Innovative Flood Impact Monitoring and Harvest Analysis in Oil Palm Plantations Utilizing Geographic Information Systems and Deep Learning
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThis paper aims to develop a method and application for inspecting oil palm plantations impacted by floods during harvesting, which is innovative and promising to a certain extent, but also has some problems:
1. The third part of the paper states “The research framework is illustrated in Figure 1.”, but the actual Figure 1 is “Study area”, please revise.
2. It is mentioned in the paper that the existing research on oil palm maturity detection is affected by factors such as changes in ambient light, inconsistency in the image background, and the small collective data volume, etc. How does this paper solve the interference of the first two factors on the detection results?
3. A software based on GIS and deep learning technology was developed to assess the impact of flooding on oil palm harvesting, how was the image data of oil palm in plantations during flooding acquired during practical use? Was it manually taken and uploaded or automatically taken through the camera? If the latter method, how was the stable transmission of image data ensured?
4. How do individual users and supervisors ensure that the acquired image data of oil palm can reflect the ripening condition of oil palm in the plantation during the flood as a whole, without presenting the partial ripening condition as a whole due to some other factors?
5. The software developed in this paper is geared towards practical engineering applications, but this paper does not provide an economic analysis of the use of the software and the installation and maintenance of the supporting facilities.
Author Response
Reviewer 1
This paper aims to develop a method and application for inspecting oil palm plantations impacted by floods during harvesting, which is innovative and promising to a certain extent, but also has some problems:
Author Response: Thank you for your valuable feedback. We appreciate your insights and recognize the importance of addressing the identified issues. Your comments will help us refine our approach and improve the clarity and impact of our research. We will carefully consider your suggestions and make the necessary revisions to enhance the quality and robustness of our study.
- The third part of the paper states “The research framework is illustrated in Figure 1.”, but the actual Figure 1 is “Study area”, please revise.
Author Response: Thank you for your valuable feedback.
Author Revised: Revised according to the provided recommendations.
- Figure 1: Research framework
- Figure 2: Study area
- It is mentioned in the paper that the existing research on oil palm maturity detection is affected by factors such as changes in ambient light, inconsistency in the image background, and the small collective data volume, etc. How does this paper solve the interference of the first two factors on the detection results?
Author Response: We appreciate this question, as it allows us to elaborate further on our approach to addressing key challenges in assessing the ripeness of oil palm bunches.
Author Revised: This additional explanation should be included in Topic 3.2
- A software based on GIS and deep learning technology was developed to assess the impact of flooding on oil palm harvesting, how was the image data of oil palm in plantations during flooding acquired during practical use? Was it manually taken and uploaded or automatically taken through the camera? If the latter method, how was the stable transmission of image data ensured?
Author Response: Thank you for your valuable feedback.
Author Revised: I have revised and provided additional explanations in Topic 3.3.
- How do individual users and supervisors ensure that the acquired image data of oil palm can reflect the ripening condition of oil palm in the plantation during the flood as a whole, without presenting the partial ripening condition as a whole due to some other factors?
Author Response: We appreciate the valuable question, which allows us to elaborate on our research approach in greater detail. We believe that the proposed method enhances the reliability of assessing flood-affected plantation areas.
Author Revised: I have revised and provided additional explanations in Discussion section. (Highlight the revised text in green.)
- The software developed in this paper is geared towards practical engineering applications, but this paper does not provide an economic analysis of the use of the software and the installation and maintenance of the supporting facilities.
Author Response: Thank you for your valuable feedback.
Author Revised: I have revised and provided additional explanations in Discussion section. (Highlight the revised text in blue.)
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThis research addresses the monitoring and harvesting of oil palm plantations during flood seasons while incorporating data from GIS and images. Overall, I have the following comments:
1. The ending of the intro could have a better focus on the contribution. You can put it in bullets with a better focus on what you are actually contributing to knowledge. Also you can add two lines at the end to mention what will be explained in each section.
2. Line 164, What findings? All the following paragraphs?? That's too long. I would go for not writing it as findings, it is expected to find the literature survey in this part.
3. The related work is too long. I would go for summarizing it in a table with the key point comparison among the references for both the machine learning and the GIS sections
4. Line 223. Is Figure really illustrating the research framework? I can see it has a too short caption mentioning the words "Study area."
