Landslide Susceptibility Assessment and Future Prediction with Land Use Change and Urbanization towards Sustainable Development: The Case of the Li River Valley in Yongding, China
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsDear authors,
Recently I reviewed a manuscript entitled by “Landslide susceptibility assessment and future prediction with land use change and urbanization towards sustainable development: A case of Li River Valley in Yongding, China” for Sustainability. Generally, its writing and organization are well and fits with the scope of the journal. I agree to consider its publication. But before this, please address the issues as below:
- The current Introduction section is too long. Please reduce it. Additionally, please add more quantitative results in the Abstract. This is also the case of the Conclusion. It is not common to see nothing quantitative in these parts.
- It is a pity that most of the figures are NOT clear in the study, including Figure 1, Figure 2, Figure 4, Figure 7 and Figure 9. Reviewer cannot see the details of legend and their color, and the Latitude and longitude. It seems a setting error when authors export the images. This is the biggest issue of the manuscript. Authors must improve this point.
- In Introduction section, some latest literature regarding the landslide susceptibility assessment model, and landslide susceptibility assessment under the LULC change are missing, for example:
https://doi.org/10.1007/s10346-021-01775-6
https://doi.org/10.1080/17499518.2023.2188465
- Please separate the results and discussion sections. This is important. In Discussion, authors need to compare your study with other similar ones, to show the strengths and limitations of this one.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThis manuscript aims to analyze the impact of land use change and urbanization on landslide susceptibility assessment and to predict future landslide susceptibility. The topic is of high significance for disaster prevention and mitigation of landslides. However, there are several major shortcomings in this manuscript. Specifically: (1) The technical methods and models in this manuscript have major deficiencies. The author mentions that the Analytic Hierarchy Process-Composite Index (AHP-CI) model is an expert experience model, which undoubtedly carries high risks. Another logistic regression (LR) model used in this manuscript is a machine learning model. However, such models lack landslide samples, or the current landslide samples cannot establish a connection with future land use type image factors. Therefore, logically, there are significant deficiencies in this manuscript 's methods. Moreover, the risk of landslides may lead to changes in land use types, which are not considered in this manuscript. (2) The data quality of this manuscript is low, and the landslide susceptibility maps of the results seem inconsistent with the actual situation or common sense. (3) This manuscript lacks in-depth scientific discussion, and the current discussion content is very superficial. (4) There are significant deficiencies in the language and grammar of this manuscript, and it is recommended to carefully polish the paper.
Comments on the Quality of English Languagerelative low.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsThis study, focusing on the Li River Valley (LRV) within the Yongding District, China, incorporates the time-varying human activities indicators (changes of land use, NDVI and traffic density) as critical evaluative indicators, Then, two common models, namely analytic hierarchy process-comprehensive index (AHP-CI) model and logistic regression (LR) model were employed to assess landslide susceptibility under varied conditions. Finally, the CA-Markov model was deployed to forecast future land use scenarios in 2025, 2030 and 2040, upon which predictions of ensuing landslide susceptibility were based. This study could help promote sustainable landslide mitigation. However, there are still the following problems in the current research:
Comment 1:
Line 64, The expression is not accurate. As described in the following paragraph, many researchers have considered land use and land cover changes and predicted future landslide susceptibility.
Comment 2:
The CA-Markov model was deployed to forecast future land use scenarios in 2025, 2030 and 2040, upon which predictions of ensuing landslide susceptibility were based. Why the lack of 2035? I think the land use scenario in 2035 should also be predicted and the corresponding landslide susceptibility map be created based on that.
Comment 3:
In the introduction or discussion section, the author should mention and cite some previous work (https://doi.org/10.1016/j.jhydrol.2024.130905; https://doi.org/10.1016/j.scitotenv.2024.170007).
Comment 4:
The authors mention that one of the specific objectives of this study was compare the ability of AHP-CI model and LR model to predict landslide susceptibility with current land use conditions and landslide data. However, the input indicators for the two models were different, and I was confused. I think the same input indicators should be used to compare the ability of two models for predicting landslide susceptibility. In addition, determine if the time-varying human activity indicators can improve the model prediction performance for landslide susceptibility assessment, the different indicators should be input into the same model.
