Utilizing Remote Sensing and Random Forests to Identify Optimal Land Use Scenarios and Address the Increase in Landslide Susceptibility

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Landslides are one of the major natural disasters that have a significant impact on regional socio-economic development and human well-being, particularly limiting urban planning and regional land resources sustainable use. This study utilized remote sensing data and 400 field-validated landslide data points to investigate the Sumber Brantas and Kali Konto sub-watersheds landslide susceptibility through cellular automata models and random forest algorithms, which is of great significance for optimizing regional land spatial layout and mitigating disaster landslides. The following suggestions are proposed for the author's reference.

• It is suggested to supplement relevant literature on landslide monitoring methods (such as high-resolution remote sensing data, InSAR technology, etc.) and the influencing factors of landslide changes (vegetation cover, precipitation, human factors, etc.) in lines 47-60, in order to provide a basis and support for the selection of remote sensing data and simulation factors in this article.
• The latitude and longitude in Figure 1 are inconsistent with the description of the research area. It is recommended to use latitude and longitude.
• The description of the three simulated scenarios in the Research Framework(lines 98-103)is relatively clear, but the explanation of the relevant roles and applications of random forests is not clear. It is recommended to supplement and improve it based on the research content. Meanwhile, it is recommended to increase the resolution of Figure 2. 
• The four scenarios in line 100 and line 355 do not match the three scenarios in the Figure 2. It is recommended to further check.
• The selection of landslide prediction parameters needs further explanation, especially for each parameter, it should be summarized into several dimensions such as geological conditions, environmental factors, and human factors, and then specifically described which factors belong to which dimension. Meanwhile, the abbreviations in Table 1 should be fully described in lines 154-158. This facilitates readers' understanding of the scientific selection of landslide prediction parameters and better reflects the complexity of landslide formation.
• In lines 171-173, landslide zones are classified into five categories based on vulnerability values: very high (>1.5), high (1.2–1.5), moderate(0.8–1.1), low (0.4–0.7), and very low (0.001–0.3).This classification is not the result of a continuous function and may miss some values, such as between the moderate (0.8–1.1) and low (0.4–0.7), the value of 0.71-0.79 may miss.
• It is suggested to supplement the accuracy/Kappa value of land use supervision classification in 1. Land Use Changes in the Sumber Brantas and Kali Konto Sub-Watersheds, and the simulation accuracy of the 2025 three land usescenariosin 3.2. Land Use Management Scenario.
• The parameters in Figure 6 are suggested to be represented by x₁, x₂, x₃,......x₂₂, and labeled accordingly in Table 1. Meanwhile, the resolution of Figure 6 is too low to see clearly.
• Suggest moving lines 206-211 to 6Accuracy Assessment.

Comments for author File: Comments.pdf

Author Response

Comments 1: It is suggested to supplement relevant literature on landslide monitoring methods (such as high-resolution remote sensing data, InSAR technology, etc.) and the influencing factors of landslide changes (vegetation cover, precipitation, human factors, etc.) in lines 47-60, in order to provide a basis and support for the selection of remote sensing data and simulation factors in this article.

Response 1: 

I have added 2 previous research articles to complete the manuscript so that it is smoother to go to paragraph 3.

1. Sinčić, M., Bernat Gazibara, S., Krkač, M., Lukačić, H., & Mihalić Arbanas, S. (2022). The use of high-resolution remote sensing data in preparation of input data for large-scale landslide hazard assessments. Land, 11(8), 1360.
2. Kalsnes, B., & Capobianco, V. (2022). Use of vegetation for landslide risk mitigation. In Climate Adaptation Modelling (pp. 77-85). Cham: Springer International Publishing.

 

Comments 2: The latitude and longitude in Figure 1 are inconsistent with the description of the research area. It is recommended to use latitude and longitude.

Response 2: We have fixed and using consistent coordinate

 

Comments 3: The description of the three simulated scenarios in the Research Framework (lines 98-103) is relatively clear, but the explanation of the relevant roles and applications of random forests is not clear. It is recommended to supplement and improve it based on the research content. Meanwhile, it is recommended to increase the resolution of Figure 2.

