Machine-Learning Approaches in N Estimations of Fig Cultivations Based on Satellite-Born Vegetation Indices

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Introduction: The introduction is very poorly written. Please restructure the whole section appropriately. It seems like all topics such as Nitrogen, remote sensing, AI etc are written in bullet form in no sequence.

Please write the introduction in proper paragraphs in  a way that there is flow of information. Please add more literature for N requirement in fig cultivation and remote sensing as they are the main premise of the paper.

Please delineate the objectives, research gaps, and research questions appropriately. They are not mentioned in the paper.

L 96: Which chlorophyll (a, b etc.) was determined. What was the unit of the values that were derived?

L 115: Why was 0.7 selected as the threshold? Also if 0.7 was selected as threshold why were values ≤ -0.7 not used.

 

How was hyperparameter tuning done for AI models?

Author Response

Introduction: The introduction is very poorly written. Please restructure the whole section appropriately. It seems like all topics such as Nitrogen, remote sensing, AI etc are written in bullet form in no sequence.

Please write the introduction in proper paragraphs in a way that there is flow of information. Please add more literature for N requirement in fig cultivation and remote sensing as they are the main premise of the paper.

Please delineate the objectives, research gaps, and research questions appropriately. They are not mentioned in the paper. THE INTRODUCTION HAS BEEN CORRECTED

L 96: Which chlorophyll (a, b etc.) was determined. What was the unit of the values that were derived? CORRECTED

L 115: Why was 0.7 selected as the threshold? Also if 0.7 was selected as threshold why were values ≤ -0.7 not used. THE INDEXES WITH THE HIGHEST RESULT, WHICH WAS 0.7, WERE TAKEN, BELOW ARE NOT SIGNIFICANT. NO RESULTS OF -0.7 WERE OBTAINED.

 How was hyperparameter tuning done for AI models? CORRECTED

YOUR REMARKS AND COMMENTS HAVE BEEN TAKEN INTO ACCOUNT

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

The authors estimated the nitrogen dynamics in the fig plant using remote sensing and machine learning techniques. Thus, this manuscript is aligned with the scope of the journal. However, the current version of the manuscript requires major revision as some parts of the manuscript lack coherence and depth, as well as typographical and grammatical errors as discussed below:

·     The abstract mentioned some terms and abbreviations (e.g. TCARI, MCARI, OSAVI) are not defined. These terms must be explained to avoid confusion. Moreover, keywords should be arranged alphabetically.

·     The introduction is incomplete as it does not clearly provide the problem statement. Moreover, the ideas are fragmented and lack coherence and depth. For example, ideas were presented in one sentence paragraphs. The first paragraph needs more elaboration to convey the need for using remote sensing and machine learning techniques. Paragraphs 2-4 could be combined as they discuss remote sensing, while paragraphs 5-9 could also be combined as they discuss machine learning techniques.

·     Moreover, the authors must discuss why they are estimating the nitrogen dynamics in the fig crop. The introduction did not even mention fig, which is the main crop in the study. Why is this not conducted for wheat or corn?

·     In the methodology, the authors characterize the soil physico-chemical properties of the two sites at the beginning, middle, and end of the 2022 production cycle (Table 1). It would be clearer if the months associated with each time of sampling were indicated to have a clear picture of when the sampling was done during the conduct of the study.

·     The Julian calendar would be better mentioned in the leaf sampling to provide information for the dates of sampling.

·     The discussion section needs major revision to provide logical structure for discussing the results and providing an explanation of the results. It is recommended that some paragraphs discussing similar results (e.g. Lines 244-254) are combined with “transitions” to guide the reader as to why certain studies were cited. Without transitions, it would be confusing why the authors mentioned the study of Xiong et al. (2019) after paragraph 1 (Lines 236-243).

·     The first sentence in the conclusion is too general and therefore irrelevant. This study was conducted on figs only and not various crops. I cannot understand why the fig crop has been omitted or not mentioned here.