Comment 5:
Line 478, Please correct “5. Conclusions” to “4. Conclusions”.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 4 Report
Comments and Suggestions for AuthorsThis manuscript is completely finished and is of interest to readers of the journal. It fully complies with the theme and design rules. The figures complement and reflect the results. The idea of the study can be clearly seen in the text. The research itself has practical significance. In conclusion, the study highlights the critical impact of human activities on landslide susceptibility in the Li River Valley, China. The findings demonstrate the significance of incorporating time-varying human activity indicators in landslide susceptibility models for more accurate predictions. The successful application of the Logistic Regression model, enhanced by the inclusion of key evaluative indicators, underscores the potential for improved forecasting and mitigation strategies. In essence, this study underscores the urgency of addressing the complex interplay between land use change, urbanization, and landslide susceptibility. I recommend this article
Comments on the Quality of English LanguageRequires a little proofreading by an English-speaking specialist
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 5 Report
Comments and Suggestions for AuthorsThe article is very interesting, but it needs a profound revision, both in the contents and the presentation of the results.
First, the description of the "indicators". They are usually called "explanatory variables" or "predisposing factors". I would avoid using the term "indicators" throughout the article.
Some of them have been transformed into discrete variables, even if they are described by a continuous variable (for example the slope gradient or the elevation), without justifying this choice and without explaining how the number of classes and the separation values between them were chosen one class and the next. In the fourth column of Table 2, you show the score that is attributed to each level of each of the variables, but it is not explained how this core was attributed. It is not clear, in particular, whether the assignment of an a priori rating can influence the statistical analysis results. To these 7 variables, you then add another 3 (only for logistic regression? Why?) and show their distribution in Figure 4, which is illegible. The subfigures are too small, and the colourmap scales are illegible (although, for the first 7 variables, they could be obtained from the first column of Table 1). I believe you should better organize the size and distribution of these maps if you consider them useful for understanding the results of the work.
From a strictly statistical point of view, it does not seem that the collinearity analysis was carried out to exclude the presence of collinearity between these variables. This is very important because two variables that seem independent of each other can carry the same information, so it is not correct to include them both in the model.
Regarding the presentation of the results, it is unclear how Table 4 was constructed, i.e. how the numbers and fractions present in columns A1...A7 were determined. In the weight column, the results are presented with 4 significant figures. Are you sure that all this precision makes sense, from a statistical point of view?
In the presentation of the results of the LR model, on lines 336-338, you say that you have carried out the significance analysis of the variables, but you do not explain which criterion was used to evaluate which of them to eliminate (p-value? at what level?). Slashes appear in Table 6, which, if I understand correctly, indicate the discarded variables, perhaps because they have too small an RC (please explain the meaning of the acronym RC: Random Coefficient? or what?). Again. the maps presented in Figure 7 are too small and are not useful for understanding the results: they must be enlarged.
Regarding the evaluation of the model with the AUROC and all the other scores (figure 8), I believe it is important to show the statistical reliability of the values you present, trying to provide an estimate of a confidence interval for each of the numbers. This can be achieved by repeating the extraction of data from the database many times, with the bootstrap method.
Comments on the Quality of English LanguageMinor editing of the English language is required, in the next review.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsDear authors,
I think all of my comments have been addressed. I agree to accept it. Congratulations.
Author Response
I am sincerely grateful for your valuable feedback and for taking the time to review our manuscript. I am delighted to hear that you believe all of your comments have been addressed, and I appreciate your agreement to accept our work.
Reviewer 2 Report
Comments and Suggestions for AuthorsAlthough the manuscript has been revised, the problems in this manuscript are fundamental, and the serious problems and shortcomings have not been improved.
Comments on the Quality of English Languageno
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsThis manuscript has been improved significantly according to my comments. I believe it should accepted in current form.
