Research on Walnut (Juglans regia L.) Classification Based on Convolutional Neural Networks and Landsat-8 Remote Sensing Imagery

Round 1

Reviewer 1 Report (Previous Reviewer 3)

I was already satisfied with the revisions done in the previous submission. I have no further comments. 

Author Response

Dear Reviewer,

Thank you for your suggestion and recognition of our related work. We are sincerely grateful. Here is the description of the revision we are presenting to you.

1. After considering the issues contributing to the inaccurate detection of walnut area identification, we closely examined our discussion and identified a lack of detailed description of this problem. In the current revision, we have supplemented the discussion with relevant details on improving accuracy. The discussion primarily focuses on (1) satellite resolution, transit period, and related factors, (2) the use of relevant algorithms, (3) the potential for improved accuracy through the integration of multi-source data, and (4) a brief mention of the primary research direction and future focus on lightweight and efficient methodologies.

2. With regard to the issue of self-citation in the references mentioned by the editor, we believe that it is a misunderstanding after double-checking.

2.1 First of all, we would like to explain the problem of citing too much literature from the submitted journals, we did not find this situation before, due to the corresponding demand, we deleted the references of 27,28,41,65 in this revision. About 26, 27,28 citation content continuity, relative 26 and our content relevance is more closely, so deleted 27,28. About 41 deletion, due to the part there are other cited references to support the description, so deleted. Regarding the deletion of 65, it was deleted because it was a comparison of the results in the other citations, which were relatively low in a similar setting.

2.2 Regarding the question raised about our excessive self-citation, to clarify to you, the references mentioned are 22,27,38,39,57,59,63,69.We made a careful review and finally found that we have 2 self-citations in Ref 38,63.Regarding Ref 38, we are following the journal's request for additional clarification, and in the introductory section, which describes the work of the article, we difference between the work of this paper and that paper, as well as reviewing it in the discussion. Regarding Ref 63, the article appeared self-citation, due to the strong correlation of the relevant content, the placement of the content position in the discussion section, the discussion of the citation is expressed from a different perspective, at the same time in the section has been part of the citation deleted, in order to avoid the lack of persuasive power of the supporting material is not deleted for the time being.

With regard to the other articles, we believe that this is due to a formatting problem, as we use "EndNote", stylized as "ACS", which leads to the abbreviation of the author's name, which is misunderstood as self-citation. The full names of the documents in Refs. 22, 27, 39, 57, 59, and 69 are, respectively “Xiangping Li”，“Xuyang Li”，“Xiang Li”，“Xuelong Li”，“Xuecao Li”，“Xingrong Li”，The author of this article appears as "Xu Li". So there is no self-citations problem. After reviewing the content of these articles and descriptions, some of which are located in the introduction, the deletion of the overall description will have little impact, and some of which are supported by several articles and can be deleted, in order to avoid misinterpretation, we have therefore deleted Ref 22,27,39,57,59. As Ref 69 is located in the Discussion section, for the same reason as Ref 63, due to the strong relevance of the relevant content, the placement of the content location in the Discussion section, the relevant citations in the Discussion are expressed from different perspectives, and in the section there has been a partial deletion of the citations, in order to avoid insufficient persuasive power of the supporting material, so for the time being, no deletion has been carried out.

Thank you for your previous advice and recommendations, as well as for recognizing our work. Your suggestions are valuable for both our writing and research.

Kind Regards,

Jingming Wu

Author Response File: Author Response.pdf

Reviewer 2 Report (Previous Reviewer 2)

I have received sufficient responses to most of the questions posed during the earlier review rounds. The authors have made an impressive effort to consider most of the suggestions I made in my previous revisions.

However, the manuscript needs minor revisions.

What elements contribute to inaccurate definitions of walnut areas? Space imagery quality? Errors in recognition algorithms? Or some other factors?

Add research limitations to the Discussion section.

Author Response

Dear Reviewer,

Thank you for your suggestion and recognition of our related work. We are sincerely grateful. Here is the description of the revision we are presenting to you.