·     The authors must carefully review the references for inconsistencies. For example, some publication years were not written in bold (e.g. Lines 335, 338, 352, 354).

·     The manuscript needs critical review by an English editor for grammatical and typographical errors.

Comments on the Quality of English Language

·     The manuscript needs critical review by an English editor for grammatical and typographical errors.

Author Response

The authors estimated the nitrogen dynamics in the fig plant using remote sensing and machine learning techniques. Thus, this manuscript is aligned with the scope of the journal. However, the current version of the manuscript requires major revision as some parts of the manuscript lack coherence and depth, as well as typographical and grammatical errors as discussed below:

• The abstract mentioned some terms and abbreviations (e.g. TCARI, MCARI, OSAVI) are not defined. These terms must be explained to avoid confusion. Moreover, keywords should be arranged alphabetically. AS THE WORDS IN THE SUMMARY ARE LIMITED, IN INTRODUCTION THESE TERMS ARE EXPLAINED. THE KEYWORDS HAVE BEEN PUT IN ALPHABETICAL ORDER.
• The introduction is incomplete as it does not clearly provide the problem statement. Moreover, the ideas are fragmented and lack coherence and depth. For example, ideas were presented in one sentence paragraphs. The first paragraph needs more elaboration to convey the need for using remote sensing and machine learning techniques. Paragraphs 2-4 could be combined as they discuss remote sensing, while paragraphs 5-9 could also be combined as they discuss machine learning techniques.
• Moreover, the authors must discuss why they are estimating the nitrogen dynamics in the fig crop. The introduction did not even mention fig, which is the main crop in the study. Why is this not conducted for wheat or corn? THE INTRODUCTION HAS BEEN CORRECTED
• In the methodology, the authors characterize the soil physico-chemical properties of the two sites at the beginning, middle, and end of the 2022 production cycle (Table 1). It would be clearer if the months associated with each time of sampling were indicated to have a clear picture of when the sampling was done during the conduct of the study. CORRECTED
• The Julian calendar would be better mentioned in the leaf sampling to provide information for the dates of sampling. JULIAN CALENDAR ARE MENTIONED IN LINES 103 AND 104
• The discussion section needs major revision to provide logical structure for discussing the results and providing an explanation of the results. It is recommended that some paragraphs discussing similar results (e.g. Lines 244-254) are combined with “transitions” to guide the reader as to why certain studies were cited. Without transitions, it would be confusing why the authors mentioned the study of Xiong et al. (2019) after paragraph 1 (Lines 236-243). CORRECTED
• The first sentence in the conclusion is too general and therefore irrelevant. This study was conducted on figs only and not various crops. I cannot understand why the fig crop has been omitted or not mentioned here. FIG IS ALREADY MENTIONED IN THE CONCLUSION
• The authors must carefully review the references for inconsistencies. For example, some publication years were not written in bold (e.g. Lines 335, 338, 352, 354). CORRECTED
• The manuscript needs critical review by an English editor for grammatical and typographical errors.

YOUR REMARKS AND COMMENTS HAVE BEEN TAKEN INTO ACCOUNT

 

 

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

Dear Authors,

 

Please, receive the review report of your manuscript titled "Nitrogen Estimation in Fig Cultivation through Remote Sensing and Machine Learning".

 

Title: I recommend changing the title since Nitrogen Estimation in Fig Cultivation was made by application of Machine Learning solutions and for this, the input dataset was Remote Sensing-based. A possible example can be 'Machine Learning approaches in N estimations of Fig Cultivations based on Satellite-born Vegetation Indices'.

 

Affiliation: Please, add the English names of your institutes.

 

Abstract: The result of the Gradient Boosting is missing from Line 18. You have left space here, you should add this information as well since it is one of your methods.

 

Keywords: Change the order of the keywords according to their importance/presence in the manuscript, please. I recommend adding Remote Sensing and Nitrogen or nitrogen concentration, as well.

