Review Reports

Round 1

Reviewer 1 Report

The manuscript reports the findings of a study using drone-basd hyperspectral imaging to determine N-status of potato plants at different growth stages. It is a well-written and interesting manuscript. I would even go as far as to say that it is one of the best examples of this type of study that I have seen in a while. The systematic evaluation of different approaches to hyperspectral data analysis with a translation into how it may be appicable to multispectral sensor use and development makes this a valuable contribution to the field.

I found no major deficiencies in the manuscript.

My only specific suggestion is to make the figure titles more complete by specifically naming the three optimal spectral indices, and to avoid the use of acronyms in the figure titles. 

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 2 Report

The scientific work is well prepared and reveals important data about how the hyperspectral bands can be used to determine the nitrogen content in different phases of potato vegetation. However, the studies must continue for at least one more year, in order to obtain relevant comparative data. 

Author Response

Reviewer 2:

Comments: The scientific work is well prepared and reveals important data about how the hyperspectral bands can be used to determine the nitrogen content in different phases of potato vegetation. However, the studies must continue for at least one more year, in order to obtain relevant comparative data.

Response: Thank you for the valuable comments of reviewer #2. Considering the model's universality, we have used two years of potato PNC data to construct the model. Our experimental treatments include different planting densities and nitrogen and potassium fertilizer treatments. We chose two independent replicates for modeling and the other for validation. As far as our current research is concerned, the data from 2 years have produced good results. The PNC estimation model constructed based on TBI5 (530,734,514) showed good applicability in different years, cultivars, and growth periods. However, whether the method proposed in this study is suitable for potato experiments in different environments and more different years still needs follow-up related research. We apologize for the lack of experimental data sources in this study. In the future, we will collect data from different places and years for verification and analysis to improve the reliability of the test. The shortcomings of this study are described in the Discussion. Thank you again for your valuable comments!

Author Response File: Author Response.docx

Reviewer 3 Report

This paper evaluates the performance of single-, two- and three-dimensional spectral indices for estimating nitrogen contents of potato crops and selects the optimal spectral index, i.e., TBI5 (530,734,514). I think that the findings of this study provide a good guide for efficient N detection of potato crops, and this paper can be considered for publication after revisions.

My comments are as follows:

1. For the introduction, in lines 79-86, these previous studies need not be itemized. It is recommended to delete them.

 

2. In the introduction (lines 92-94), this paper indicates, "The existing VIs are prone to saturation when used with high vegetation cover, and thus cannot accurately reflect the N nutrient status…”.

And the question about “saturation” is proposed many times in this paper, however, I cannot see the solution to the saturation question in this paper.  Please demonstrate how this study's spectral indices (or optimal spectral index) avoid the saturation problem.

 

3. In the introduction (lines 105-107), the paper points out that the complexity of machine learning methods hinders the development and integration of real-time PNC detection devices. However, this study does not seem to achieve real-time detection of nitrogen content in potatoes by combining drones, spectral sensors, and selected spectral indicators. Please illustrate the fact or the potential of spectral indices for real-time and accurate detection of potato nitrogen.

 

4. These spectral indices used in this study have been proposed by other studies, as listed in Table 1. Please point out the improvement or innovation of these spectral indicators in this study.

 

5. In Section 2.5 (lines 228-229), please indicate what are "replicates 2 and 3" and "replicate 1". I think they are related to experimental design (section 2.1). However, please describe the dataset in detail.

 

6. In Table 2, it is clear that some validation sets have a larger PNC range than the corresponding calibration set, for example, the validation set of S1 in 2018 is 2.70-4.59, and the corresponding calibration set is 2.81-4.57. I don't think this is appropriate for modeling analysis, as it will decrease the stability and accuracy of the established model.

 

7. The results of Section 3.2.1 shows that the FDR, SDR, and CR spectra have higher correlation coefficients than OR spectra. So why not use the processed spectra to select/calculate the spectral indices, but use the original spectra to select/calculate the spectral indices?

 

8. For figures 5-8, to improve the readability of these graphs, please modify the text color (i.e., R², RMSE, NRMSE) in each graph to be consistent with the color of the corresponding scatters.

 

9. In Section 4.3, this paper indicates that three-dimensional spectral indices have higher stability and accuracy for detecting the N content of potato crops. This is mainly attributed to the three-dimensional spectral indices contain more spectral bands and can reflect more physiological information about potato canopy than two- and one-dimensional spectral indices.

According to this idea, if multiple characteristic wavelengths are selected for modeling, will the accuracy of the model be higher than the model established by the spectral indices?

It is therefore necessary to emphasize/illustrate the superiority of spectral indices over characteristic variables. Please complement it by citing some previous studies detecting the nitrogen or chlorophyll content of potato crops.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Round 2

Reviewer 3 Report

 Good revision! I recommend acceptance of this paper.