Evaluation of Geological Hazard Susceptibility Based on the Regional Division Information Value Method
Round 1
Reviewer 1 Report (Previous Reviewer 3)
The paper can be accepted in the current version.
Author Response
Dear Expert,
Thank you very much for your time involved in reviewing the manuscript and your very encouraging comments on the merits.
Sincerely,
The Authors
Reviewer 2 Report (New Reviewer)
The research workshop was presented correctly. The model validation criterion was met, which is particularly important in studies based on statistical analyses. In the multi-criteria analysis, the basis is the simultaneous study of the influence of factors on the course of the phenomenon. In the presented regional groups, however, one can distinguish constant features, e.g. height, curvature, roughness, and “variable features”, e.g. precipitation, lithology. Therefore, it may be a good solution to perform an analysis for sets with similar "constant features" with different "variable features". Such research may be more useful on a local scale. In the assessment of real risks, information on the local scale is most useful.
Author Response
dear expert
Merry Christmas!
Thank you for your positive comments and valuable suggestions to improve the quality of our manuscript.
My research focuses on providing a method to improve the accuracy of the previous method of vulnerability assessment without differentiating the types of hazards, and to study the significance of zoning and the impact on the contribution of the assessment factors, both of which have been achieved.
With regard to your suggestion, In the presented regional groups, however, one can distinguish constant features, e.g. height, curvature, roughness, and “variable features”, e.g. precipitation, lithology. Therefore, it may be a good solution to perform an analysis for sets with similar "constant features" with different "variable features". I will certainly conduct an in-depth study in my future research.
Sincerely,
The Authors
Reviewer 3 Report (New Reviewer)
Dear authors,
This study set out to estimate the differences between global and local measures of a model. While global measures summarize the overall spatial autocorrelation of the study area in one single value, local measures of spatial association identify local clusters. This topic is interesting and relevant. Unfortunately, it is not novel and the method suffers from serious deficiencies in the following comments. First, the spatial resolutions of the data used were not provided enough and clear in this paper. Do they have different spatial resolutions? If so, how does the difference impact on the model? Second, although results show that the regional model has the higher accuracy in comparison to the global model, without uncertainty estimation, it cannot be decided which on is better. It is also hard to understand the conclusion of the author about “better robustness” in this paper. Third, some data were provided without an official and reliable source, for example the shapefile of the study area. Is the dashline official and accepted international community? Finally, authors do likely not carefully pay attention to concise writing. Many parts are hard for reader to understand and non-standard writing appears many place like suddenly using capital in sentences. More importantly, it is better to clarify what is the novelty of this research.
These above-mentioned issues must be resolved before the paper may be considered for publication. To improve the clarity of the manuscript, specific comments are as follows.
L (line) 9: Why do authors start “Susceptibility” with a capital?
L 13 and 15: Must there be a space after a comma.
L 20: Must there be a space between words, specifically AUC = 0.893.
L 34: What is DEM? Do authors mean “digital elevation model”, “demurrage charge” or what?
L 47: Why do authors keep starting some words with a capital?
L 50 to 69: Author must explain more about their discussion and should provide obvious clues about each discussion.
L 76: Again! Why is there a capital in the word “Information”?
L 70: What do authors mean “Scholars in China and abroad”? Does it mean “Scholar”. There are some redundancies, please go through the paper and make it more concise.
Authors must elaborate what is “global and local model”?
L 96: What does the word “split” exactly mean? I suggest to use a simple and clear word in stead of an ambiguous word.
L 101: This sentence lasts for 7 lines. It is too long and to complex. There are many other long sentences. It is necessary to write the paper more simple and concise.
L 105: Province is a common noun so that it is not need a capital. Authors need to check the other places such as L 113 and 114.
L 118: What do authors mean “heat” and “the light is abundant? This sentence seems not complete.
L 124: It seems not common to start a sentence with “And”.
L 101: This paragraph is wordy. Author mentioned many things before talking about the study area. It seems the paragraph is not structured well. It is better to make it short and clear. Instead of wordy description, it is better to provide with numbers. For example, when talking about rainfall, authors may provide with annual average rainfall values.
L 147-151: Confusions are common in this paper. Line 147-1521 are examples. “Up to now” seems not a suitable phrase here. If readers read this paper ten year later, they may have question about the logic of “up to now” with the other parts of the sentence. In addition, author said “mainly medium-small”, but only 6 are medium-sized and 82 are small. These words “medium-small, small and medium-sized are confused readers. Medium-sized looks like an adjective so it is not suitable to use a single phrase like “one medium-sized”.