 

Introduction: add more literature references which are relevant. Widen this part, please, write a more complex intro part with decent paragraphs longer than 3-4 rows, basically 1 or 2 sentences. It cannot be left this way. Please, also provide literature examples about soil-plant relationships and related studies.

 

Materials and Methods:
2.1.: Line 68: X and Y coordinates cannot be extreme, altitutude can be, especially from an agricultural perspective. Please modify the sentence here. You provide hydrometeorological details but PREC is for summer, only, but Temp mean is for the whole year. Please, improve this part to show more coherence for the readers. Figure 1: well-edited in general, but there are some issues. Please, decrease the size of the North Arrow, since it is unreasonably big, km can be used instead of Km. You should not apply the minutes and seconds on the first submap since the scale is set to 10°. On the second submap, what is this greyscale raster layer visible within the plots? Please, explain it with proper Legend or remove it. If this is supposed to show the Sentinel2 grid net, then use a proper vector layer for this reason. The title of Figure 1 contains the name of the farm, but it does not appear in the text in this section. The plots should be introduced a bit "richer". E.g. write about the two plots here, why two plots? why these sampling points were selected, etc. Provide DEM here, write about the elevation and slope characteristics here.

2.2.: This part is very short, as well. I recommend changing the title here since it is not only about the fieldwork, this section also mentions the soil characteristics and the laboratory measurements, too. Please, restructure this part. Maybe, you add most of the information here to the Study area description section. Table 1 needs modifications: Beginning, Middle and End are relative terms, please, in a scientific paper, provide date intervals. It is extremely important to clarify in the case of Sentinel2 "sampling" as well. Please, define what P1 and P2 are. The plots, probably, but here, solely from the table, it is not obvious. What were the soil sampling "circumstances": how many soil samples, how many depths? etc. I miss this information from here. Please, add a reference to Kjeldahl as well.

2.3.: is incorrectly written as 2.4. in the manuscript. Please, correct such mistakes. This part is not about data analysis but rather RS data gathering and the VI generation in GIS. Please, specify these details, and also add how many days were non-cloudy. Which product did you use? TOA or SR, or "mixed"? It is crucial to clarify. 

2.4.: This is also incorrectly written, since it cannot the 2.6., in the manuscript. Please, correct such mistakes. Change the title, please. You also made it too simple: 4 rows and a Table, only. Please, develop this part, as well. If you use Sentinel2 for the calculation in the GIS, then you shall provide the bands as well, you used beside the wavelength values. It can be wiser to combine the  2.3. and 2.4.. Table 2 contains formulas not correctly: MCARI formula and TCARI formula are the same here. Later, these indices will have huge importance, but I detected serious mistakes, which I write here at its place.

2.5. (not 2.3.) Please, provide a more consistent section and since it is supposed to be methodology, the readers must see the steps you made. It is true for all the steps you have made and you present the results of them. In general, these are missing from this manuscript, therefore I strongly recommend writing about them properly after revision.

2.6., actually not 2.5. (line 167): Data validation but for the statistical analysis. This must be clarified. You generally write about the K-fold cross validation method, but the readers cannot see how the 5 subsets were employed. What data was involved here?

2.7. which is not 2.6. as it is written in the manuscript. For the performance evaluation, I would recommend keeping R² and RMSE. Does MSE provide more information than RMSE, does it add something new? I recommend adding a better formula explanation here, in terms of the members (O_i, S_i, etc.), try to correct Spanish words into English everywhere (and instead of y).

 

Results:

In general, it is very hard to follow how the results were generated.
3.1.: The title is not good here: in the laboratory, you measure, not estimate.