Author Response
I am sincerely grateful for your valuable feedback and for taking the time to review our manuscript. I am delighted to hear that you believe all of your comments have been addressed, and I appreciate your agreement to accept our work.
Reviewer 5 Report
Comments and Suggestions for AuthorsThe revised version of your article is much improved over the first draft.
I appreciate the attempt to improve the quality of the figures, but I say straight away that you have to make an effort to make them even more understandable.
In Figure 1, the legend in B is incomprehensible, and the font size in D and E is too small. We all know that it is possible to enlarge the figures in the electronic version of the work, but it is more correct to ensure that the figures are also understandable in the version printed on paper.
In Figure 2 you can place a single legend outside the boxes, common to all three subfigures; it is useless to repeat the longitude (or latitude) scale twice; by inserting it only once, you gain space to enlarge the three subfigures.
The same considerations apply to Figures 3, 7, 9, and 10: the edges of 2 subfigures can be "attached" by eliminating the internal stairs; the font sizes in the legends are too small.
For all these changes, I believe it is useful to have the help of editorial support.
Now let's move on to comments on the content. I am referring to the points on the cover letter.
1) OK, nothing to say.
2) Now the issue of scores is clearer. However, I have two recommendations:
2a) Insert references on line 185 of page 6 to explain (at least in part) the "extensive literature review".
2b) Insert (perhaps after line 187?) the sentence in the cover letter: "These scores are used exclusively in the AHP-CI model and not in the LR model. Given that the AHP-CI model is empirical, it is acceptable to assign ratings based on prior understanding."
The issue of transformation into categorical variables is also now clearer. You must include among the references some of the works in which the use of the same transformation is shown. It would also be good to highlight the limits of using discrete variables (for example the fact that each level corresponds to a fit parameter, so the number of parameters can become high if there are many levels) and also to point out that there are other methods for the treatment of nonlinearities, such as the Generalized Additive Model (GAM).
The sentence "The AHP-CI model is empirically based. Its advantage lies in the fact that it can be constructed without a dataset and applied at various scales, although it is highly subjective. On the other hand, the LR model is a statistical model that requires a training set. The two models represent heuristic and statistical approaches, respectively, in assessing landslide susceptibility, which is why we have chosen to utilize both." it is too important to be left only to the delight of the reviewer. I think it is more correct to insert it somewhere in the article (in the introduction if not even in the abstract).
3) In the cover letter you write that you do not want to compare the performance of AHP-CI and LR. In reality, unfortunately, this comparison comes out in the end. However, you did well in presenting two different types of models. Perhaps it would be better to insert the phrase somewhere in the work
"Our intention is not to compare the relative performance of the AHP-CI and LR models. Instead, we aim to demonstrate how each model addresses landslide susceptibility in the context of changing environments (land use and urban expansion) from different perspectives. The AHP-CI model focuses on basic geological conditions and reflects “changes” across different landslide validation datasets. In contrast, the LR model considers changes within its predisposing factors and uses landslides from different periods as training and validation sets."
4) OK, nothing to say.
5) I therefore imagine that you have eliminated A1, A2, A3 and B2. For the others, even if the CR of one or two levels is not significant, you must maintain the predisposing factor (at most you could try to merge the other levels). The final results are obtained by refitting the model without the discarded predisposing factors, right? Otherwise, why did you conduct the p-value analysis?
6) At this point, you can see the comparison between the two methods in Table 7. You must explicitly say that the AUROC values for AHB-CI are very low, bordering on unacceptable (even if the confidence interval does not include 0.5, the lower limit is very close to it). However, from the point of view of temporal comparison, it is right to present both methods, and this point should also be appropriately underlined.
There is a typo both in the abstract and the conclusions: you wrote 0.477 instead of the correct value of 0.744. Please correct.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 3
Reviewer 2 Report
Comments and Suggestions for AuthorsDespite the revisions made by the authors, the fundamental issues persist in this paper, with serious problems and deficiencies remaining unaddressed.
Comments on the Quality of English Languageno.