Integrating Satellite and UAV Data to Predict Peanut Maturity upon Artificial Neural Networks

Round 1

Reviewer 1 Report

Dear Authors,

I revised the manuscript "Artificial Neural Networks applied to remote sensing: a new method for predicting peanut maturity" submitted to the Agronomy journal. The manuscript is very interesting. However, I have some concerns, which need to be addressed.

Minor comments:

1. Section "1. Introduction". In my opinion, it is worth citing other papers that deal with the application of remote sensing methods in peanut cultivation. In addition, it is worth pointing out what independent factors (network inputs) influence the yield and maturity of peanuts. For ease of reference, see the paper: https://doi.org/10.3390/land10060609
2. Line 55. Add the authors' names before the reference number, e.g. Li et al. [11].
3. Line 83. Add in brackets the geographical coordinates of the commercial plot where the experiments were conducted.
4. Line 124-127. Add references for the software used.
5. Line 166. Compare the abbreviation ADG used, with the abbreviation in equation 1 (GDA). Use the correct abbreviation.
6. Line 167-168. Figure 2 is missing from the supplementary materials.
7. Line 172. Expand the abbreviation aGDD.
8. Subsection "2.8 Multilayer Percptron Neural Networks (MLP)". It is not clear how many hidden layers have been applied to the MLP network. Please specify this.
9. Line 191. The correct name for StatSoft's software is Statistica, with an "a" at the end. Add the reference here.
10. Line 218. Figure 3 is missing in the manuscript.
11. Line 235. Justify the method of splitting sets adopted. Why do you not use a test set?
12. Line 237. The correct name for StatSoft's software is Statistica, with an "a" at the end. Add the reference here.
13. Line 240, 245. Add MAPE error indicator as it is expressed as a percentage.
14. Line 344, 355. MIP or PMI?
15. Line 368. Add the authors' names before the reference number, e.g. Santos et al. [9].
16. Section "References". The style of the references is not in accordance with the requirements of Agronomy. I suggest using a bibliography software package like Mendeley, Zotero, EndNote etc. Check out the instructions for authors at https://www.mdpi.com/files/word-templates/agronomy-template.dot.

Author Response

Reviewer 1:

Dear Editor and reviewer:

We greatly appreciate the thorough and thoughtful comments provided on our submitted article. It has taken us a rather long time to complete the final revision. We made sure that each one of the reviewer’s comments has been addressed carefully and the paper is revised accordingly.

Please let us know if you still have any questions or concerns about the manuscript. We will be happy to address them, now in a timely manner.

Minor comments:

1. Section "1. Introduction". In my opinion, it is worth citing other papers that deal with the application of remote sensing methods in peanut cultivation. In addition, it is worth pointing out what independent factors (network inputs) influence the yield and maturity of peanuts. For ease of reference, see the paper: https://doi.org/10.3390/land10060609.

We appreciate the suggestions. Regarding peanut works using remote sensing, we present some works in the introduction. As can be seen in lines 52 to 59. Regarding the variables that influence productivity, our work seeks to generate simple prediction models. If we addressed the various variables that influence productivity, we would have to use them as input to the model. For example, weather data. The objective is to use only data from the platforms (vegetation indices and spectral bands) to make the prediction model as simple as possible.

1. Line 55. Add the authors' names before the reference number, e.g. Li et al. [11].

The citation has been modified. Thank you.

1. Line 83. Add in brackets the geographical coordinates of the commercial plot where the experiments were conducted.

Thanks for your suggestion. We decided to insert a map with the coordinates of the area

1. Line 124-127. Add references for the software used.

Other articles in this journal only do the citation of the company, city, and country of the software like this one https://doi.org/10.3390/agronomy10070959

1. Line 166. Compare the abbreviation ADG used, with the abbreviation in equation 1 (GDA). Use the correct abbreviation.

Thanks for your suggestion. Corrections made.

1. Line 167-168. Figure 2 is missing from the supplementary materials.

The temperature graph has been inserted. thanks

1. Line 172. Expand the abbreviation aGDD.

Thanks for your suggestion. Corrections made.