Table 3 can be a very useful figural element, but it strongly needs serious modifications: technically the title is not good here (P. corr. with VIs and the Nitrogen%), and numbers on the x-axis overlap each other; the grey boxes are too thin for longer VI names, and we do not know what *, ** and *** stand for? Which significant level? Please, clarify these here. There are issues which must be solved from a mathematical point of view: MCARI and TCARI must correlate with each other since the formula is almost (only almost) the same (a multiplication step is the difference between them), and it must be seen on the correlation parameters. There are issues about the OSAVI values as well, on similar mathematical relations. So, please, check your calculation from the beginning and modify the values accordingly. I am aware that this demands a greater amount of work, but it must be invested.

 

3.2. There is no description earlier about how the regression model was made here. It should be presented in the methodology main section. Please, correct it. Multilinear regression results are not founded well, hence it is not acceptable in this form. Fig. 3 (b) is very strange. I strongly recommend to check again the steps, and the input data. Please, also explain why 14 data points are plotted. What is the optimum input data for Random Forest?
I see issues about Fig. 4 (b), too. Technically: what are the x and y-axis? There is no explanation, . Please, provide them. From a content perspective: TCARI and MCARI should show similar results, as I explained earlier, and this is a great problem. You must check it and revise it. Otherwise, this work is vulnerable. 

Line 230-233 should be in the Discussion.

Discussion: Line 236-238 should not be in Discussion. Line 241: What banana? This research was done on Fig cultivation, according to the title and what you have written above. Please, explain it or correct it.

Line 244-248 and 255-263 belong to a state of art, hence please, put them into the Introduction.

If you checked and corrected your data analysis and results, then you will be able to re-structure and re-write the Conclusion part, as well.

You used 48 citations, which seem to be relevant and sufficient in number. Please, check the formatting/editings here.

 

Please, check the comments, suggestions, and recommendations carefully and improve the presentation of the valuable work you have carried out.

 

Kind regards,
   Reviewer

 

 

 

Comments on the Quality of English Language

I have detected only a few mistakes. In general, there are no serious quality issues in this regard.

Author Response

Please, receive the review report of your manuscript titled "Nitrogen Estimation in Fig Cultivation through Remote Sensing and Machine Learning".

 Title: I recommend changing the title since Nitrogen Estimation in Fig Cultivation was made by application of Machine Learning solutions and for this, the input dataset was Remote Sensing-based. A possible example can be 'Machine Learning approaches in N estimations of Fig Cultivations based on Satellite-born Vegetation Indices'. THE SUGGESTION TO CHANGE THE TITLE WAS TAKEN INTO ACCOUNT.

 Affiliation: Please, add the English names of your institutes. THE AFFILIATIONS WERE WRITTEN IN ENGLISH, HOWEVER, SINCE THEY ARE NAMES OF OFFICIAL INSTITUTIONS

 Abstract: The result of the Gradient Boosting is missing from Line 18. You have left space here, you should add this information as well since it is one of your methods. CORRECTED

 Keywords: Change the order of the keywords according to their importance/presence in the manuscript, please. I recommend adding Remote Sensing and Nitrogen or nitrogen concentration, as well. THE KEYWORDS WERE ORDERED IN ALPHABETICAL ORDER.

 Introduction: add more literature references which are relevant. Widen this part, please, write a more complex intro part with decent paragraphs longer than 3-4 rows, basically 1 or 2 sentences. It cannot be left this way. Please, also provide literature examples about soil-plant relationships and related studies. THE INTRODUCTION HAS BEEN MODIFIED

Materials and Methods:
2.1.: Line 68: X and Y coordinates cannot be extreme, altitutude can be, especially from an agricultural perspective. Please modify the sentence here. You provide hydrometeorological details but PREC is for summer, only, but Temp mean is for the whole year. THE PARAGRAPH HAS BEEN REDACTED