Figure 1: Where do author take the shapefile of China border? Is it officially accepted by international community? If so, please provide such a document. The disaster points must provide more detail, including how to extract it, the location (lat and lon), time, and so on. If it is too long, authors can add it in supplementary. The map needs to add lat and lon and it should have the same and readable font.
Figure 2: What is the “modle” mean?
L 181: What are “ith” and “kth”? Must they be written in an italic form? Authors must go through the paper to check whether mathematical notations are written in right format as the SI unit standard?
Section 3.1.1: This section is hard to understand. Readers may not understand it so that it needs more elaboration.
L 195-197: It is not common to start a sentence with “And”. Author must explain clearly what are AIC, BIC, and AC. If it is too long, a supplementary is necessary.
Table 2. What do authors mean “Data sources” here? A “data source” means where readers can refer and may take the data. Do authors think readers can do such a thing? I cannot. Also, authors must provide detailed information about the data such as spatial resolution, time collection, how authors have pre-processed the data, which software and tools were used, and so on.
Equations 6 and 7: Authors are careless or may not know how to write mathematical notation in a standard form (SI Units). For example, ith or jth must be in italic forms.
Table 1: Using Pearson correlation coefficient to detect high correlation features and remove them is good. However, in many cases, many features highly correlate to each other. As such, how do authors handle for this problem?
L 233-234: Authors should explain clearly with scientific evidence why a correlation coefficient greater than 0.3 is considered to have high correlation?
L 267: What does the “determining slope stress” mean?
Figure 3. In figure 3 (I), there is a gap in the north part. What’s the problem?
Figure captions should be standalone, i.e., descriptive enough to be understood without having to refer to the main text. Please check all the figure captions in this paper.
Table 4. How samples for training and validating have been designed and collected is not explained clearly in this study.
Author Response
Thanks for providing us with this great opportunity to submit a revised version of our manuscript. We appreciate the detailed and constructive comments provided by the reviewers. We have carefully revised the manuscript by incorporating all the suggestions by the review panel.
We hope this revised manuscript has addressed your concerns, and look forward to hearing from you.
Please see the attachment
Author Response File: 
Author Response.docx
Round 2
Reviewer 3 Report (New Reviewer)
Please contact Editors for more details. Thanks!
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
The paper resented an interesting work to me.
However, some revisions should be made before publication consideration.
(1) In fact, there are many models (such as neural networks, support vector machines, random forests) can be applied to regional geological disaster assessment. The author uses information value model to evaluate. It is suggested some necessary comparisons between different models in introduction section to enhance the readability of the article;
(2) The second formula (1) should be (4);
(3) When the Pearson correlation coefficient is greater than 0.3, there is already a moderate correlation. It is suggested that the author should have more discussion on Table 1;
(4) It is suggested that the author explain the data source of evaluation factors and the principle of interval division, like '35°~45°' and '35°~78°' in section 4.1.5,;
(5) '45°~80°' appears in Table 2. Please specify the true range of slope gradient factor;
(6) In Table 2, please explain the data units (m, mm, °, etc) of the evaluation factors;
(7) The author chooses slope and NDVI as the basis for division, however, Fig 8 shows that there are obvious differences in the weight of influencing factors in different regions, please make a further discussion;
(8) The ROC curve results show that there is little difference in the evaluation accuracy between the two models (0.02), please further explain the validity of the model.
Author Response
Dear expert,
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 2 Report
Comments and Suggestions for Authors
1. Line 63-69, there is no references to support this passage.
2. Line 126. It is suggested to add an overall flowchart of the method section, so that readers can understand the methodology clearly.
3. line 184, “with reference to relevant research results in China and abroad”, it is suggested that the author add relevant references which are referred to.
4. The legend of geological hazards in Figure 1 cannot be seen clearly. It is suggested to use darker colors.
5. Line 201, it is difficult to find the highly correlated factors in the Table 1 at a glance, so it is recommended to highlight them.
6. "4 selection of sensitivity factors" is not the main part of the article, and it is unnecessary to describe it with so many texts. It is suggested to shorten sections 4.1-4.3.
7. Line 342, What is “the neural network model”? How to do it?
8. Section 5.4, because of lacking a method for calculating errors, it is suggested that more evaluation methods be considered to be used here.
