Robustness of Optimized Decision Tree-Based Machine Learning Models to Map Gully Erosion Vulnerability

Round 1

Reviewer 1 Report

The work developed by the authors is potentially significant and could be a good contribution to the knowledge of Gully Erosion vulnerability. This work constitutes a basic tool to help managers and planners ensure the effective and sustainable management of areas affected by soil erosion.

The manuscript is well written and well-illustrated. However, some figures are unreadable and need to be changed with ones of high quality.

Further comments are indicated in the annotated manuscript.

 

Additional comments

 

1. What is the main question addressed by the research?

Gully erosion is a global threat with numerous negative effects on the ecosystem, society, and the economy.

In the literature, many are used to create Erosion Susceptibility Maps. This work tested six models in the semi-arid area. Moreover, it discusses the main factors among the 17 that control gully erosion.

2. Do you consider the topic original or relevant in the field? Does it
address a specific gap in the field?

Yes, this work constitutes a basic tool to help managers and planners ensure the effective and sustainable management of areas affected by soil erosion.

3. What does it add to the subject area compared with other published
material?

The methodology of the present research consists of several key steps in order to create accurate Gully Erosion Susceptibility Maps using decision tree models. Six decision tree models based on ML algorithms were tested to determine the role of 17 parameters in the gully formation in a semi-arid environment.

The results are six Gully Erosion Susceptibility Maps that shows that all the utilised models are robust and extremely reliable at predicting and identifying the sensitivity to gully erosion and that the most influencing factors are Lithology, LULC, Geomorphons, and Elevation factors.

4. What specific improvements should the authors consider regarding the
methodology? What further controls should be considered?

The methodology is ill described and control points were provided and supported by photos.   

5. Are the conclusions consistent with the evidence and arguments presented
and do they address the main question posed?

Yes, they conclude that the tested models are very useful and they discussed the main factors that control the gully erosion in such environment.

6. Are the references appropriate?

Yes

7. Please include any additional comments on the tables and figures.

The comment on the figure was indicated in the annotation on the PDF.

Comments for author File: Comments.pdf

Author Response

Please see the attached file.

Author Response File: Author Response.pdf

Reviewer 2 Report

 

Based on the abstract of the article [1], the research aims to evaluate the performance and robustness of six machine learning ensemble models based on the decision tree principle to map and predict gully erosion-prone areas in a semi-arid mountain context. The study used 217 gully points and several geo-environmental variables as potential controlling factors. The six models used in the study are Random Forest (RF), C5.0, XGBoost, treebag, Gradient Boosting Machines (GBM) and Adaboost. The inventory data was randomly subdivided into five percentages Train/Test to assess the stability and robustness of the models. The study found that all of the models performed well in terms of predicting vulnerability to gully erosion. However, the C5.0 and RF models had the best prediction performance, with AUC values of 90.8 and 90.1, respectively. The study also found that the models exhibited small but noticeable instability, with high performance for the 80/20% and 70/30% subdivisions. The study highlights the importance of database refining and the need to test various splitting data to ensure efficient and reliable output results.

Overall, the study presents a comprehensive evaluation of six machine learning ensemble models for predicting gully erosion-prone areas in a semi-arid mountain context. The study findings suggest that the C5.0 and RF models are the most effective in predicting vulnerability to gully erosion. However, it is important to note that the models exhibited some instability, indicating the need for further testing and refining of the database. Overall, the study contributes to the development of effective tools and methods for gully erosion mapping and prediction.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

Gully erosion is related to geological hazards of slope. There are many factors involved in the formation of gully erosion, such as terrain, precipitation, soil quality, and so on. The purpose of this manuscript is to evaluate the performance and robusnessof six machine learning methods: Randon Forest, C5.0, XGboost, Treebag, Gradient Boosting Machines, and Adaboost, for mapping and predicting gully erosion-prone areas. On the basis of 217 gully Train/Test points and 17 geo-environmental variables, The stability and robustness of the models are assessed. The manuscript shows that a lot of work has been done, and the results revealed that all of the models used performed well in terms of predicting vulnerability to gull erosion. The research results maybe provide information for gully erosion conservation. This is an interesting and useful topic. However, there are still doubts to be clarified as follows:

1) Line 153: Figure 2 lacks a title and is unclear.

2) Line 162-230: The figures cited in the manuscript do not correspond to the figure No., Such as Fig3-1 and Figure 3.1, Fig 3-2 and figure3.2. Readers cannot understand whether it is a small image in the figure or the entire figure.

3) Line 349: There are overlaps in the text in Table2.

4) Line 377: What does Fig3 refer to. Figure3.1 or Figure3.2?

5) Line 412-413: Where to cite Figure 4 ?

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 4 Report

this MS collect plentiful data on gully erosion, and analyzed the potential factors impacting gully erosion. my only suggestions is that the figures in this MS need improvement. the readers can not clearly find the information on these figures.  I have some minor suggestions:

line 75: there existed some models of gully erosion, so I suggest that authors can give some explanations to the advantage of this research.

line 153: please add the title of fig.2. the readers needs a clearer figure.  

line 157-159: I suggest the authors listed a table which give explanation to each variables, and cited the related references. 

line 578-579: can the authors give the detailed impact factors, such as slope steepness.

Author Response

Please see the attachment

Author Response File: Author Response.pdf