Please, improve this part to show more coherence for the readers. Figure 1: well-edited in general, but there are some issues. Please, decrease the size of the North Arrow, since it is unreasonably big, km can be used instead of Km. You should not apply the minutes and seconds on the first submap since the scale is set to 10°. On the second submap, what is this greyscale raster layer visible within the plots? Please, explain it with proper Legend or remove it. If this is supposed to show the Sentinel2 grid net, then use a proper vector layer for this reason. The title of Figure 1 contains the name of the farm, but it does not appear in the text in this section. SCALE AND WIND ROSE MODIFICATIONS HAVE BEEN MADE; IN SUBFIGURE 2 THE PLOTS ARE HIGHLIGHTED WITH THE COMBINATION OF NATURAL COLOR BANDS OF SENTINEL-2, AND THE GRAYISH PART IS THE BASE MAP OF THE SOFTWARE. TO AVOID DUPLICATION OF INFORMATION, POINT 2.2 MENTIONS THE REGION AND POINT 2.2 SPECIFIES THE WORKING AREA, WHICH IS THE ANA FARM.

The plots should be introduced a bit "richer". E.g. write about the two plots here, why two plots? why these sampling points were selected, etc. Provide DEM here, write about the elevation and slope characteristics here. CORRECTED

2.2.: This part is very short, as well. I recommend changing the title here since it is not only about the fieldwork, this section also mentions the soil characteristics and the laboratory measurements, too. Please, restructure this part. Maybe, you add most of the information here to the Study area description section. Table 1 needs modifications: Beginning, Middle and End are relative terms, please, in a scientific paper, provide date intervals. It is extremely important to clarify in the case of Sentinel2 "sampling" as well. Please, define what P1 and P2 are. The plots, probably, but here, solely from the table, it is not obvious. What were the soil sampling "circumstances": how many soil samples, how many depths? etc. I miss this information from here. Please, add a reference to Kjeldahl as well. CORRECTED

2.3.: is incorrectly written as 2.4. in the manuscript. Please, correct such mistakes. This part is not about data analysis but rather RS data gathering and the VI generation in GIS. Please, specify these details, and also add how many days were non-cloudy. Which product did you use? TOA or SR, or "mixed"? It is crucial to clarify. IN THIS WORK WE ARE TALKING ABOUT SPECTRAL IMAGES, SO THE IMPORTANT THING IS THE WIDTH OF EACH BAND AND NOT THE RS.  

2.4.: This is also incorrectly written, since it cannot the 2.6., in the manuscript. Please, correct such mistakes. Change the title, please. You also made it too simple: 4 rows and a Table, only. Please, develop this part, as well. If you use Sentinel2 for the calculation in the GIS, then you shall provide the bands as well, you used beside the wavelength values. It can be wiser to combine the  2.3. and 2.4.. Table 2 contains formulas not correctly: MCARI formula and TCARI formula are the same here. Later, these indices will have huge importance, but I detected serious mistakes, which I write here at its place. THE ERROR IN TCARI'S FORMULA HAS BEEN CORRECTED. THE BANDS WERE NOT PUT AS SUCH, BECAUSE THE LITERATURE MARKS THEM IN NM, SINCE IT SERVES AS A REFERENCE FOR THE CALCULATIONS WITH BANDS FOR DIFFERENT SATELLITE IMAGES.

2.5. (not 2.3.) Please, provide a more consistent section and since it is supposed to be methodology, the readers must see the steps you made. It is true for all the steps you have made and you present the results of them. In general, these are missing from this manuscript, therefore I strongly recommend writing about them properly after revision. CORRECTED

2.6., actually not 2.5. (line 167): Data validation but for the statistical analysis. This must be clarified. You generally write about the K-fold cross validation method, but the readers cannot see how the 5 subsets were employed. What data was involved here? FROM LINE 227 IT IS EXPLAINED THAT THE K-FOLD METHOD IS USED TO VALIDATE ARTIFICIAL NEURAL NETWORKS, RANDOM FOREST AND GRADIENT BOOSTING, AND A FIGURE HAS BEEN ADDED TO BETTER REPRESENT THE METHOD.

