Crop Canopy Nitrogen Estimation from Mixed Pixels in Agricultural Lands Using Imaging Spectroscopy

Round 1

Reviewer 1 Report

I'm reviewing the manuscript "Crop canopy nitrogen estimation from mixed pixels in agricultural lands using imaging spectroscopy".

 The proposed paper focus on handling the mixed information derived from each single pixel where multiple fractions are settled.

 The fractional cover effects on nitrogen retrieval by PLSR model is the key goal of the proposed research.

 The manuscript is well written and all the needed information are easily available to the reader.

 The mentioned companion paper does not include the cover fraction effect on nitrogen retrieval practice.

 The manuscript is almost publication-ready.

 I have to point out that something mismatch in figure 9.

 The authors declare no effect is due to crop. It looks like a single crop "pink triangle" stands out of the chore and it is not mentioned in the legend.

Author Response

Dear Editor and reviewers,

We appreciate the time and effort you have dedicated to reviewing our manuscript. Your insightful comments and constructive criticisms have been invaluable in guiding our revisions and improving the quality of our work. We have carefully considered each point raised and have made corresponding modifications to the manuscript to address these concerns. Enclosed to this message (the pdf file), we provide detailed responses (in color blue) to each of the comments and explain how we have incorporated your feedback into our revised manuscript.

We are sincerely thankful for the opportunity to improve our work with your guidance and look forward to any further suggestions you may have.

Kind regards. 

Author Response File: Author Response.pdf

Reviewer 2 Report

In genral,  the manuscript is simple and have not provide innovations about N estimation. Besides,  airborne observations generally provide a fine spatial resolution, so studying the mixed pixel is really necessray? 

1. Introduction is too simple, not clearly stating the current research status and the entry point of this study.

2. Section 2.1, the sampling method should be described in detail, e.g., how many smaples were collected in aeach sampling points;

3. Section 2.2, the spatial resolution should be illustrated;

Author Response

Dear Editor and reviewers,

We appreciate the time and effort you have dedicated to reviewing our manuscript. Your insightful comments and constructive criticisms have been invaluable in guiding our revisions and improving the quality of our work. We have carefully considered each point raised and have made corresponding modifications to the manuscript to address these concerns. Enclosed to this message (the pdf file), we provide detailed responses (in color blue) to each of the comments and explain how we have incorporated your feedback into our revised manuscript.

We are sincerely thankful for the opportunity to improve our work with your guidance and look forward to any further suggestions you may have.

Kind regards. 

Author Response File: Author Response.pdf

Reviewer 3 Report

Introduction 

The project aims to investigate the impact of spectral mixing on canopy nitrogen estimation based on the known VSWIR-spectral regions. 

The authors have shown that the endmember fractions (PV, NPV, and BS) as derived from hyperspectral data using the Automatic Monte Carlo spectral Unmixing model play a role in canopy nitrogen estimation and that PV is the most important variable.

There is limited literature that quantified the effect of changes in fraction cover on N estimation, especially covering different crop types. This study quantifies this gap on more than 10 different crops.

The study improves on canopy N estimation by taking into account the cover fractions. This can be applied/tested in other studies or crops to improve N estimations.

methodology

Other studies should consider the application of this method on multispectral datasets.

The authors concluded that they have shown that PV-photosynthetic vegetation had the strongest effect on the PLSR model, with increase in PV values associated with low variations in residuals of the model.

The experiment conducted across different farm areas in the USA showed that the results are consistent, and the main question posed of “ how nutrient content estimates perform when applied to pixels not dominated by photosynthetic vegetation (PV)” was addressed by the conclusion.

Tables, figures and data are of high quality

Line 43: the authors make references of previous studies [25-28]. I suggest that the authors give more details of these previous studies. This can assist readers that are not familiar with these studies. 

Methods 

Line 135: the link is broken and not working. I was able to locate the files on: AutoMCU (zenodo.org). This must be corrected. 

 Line 136: The reference 3 has ‘?’ which must be removed. 

Figure 8: figure a, b and c must be named correctly on the caption.