Response 3: I have added the explanation of RF function/overview in the paragraph

Figure 2: the figure improve for 400 dpi resolution

 

Comments 4: The four scenarios in line 100 and line 355 do not match the three scenarios in Figure 2. It is recommended to further check.

Response 4: We have replaced four scenarios with three

 

Comments 5: The selection of landslide prediction parameters needs further explanation, especially for each parameter, it should be summarized into several dimensions such as geological conditions, environmental factors, and human factors, and then specifically described which factors belong to which dimension. Meanwhile, the abbreviations in Table 1 should be fully described in lines 154-158. This facilitates readers' understanding of the scientific selection of landslide prediction parameters and better reflects the complexity of landslide formation.

Response 5: We have revised it by adding explanations and categorization below Table 1

 

Comments 6: In lines 171-173, landslide zones are classified into five categories based on vulnerability values: very high (>1.5), high (1.2–1.5), moderate (0.8–1.1), low (0.4–0.7), and very low (0.001–0.3). This classification is not the result of a continuous function and may miss some values, such as between the moderate (0.8–1.1) and low (0.4–0.7), the value of 0.71-0.79 may miss.

Response 6: Sorry for this error, i have fixed it by replace the true categorize from Nugroho & Nugroho [41]

 

Comments 7: It is suggested to supplement the accuracy/Kappa value of land use supervision classification in 3.1. Land Use Changes in the Sumber Brantas and Kali Konto Sub-Watersheds, and the simulation accuracy of the 2025 three land use scenarios in 3.2. Land Use Management Scenario.

Response 7: I have added “The kappa analysis result comparing the ground check data from the field with the produced map is 81%, surpassing the minimum threshold established by Rwanga and Ndambuki (2017).”

 

Comments 8: The parameters in Figure 6 are suggested to be represented by x1, x2, x3,...... x22, and labeled accordingly in Table 1. Meanwhile, the resolution of Figure 6 is too low to see clearly.

Response 8: We have labeled all of the map and also have been increasing the resolution.

 

Comments 9: Suggest moving lines 206-211 to 3.6Accuracy Assessment.

Response 9: We have moved it

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

This paper integrates remote sensing technology, the Cellular Automata-Markov (CA-Markov) model, and the Random Forest (RF) model to identify optimal land use scenarios for reducing landslide risks in the Sumber Brantas and Kali Konto sub-watersheds of Indonesia. The research provides valuable insights into the relationship between land use changes and landslide susceptibility, demonstrating an innovative approach. However, there are some issues in certain details and descriptions that require revision. Below are my specific comments and suggestions:

1. The abstract is concise and clear but lacks quantitative results. For example, the changes in landslide risk percentages under different scenarios (BAU, LCC, RSP) or the accuracy evaluation of the models should be briefly mentioned to help readers better understand the study’s contributions and significance.
2. The introduction provides a broad background discussion, but the review of existing studies could be further condensed. Specifically, the discussion on remote sensing and CA-Markov methods is somewhat lengthy. It is recommended to streamline this section and focus on the specific research gap this study addresses.
3. The model combination used in this study is innovative, but there is insufficient discussion on why this specific combination (remote sensing data + CA-Markov + RF) was chosen over other potential models, such as deep learning approaches. A brief explanation of this choice would strengthen the justification of the methodology.
4. The description of "ANN – CA Markov" is unclear, lacking detailed explanations of its working principles and parameter settings. It is recommended to supplement the detailed description of this model, including how it integrates with the Random Forest algorithm and its specific application in predicting land use changes.
5. While the paper mentions the use of remote sensing data, it lacks details on data preprocessing steps such as image classification, data cleaning, and validation. Providing more details on these aspects would enhance the study’s transparency and reproducibility.
6. In the experiments, the paper lacks qualitative expression and analysis regarding the recognition accuracy of different models, and also lacks ablation experiments.
7. In predicting land use changes, only data from 2017-2022 are used to predict the 2025 scenario, which is a relatively short time span. It is recommended to supplement historical data or validate the model's long-term applicability. Additionally, the prediction accuracy of the CA-Markov model is not mentioned. It is suggested to include relevant validation results (e.g., Kappa coefficient or other accuracy metrics).
8. The study compares landslide risk across different scenarios, but it does not explicitly state the statistical significance of these results. It is recommended to include significance tests in the conclusion or appendix to verify whether the differences between scenarios are statistically meaningful.
9. The discussion section primarily restates the experimental results. A deeper analysis is recommended. For example, why does the LCC scenario result in the lowest landslide risk? Are there other potential influencing factors? Additionally, does the model’s performance vary significantly across different geographical areas or scenarios, and is there room for further optimization?