1. After considering the issues contributing to the inaccurate detection of walnut area identification, we closely examined our discussion and identified a lack of detailed description of this problem. In the current revision, we have supplemented the discussion with relevant details on improving accuracy. The discussion primarily focuses on (1) satellite resolution, transit period, and related factors, (2) the use of relevant algorithms, (3) the potential for improved accuracy through the integration of multi-source data, and (4) a brief mention of the primary research direction and future focus on lightweight and efficient methodologies.

2. With regard to the issue of self-citation in the references mentioned by the editor, we believe that it is a misunderstanding after double-checking.

2.1 First of all, we would like to explain the problem of citing too much literature from the submitted journals, we did not find this situation before, due to the corresponding demand, we deleted the references of 27,28,41,65 in this revision. About 26, 27,28 citation content continuity, relative 26 and our content relevance is more closely, so deleted 27,28. About 41 deletion, due to the part there are other cited references to support the description, so deleted. Regarding the deletion of 65, it was deleted because it was a comparison of the results in the other citations, which were relatively low in a similar setting.

2.2 Regarding the question raised about our excessive self-citation, to clarify to you, the references mentioned are 22,27,38,39,57,59,63,69.We made a careful review and finally found that we have 2 self-citations in Ref 38,63.Regarding Ref 38, we are following the journal's request for additional clarification, and in the introductory section, which describes the work of the article, we difference between the work of this paper and that paper, as well as reviewing it in the discussion. Regarding Ref 63, the article appeared self-citation, due to the strong correlation of the relevant content, the placement of the content position in the discussion section, the discussion of the citation is expressed from a different perspective, at the same time in the section has been part of the citation deleted, in order to avoid the lack of persuasive power of the supporting material is not deleted for the time being.

With regard to the other articles, we believe that this is due to a formatting problem, as we use "EndNote", stylized as "ACS", which leads to the abbreviation of the author's name, which is misunderstood as self-citation. The full names of the documents in Refs. 22, 27, 39, 57, 59, and 69 are, respectively “Xiangping Li”，“Xuyang Li”，“Xiang Li”，“Xuelong Li”，“Xuecao Li”，“Xingrong Li”，The author of this article appears as "Xu Li". So there is no self-citations problem. After reviewing the content of these articles and descriptions, some of which are located in the introduction, the deletion of the overall description will have little impact, and some of which are supported by several articles and can be deleted, in order to avoid misinterpretation, we have therefore deleted Ref 22,27,39,57,59. As Ref 69 is located in the Discussion section, for the same reason as Ref 63, due to the strong relevance of the relevant content, the placement of the content location in the Discussion section, the relevant citations in the Discussion are expressed from different perspectives, and in the section there has been a partial deletion of the citations, in order to avoid insufficient persuasive power of the supporting material, so for the time being, no deletion has been carried out.

Thank you for your previous advice and recommendations, as well as for recognizing our work. Your suggestions are valuable for both our writing and research.

Kind Regards,

Jingming Wu

Author Response File: Author Response.pdf

Reviewer 3 Report (Previous Reviewer 1)

Author Response

Dear Reviewer,

Thank you for your suggestion and recognition of our related work. We are sincerely grateful. Here is the description of the revision we are presenting to you.

1. After considering the issues contributing to the inaccurate detection of walnut area identification, we closely examined our discussion and identified a lack of detailed description of this problem. In the current revision, we have supplemented the discussion with relevant details on improving accuracy. The discussion primarily focuses on (1) satellite resolution, transit period, and related factors, (2) the use of relevant algorithms, (3) the potential for improved accuracy through the integration of multi-source data, and (4) a brief mention of the primary research direction and future focus on lightweight and efficient methodologies.

2. With regard to the issue of self-citation in the references mentioned by the editor, we believe that it is a misunderstanding after double-checking.

2.1 First of all, we would like to explain the problem of citing too much literature from the submitted journals, we did not find this situation before, due to the corresponding demand, we deleted the references of 27,28,41,65 in this revision. About 26, 27,28 citation content continuity, relative 26 and our content relevance is more closely, so deleted 27,28. About 41 deletion, due to the part there are other cited references to support the description, so deleted. Regarding the deletion of 65, it was deleted because it was a comparison of the results in the other citations, which were relatively low in a similar setting.

