Fresh Yield Estimation of Spring Tea via Spectral Differences in UAV Hyperspectral Images from Unpicked and Picked Canopies

Round 1

Reviewer 1 Report

This manuscript focuses on the research on spring tea fresh yield estimation by using UAV hyperspectral images from the unpicked and picked Anji white tea tree canopy. Several indices, including chlorophyll spectral indices (CSIs) or leaf area indices (LAIs), have been analyzed for this estimation work. This application work is interesting.  I have the following concerns with the manuscript:

1)     In the abstract, the full name of CUR (Line 27) should be declared at the first time. So is the full name of RMSE\VR² (line 30);

2)     In the FigureA5b- Figure A6d, the “Unselected” models should be explained;

3)     I recommend that more hyperspectral images should be displayed in the paper.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Dear authors,

the manuscript deals with hyperspectral remote sensing. It is relatively new and the literature about is topic is still needed. Unfortunately, the manuscript is difficult to read. Deep English revisions are needed before considering it for publication. Moreover, the structure of the manuscript should be improved: Introduction should be more focused on on the methods and the gaps of knowledge. M&M have sub-sections in form of bullet points and it should be avoided. Moreover, M&M are present in the results section and should be separated. I think that the difference among CSIs and LAIs indices is inappropriate and should be better explained because LAI is a specific crop variable so another name for the vegetation indices related to LAI should be used.  More about data analysis should be described in the M&M section to better understand the results. The potential of hyperspectral remote sensing should be better explored by the authors

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

The authors present a method to estimate fresh tea yield from hyperspectral imagery of the canopy that involves vegetation indexes and a piece-wise linear combination of indices based on canopy cover.

The paper structure is messy, only until the Results and Analysis one can get a grasp of what they mean in the previous sections. A methodology section should be added to describe the methods. The introduction spends a lot of time and one Figure on the obvious difference between young and mature leaves. There is no explanation whatsoever as to why and how RGB images must be transformed to HSI since canopy cover can be directly detected with RGB images. There is a confusing and unnecesary use of non-piecewise function, since the linear function is non-piecewise as well as nonlinear functions. According to the results there are two models linear and a piece-wise linear function nothing more. 

There is an artificial division of vegetation indexes into Clorophyll indexes and LAI indexes as if those LAI indexes could only be used to compute LAI. The cited LAI vegetation indexes are used for many other applications than just LAI.

The methodology for the single index linear models and single index piece-wise linear model is adequate dividing the 24 samples into training and validation (Tables 3 and 4), but the combined indexes piece-wise linear model uses all 24 samples (Table 5) and there are no validation samples and cannot be compared with the other models since it only contains training data, which is a serious methodology problem. In addition, there are many combinations of vegetation indexes possible and no explanation as to how the authors chose the best index combinations to avoid saturation.

Figure 7 (which should be Figure 8) on page 7 shows the overall methodology is hard to read and does not help clarify the methodology with small figures that don't say really much. Figure 7 on page 6 doesn't help to understand the saturation problem covered in the "Green coverage change acquisition of unpicked and picked tea tree canopy".

Figure 6 shows a very good linear relationship between fresh tea area and and fresh tea yield, however there is no discussion as to why the so-called LAI indexes failed to exhibit this behavior.

The literature review is adequate and the gap in knowledge covered is relevant to the scientific community.   

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 4 Report

It was not clear why a Hyperspectral image was needed in this study.  The multispectral image should fulfill the application if the prediction model uses only VIs as input.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

I think this manuscript can be published as it is .

Reviewer 2 Report

Thank you for your revisions.

Reviewer 3 Report

I am satisfied with the reviewed mansucript. However, I'd suggest to change the use of the "ideal" word in the manuscript. The R² and RMSE for some methods may the best, but no ideal (R² ~ 1, RMSE ~ 0g). The "ideal" word in the abstract should be changed to report the actual best R² and RMSE values. 

I'm not requesting "accept after minor revision" to avoid delays in publications, so I request the authors to please change the "ideal" word that it is missused in the manuscript, specially in the abstract.