Rainfall Induced Landslide Susceptibility Mapping Based on Bayesian Optimized Random Forest and Gradient Boosting Decision Tree Models—A Case Study of Shuicheng County, China

Round 1

Reviewer 1 Report

Dear Authors,

The article is interesting, but it has some mistakes.

1. There is no clearly defined research goal in the introduction section
2. There are not enough references to literature in the Discussion section. It can be said that this is a discussion of the results, not a discussion.

3.The choice of the factor determining the impact of precipitation in the form of average annual precipitation (line 157-158) seems incorrect, the more so that the authors state that in the case of landslides, one of the most important triggers is rainfall, especially short-term and momentary extreme precipitation (line 74-75).

4. Where do the rainfall data come from and what period are they from? From 10 years, from 30 years ???
5. In line 95, you write that on July 23, 2019, a landslide formed. What were the weather conditions like before that day? How much rain fell before the landslide formed? Was it extreme rainfall?
6. Line 141: there is a figure 3m, there should be a figure 2m,
7. In Figure 2 there is 2 times (k) SPI, there should be (k) SPI and (l) STI.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

A review report for water-965653-peer-review-v1

Summary

The study shows a landslide susceptibility mapping based on Bayesian Optimized Random Forest and Gradient Boosting Decision Tree Models. In general, the manuscript is well documented and up to the standards of an international journal. Some revisions need to clear the issues of the manuscript.

Major issues

1. L350, please show quantitative results for the comparison of historical landslides and high risk areas.
2. Why not use historical areas of landslide as one of the conditioning factors?

More minor issues

1. L109, the authors used 30x30m DTM (L112), should the pixel size resampled to 30 m grid? What is the original size of DTM used in the study?
2. 2(j), the authors used two rivers for the study. In most of the study cases, river watershed created by DTM was used instead of river solely. Should the study evaluate the differences of the results?
3. 2(k), the legend text should be STI not SPI.
4. In Fig. 4, adjust the text size in the title (e.g. Importance).
5. Figs 4 and 5 are ugly; I suggest to modify both of the figures.
6. In Fig. 7, there shows a green color bar in the GBDT_B model, what is it? Also, the very low and very high categories have a close color, I suggest to change one of the color.
7. In Fig. 7, here shows a big difference for the GBDT_B model for the high percentage of “very low” but its precision is better than and closes to other models. Can the models evaluated use the ROC curve? The curve seems cannot distinguish different models.

Opinion

In general, the manuscript could be interesting to readers of the journal. I suggest a major revision of the manuscript.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Dear authors,

the article was corrected according to my comments.