2.2 Regarding the question raised about our excessive self-citation, to clarify to you, the references mentioned are 22,27,38,39,57,59,63,69.We made a careful review and finally found that we have 2 self-citations in Ref 38,63.Regarding Ref 38, we are following the journal's request for additional clarification, and in the introductory section, which describes the work of the article, we difference between the work of this paper and that paper, as well as reviewing it in the discussion. Regarding Ref 63, the article appeared self-citation, due to the strong correlation of the relevant content, the placement of the content position in the discussion section, the discussion of the citation is expressed from a different perspective, at the same time in the section has been part of the citation deleted, in order to avoid the lack of persuasive power of the supporting material is not deleted for the time being.

With regard to the other articles, we believe that this is due to a formatting problem, as we use "EndNote", stylized as "ACS", which leads to the abbreviation of the author's name, which is misunderstood as self-citation. The full names of the documents in Refs. 22, 27, 39, 57, 59, and 69 are, respectively “Xiangping Li”，“Xuyang Li”，“Xiang Li”，“Xuelong Li”，“Xuecao Li”，“Xingrong Li”，The author of this article appears as "Xu Li". So there is no self-citations problem. After reviewing the content of these articles and descriptions, some of which are located in the introduction, the deletion of the overall description will have little impact, and some of which are supported by several articles and can be deleted, in order to avoid misinterpretation, we have therefore deleted Ref 22,27,39,57,59. As Ref 69 is located in the Discussion section, for the same reason as Ref 63, due to the strong relevance of the relevant content, the placement of the content location in the Discussion section, the relevant citations in the Discussion are expressed from different perspectives, and in the section there has been a partial deletion of the citations, in order to avoid insufficient persuasive power of the supporting material, so for the time being, no deletion has been carried out.

Thank you for your previous advice and recommendations, as well as for recognizing our work. Your suggestions are valuable for both our writing and research.

Kind Regards,

Jingming Wu

Author Response File: Author Response.pdf

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

The manuscript entitled Research on Walnut Classification Based on Convolutional Neural Networks and Landsat-8 Remote Sensing Imagery after Major revision

The abstract

This section could be expanded a bit. In this section, the authors only need to summarize the main results of this research in two sentences. In this section, the main results of the manuscript need to be added. Authors need to add the main methods used in this research at the end.

Introduction

Lines between 47 and 58. This paragraph could be rewritten again in the scientific sense. This means that the authors need to add more explanations about the walnut species and its distribution in the world. The most important thing in this research is the connection between the distribution of walnut and the influence of climate on the area with walnuts.

In lines 109-123 in this part, the authors need to explain in more detail how the Google Earth Engine did this work with the forest analyzes. The area with Orhard is very difficult to determine even with the super pixel analysis. So the authors need to explain the methodology better.

Figure 1: The geographic coordinates are really hard to see. So it is necessary to use larger numbers.

As the authors deal with the exact geographical analysis, I strongly recommend reading and citing two valuable references

-Lv, X., Ming, D., Chen, Y., & Wang, M. (2019). Very high resolution remote sensing image classification with SEEDS-CNN and scale effect analysis for superpixel CNN classification. International Journal of Remote Sensing, 40(2), 506-531.

- Valjarević, A., Algarni, S., Morar, C., Grama, V., Stupariu, M., Tiba, A., & Lukić, T. (2023). The coastal fog and ecological balance for plants in the Jizan region, Saudi Arabia. Saudi Journal of Biological Sciences, 30(1), 103494.

Materials and methods

The first part of this section must consist of more text. This text must summarize the main methods as well as the methodology used in this study.

In this part of the manuscript, it is not clear what the main methodology is, so it should be better explained.

Results

This section should be divided into five parts.

The first part could be a numerical part,

The second statistical part

The third theoretical part

The fourth part is remote sensing

The fifth GIS.

Discussion

In this section, the authors add more research results already published.

They also explain the advantages and disadvantages of this research.

Conclusion

This section can be expanded with a few sentences.

