Estimation of Bale Grazing and Sacrificed Pasture Biomass through the Integration of Sentinel Satellite Images and Machine Learning Techniques

Round 1

Reviewer 1 Report

The paper entitled "Bale Grazing and Sacrificed Pasture Biomass Estimation Using Integration of Sentinel Satellite Images and Machine Learning Techniques" studies the temporal variation of spectral and synthetic aperture radar.

I found the article interesting, insightful, and particularly well-written. I think the authors have tried to put their findings into context.

Specific Comments:

1—The introduction section is presented very nicely. However, the contribution section is missing. It is nice if the author contributes in list form. This will help the journal readers understand the manuscript quickly and easily.

2— In section 3.2, the author introduces machine learning algorithms, including Random Forest, Support Vector Regression, and Artificial Neural Networks, which are previously established models in the field. However, it is essential to delineate the specific contributions made by the authors to these existing models. While the utilized model is pre-existing, the manuscript should emphasize the novel aspects that the authors brought to these models. A mere data analysis using existing models might not suffice as innovative work. Therefore, it is crucial to highlight the distinct and noteworthy contributions the authors introduced to enhance these models.

3—In Figure 3, the author must add (a), as it is missing. Furthermore, re-arrange the figure.

4—The author must add theoretical and practical implications.  

5—This research is based on deep learning. It is suggested to provide the hyper-parameters, such as learning rate, epochs, etc.

6—It is suggested to provide a comparison of this manuscript with previous results.

 

 

Satisfactory 

Author Response

Dear Reviewer,

We are grateful for your valuable feedback and thoughtful comments on our manuscript titled "Bale Grazing and Sacrificed Pasture Biomass Estimation Using Integration of Sentinel Satellite Images and Machine Learning Techniques." Your insights are greatly appreciated, and we have addressed your comments and suggestions as follows:

 

1. The introduction section is presented very nicely. However, the contribution section is missing. It is nice if the author contributes in list form. This will help the journal readers understand the manuscript quickly and easily.

We appreciate your positive feedback on the introduction. We have revised the manuscript to include a dedicated "Contribution" section, presented in list form, to convey our study's unique contributions better. This addition will assist readers in quickly understanding the manuscript's distinctive aspects.

 

2. In section 3.2, the author introduces machine learning algorithms, including Random Forest, Support Vector Regression, and Artificial Neural Networks, which are previously established models in the field. However, it is essential to delineate the specific contributions made by the authors to these existing models. While the utilized model is pre-existing, the manuscript should emphasize the novel aspects that the authors brought to these models. A mere data analysis using existing models might not suffice as innovative work. Therefore, it is crucial to highlight the distinct and noteworthy contributions the authors introduced to enhance these models.

You are correct in pointing out that we employed well-established algorithms within the Python programming language. However, it's essential to highlight that our approach involved adapting and training these algorithms to suit the specific context of biomass estimation in pasture settings. This adaptation involved supervised ground sampling and image value manipulation to create a customized model tailored to the integration of Sentinel products.

Furthermore, our study focused on the innovative aspect of utilizing a combination of Sentinel-1 and Sentinel-2 data for biomass mapping. While established algorithms served as the foundation, they had not previously been trained using this unique combination of Sentinel datasets. This integration allowed us to provide more detailed and accurate biomass volume estimates for each paddock, thereby advancing the field's practical applicability.

Regarding your suggestion to explore the mathematical intricacies of the algorithms further, we understand the importance of this aspect. While our primary emphasis was on the practical application of these algorithms and assessing variations in Sentinel datasets, we recognize the value of delving into the mathematical details. We will certainly consider this recommendation for future research, as it can enhance biomass estimation precision, especially when dealing with variables of differing distributions, such as backscattering and optical data.

 

 

 

3—In Figure 3, the author must add (a), as it is missing. Furthermore, re-arrange the figure.

We added the missing subfigure "(a)" to Figure 3 and re-arranged the figure for improved clarity and readability.

 

4—The author must add theoretical and practical implications.  

We have included a section discussing the theoretical and practical implications of our research findings in the last paragraph of the conclusion section. This addition to previous work's significance and potential applications of our work.