Author Response

Comments 1: The abstract is concise and clear but lacks quantitative results. For example, the changes in landslide risk percentages under different scenarios (BAU, LCC, RSP) or the accuracy evaluation of the models should be briefly mentioned to help readers better understand the study’s contributions and significance.

Response 1: 

Thank you very much

I have fixed it by “From 2017 to 2022, landslide risk increased as the "High" category rose from 33.95% to 37.59% and "Very High" from 10.24% to 12.18%; under BAU 2025, they reach 40.89% and 12.48%, while RSP and LCC reduce "High" to 44.12% and 34.44%, respectively”

 

Comments 2: The introduction provides a broad background discussion, but the review of existing studies could be further condensed. Specifically, the discussion on remote sensing and CA-Markov methods is somewhat lengthy. It is recommended to streamline this section and focus on the specific research gap this study addresses.

Response 2: I have reduced the paragraph and focused on the research gap.

 

Comments 3: The model combination used in this study is innovative, but there is insufficient discussion on why this specific combination (remote sensing data + CA-Markov + RF) was chosen over other potential models, such as deep learning approaches. A brief explanation of this choice would strengthen the justification of the methodology.

Response 3: I have added “Remote sensing, ANN-CA-Markov, and Random Forest are combined to effectively analyze land use changes and landslide susceptibility while balancing accuracy and computational efficiency. Unlike deep learning, this approach requires less data and processing power, making it more practical for GIS-based applications.”

 

Comments 4: The description of "ANN – CA Markov" is unclear, lacking detailed explanations of its working principles and parameter settings. It is recommended to supplement the detailed description of this model, including how it integrates with the Random Forest algorithm and its specific application in predicting land use changes.

Response 4: I added information about CA-Markov in the last paragraph of the 2.3.1.

 

Comments 5: While the paper mentions the use of remote sensing data, it lacks details on data preprocessing steps such as image classification, data cleaning, and validation. Providing more details on these aspects would enhance the study’s transparency and reproducibility.

Response 5: I have fixed it by additional information in the 2.3

 

Comments 6: In the experiments, the paper lacks qualitative expression and analysis regarding the recognition accuracy of different models, and also lacks ablation experiments.

Response 6: I have added in 4.4

 

Comments 7: In predicting land use changes, only data from 2017-2022 are used to predict the 2025 scenario, which is a relatively short time span. It is recommended to supplement historical data or validate the model's long-term applicability.

Response 7: Thank you for your valuable suggestion. In this study, we have chosen to apply the prediction model within a relatively short time frame (2017–2022 to 2025) to minimize potential errors. This approach allows us to evaluate the model's performance in a more controlled manner before extending it to a longer period. However, we fully acknowledge the importance of long-term validation and incorporating a broader historical dataset. Therefore, your suggestion is highly appreciated, and in the future, we will explore scenarios with a more significant time range to enhance the reliability of our prediction model

 

Comments 8: Additionally, the prediction accuracy of the CA-Markov model is not mentioned. It is suggested to include relevant validation results (e.g., Kappa coefficient or other accuracy metrics).

Response 8: I have fixed it

 

Comments 9: The study compares landslide risk across different scenarios, but it does not explicitly state the statistical significance of these results. It is recommended to include significance tests in the conclusion or appendix to verify whether the differences between scenarios are statistically meaningful.

Response 9: I have added the information in abstract and also in discussion

 

Comments 10: The discussion section primarily restates the experimental results. A deeper analysis is recommended. For example, why does the LCC scenario result in the lowest landslide risk? Are there other potential influencing factors? Additionally, does the model’s performance vary significantly across different geographical areas or scenarios, and is there room for further optimization?