Author Response
Dear expert,
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report
The introduction is general and there is no proper literature gap to focus on. Lack of introduction to GWR method and why it can solve this problem you pointed out. Authors should clearly point out the motivation of this work. In addition, several papers related to this study are missing.
- 10.5194/nhess-19-1973-2019
- 10.3390/app10031107
Fig.1: Anji County, not Anji Country. The geological disasters cannot be clearly seen from the figure. In addition, an overall map of the location of the study area on Earth is missing here.
The quality of figures should be improved.
The information of different data should be shown in a table, so that readers can clearly see them, such as data source, resolution, time, etc.
Rainfall is important for slope instability and should be considered in the study.
A description of geological disasters in the study area is missing, including the type, number, size, source, etc. It is unclear the geological disasters are points or polygons.
Line 402, “The top two important evaluation factors are slope direction and NDVI, were selected as this regional split factor …”, Why?
Lines 406-408, “The total information value of each slope unit was 406 calculated for each region separately using the information value model …”, slope unit? Lines 455-456, “With Changxing County as the study area and the raster cell as the evaluation unit, …”, raster cell? As we know, the slope unit and raster cell are different. The susceptibility maps driven by them are also different. So, which evaluation unit do you use? This must be described clearly.
A substantive discussion is missing, including the applicability of the GWR method, the limitation of this study, as well as the comparison.
The English must be improved. Some spelling or grammatical errors need to be checked carefully, e.g., Lines 394-396.
Author Response
Dear expert,
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 4 Report
The paper treats the susceptibility evaluation of the Geological Hazard, and in particular face the problem of the differences between local areas that influence the results. Moreover, aims to model geological hazard impact factors in different regions on geological hazard susceptibility, reducing the influence of some factors that have higher influence in the global model but lower influence in local areas on the evaluation results.
The paper is written in a good English and the methodology is well described.
The use of logistic regression is commonly used to susceptibility maps related to landslide.
The main problem that needs a deep revision of the application is the attempt to give answer to the generic geological hazard. I want to underline that the influence of each evaluation factor mentioned in the text vary in relation with the investigated phenomena. For instance the slope factor play a different role if I am investigating the hazard related with a debris flow or a earth flow or a rock-fall. This concept is completely missing in the paper and even if the methodology is well described and formally correct the result is to have a too generic susceptibility evaluation zone map. The weight to give to each parameter needs to be calibrate to a data set of inventoried (for instance rock fall) elements, absent in this paper.
For this my suggestion is to rewrite the paper focusing the attention on two, three type of landslides, or other geological hazards, defining the weight of each factor on the base of the statistic relevance of each occurrence and then generate different susceptibility model for each type.
If the author wants to produce a general map of susceptibility would make a synthesis merging the value of each susceptibility map.
I think that without these modification the results are far from the true evaluation of geological hazard
Author Response
Dear expert,
Please see the attachment.
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
Comments and Suggestions for Authors
1. Figure 9 can't be clearly seen .
2. How do you explain the low values of recall and F1-score?
Author Response
We feel great thanks for your professional review work on our article. As you are concerned, there are several problems that need to be addressed. According to your nice suggestions, we have made extensive corrections to our previous draft, the detailed corrections are listed below.
Please see the attachment
Author Response File: Author Response.docx
Reviewer 3 Report
In general, the authors have attended most of my suggestions. The replies and revision are both a good effort from the authors.
The analysis in which the different landslide typologies occur are very relevant to found the physical significance of the failures. Rock falls do not occur in the same conditions (geology, slope, etc.) as rock slides.
Although rainfall is a predisposing factor, it should be considered. Because human engineering activities are predisposing factors, which have been considered in the study. So, why not consider rainfall?
Figure 3: Of all the factors, aspect is the most important, while slope is the last. In my opinion, some explanation about this is required (about why some factors influence or not the landslides occurrence), that can be related to the landslide typology and dimensions, the relation to geology, etc.
Does the division of factor categories affect the results?
Line 863, Table x?
As show in Table 4, the values of Recall and F1 are very low. In my opinion, the higher the value, the better the robustness of the model. Such a low value only indicates that both results are not good.
Author Response
We feel great thanks for your professional review work on our article. As you are concerned, there are several problems that need to be addressed. According to your nice suggestions, we have made extensive corrections to our previous draft, the detailed corrections are listed below.
Please see the attachment