 

5—This research is based on deep learning. It is suggested to provide the hyper-parameters, such as learning rate, epochs, etc.

Thank you for your insightful comment. The learning rate has been added (Figure 6-c)

 

6—It is suggested to provide a comparison of this manuscript with previous results.

We would like to address this comment regarding the comparison of our results with previous studies that have focused solely on Sentinel-1 or Sentinel-2 data for biomass estimation. Our research indeed represents a unique approach that integrates both Sentinel-1 and Sentinel-2 datasets and employs advanced machine learning techniques. This integrated approach allows us to harness the complementary strengths of these satellite datasets and significantly enhance the accuracy and reliability of biomass estimation.

Given the distinctiveness of our methodology, it is essential to recognize that direct comparisons with previous studies focused on individual Sentinel datasets may not be entirely fair or informative. The combination of optical and SAR data, coupled with machine learning, introduces a novel dimension to biomass estimation in pasture settings. Our primary objective is to evaluate the effectiveness of this integrated approach and its practical applicability, rather than engaging in a comparative analysis with studies that utilized single datasets or methodologies.

While we acknowledge the importance of building upon existing research, we believe that our study's unique contribution lies in its holistic approach to biomass estimation, which holds significant potential for improving the accuracy and robustness of such assessments. We aim to advance the state of the art by exploring new avenues for combining and utilizing Sentinel-1 and Sentinel-2 data alongside machine learning techniques.

We appreciate the reviewer's consideration of this aspect and welcome any further guidance or insights on how to best communicate the novelty and significance of our research within the context of existing literature.

 

We believe that these revisions have strengthened our manuscript and addressed your valuable comments. We sincerely appreciate your time and effort in reviewing our work, which has undoubtedly improved the quality of our research.

Thank you once again for your valuable feedback and support.

Reviewer 2 Report

The research is well-designed and written. However, there are still some questions that should be clarified. Details are as follows:

1.    Line 57. It should be “AVHRR”, not “AVHR”.

2.    Line 110 and 115. A small map indicating the location of the study within the USA could be put in Figure 1 to let the readers directly understand where the study is.

3.    Line 203. There are about 20 kinds of vegetation indexes for the researchers. The most widely used VI indexes such as NDVI, EVI, and EVI2. Why only the NDVI, MNDVI, GNDVI, AVI, GCI, and SIPI were selected? I suggest a reasonable and convincing selection principle of the VI index should be supplemented.

Zeng, Y. et al. Optical vegetation indices for monitoring terrestrial ecosystems globally. Nature Reviews Earth & Environment 3, 477-493, doi:10.1038/s43017-022-00298-5 (2022).

4.    Line 245. Three machine-learning methods (SF, SVR, and ANN) were used for biomass estimation. The reason why the three methods were chosen should also be supplemented.  

Morais, T. G., Teixeira, R. F. M., Figueiredo, M. & Domingos, T. The use of machine learning methods to estimate aboveground biomass of grasslands: A review. Ecol. Indic. 130, doi:10.1016/j.ecolind.2021.108081 (2021).Line 419-450.

5.    Line 359. Figure 3 could be improved to look better. Such as text enlargement, line bolding, and legend adjustment.

Good.

Author Response

Dear Reviewer,

 

We appreciate your thorough review of our manuscript, "Bale Grazing and Sacrificed Pasture Biomass Estimation Using Integration of Sentinel Satellite Images and Machine Learning Techniques." Your constructive comments are invaluable in enhancing the quality of our work. We have addressed your questions and suggestions as follows:

 

1. Line 57. It should be “AVHRR,” not “AVHR

We have corrected the typo on Line 57, changing "AVHR" to "AVHRR" to represent the Advanced Very High-Resolution Radiometer accurately.

 

2. Line 110 and 115. A small map indicating the location of the study within the USA could be put in Figure 1 to let the readers directly understand where the study is.

We have added a small location map to Figure 1, as suggested. This edition provides readers with an apparent visual reference to the study's location within the USA.

 

3. Line 203. There are about 20 kinds of vegetation indexes for the researchers. The most widely used VI indexes such as NDVI, EVI, and EVI2. Why only the NDVI, MNDVI, GNDVI, AVI, GCI, and SIPI were selected? I suggest a reasonable and convincing selection principle of the VI index should be supplemented.