I recommend a Major revision

The work is good and scientifically correct

Good luck to the authors

Reviewer#2

Author Response

Dear Reviewer,

Thank you for your careful reading, valuable comments, and constructive suggestions. They have greatly improved our manuscript. Below is a detailed description of the modifications we made:

1. In the abstract, we found your suggestions to be very insightful. In this revision, we summarized the research findings, added the final relevant research results, and emphasized the main methods used in this study at the end of the abstract. If you are reviewing the revised version, this section can be found in L17-21 and L36-44, or in the accepted version, it is in L17-19 and L33-39.

2. Regarding the introduction, we greatly appreciate the two articles you recommended. Both of them were excellent, and I had read them recently, which made rereading them very beneficial. As for the detailed modifications in the introduction, we provided further descriptions of walnut-related species and habits, as well as the work related to Google Earth Engine and satellite imagery. We also made additional clarifications in the Materials and Methods section. However, it's worth noting that in a previous revision, we provided detailed descriptions in this part, but the academic editor advised to remove some non-essential descriptions, which we have done. If you are reviewing the revised version, this section can be found in L59-66 and L69-82, or in the accepted version, it is in L57-70.

3. In the Materials and Methods section, we acknowledge that the geographical map-related figures were difficult to read. As a result, we have redrawn the geographical maps. Additionally, we rephrased many parts of the first section of this part and provided detailed supplements regarding the methods and main techniques used in this revision. If you are reviewing the revised version, this section can be found in L199, L265-277, or in the accepted version, it is in L186, L248-259.

4. Concerning the Results section, we agree with your suggestions. However, in a previous revision, we were advised by the academic editor to omit these parts, as they were believed to affect readability. We retained some and moved most of this content to the Materials and Methods section. We also added supplemental remote sensing images in this revision. If you are reviewing the revised version, this section can be found in L583-584, or in the accepted version, it is in L563-564.

5. We appreciate your recognition of the Discussion section.

6. Regarding the Conclusion section, we referenced relevant literature from the journal and provided a more concise summary in this revision. If you are reviewing the revised version, this section can be found in L699-743, or in the accepted version, it is in L679-692.

In addition, we would like to report on other work we conducted in this revision:

7. We added electronic links to the original code of relevant model modifications in this revision.

8. In this revision, we provided further explanations about the satellite images used and described some of the early unsuccessful experiments. We also provided additional information about the data used. If you are reviewing the revised version, this section can be found in L203-210, L214-222, L226-232, L247-258, or in the accepted version, it is in L189-196, L200-208, L212-219, L234-243.

9. We further elaborated on the advantages of Google Earth Engine (GEE), keeping the content of the description under control. The rationale for this is the same as the one we explained earlier regarding the additional information about walnut-related species and habits. If you are reviewing the revised version, this section can be found in L133-138, L141-143, or in the accepted version, it is in L120-124, L128-130.

10. We provided further explanations of the image data and supplemented information about relevant parameters in this revision. If you are reviewing the revised version, this section can be found in L306-322, L334-340, L358-364, L394-403, L426-435, or in the accepted version, it is in L287-303, L315-321, L339-345, L375-384, L407-416.

All of your suggestions have been incredibly valuable and have provided important guidance for my research and paper writing. Thank you once again for your suggestions.

Kind Regards,

Jingming Wu

Author Response File: Author Response.pdf

Reviewer 2 Report

I appreciate that the authors have considered most of the comments and made significant changes to the manuscript.

However, the manuscript needs minor revisions.

Please state where the neural network computer code originated (e-link). For each convolutional neural network, please include the e-links directly in the manuscript. The original codes for each programme are as follows:

AlexNet https://paperswithcode.com/method/alexnet

VGG https://paperswithcode.com/method/vgg

GoogLeNet https://paperswithcode.com/method/googlenet

ResNets https://paperswithcode.com/method/resnet

EfficientNet https://paperswithcode.com/method/efficientnet

If you obtained the code from these sources, please include this information in the manuscript. Please include other links to the datasets that you used if you obtained the code from other sources.

Author Response

Dear Reviewer,

We extend our gratitude for your meticulous review, valuable comments, and constructive recommendations, which have significantly enhanced our manuscript. Below, we provide a comprehensive account of the revisions we've made:

1. We greatly appreciate your suggestions. In this revision, we have included electronic links to the original code, as you recommended.