5. The reasoning behind the choice of the study area needs to be based on references.
6. It is very confusing for me to see that the RF and SVM have a learning rate... They are not deep learning networks
7. Table 3 equations need references
8. How you apply mobilenet, RF, and SVM? on the same data? RF and SVM are not image classifiers unless you extracted features yet you explicity said images...
9. What models did you employ for the flood image classification??? Same comment applies for the RF and SVM
10 you mentioned that both datasets were available publicly, then why didnt you show in the literature who worked on them and why not compare to them?
11. This is an overall comment; you tend to have very long paragraphs over the article. It is not reader-friendly to have paragraphs that are half a page long. Kindly split the long paragraphs to be of reasonable length.
Overall, I hardly see a contribution; the title and abstract promised innovation, but from the deep learning point of view, only employing mobilenet on images is not innovative. This is more of creating a software tool than a research article
Comments on the Quality of English Language
paragraphs too long
Author Response
Reviewer 2
This research addresses the monitoring and harvesting of oil palm plantations during flood seasons while incorporating data from GIS and images. Overall, I have the following comments:
Author Response: Thank you for your valuable feedback. We appreciate your insights and recognize the importance of addressing the identified issues. Your comments will help us refine our approach and improve the clarity and impact of our research. We will carefully consider your suggestions and make the necessary revisions to enhance the quality and robustness of our study.
- The ending of the intro could have a better focus on the contribution. You can put it in bullets with a better focus on what you are actually contributing to knowledge. Also you can add two lines at the end to mention what will be explained in each section.
Author Response: Thank you for the valuable suggestions. We agree that the conclusion of the introduction should focus more on clearly identifying the advantages and contributions of this research. Therefore, we will revise it by using bullet points for better readability and add two final sentences to outline the structure of each section of the paper.
Author Revised: The revisions have been made according to the suggestions by adding content at the end of the introduction. (Highlight the revised text in green.)
- Line 164, What findings? All the following paragraphs?? That’s too long. I would go for not writing it as findings, it is expected to find the literature survey in this part.
Author Response: Thank you for the valuable suggestions.
Author Revised: for the valuable Reviewed and revised the Related Works section and other relevant sections according to the provided suggestions. (Highlight the revised text in green.)
- The related work is too long. I would go for summarizing it in a table with the key point comparison among the references for both the machine learning and the GIS sections.
Author Response: Thank you for the valuable suggestions.
Author Revised: Rewrite the Related Works section and summarize the comparison in a tabular format.
- Line 223. Is Figure really illustrating the research framework? I can see it has a too short caption mentioning the words “Study area.”
Author Response: Thank you for the valuable suggestions.
Author Revised: Revised by rearranging the images correctly and enhancing the caption descriptions for better clarity. (Highlight the revised text in green.)
- The reasoning behind the choice of the study area needs to be based on references.
Author Response: Thank you for your valuable suggestion. We recognize the importance of justifying the selection of the study area with references to relevant sources. Therefore, we will improve this section by incorporating supporting references that highlight the frequency of flooding, its impact on the agricultural sector, and the significance of oil palm plantations in Thailand. We appreciate this recommendation, as it helps enhance the clarity and credibility of the study area selection in this research.
Author Revised: Revised the Study Area section with appropriate references. (Highlight the revised text in green.)
- It is very confusing for me to see that the RF and SVM have a learning rate… They are not deep learning networks. Thank you for this suggestion.
Author Response: Thank you for this suggestion, which has helped us improve the accuracy of the explanation in the Methodology section.
Author Revised: Revised to correct the error by removing the learning rate for RF and SVM. (Highlight the revised text in green.)
- Table 3 equations need references.
Author Response: Thank you for this suggestion.
Author Revised: Added references for the equations in the table. (Highlight the revised text in green.)
- How you apply mobilenet, RF, and SVM? On the same data? RF and SVM are not image classifiers unless you extracted features, yet you explicitly said images…
Author Response: Thank you for this suggestion, which has helped us improve the accuracy of the explanation in the Methodology section.
Author Revised: Added explanations based on the suggestions in Topic 3.5. (Highlight the revised text in green.)
- What models did you employ for the flood image classification??? Same comment applies for the RF and SVM.
Author Response: Thank you for this suggestion, which has helped us improve the accuracy of the explanation in the Methodology section.
Author Revised: Added explanations based on the suggestions in Topic 3.5. (Highlight the revised text in green.)
- You mentioned that both datasets were available publicly, then why didn’t you show in the literature who worked on them and why not compare to them?