1. Subsection "2.8 Multilayer Perceptron Neural Networks (MLP)". It is not clear how many hidden layers have been applied to the MLP network. Please specify this.

We added in the manuscript

1. Line 191. The correct name for StatSoft's software is Statistica, with an "a" at the end. Add the reference here.

We added the reference of the software.

1. Line 218. Figure 3 is missing in the manuscript.

This figure was taken from the manuscript, but was forgotten in the text. We corrected the text, thanks.

1. Line 235. Justify the method of splitting sets adopted. Why do you not use a test set?

Our approach is actually based on a train/test split. A validation step was not performed. Prior to running the training, 80% of the data was randomly selected and 20% was used only for the testing step.

We understand that several approaches exist on how to train a neural network. The approach mentioned by the reviewer takes into consideration the error on the validation portion of the dataset during the training process to select the best hyperparameters. Our approach during the training process uses the algorithm Balance error against diversity to retain the models. According to the Statistica software guidelines, “this algorithm is based on three criteria: the first is Try to maintain diversity, in this criterium, the algorithm will attempt to preserve a balanced mix of network and ensemble types and architectures, including diverse numbers of input variables and network sizes. Secondary to maintaining diversity, poor performance models will be selected for replacement. Second is replace the oldest model, models are selected according to their index, which indicates the creation order, with oldest models replaced first. And third is replace the highest error model, models are selected for replacement according to the training subset error; the highest error models are replaced first.

The algorithm first "adds" the new network to the network set. It then looks for a network to delete, which may in fact be the "added" network. As a consequence of this approach, the new network is treated on an equal footing with all other networks. A list of candidates for removal is maintained. This is progressively narrowed down as described below.

Only unlocked networks are candidates for removal. Locked networks are not placed in the candidate list. Underrepresented network types are removed from the candidate list. Only the most numerous network types (that is, the joint maximums) are considered. As a consequence, if the new network is of the most numerous types, only that type will be considered as candidates for removal.

The last stage attempts to maintain an "interesting" performance/complexity trade-off. If a given network has both better performance and lower complexity than another, then it is said to dominate that other network. If there exist any networks that are dominated by others, then the candidate list is reduced to the least domineering (networks which are dominated by others, and do not themselves dominate any).

The algorithm finally considers a list of non-dominating networks (networks which, if listed in order of increasing complexity, also have improving performance). The network with the best performance is never removed. All the remaining networks might be considered valuable, given that they present a genuine trade-off between reduced complexity and reduced performance. The algorithm attempts to maintain diversity, by calculating the performance versus complexity trade-off of each such network with respect to the next most complex. The network with the worst trade-off is removed. The effect is to prevent networks "bunching" with similar complexity and marginally different performance. In particular, if there are two candidates with the same complexity, the inferior of the two will always be replaced rather than a less complex network with lower performance.

In judging "complexity", the primary determinant is the number of input variables, as a reduction in the number of input variables makes a network more practical (less need of gathering data) and more informative (a better idea of which variables are important) in addition to being more efficient and less prone to over-learning. The secondary determinant of complexity is the number of hidden units, and this is used if networks have the same number of hidden units”.

For the approach proposed in the manuscript, the authors believe that the conclusions are valid since a different dataset was used during the training phase to test the models.

1. Line 237. The correct name for StatSoft's software is Statistica, with an "a" at the end. Add the reference here

Other articles in this journal only do the citation of the company, city, and country of the software like this one https://doi.org/10.3390/agronomy10070959

1. Line 240, 245. Add MAPE error indicator as it is expressed as a percentage.

We did not use MAPE. We used only mean absolute error (MAE). 

1. Line 344, 355. MIP or PMI?

Corrections made. Thank you.

1. Line 368. Add the authors' names before the reference number, e.g. Santos et al. [9].

Corrections made. Thank you.

1. Section "References". The style of the references is not in accordance with the requirements of Agronomy. I suggest using a bibliography software package like Mendeley, Zotero, EndNote etc. Check out the instructions for authors at https://www.mdpi.com/files/word-templates/agronomy-template.dot.

Corrections made. Thank you.