In addition, we wish to apprise you of other significant modifications made during this revision:

2. In the abstract, we have summarized our research findings, incorporated the latest relevant research outcomes, and highlighted the primary methodologies employed in our study. If you are reviewing the revised version, you can locate this section at L17-21 and L36-44, or in the accepted version, it is found at L17-19 and L33-39.

3. In the introduction, we have provided further descriptions of walnut-related species and their habits, as well as elaborated on the work related to Google Earth Engine and satellite technology. If you are reviewing the revised version, these sections can be found at L59-66, L69-82, L133-138, and L141-143. In the accepted version, they are at L57-70, L120-124, L128-130.

4. In the Materials and Methods section, we have redrawn the geographical maps, made additional modifications to numerical data, provided further explanations about image data, and supplemented details about relevant parameters. If you are reviewing the revised version, these updates are located at L199, L265-277, L306-322, L334-340, L358-364, L394-403, and L426-435. In the accepted version, they can be found at L186, L248-259, L287-303, L315-321, L339-345, L375-384, and L407-416.

5. In the Results section, we have introduced a new set of classification figures. If you are reviewing the revised version, these figures can be found at L583-584, or in the accepted version, they are at L563-564.

6. Regarding the Conclusion section, we have consulted relevant literature from the journal to provide a more concise summary. If you are reviewing the revised version, this section is available at L699-743, or in the accepted version, it can be found at L679-692.

Your insights and suggestions are of utmost importance and have significantly guided our academic and research endeavors. Once again, we express our gratitude for your valuable input.

Kind Regards,

Jingming Wu

Author Response File: Author Response.pdf

Reviewer 3 Report

Thank you for the invitation. The manuscript in question requires some revisions before it proceeds to further review processes. Here are my comments:

- The authors should explain why they chose Landsat-8 images over Sentinel-2, which has better spatial resolution. 

- The authors mention using seven different spectral bands, but they haven't specified which bands or indices were used for generating the final sample images. It's important to clarify the reasons for merging different bands. Additionally, sharing insights on the spectral characteristics of walnut crops would help understand which bands/indices work best for detecting walnuts.

- You should emphasize the advantages of using GEE, the effectiveness of cloud masks, and the significance of geometric corrections. Take a look at some example articles for reference: https://doi.org/10.1007/s10661-022-10437-6 and https://doi.org/10.1016/j.compag.2023.108105

-The paper should provide more details about the manual field survey used to delineate walnut crop areas. Explain how, when, and with what precision the ground samples were collected.

-Specify the number of images collected during August and September 2022. Also, provide the tile and row information for these input images, not just the total number of partitioned images.

- Do the training and validation datasets have any overlapping areas among the partitioned images?

-Describe and provide the hyperparameters and their values optimized for all your models. This is crucial for AI-based studies.

- Consider alternative ways to present the architecture of your models. The current figures may strain readers' eyes.

- The most significant issue in the paper is the absence of output classified images. The authors should include sample classification results for test images, showing examples of TP, FP, TN, etc. For each sample scene, provide the annotated class and the predicted class underneath the images.

Author Response

Dear Reviewer,

We extend our sincere appreciation for your thorough examination, valuable insights, and constructive recommendations. Your input has significantly enhanced the quality of our manuscript. Below, we provide a detailed account of the revisions made:

1. We genuinely thank you for your guidance, especially regarding the issue of Landsat-8 and Sentinel-2 images. In relation to the classification problem, it is indeed ideal to employ higher-resolution images. In the initial stages of our research, we followed this path. However, during the image processing, we encountered persistent issues that led to suboptimal results, despite various attempts to rectify them. Regrettably, we did not explicitly mention these challenges in the manuscript. In this revision, we have highlighted the use of the relevant satellites in the title and further elaborated on this aspect. If you are reviewing the revised version, these changes can be found at L214-222. In the accepted version, they appear at L200-208.

2. In this revision, we have provided additional explanations regarding the specific spectral bands we used. As you correctly pointed out, our initial band selection was intended to understand the spectral characteristics of walnut crops and identify parameters suitable for detecting walnuts. We greatly appreciate your suggestion. If you are reviewing the revised version, these updates are available at L226-232. In the accepted version, they can be found at L212-219.