Thanks a lot for your comment. The selection of these indices has been informed by an extensive literature review in our introduction section. They have been chosen based on their established relevance in prior research related to biomass estimation. In future studies, we acknowledge the potential for incorporating additional indices and conducting feature selection to further refine and enhance our biomass estimation methodology.

 

4. Line 245. Three machine-learning methods (RF, SVR, and ANN) were used for biomass estimation. The reason why the three methods were chosen should also be supplemented.  

We have provided additional context and justification for selecting our study's three machine learning methods (SF, SVR, and ANN). The rationale for choosing these methods is now explained in relation to the biomass estimation context. We have also referenced Morais et al. (2021) to support this choice.

 

5. Line 359. Figure 3 could be improved to look better. Such as text enlargement, line bolding, and legend adjustment.

We have enhanced Figure 3 to improve its visual clarity. Text has been enlarged, lines have been bolded for emphasis, and legends have been adjusted for better readability. These modifications aim to provide readers with a more user-friendly and informative visualization of our findings.

 

We genuinely appreciate your time and effort in reviewing our work. Your suggestions have significantly contributed to refining our manuscript, and we believe these revisions have addressed your concerns effectively.

 

Thank you for your valuable feedback and support

Reviewer 3 Report

In this paper, the authors presented how to measure forage biomass based on the use of integration of Sentinel-1 and Sentinel-2 satellite images with machine learning techniques. This topic is important to a wide audience, but there are minor comments that need to be addressed.

Minor comments:

(1) Please complete the DOI references of publications (where possible) in the references. 

(2) Please, in Figure 3 reorganize well the symbol of legends beside the words.

(3) In the line 449-451, please add the reference.

(4) In the line (459 and 491), the unit of RMSE is missing g/??.

(5) Table 4 and 5, Instead of writing correlation use symbol for consistency.

(6) Line 545, add other references (not only one reference) if the authors wrote in the text that previous studies have been reported.

(7) Line 553, reference is missing developed by Chen et al.

 

(8) Line 590, please change the Bold of the letter R to R.

Kind Regards, 

Reviewer 

Dear Authors:

No comment, the manuscript is written in a good way.

Good luck, 

Reviewer

 

 

Author Response

Reviewer 3:

Dear Reviewer,

 

We appreciate your thoughtful review of our manuscript titled "Bale Grazing and Sacrificed Pasture Biomass Estimation Using Integration of Sentinel-1 and Sentinel-2 Satellite Images with Machine Learning Techniques." Your valuable feedback is instrumental in improving the overall quality of our work. We have addressed your comments as follows:

 

1. Please complete the DOI references of publications (where possible) in the references. 

Thanks a lot for your comment. We have provided the complete DOI references for publications in the references section wherever possible.

 

2. Please, in Figure 3 reorganize well the symbol of legends beside the words.

Thanks a lot for your comment. We have reorganized the legends in Figure 3 to ensure that symbols are appropriately aligned with their corresponding words for improved clarity and readability.

 

1. In the lines 449-451, please add the reference.

We have added the missing reference as per your suggestion in lines 449-451 (references 32, 68-70)

 

4. In the line (459 and 491), the unit of RMSE is missing g/??.

For clarity, we have included the unit for RMSE (root mean square error) as Kg/100 m2 in lines 459 and 491.

 

5. Table 4 and 5, Instead of writing correlation use symbol for consistency.

We have deleted the correlation column as it was related to R² and caused confusion.

 

6. Line 545, add other references (not only one reference) if the authors wrote in the text that previous studies have been reported.

Thanks for your comment. In that section, we have indeed referenced three key sources (74-77), and then focused on reference No.77 and looked at it in more details.

 

 

7. Line 553, reference is missing developed by Chen et al.

We have included the missing reference developed by Chen et al. in line 553.

 

8. Line 590, please change the Bold of the letter R to R.

We have changed the formatting of the letter "R" to uppercase "R" in line 590.

 

We genuinely appreciate your attention to detail and your dedication to improving our manuscript. Your comments have been invaluable in enhancing the accuracy and presentation of our research.