Reviewer 2 Report

This paper introduces a method for predicting peanut fruit maturity using artificial neural network-driven wide-band vegetation index data. however, It may need to be rewritten because of a large number of unnecessarily long sentences, grammatical errors, and incorrect abbreviations. Portuguese was used in some parts of the present paper. It is worth noting that the research contents and methods of this article are similar to those of a paper published in Remote Sensing (Adão F. Santos et al., 2022, DOI: 10.3390/rs14010093), which was not been quoted and discussed. Why?
Overall, I do not recommend it be published in Agronomy in the current version.
Highlights:
L3-4: the title of the paper is problematic
L13: "is fundamental for reduce losses", grammatical error
L15-16: "by unmanned aerial vehicle (UAV) and satellite", grammatical error
L17-23: ANN, RBF, MLP, NDRE, MAE, etc., abbreviations need to be given their full names when they first appear.
L30-33: A long sentence with grammatical error
L35: "make easier decision making and resource management in the field", how to understand?
L57-60: Lack of references
L61-64: "an ... method ... generate ...", grammatical error
L63: Artificial Neural Networks (ANN's), why is "ANN's"?
L71: what is SR?
Table 1: What do the star of VI and the double-star of NDRE mean? 
Why did the authors select these vegetation indices? Is there a clear mechanism or supporting evidence for the association between these indices and PMI?
L166-174: are "accumulated degree days (ADG)", "GAD", and "aGDD" the same variable?
are "maturation index (MIP)" in Line 172 and "Peanut Maturity Index (PMI)" in L108-109 the same variable?
L185: "Percptron", Misspelling
L200: what is "IVs"?
L205-206: please check Equation 2.
L310-311: "B/IV_sat = bandas/índices de vegetação do satélite; **B/IV_UAV = bandas/índices de vegetação do UAV.", Portuguese?
This article claimed that the Hull-Scrape method is "extremely subjective" in Line 47, but it also used the Hull-Scrape method to collect the maturity of peanuts. What do the authors think of this logic? Do the authors consider how to reduce this subjectivity?

Author Response

Reviewer 2:

Minor comments:

This paper introduces a method for predicting peanut fruit maturity using artificial neural network-driven wide-band vegetation index data. however, It may need to be rewritten because of a large number of unnecessarily long sentences, grammatical errors, and incorrect abbreviations. Portuguese was used in some parts of the present paper. It is worth noting that the research contents and methods of this article are similar to those of a paper published in Remote Sensing (Adão F. Santos et al., 2022, DOI: 10.3390/rs14010093), which was not been quoted and discussed. Why?

We added this paper to the manuscript.

Overall, I do not recommend it be published in Agronomy in the current version.

Highlights:

The English were corrected, we fixed the grammar problems and improved the manuscript's readability. Thank you.

We appreciate the suggestion. Line 378 to 384, I discussed the work in which it was suggested.

L3-4: the title of the paper is problematic

The title was changed.

L13: "is fundamental for reduce losses", grammatical error

Correction done.

L15-16: "by unmanned aerial vehicle (UAV) and satellite", grammatical error

Correction done.

L17-23: ANN, RBF, MLP, NDRE, MAE, etc., abbreviations need to be given their full names when they first appear.

Corrections made. Thank you.

L30-33: A long sentence with grammatical error

Correction done.

L35: "make easier decision making and resource management in the field", how to understand?

Correction done.

L57-60: Lack of references

Corrections done. Thank you.

L61-64: "an ... method ... generate ...", grammatical error

Correction done.

L63: Artificial Neural Networks (ANN's), why is "ANN's"?

Corrections done. Thank you.

L71: what is SR?

The SR was not supposed to be in the text. Corrections made. Thank you.

Table 1: What do the star of VI and the double-star of NDRE mean?

The star in the VI was a clerical error. The two stars are not referring to the NDRE but the constant L in the MNLI and SAVI equation. Thank you for the correction.

Why did the authors select these vegetation indices? Is there a clear mechanism or supporting evidence for the association between these indices and PMI?

Thanks for the consideration. Some of these indices were selected because they present a good correlation with the PMI in other studies. And some other indices have already been studied in other crops, and we are testing these indices for the peanut crop.