Author Response: Thank you for this suggestion, which has helped us improve the accuracy of the explanation in the Methodology section.
Author Revised: Added explanations based on the suggestions in Topic 3.2. (Highlight the revised text in blue.)
- This is an overall comment; you tend to have very long paragraphs over the article. It is not reader-friendly to have paragraphs that are half a page long. Kindly split the long paragraphs to be of reasonable length.
Author Response: Thank you for the valuable suggestions.
Author Revised: Rewrite the Related Works section and summarize the comparison in a tabular format.
Overall, I hardly see a contribution; the title and abstract promised innovation, but from the deep learning point of view, only employing mobilenet on images is not innovative. This is more of creating a software tool than a research article.
Author Response: Thank you for the valuable observations. We acknowledge the concerns regarding the academic contribution of this research, particularly in the aspects of Deep Learning and the innovation presented in the article.
Author Revised: Add The key contributions of this research at the end of the introduction. (Highlight the revised text in blue.)
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsThe article addresses a critical challenge in agricultural management: mitigating the effects of flooding on oil palm plantations, a significant economic sector in flood-prone regions like southern Thailand. Utilizing GIS and deep learning, the study presents a novel application for real-time flood monitoring and ripeness classification of oil palm bunches. The results demonstrate impressive classification accuracies of 96.80%–98.29% for flood impact and 97.60%–99.75% for fruit ripeness. This research has strong practical implications for disaster management, providing timely support to farmers while advancing geospatial and AI-driven agricultural tools. Its innovative integration of crowdsourced data and satellite imagery underscores its utility in scalable, real-world applications.
My Comments and Questions for Authors
1. Introducrion. The application includes a machine learning model for ripeness classification, which, while technologically advanced, appears to duplicate a task that experienced farmers can perform visually. Could the authors elaborate on the added value of this feature, particularly in cases where farmers already possess the expertise to evaluate ripeness manually?
2. Section 3.2. The ripeness model relies on a dataset of approximately 9,223 images. Given the variability in natural lighting and environmental conditions, how robust is this dataset for generalization to other regions or scenarios outside the study area? Were measures like data augmentation employed to address potential biases?
3. Section 3.2. The application integrates crowdsourced images and data. How do the authors ensure the quality and reliability of these inputs, especially considering that they significantly influence the classification outcomes?
4. Section 3.6. The methodology mentions the integration of satellite imagery for flood verification. While this enhances accuracy, how does it address limitations like cloud cover or delays in satellite data availability during real-time flood events?
5. Section 3.7. The study briefly mentions the application's user interface and functionality. Were usability tests conducted with end-users (e.g., farmers or local administrators)? If so, what were the key findings, and how were they incorporated into the application design?
6. Results and Discussion. The application demonstrates promise for flood management, but its practicality in diverse socioeconomic contexts remains unclear. Could the authors discuss the scalability and affordability of this system, particularly for smallholder farmers or regions with limited technological infrastructure?
The research presents a well-conceptualized and meticulously executed solution for addressing the impact of flooding on oil palm plantations. However, the inclusion of machine learning for ripeness classification – despite its novelty – may not add significant practical value compared to manual evaluation by farmers. Overall, the article represents a high-quality academic contribution and could be suitable for publication in AgriEngineering.
Comments for author File: Comments.pdf
Author Response
Reviewer3
Original:
The article addresses a critical challenge in agricultural management: mitigating the effects of flooding on oil palm plantations, a significant economic sector in flood-prone regions like southern Thailand. Utilizing GIS and deep learning, the study presents a novel application for real-time flood monitoring and ripeness classification of oil palm bunches. The results demonstrate impressive classification accuracies of 96.80%–98.29% for flood impact and 97.60%–99.75% for fruit ripeness. This research has strong practical implications for disaster management, providing timely support to farmers while advancing geospatial and AI-driven agricultural tools. Its innovative integration of crowdsourced data and satellite imagery underscores its utility in scalable, real-world applications.
Author Response: Thank you for your valuable feedback. We appreciate your insights and recognize the importance of addressing the identified issues. Your comments will help us refine our approach and improve the clarity and impact of our research. We will carefully consider your suggestions and make the necessary revisions to enhance the quality and robustness of our study.
- Introduction. The application includes a machine learning model for ripeness classification, which, while technologically advanced, appears to duplicate a task that experienced farmers can perform visually. Could the authors elaborate on the added value of this feature, particularly in cases where farmers already possess the expertise to evaluate ripeness manually?