Response 10: 

I have added information in 4.1

for the model performance and further optimization I have add information in 4.4

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Dear authors,

Recently I was asked to review a manuscript entitled by “Utilizing remote sensing and random forests to identify optimal land use scenarios to address the increase in landslide susceptibility”. It analyzed the dynamic changes of LULC at regional scales, and the corresponding changes of landslide risk, and authors claimed that integrating geospatial analysis and machine learning in regional planning can promote sustainable land use, and enhance watershed resilience. Overall, the reviewer thinks the manuscript fits with the scope of the journal. It can be considered as a publication after some revisions.

- Abstract: What do you mean “landslide risk” in the whole manuscript? This should be clarified. I didn’t see relevant definitions.

- “Research location” should be “study area”

- Authors made some mistakes when mentioning some professional items, such as “landslide susceptibility”, “landslide vulnerability”, “landslide risk”. Please don’t use “landslide hazard susceptibility”, “landslide hazard vulnerability”, or “landslide disaster”, etc.

-Figure 6 is not clear. Please improv its resolution. The labels of subfigures should also be added, and the names for each subfigure should be provided.

- All the figures. Please add more details in captions of figures.

- Discussion. The comparison among the present work with previous literature should be mentioned. For example:

https://doi.org/10.1007/s10346-021-01775-6

https://doi.org/10.5194/nhess-21-147-2021

Author Response

Comments 1: Abstract: What do you mean “landslide risk” in the whole manuscript? This should be clarified. I didn’t see relevant definitions.

Response 1: We have fixed the “landslide risks” into “landslide hazards” because it is more relevant with the aims of this research.

 

Comments 2: “Research location” should be “study area”

Response 2: We have fixed it into “study area”

 

Comments 3: Authors made some mistakes when mentioning some professional items, such as “landslide susceptibility”, “landslide vulnerability”, “landslide risk”. Please don’t use “landslide hazard susceptibility”, “landslide hazard vulnerability”, or “landslide disaster”, etc.

Response 3: We have fixed is based on your suggestion 

 

Comments 4: Figure 6 is not clear. Please improve its resolution. The labels of subfigures should also be added, and the names for each subfigure should be provided.

Response 4: We have fixed it by improving the resolution to 400 dpi

 

Comments 5: All the figures. Please add more details in captions of figures.

Response 5: We have fixed it by adding more detail caption in each figure

 

Comments 6: 

Discussion. The comparison among the present work with previous literature should be mentioned. For example:

• https://doi.org/10.1007/s10346-021-01775-6
• https://doi.org/10.5194/nhess-21-147-2021

Response 6: We have added previous results from another relevant article as it mentions in the 3rd and 4th paragraph (4.1)

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript presents a strong and relevant contribution to landslide risk management through the integration of remote sensing and machine learning models. The authors have addressed  many of the previous concerns, and the revisions have significantly improved the quality and clarity of the manuscript, but additional clarity in terms of scenario assumptions, model validation, and the practical application of results would enhance the manuscript further. With these additional clarifications, the study has the potential to provide valuable insights for both academic and policy-making communities.

Author Response

Comments 1: The abstract is concise and clear but lacks quantitative results. For example, the changes in landslide risk percentages under different scenarios (BAU, LCC, RSP) or the accuracy evaluation of the models should be briefly mentioned to help readers better understand the study’s contributions and significance.

Response 1: 

Thank you very much

I have fixed it by “From 2017 to 2022, landslide risk increased as the "High" category rose from 33.95% to 37.59% and "Very High" from 10.24% to 12.18%; under BAU 2025, they reach 40.89% and 12.48%, while RSP and LCC reduce "High" to 44.12% and 34.44%, respectively”

 

Comments 2: The introduction provides a broad background discussion, but the review of existing studies could be further condensed. Specifically, the discussion on remote sensing and CA-Markov methods is somewhat lengthy. It is recommended to streamline this section and focus on the specific research gap this study addresses.

Response 2: I have reduced the paragraph and focused on the research gap.