3. Indeed, it is crucial to emphasize the advantages of using Google Earth Engine (GEE) and its relevance. In this revision, we have added further explanations in response to your feedback. It is worth noting that we had initially included descriptions related to GEE and crop introductions, among other details. However, these were removed based on the recommendations of the academic editor. We have attempted to strike the right balance in this revision. Once again, we appreciate your guidance. If you are reviewing the revised version, these changes can be found at L133-138 and L141-143. In the accepted version, they are located at L120-124 and L128-130.

4. We appreciate your recommendation of the articles, which we have duly cited. It might interest you to know that these articles were already saved on my computer, and their content not only inspired this manuscript but also future related work.

5. In this revision, we have made further modifications to the descriptions of the manual survey, sample data, and have provided additional details about image acquisition, including time, quantity, and row-column data. If you are reviewing the revised version, these changes can be found at L203-210. In the accepted version, they appear at L189-196.

6. Regarding the training and validation sets, it is important to clarify that there were no overlapping parts in the original images. The potential overlap you noted could occur when combining data from different spectral bands. However, it's essential to emphasize that the selection of overlapping areas was entirely random and varied with each experiment.

7. In this revision, we have provided comprehensive explanations for all model parameters used, detailing the specific parameters employed in our study. If you are reviewing the revised version, you can find these additions at L306-322, L334-340, L358-364, L394-403, and L426-435. In the accepted version, they are located at L287-303, L315-321, L339-345, L375-384, and L407-416.

8. From my perspective, I concur with your suggestion that the extensive numerical data can be overwhelming. I'm not certain about the version of the document you have, but in the document I have, some of the images have been moved to the appendix, and certain result figures were removed based on the feedback from the academic editor. If you are reviewing the revised version, these changes can be found at L583-584. In the accepted version, they appear at L563-564.

9. Regarding the images, initially, they were visible in the main body of the manuscript. You mentioned "confusion matrix" parameters. I'm not sure if the document provided to you includes an appendix, but if it does, you may find some related images there. As for other images, their inclusion was subject to the requirements of the academic editor. Additionally, in this revision, we have introduced a new set of classification figures showcasing different methods in different study areas. I'm uncertain if these figures are included in the appendix or elsewhere in the document.

In addition to the above, we would like to inform you of the following work conducted in this revision:

10. In the abstract, we have summarized our research findings, included the latest relevant research outcomes, and highlighted the primary methodologies employed in our study. If you are reviewing the revised version, these changes are available at L17-21 and L36-44. In the accepted version, they appear at L17-19 and L33-39.

11. In the introduction, we have further described walnut-related species and their characteristics. However, we have exercised caution in the level of detail provided, consistent with our approach to supplementing information on GEE and related matters. If you are reviewing the revised version, these changes can be found at L59-66 and L69-82. In the accepted version, they appear at L57-70.

12. We have redrawn the geographical maps in this revision, emphasizing the visibility of previously unclear numerical data. Additionally, we have provided detailed supplements about the methods used and the primary techniques employed. If you are reviewing the revised version, these updates are located at L199 and L265-277. In the accepted version, they can be found at L186 and L248-259.

13. In the Conclusion section, we have provided a more comprehensive summary, referring to relevant literature from the journal. If you are reviewing the revised version, this section is available at L699-743. In the accepted version, it can be found at L679-692.

We value your feedback, which has proven to be invaluable in guiding our research and manuscript development. Once again, we express our gratitude for your valuable input.

Kind Regards,

Jingming Wu

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The paper titled Research on Walnut Classification Based on Convolutional Neural Networks and Landsat-8 Remote Sensing Imagery can be accepted in my opinion.

The authors have greatly improved the previous version of the manuscript and corrected all errors in the text.

Therefore, I recommend the editor to accept this work

Reviewer 2

Author Response

Dear Reviewer,

Thank you very much for taking the time to read and revise my paper. I am grateful for your valuable suggestions. Your comprehensive corrections to the structure, content, research methods, and results of my paper have played a crucial role in improving its quality. Below is a detailed description of the revisions we made to the manuscript.