 

Thank you for your time and valuable input.

Round 2

Reviewer 1 Report

The manuscript entitle Estimation of Bale Grazing and Sacrificed Pasture Biomass Through Integration of Sentinel Satellite Images and Machine Learning Techniques, proposed  the temporal variation in spectral and synthetic aperture radar variables derived from Sentinel (1-2) time series images. Further, this manuscript proceeded to assess how these variables impact the estimation of grassland biomass through the utilization of three machine learning algorithms: support vector regression, random forest, and artificial neural network.

• (1) Generally speaking, the innovation of this paper is limited, it is only the application of grassland regions have experienced significant degradation.  
  (2) Too few experiments and tasks can not verify the effectiveness of this unified model.
  (3) Many words in Figure 2 are abbreviated and the flowchart is not explained, which makes the paper less readable.
  (4) All figures in this paper should be improved to make it have high resolution and be artistic.
  (5) None of the equations in the paper are numbered.
• (6) Justify the results, It is recommended to compared the results with the previous results. 

Proofread is required

Author Response

Dear Reviewer,

We are grateful for your valuable feedback and thoughtful comments on our manuscript titled "Bale Grazing and Sacrificed Pasture Biomass Estimation Using Integration of Sentinel Satellite Images and Machine Learning Techniques." Your insights are greatly appreciated, and we have addressed your comments and suggestions as follows:

 

1. Generally speaking, the innovation of this paper is limited, it is only the application of grassland regions that have experienced significant degradation.  

Thanks a lot for your feedback. While it may appear that the innovation in our paper is limited to the application in grassland regions with significant degradation, we believe that this focus is precisely what sets our work apart. By concentrating on areas that have undergone substantial environmental challenges, we aim to demonstrate the practical utility and effectiveness of our approach in addressing critical ecological issues.

Moreover, it's important to note that our research often serves as a starting point for further exploration and adaptation in various environmental contexts. We hope that by highlighting the potential of our methodology in grassland regions, we can inspire other researchers to apply similar principles to different ecosystems and contribute to broader environmental conservation efforts.

We value your input, and we're continuously working to expand the scope and applicability of our research.

 

2. Too few experiments and tasks can not verify the effectiveness of this unified model.

      Thank you for your comment and for raising the concern about the number of experiments and tasks conducted in our study. We appreciate your feedback, and we'd like to provide some context for our research approach.

While it is true that our study involved a limited number of experiments and tasks, we want to emphasize that our primary goal was to establish the feasibility and foundational effectiveness of our unified model in a controlled setting. Our intention was to lay the groundwork for future research that can explore a broader range of applications and tasks.

We acknowledge that expanding the scope of experiments and tasks is a valuable direction for future work. Based on your feedback, we are actively planning additional experiments and tasks to further validate and showcase the versatility of our unified model. These forthcoming experiments will include a more diverse set of challenges and benchmarks to comprehensively evaluate its performance.

Your insights have been instrumental in guiding the next steps of our research, and we are committed to addressing the concerns you've raised. We appreciate your valuable input and look forward to sharing our progress with you in future publications.

If you have any specific suggestions or recommendations regarding the types of experiments or tasks you believe would be most informative for evaluating our unified model, we would welcome your input.

3. Many words in Figure 2 are abbreviated and the flowchart is not explained, which makes the paper less readable.

Thanks a lot for your comment. We added the full words to the flowchart in Figure 2.

 

 

1. All figures in this paper should be improved to make it have high resolution and be artistic.

Thanks a lot for your comment. We have added the high-resolution files for each figure.

 

 

5. None of the equations in the paper are numbered.

Thanks a lot for your comment. We have three equations in the manuscript (Section 3.3), and they are numbered

 

6. Justify the results, It is recommended to compared the results with the previous results.

We appreciate your valuable suggestion to justify our results and compare them with previous studies. It's essential to emphasize that our paper goes to great lengths to both justify and comprehensively compare our findings with the existing body of literature. Throughout our manuscript, we have taken care to not only provide justifications for our results but also to draw extensive comparisons with prior research, as documented in the references(reference 51 to 71). This approach not only enhances the robustness and relevance of our findings but also positions them within the broader scientific context."