L166-174: are "accumulated degree days (ADG)", "GAD", and "aGDD" the same variable?

Corrections done. Thank you.

are "maturation index (MIP)" in Line 172 and "Peanut Maturity Index (PMI)" in L108-109 the same variable?

Corrections done. Thank you.

L185: "Percptron", Misspelling

Corrections done. Thank you.

L200: what is "IVs"?

Writing error. Thank you.

L205-206: please check Equation 2.

The equation was modified. Thank you.

L310-311: "B/IV_sat = bandas/índices de vegetação do satélite; **B/IV_UAV = bandas/índices de vegetação do UAV.", Portuguese?

Corrections done. Thank you.

This article claimed that the Hull-Scrape method is "extremely subjective" in Line 47, but it also used the Hull-Scrape method to collect the maturity of peanuts. What do the authors think of this logic? Do the authors consider how to reduce this subjectivity?

In lines 361 to 364, we discussed the subjectivity of the method. This subjectivity is because several people apply the method throughout the seasons. For our work, peanut maturation using the hull-scrape method was performed with the same researcher in all evaluations, and this evaluator has extensive experience in identifying maturation through the color chart.

Round 2

Reviewer 2 Report

I strongly doubt whether the author has made serious changes. I still think extensive editing of the English language and style is necessary. I hope the authors' reply is honest rather than perfunctory. The authors can provide suitable rebuttals for any comments with which they disagree. Responses to some questions addressed "Correction done", but there were no corresponding changes in the modified version, such as: L15-16: "by unmanned aerial vehicle (UAV) and satellite", grammatical error L17-23: ANN, RBF, MLP, NDRE, MAE, etc., abbreviations need to be given their full names when they first appear. L205-206: please check Equation 2. It is different in this manuscript and the paper authored by Adão F. Santos et al. (2022). In particular, the paper (Adão F. Santos et al., 2022, DOI: 10.3390/rs14010093) was yet not adopted in the modified manuscript.   In addition, there are some new errors in the modified version, such as: L46: "The Hull-Scrape method[6] s the most widely used", clerical error

Author Response

Dear reviewer,

Thank you for your comments concerning our manuscript entitled “Integrating satellite and UAV data to predict peanut maturity upon artificial neural networks”. Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied comments carefully and have made correction which we hope meet with approval. Revised portion are marked in red in the paper. The main corrections in the paper and the responds to your comments are as follow:

I still think extensive editing of the English language and style is necessary.

We tried to improve the sentences that were confused with the help of a native English speaker. Extensive changes were made to the sentences to better understand them. Please let us know if is okay now.

Responses to some questions addressed "Correction done", but there were no corresponding changes in the modified version, such as: L15-16: "by unmanned aerial vehicle (UAV) and satellite", grammatical error.

We rewrite the sentence and correct the grammatical error. “[…] However, the methods currently used are laborious and subjective. To solve this problem, we developed models to access peanut maturity using images from unmanned aerial vehicle (UAV) and satellite. […]”

 L17-23: ANN, RBF, MLP, NDRE, MAE, etc., abbreviations need to be given their full names when they first appear

Thank you for pointing out the problem. We have modified it and marked it in red font (abstract section).

L205-206: please check Equation 2. It is different in this manuscript and the paper authored by Adão F. Santos et al. (2022).

We checked the equation and found the problem. Thanks for your careful work.

In particular, the paper (Adão F. Santos et al., 2022, DOI: 10.3390/rs14010093) was yet not adopted in the modified manuscript.

We agreed if the reviewer and we would like to report a failure from our team at the moment to insert the reference. Thanks for your careful work. Now we have added a discussion about both papers published by Adão F Santos in 2021 and 2022 using satellites and UAV images, respectively (L379-388).

   In addition, there are some new errors in the modified version, such as: L46: "The Hull-Scrape method[6] s the most widely used", clerical error

Thank you very much for your meticulous work. We have revised all paper and fixed this kind of clerical error.

Finally, we tried our best to improve the manuscript and made some changes in it. We appreciate for your warm work earnestly, and hope that the correction will meet with approval.

Once again, thank you very much for your comments and suggestions.

Author Response File: Author Response.pdf