Author Response: Thank you for the valuable suggestions.
Author Revised: Add The key contributions of this research at the end of the introduction. (Highlight the revised text in blue and green.)
- Section 3.2. The ripeness model relies on a dataset of approximately 9,223 images. Given the variability in natural lighting and environmental conditions, how robust is this dataset for generalization to other regions or scenarios outside the study area? Were measures like data augmentation employed to address potential biases?
Author Response: Thank you for the valuable suggestions.
Author Revised: The explanation has been expanded in Topic 3.2 and 3.3. (Highlight the revised text in green.)
- Section 3.2. The application integrates crowdsourced images and data. How do the authors ensure the quality and reliability of these inputs, especially considering that they significantly influence the classification outcomes?
Author Response: Thank you for the valuable suggestions.
Author Revised: The explanation has been expanded in Topic 3.2 and 3.3. (Highlight the revised text in blue.)
- Section 3.6. The methodology mentions the integration of satellite imagery for flood verification. While this enhances accuracy, how does it address limitations like cloud cover or delays in satellite data availability during real-time flood events?
Author Response: Thank you for the valuable suggestions.
Author Revised: The explanation has been expanded in Topic 3.7 (Highlight the revised text in blue.)
- Section 3.7. The study briefly mentions the application’s user interface and functionality. Were usability tests conducted with end-users (e.g., farmers or local administrators)? If so, what were the key findings, and how were they incorporated into the application design?
Author Response: Thank you for the valuable suggestions.
Author Revised: The explanation has been expanded in Discussion Sections (Highlight the revised text in blue.)
- Results and Discussion. The application demonstrates promise for flood management, but its practicality in diverse socioeconomic contexts remains unclear. Could the authors discuss the scalability and affordability of this system, particularly for smallholder farmers or regions with limited technological infrastructure?
Author Response: Thank you for the valuable suggestions.
Author Revised: The explanation has been expanded in Discussion Sections (Highlight the revised text in blue.)
The research presents a well-conceptualized and meticulously executed solution for addressing the impact of flooding on oil palm plantations. However, the inclusion of machine learning for ripeness classification – despite its novelty – may not add significant practical value compared to manual evaluation by farmers. Overall, the article represents a high-quality academic contribution and could be suitable for publication in AgriEngineering.
Author Response: Thank you for your thoughtful feedback. We appreciate your recognition of our research and its potential contribution. Regarding the use of machine learning for ripeness classification, we acknowledge that experienced farmers can manually assess ripeness. However, our approach aims to enhance flood impact assessment by automating data collection and integrating it with spatial analysis. Your insights are highly valuable, and we will carefully consider them to improve the clarity and practical relevance of our study.
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsThe reviewers' comments and suggestions have been revised, and the paper is ready for publication.
Author Response
Thank you for your valuable feedback and constructive suggestions. We have carefully revised the manuscript based on the reviewers’ comments, ensuring that all suggested improvements have been addressed. We appreciate your time and effort in reviewing our work, and we are pleased that the paper is now ready for publication.
Reviewer 2 Report
Comments and Suggestions for AuthorsThe authors did modify the article and correct the major misconcepts that were previously presented.
On a minor note:
1. remove duplicated Figure 2 in caption
2. Add the different dataset sizes in Table1
Author Response
Reviewer 2
Comments and Suggestions for Authors
The authors did modify the article and correct the major misconcepts that were previously presented.
On a minor note:
Author Response: We sincerely appreciate your valuable comments and suggestions, which have helped improve the clarity and accuracy of our manuscript.
- remove duplicated Figure 2 in caption
Author Action: Removal of duplicated Figure 2 in the caption
We have corrected this issue by ensuring that Figure 2 is referenced appropriately and that any duplicated mention in the caption has been removed.
- Add the different dataset sizes in Table1
Author Action: Addition of different dataset sizes in Table
- We have revised Table by including a new column specifying the dataset size for each data type. This addition provides greater transparency regarding the volume of data used in our research. The updated table now explicitly states the number of images for each dataset and clarifies spatial data availability.
Additionally, to ensure transparency in our revisions, all modifications in the manuscript have been highlighted in yellow for easy identification. This includes the corrections made in the caption of Figure 2 and the updated dataset sizes in Table 1.
We appreciate your constructive feedback, which has contributed to refining our manuscript. Thank you for your time and effort in reviewing our work.