Author Response

Dear Editor and reviewers,

We appreciate the time and effort you have dedicated to reviewing our manuscript. Your insightful comments and constructive criticisms have been invaluable in guiding our revisions and improving the quality of our work. We have carefully considered each point raised and have made corresponding modifications to the manuscript to address these concerns. Enclosed to this message (the pdf file), we provide detailed responses (in color blue) to each of the comments and explain how we have incorporated your feedback into our revised manuscript.

We are sincerely thankful for the opportunity to improve our work with your guidance and look forward to any further suggestions you may have.

Kind regards. 

Author Response File: Author Response.pdf

Reviewer 4 Report

This paper deals with the estimation of nitrogen in plants using satellite images in the visible and near infrared spectra, applying the PLSR and AutoMCU statistical methods. Overall, the article is well written and organized, the experimentation is well designed and scientifically sound, and the results may be of interest to other researchers in the field. However, some aspects need to be corrected or clarified:

1. In the abstract, it would interesting to include some numerical results that make the article more attractive to potential readers.

2. In the introduction, self-citations should be eliminated as much as possible, because it is not a good practice in scientific research. For example, 15 of the references are self-citations by Prof. G.P. Asner. Some of them are not necessary, as in ".... Bare Soil (BS) fractions are of particular importance [1-4]...". A reference would suffice to exemplify what is said. In my opinion, I believe that it is necessary to reduce the number of self-citations to make the article acceptable.

3. In section 2, in leaf collection, is the number of leaves collected sufficient for nitrogen estimates to be statistically representative of the total? For example, for many species only 15 leaves have been collected. An analysis of the consistency and variability of the nitrogen measurements obtained for the leaves collected would be useful.

4. In Table 1, what does "7*15 +5" mean? I assume that from some fruit tree only 5 leaves were taken. But from which one?

5. Clarify the concept of "non-photosynthetic vegetation". What type of vegetation is considered to be non-photosynthetic? 

6. In page 4, the link to: https://github.com/CMLandOcean/AutoMCU, is broken. It does not work, it gives a 404 error.

7. Line 136 reads: "in previous studies [3? ]". So, is it [3] or not?

8. Figure 4 would be better placed on page 6, closer to where it is mentioned.

9. In section 2.5, why do you use Random Forest (RF) regressor algorithm and not some other more modern and accurate type of regressor? The grid optimization of the hyperparameters would not be needed by other better regressors.

10. At the beginning of section 3, you say: "we found that the AutoMCU algorithm produced values of endmember cover fraction that were highly consistent with field observations (Fig. 5)". But the images in Figure 5 represent a very small area of land, perhaps less than 1x1m. Do these images correspond (even roughly) to the pixel size of the aerial images? That is, if the resolution of the aerial images is for example 5x5 m, would it make sense to put in Figure 5 a photograph of a slice of about 5x5 m?

11. In Figure 6, it is difficult to see the percentage obtained for each class. It would be better to put three images in gray, one to represent the percentage obtained for each class.

12. Please arrange Figure 9 in a different way, so that it does not appear rotated, because it is very uncomfortable to read.

13. What is the unit of measurement of RMSE? If it is not known, it is difficult to know whether an RMSE of 0.5 or 2.25 is good or bad. The Rnorm measurement is easier to interpret, because it is a relative  value. It would be interesting to use Rnorm in Figure 8, instead of RMSE, in order to better interpret the results.

14. The conclusions lack a discussion of the drawbacks, limitations and possible improvements of the proposed method. The R2 values obtained are not good, with a maximum of only around 0.65. Is this low precision considered sufficient to be of practical use?

15. The writing is correct and only needs the correction of some small errors.

Author Response

Dear Editor and reviewers,

We appreciate the time and effort you have dedicated to reviewing our manuscript. Your insightful comments and constructive criticisms have been invaluable in guiding our revisions and improving the quality of our work. We have carefully considered each point raised and have made corresponding modifications to the manuscript to address these concerns. Enclosed to this message (the pdf file), we provide detailed responses (in color blue) to each of the comments and explain how we have incorporated your feedback into our revised manuscript.

We are sincerely thankful for the opportunity to improve our work with your guidance and look forward to any further suggestions you may have.

Kind regards. 

Author Response File: Author Response.pdf

Round 2

Reviewer 4 Report

The authors have revised the article according to the suggested indications, making the requested changes. Therefore, in my opinion, the article is acceptable for publication.