2.7. which is not 2.6. as it is written in the manuscript. For the performance evaluation, I would recommend keeping R² and RMSE. Does MSE provide more information than RMSE, does it add something new? I recommend adding a better formula explanation here, in terms of the members (O_i, S_i, etc.), try to correct Spanish words into English everywhere (and instead of y). THESE PARAMETERS WERE USED BECAUSE THEY ARE THE ONES USED TO VALIDATE THE MODELS IN ARTICLES SUCH AS MARTINEZ-SIFUENTES ET AL., (2024), BEI CUI ET AL., (2019) AND OTHERS.

Martínez-Sifuentes, Aldo Rafael; Trucíos-Caciano, Ramón; López-Hernández, Nuria Aíde; Miguel-Valle, Enrique; Estrada-Ávalos, Juan. Spectral Index-Based Estimation of Total Nitrogen in Forage Maize: A Comparative Analysis of Machine Learning Algorithms. Nitrogen 2024, 5(2), 468-482; https://doi.org/10.3390/nitrogen5020030

Bei Cui; Qianjun Zhao; Wenjiang Huang; Xiaoyu Song;Huichun Ye; Xianfeng Zhou. A New Integrated Vegetation Index for the Estimation of Winter Wheat Leaf Chlorophyll Content. Remote Sens. 2019, 11(8), 974; https://doi.org/10.3390/rs1108097

Results:

In general, it is very hard to follow how the results were generated.
3.1.: The title is not good here: in the laboratory, you measure, not estimate. CORRECTED

Table 3 can be a very useful figural element, but it strongly needs serious modifications: technically the title is not good here (P. corr. with VIs and the Nitrogen%), and numbers on the x-axis overlap each other; the grey boxes are too thin for longer VI names, and we do not know what *, ** and *** stand for? Which significant level? Please, clarify these here. There are issues which must be solved from a mathematical point of view: MCARI and TCARI must correlate with each other since the formula is almost (only almost) the same (a multiplication step is the difference between them), and it must be seen on the correlation parameters. There are issues about the OSAVI values as well, on similar mathematical relations. So, please, check your calculation from the beginning and modify the values accordingly. I am aware that this demands a greater amount of work, but it must be invested. THE EXPERIMENT WAS REPLICATED WITH THE SAME PARAMETERS AND THE RESULTS WERE THE SAME. THE COMMENTS IN THE FIGURES HAVE BEEN TAKEN INTO ACCOUNT.

 3.2. There is no description earlier about how the regression model was made here. It should be presented in the methodology main section. Please, correct it. Multilinear regression results are not founded well, hence it is not acceptable in this form. Fig. 3 (b) is very strange. I strongly recommend to check again the steps, and the input data. Please, also explain why 14 data points are plotted. What is the optimum input data for Random Forest?
I see issues about Fig. 4 (b), too. Technically: what are the x and y-axis? There is no explanation, . Please, provide them. From a content perspective: TCARI and MCARI should show similar results, as I explained earlier, and this is a great problem. You must check it and revise it. Otherwise, this work is vulnerable.  THE EXPERIMENT WAS REPLICATED WITH THE SAME PARAMETERS AND THE RESULTS WERE THE SAME.

Line 230-233 should be in the Discussion.

Discussion: Line 236-238 should not be in Discussion. Line 241: What banana? This research was done on Fig cultivation, according to the title and what you have written above. Please, explain it or correct it.

Line 244-248 and 255-263 belong to a state of art, hence please, put them into the Introduction.

If you checked and corrected your data analysis and results, then you will be able to re-structure and re-write the Conclusion part, as well.

You used 48 citations, which seem to be relevant and sufficient in number. Please, check the formatting/editings here.

 Please, check the comments, suggestions, and recommendations carefully and improve the presentation of the valuable work you have carried out.

I have detected only a few mistakes. In general, there are no serious quality issues in this regard.

YOUR REMARKS AND COMMENTS HAVE BEEN TAKEN INTO ACCOUNT

 

 

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Thank you for improving the paper based on the reviewers' suggestions. It is in good shape for publication.