 

Comments 3: The model combination used in this study is innovative, but there is insufficient discussion on why this specific combination (remote sensing data + CA-Markov + RF) was chosen over other potential models, such as deep learning approaches. A brief explanation of this choice would strengthen the justification of the methodology.

Response 3: I have added “Remote sensing, ANN-CA-Markov, and Random Forest are combined to effectively analyze land use changes and landslide susceptibility while balancing accuracy and computational efficiency. Unlike deep learning, this approach requires less data and processing power, making it more practical for GIS-based applications.”

 

Comments 4: The description of "ANN – CA Markov" is unclear, lacking detailed explanations of its working principles and parameter settings. It is recommended to supplement the detailed description of this model, including how it integrates with the Random Forest algorithm and its specific application in predicting land use changes.

Response 4: I added information about CA-Markov in the last paragraph of the 2.3.1.

 

Comments 5: While the paper mentions the use of remote sensing data, it lacks details on data preprocessing steps such as image classification, data cleaning, and validation. Providing more details on these aspects would enhance the study’s transparency and reproducibility.

Response 5: I have fixed it by additional information in the 2.3

 

Comments 6: In the experiments, the paper lacks qualitative expression and analysis regarding the recognition accuracy of different models, and also lacks ablation experiments.

Response 6: I have added in 4.4

 

Comments 7: In predicting land use changes, only data from 2017-2022 are used to predict the 2025 scenario, which is a relatively short time span. It is recommended to supplement historical data or validate the model's long-term applicability.

Response 7: Thank you for your valuable suggestion. In this study, we have chosen to apply the prediction model within a relatively short time frame (2017–2022 to 2025) to minimize potential errors. This approach allows us to evaluate the model's performance in a more controlled manner before extending it to a longer period. However, we fully acknowledge the importance of long-term validation and incorporating a broader historical dataset. Therefore, your suggestion is highly appreciated, and in the future, we will explore scenarios with a more significant time range to enhance the reliability of our prediction model

 

Comments 8: Additionally, the prediction accuracy of the CA-Markov model is not mentioned. It is suggested to include relevant validation results (e.g., Kappa coefficient or other accuracy metrics).

Response 8: I have fixed it

 

Comments 9: The study compares landslide risk across different scenarios, but it does not explicitly state the statistical significance of these results. It is recommended to include significance tests in the conclusion or appendix to verify whether the differences between scenarios are statistically meaningful.

Response 9: I have added the information in abstract and also in discussion

 

Comments 10: The discussion section primarily restates the experimental results. A deeper analysis is recommended. For example, why does the LCC scenario result in the lowest landslide risk? Are there other potential influencing factors? Additionally, does the model’s performance vary significantly across different geographical areas or scenarios, and is there room for further optimization?

Response 10: 

I have added information in 4.1

for the model performance and further optimization I have add information in 4.4

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

I am haappy to see all my concerns have been addressed well. I agree to accept it.

Author Response

Comments 1: Abstract: What do you mean “landslide risk” in the whole manuscript? This should be clarified. I didn’t see relevant definitions.

Response 1: We have fixed the “landslide risks” into “landslide hazards” because it is more relevant with the aims of this research.

 

Comments 2: “Research location” should be “study area”

Response 2: We have fixed it into “study area”

 

Comments 3: Authors made some mistakes when mentioning some professional items, such as “landslide susceptibility”, “landslide vulnerability”, “landslide risk”. Please don’t use “landslide hazard susceptibility”, “landslide hazard vulnerability”, or “landslide disaster”, etc.

Response 3: We have fixed is based on your suggestion 

 

Comments 4: Figure 6 is not clear. Please improve its resolution. The labels of subfigures should also be added, and the names for each subfigure should be provided.

Response 4: We have fixed it by improving the resolution to 400 dpi

 

Comments 5: All the figures. Please add more details in captions of figures.

Response 5: We have fixed it by adding more detail caption in each figure

 

Comments 6: 

Discussion. The comparison among the present work with previous literature should be mentioned. For example:

• https://doi.org/10.1007/s10346-021-01775-6
• https://doi.org/10.5194/nhess-21-147-2021

Response 6: We have added previous results from another relevant article as it mentions in the 3rd and 4th paragraph (4.1)

Author Response File: Author Response.docx