Landslide Susceptibility Mapping Using Machine Learning: A Literature Survey

Round 1

Reviewer 1 Report

Dear Authors,

This manuscript illustrated landslide susceptibility mapping using Machine Learning. This article presents a literature review machine learning algorithms, detailed analysis and in-depth discussion. Overall, this study could provide valuable guidance for mitigating and managing landslides. In my point of view, this manuscript only need some minor clarifications to be published.

 

-          Line 19: Why use the short form of landslide? It is only one word, so it is not necessary to use abbreviation.

-          Line 36: The authors write that many literature … however, use only one reference.

-          Line 45: what are the ML models?

-          Lines 46-49: Whose study? Merghadi et al., 2020? If yes, please avoid writing short sentences combine the related sentences or use appropriate reference for the sentence.

-          Table 1: You indicated the references from 2018, so there is no literature review exist on landslide susceptibility mapping using machine learning before 2018? There is lots of papers! Why you only focus on review papers?

-          The figures must be inserted immediately after the first time that authors mention it, however, you inserted the figures at first and then write about the figure. That’s not right.

-          Also same as the previous comment please consider it for tables. Please revise all the figures and tables.

    Best regards,    

Author Response

Dear Reviewer,

Thank you for your insightful remarks. We have incorporated all the suggested changes in the revised manuscript. Please see the attachment for the detailed response to each point.

Author Response File: Author Response.pdf

Reviewer 2 Report

It is not a topic for Remote Sensing journal.

Author Response

Dear reviewer,

Please see the attachment for the response.

Author Response File: Author Response.pdf

Reviewer 3 Report

This manuscript presents a survey on the landslide susceptibility mapping using machine learning. Until recently, many literature surveys and reviews on the topic of landslide susceptibility mapping are available, such as ‘Deep learning for geological hazards analysis: Data, models, applications, and opportunities’. Some other related reviews were also listed in Table 1. Thus, the novelty or contribution of this manuscript needs to be better highlighted throughout the main text. Besides, it is suggested to focus more on the implications of this review manuscript for geotechnical researchers and practitioners in the landslide susceptibility mapping.  

This manuscript presented an interesting study on characterizing the commonly used smart soil-moisture sensors by integrating ensemble learning techniques. In general, the manuscript is well-written and revision is suggested. Some technical issues can be addressed or clarified to improve the quality of the manuscript:

(1) The introduction containing 8 paragraphs tends to be relatively fragmented, which poses difficulties in capturing the necessity of this research. It is advisable to reorganize this section and highlight the contribution of this study compared with previous studies.

(2) Subsection 3.2 compared the predictive performance of the four machine learning algorithms, namely XGB, LGBM, RF, and extra trees. Whether the results presented in Table 3 calculated from the training dataset or testing dataset. Both the results on the training dataset and testing dataset are suggested to be provided.

(3) The detailed descriptions of the four machine learning algorithms were omitted in the manuscript probably for brevity. To improve the readability of these algorithms (e.g., RF and XGB), it is advisable to provide some related references, and thus interesting readers can refer to the cited references, such as ‘Efficient reliability analysis of earth dam slope stability using extreme gradient boosting method’ and ‘Slope stability prediction using ensemble learning techniques: A case study in Yunyang County, Chongqing, China’.

(4) The machine learning algorithms generally contain several hyperparameters, and several hyperparameter optimization techniques have been developed, such as the Bayesian optimization technique in ‘Prediction of undrained shear strength using extreme gradient boosting and random forest based on Bayesian optimization’. Please specify which optimization technique was used in this study.

(5) Table 3 compared the predictive performance of different modesl with different features. It can be observed that extra tree model performed the best among them, please explain the underlying reasons.

(6) A total of 11 features were used in the construction of extra tree model (i.e., the last row of the last column in Table 3). In machine learning, feature important analysis is frequently applied to explore the relative contribution of each input to the output, such as. It is suggested to add the feature importance analysis result, further helping engineers focus on the more influential parameters in practical applications.

 

(7) The section ‘5. Conclusions’ is relatively lengthy, it is suggested to focus more on the implications of this study for geotechnical researchers and practitioners in the soil moisture measurement.

Comments for author File: Comments.pdf

Author Response

Dear reviewer,

Please see the attachment for the response.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

It is not a topic for Remote Sensing journal.

Author Response

Point 1: It is not a topic for Remote Sensing journal.

Dear Reviewer, thank you for your remark. 
Response 1:  The paper introduces the essential elements in landslide susceptibility studies, including landslide causative factors, datasets, landslide-inventory maps, machine learning models, and evaluation methods. We have discussed state-of-the-art and recently proposed conventional, hybrid, ensemble, and deep learning models. Based on the literature survey, a few recommendations and future works which may help the new researchers in the field are also presented.

Our paper covers the aspects of environmental sciences, ecology and civil engineering, change detection, and remote sensing applications mentioned in the Aims and Scope of the Remote Sensing journal.

We sincerely hope our clarification can bring a positive assessment of our paper.

Thank you.

Reviewer 3 Report

Thanks to the contribution of all the authors, the manuscript has been carefully revised according to the reviewer comments. In general, it can be accepted for publication after minor corrections. Specifically, the section 5 summarized the machine learning techniques used in landslide susceptibility mapping, including RF, SVM, LGR, and ANN. Besides them, some advanced machine learning algorithms (e.g., XGBoost) have also been successfully applied, such as ‘An interpretable model for the susceptibility of rainfall-induced shallow landslides based on SHAP and XGBoost’ and ‘Slope stability prediction using ensemble learning techniques: A case study in Yunyang County, Chongqing, China’. These latest research may help to gain a better understanding of the recent development for interesting readers.

Author Response

Dear Reviewer,
Thank you for your remarks. We have incorporated the suggested changes in the revised manuscript. The detailed response is as follows:

Point 1: Thanks to the contribution of all the authors, the manuscript has been carefully revised according to the reviewer comments. In general, it can be accepted for publication after minor corrections. Specifically, the section 5 summarized the machine learning techniques used in landslide susceptibility mapping, including RF, SVM, LGR, and ANN. Besides them, some advanced machine learning algorithms (e.g., XGBoost) have also been successfully applied, such as ‘An interpretable model for the susceptibility of rainfall-induced shallow landslides based on SHAP and XGBoost’ and ‘Slope stability prediction using ensemble learning techniques: A case study in Yunyang County, Chongqing, China’. These latest research may help to gain a better understanding of the recent development for interesting readers.

Response 1: We have included the discussion of both papers in section 5 of the updated manuscript.  
The papers are mentioned/cited in line no. 1093-1097 introduction to XGBoost, included in the Literature on ensemble-based landslide susceptibility mapping in Table 4 and discussed in line no. 1355-1365. To accommodate the changes, the statistics of ensemble methods are updated in Table 6, line no. 1609, and line no 1639.

Thank you