation results for the predictions made using the two methods employed in the study.
Please provide information on the sources and reliability of the influencing factors considered in the study, including NDVI, distance from the fault, slope, and distance from the river.
Author Response
Reviewer：We have received your suggestions and thank you for your valuable comments on our manuscript.We have responded to each of them.Please see the attachment.
Author Response File: Author Response.docx
Reviewer 4 Report
Comments and Suggestions for AuthorsThis manuscript proposes a method for evaluating landslide susceptibility using integrated learning and Bayesian hyperparameter optimization. The results show that after removing the terrain undulation factor, the four most significant factors related to landslide susceptibility are NDVI, fault distance, slope, and river distance. Additionally, the BO-XGBoost model demonstrates better generalization ability and higher prediction accuracy compared to the BO-LightGBM and BO-RF models. These findings are significant for further improving landslide susceptibility evaluation. Overall, the reviewer believes the manuscript could be published in the Sustainability Journal after major revision. Specific comments are provided below:
1. It is suggested that the authors rewrite certain sentences to enhance readability. For instance, in the abstract, the phrase "the articulation part between the Anning River Fault and the Zemu River Fault" could be simplified for better understanding.
2. The first paragraph of the introduction primarily discusses the impact of landslides triggered by strong earthquakes. It is recommended that the author include recent data on landslide caused by other natural factors in Xichang City in Xichang City to better highlight the importance of landslide susceptibility assessment.
3. While traditional population optimization algorithms are prone to local optima, Bayesian optimization algorithms also face this issue. The author is requested to explain why Bayesian optimization was chosen and to discuss its advantages. Additionally, further investigation into strategies to avoid local optima when using Bayesian optimization algorithms is recommended.
4. It is suggested that the authors add annotations to the figures for improved clarity. For example, in Figure 1, arrows or other markers could be included to more clearly indicate the geographical location of the study area.
5. The text size in some images is relatively small. It is recommended to increase the font size to improve readability, particularly in Figures 3 and 4.
6. Boosting and bagging methods can increase model complexity and consume significant computational resources. The author should address how these issues are managed in the study.
7. The titles in sections 3.1.1 and 3.1.2 are duplicated. The author should verify and correct these issues.
8. It is suggested that the author refine the content of the article, such as simplifying and optimizing the evaluation indicators in section 3.2.6 to enhance clarity.
9. In section 4.1, it is recommended that the author remove several unnecessary evaluation factors and provide a brief description of the basic principles of the remaining factors to improve the article’s clarity and eliminate redundant content.
10. References 12 and 16 are duplicated. The author should verify and correct these issues.
11.The references in this paper are relatively few and do not adequately reflect the latest progress and developments in related fields at home and abroad. It is suggested to add the following references to reflect the timeliness and depth of the paper.
https://doi.org/10.1016/j.asoc.2023.110324 https://doi.org/10.1007/s11629-023-8484-9Comments on the Quality of English Language
Moderate editing of English language required.
Author Response
Reviever：We have received your suggestions and thank you for your valuable comments on our manuscript.We have responded to each of them.Please see the attachment.
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
Comments and Suggestions for AuthorsThe authors have modified them. Pay attention to the relation landslide research. (e.g., https://doi.org/10.1016/j.enggeo.2024.107563 ; https://doi.org/10.1016/j.gsf.2022.101493)
Author Response
Dear Reviewer：We have received your suggestions The research you mentioned makes us pay attention to the relationship slide. We have actively absorbed your suggestions and read the two literatures you gave. In the future, our team will pay attention to and attach importance to the research related to landslide in this direction. Thank you again for your valuable comments on our article.
best regards to you
Reviewer 3 Report
Comments and Suggestions for AuthorsThe author's response does not address some of my concerns, so I recommend rejecting the paper!
Author Response
Reviewer：We have received your suggestions and thank you again for your valuable comments on our manuscript. We have again responded to your questions raised in point 6 and 8 of the first review comments. Please see the attachment.
Author Response File: 
Author Response.pdf
Round 3
Reviewer 3 Report
Comments and Suggestions for AuthorsI think this paper can be accepted for publication now.
