A Comprehensive Assessment of XGBoost Algorithm for Landslide Susceptibility Mapping in the Upper Basin of Ataturk Dam, Turkey
Round 1
Reviewer 1 Report
This paper was interesting, well written and pleasant to read.
Despite this, the methodology addressing this problem can be significantly be improved.
It lacks a section introducing the related works in the field and another one discussing and comparing the results to the related work.
Author Response
Dear Reviewer,
First of all, we would like to thank for your valuable contributions to our manuscript. We have modified the paper according to your suggestions as well as the inputs of other reviewers. Please also see our responses to your comments in the attached file. We would be grateful if you could review the modified version.
Kind regards,
Authors
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The draft of “A comprehensive assessment of XGBoost….Basin of Ataturk Dam, Turkey” discussed the landslide susceptibility mapping using XGBoost algorithm. This is new and interesting. The following unclear parts must be clarified to improve the quality the paper:
- The authors must check size of the number or words in each figure. In addition, many figure are hard to distinct in black and white print.
- Line 62, “Among the commonly used ML algorithm … (ANN), (AHP) can be listed”, the following references are also related to the algorithm for the slope stability analysis. They should also be cited:
- Rainfall-based criteria for assessing slump rate of mountainous highway slopes: A case study of slopes along Highway 18 in Alishan, Taiwan, Engineering Geology. 118, 3-4, p. 63-74
- Neural network-based model for assessing failure potential of highway slopes in the Alishan, Taiwan Area: Pre- and post-earthquake investigation, Engineering Geology. 104, 3-4, p. 280-289
- Line 130, “the geological structure is highly complex…”, additional statements must be added to explain why it is highly complex.
- Line 263, “…was split as training (80%) and testing (20%) sets”, technically, when the number of training data increases, the final results can get better. Is there any criteria for the readers to choose the ratio of training and testing sets?
- The TP, FP, TN, FN in Table 4 are not defined.
- In Figs. 15 and 16, why the authors choose these factors to the investigations? In addition, are they independent factors?
Author Response
Dear Reviewer,
First of all, we would like to thank for your valuable contributions to our manuscript. We have modified the paper according to your suggestions as well as the inputs of other reviewers. Please also see our responses to your comments in the attached file. We would be grateful if you could review the modified version.
Kind regards,
Authors
Author Response File: Author Response.pdf
Reviewer 3 Report
This manuscript describes the production of a LSM in southern Turkey. The authors used a popular machine-learning tool to develop a relationship between mapped landslides and primarily topographic variables. The methodology section is admirably detailed. I think this is one of the more replicable papers on the topic. However, I think there are a couple of possible weaknesses with this approach. First, the heavy reliance on DEM-derived factors is not really justified. Furthermore, it may put into doubt the conclusion that “the main factors controlling the landslides in the study area are the lithology, altitude and the TWI.” It could be fine, but it looks a little lopsided to me. More importantly, I wonder about the apparent difference in size between the landslides in the training and test sets. Why were these chosen? Are they really representative of landslides in this area? That said, the quality of this work seems in line with similar studies, and I don’t see any important barriers to publication.
It’s hard to read the text in the figures, but I assume this is an artifact of the preprint, and it will be more visible in the proofed version.
Comments on line:
65: AHP is not machine learning. LR could be classified either way. I would just drop the mention of AHP, it’s not necessary here.
118: “exists” should be “exist”
196: why were these 5 chosen? They look bigger.
198: missing “.”
Author Response
Dear Reviewer,
First of all, we would like to thank for your valuable contributions to our manuscript. We have modified the paper according to your suggestions as well as the inputs of other reviewers. Please also see our responses to your comments in the attached file. We would be grateful if you could review the modified version.
Kind regards,
Authors
Author Response File: Author Response.pdf
Reviewer 4 Report
The authors apply XGBoost algorithm to obtain the landslide susceptibility map of a region in southern Turkey where the Ataturk dam, one of the largest in Turkey, is located.
The authors use freely available high quality DEM data and apply appropriate parameterization strategies to extract significant features.
Thanks to appropriate pre-processing strategies, the authors obtain a good performance of the method, higher than that obtained in other works that exploit the same classification algorithm.
The article is very well written and can be accepted for publication. I have only a few very small comments that I report below.
Minor comments:
Lines 14-15 rephrase, please.
Line 62 – “Among the commonly used ML algorithms used for the LS mapping” Remove the first “used”, please
Line 80 - remove the semicolon after "Turkey", please.
Lines 173-175 Please, clarify.
Lines 195-197 Why? Please, clarify.
Figure 8 The color scale is not very readable, perhaps it can be improved.
Figure 17 I don't quite distinguish the landslides for training and the landslides for assessment, perhaps their representation can be improved.
Author Response
Dear Reviewer,
First of all, we would like to thank for your valuable contributions to our manuscript. We have modified the paper according to your suggestions as well as the inputs of other reviewers. Please also see our responses to your comments in the attached file. We would be grateful if you could review the modified version.
Kind regards,
Authors
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
The authors did a good job modifying the draft. It can be accepted in the current format.