1. We carefully reviewed the entire paper and made various grammar and language edits.

2. We provided further clarification on the concept of 'efficiency,' which refers to the time consumed by the algorithm used in this study to produce the final results. If you are looking at the revised version, this information will appear in L35-37; if you are reviewing the accepted revision, it will be in L33-35.

3. Our research is aimed at enhancing efficiency, but we realized that the previous description could be misleading during our grammar revisions. As you suggested, we made additional explanations in the abstract to minimize any leaps in the results. We first describe the importance of forest products, followed by the significance of relevant technologies, feasibility, the platform used, the methodological model, and evaluation metrics. We then present specific improved results and finally engage in a discussion of the outcomes. If you are looking at the revised version, this information will appear in L13-15 and L35-45; for the accepted revision, it will be in L12-15 and L33-38.

4. We made extensive modifications and deletions to the description of walnuts and supplemented information about deep learning for forest mapping. If you are looking at the revised version, this information will appear in L59-101 and L120-129; for the accepted revision, it will be in L52-73 and L91-100.

5. We redefined GEE as a platform, recognizing that our previous description was not accurate. In this revision, we provided a more accurate account of this aspect. If you are looking at the revised version, this information will appear in L138-152; for the accepted revision, it will be in L108-119.

6. We discussed our research in the context of your recommended related study at https://www.mdpi.com/2076-3417/13/9/5666 and made detailed distinctions between the two studies. We also elaborated on the potential future integration of these two lines of research. If you are looking at the revised version, this information will appear in L700-715; for the accepted revision, it will be in L663-674.

7. We have moved Figure 6 to the appendix, but if you still find it inappropriate, we are open to its removal. Figure 6 pertains to classification-related graphics.

Once again, I appreciate your guidance and your thorough review and revisions of my paper. With your help, I hope to produce an outstanding paper and sincerely hope for its publication in your esteemed journal.

Jingming Wu

Author Response File: Author Response.pdf

Reviewer 3 Report

Thank you for the revisions. I have no further comments. 

Author Response

Dear Reviewer,

Thank you very much for taking the time to read and revise my paper. I am grateful for your valuable suggestions. Your comprehensive corrections to the structure, content, research methods, and results of my paper have played a crucial role in improving its quality. Below is a detailed description of the revisions we made to the manuscript.

1. We carefully reviewed the entire paper and made various grammar and language edits.

2. We provided further clarification on the concept of 'efficiency,' which refers to the time consumed by the algorithm used in this study to produce the final results. If you are looking at the revised version, this information will appear in L35-37; if you are reviewing the accepted revision, it will be in L33-35.

3. Our research is aimed at enhancing efficiency, but we realized that the previous description could be misleading during our grammar revisions. As you suggested, we made additional explanations in the abstract to minimize any leaps in the results. We first describe the importance of forest products, followed by the significance of relevant technologies, feasibility, the platform used, the methodological model, and evaluation metrics. We then present specific improved results and finally engage in a discussion of the outcomes. If you are looking at the revised version, this information will appear in L13-15 and L35-45; for the accepted revision, it will be in L12-15 and L33-38.

4. We made extensive modifications and deletions to the description of walnuts and supplemented information about deep learning for forest mapping. If you are looking at the revised version, this information will appear in L59-101 and L120-129; for the accepted revision, it will be in L52-73 and L91-100.

5. We redefined GEE as a platform, recognizing that our previous description was not accurate. In this revision, we provided a more accurate account of this aspect. If you are looking at the revised version, this information will appear in L138-152; for the accepted revision, it will be in L108-119.

6. We discussed our research in the context of your recommended related study at https://www.mdpi.com/2076-3417/13/9/5666 and made detailed distinctions between the two studies. We also elaborated on the potential future integration of these two lines of research. If you are looking at the revised version, this information will appear in L700-715; for the accepted revision, it will be in L663-674.

7. We have moved Figure 6 to the appendix, but if you still find it inappropriate, we are open to its removal. Figure 6 pertains to classification-related graphics.

Once again, I appreciate your guidance and your thorough review and revisions of my paper. With your help, I hope to produce an outstanding paper and sincerely hope for its publication in your esteemed journal.

Jingming Wu

Author Response File: Author Response.pdf