Yield Estimates by a Two-Step Approach Using Hyperspectral Methods in Grasslands at High Latitudes

Round 1

Reviewer 1 Report

Dear Authors,

thank you for your improvement, it was a good step, but there are a lot of other things, which should be improved. The most important:

- you focus on the EnMAP, it isn't a problem, but why you don't try to validate the results on the Sentinel2?

- your Introduction is significantly too general, you don't present a theoretical background of your issue, please add much more details oriented on similar solutions, which are used by other researchers,

- you Methods are described too general, please add much more details.

- your results are presented clearly, you add a lot of table data, but please present a graph to show better and worst results. A lot of your graphs aren't interpreted, there are non-linear relationships, please, add R2 on the graphs.

Much more detailed comments you can find in the manuscript.

Best regards

Reviewer

.

Comments for author File: Comments.pdf

Author Response

Dear Authors,

thank you for your improvement, it was a good step, but there are a lot of other things, which should be improved. The most important:

Dear Reviewer:

Thank you for a thorough review of the article. We have followed your advice, both here and in the PDF file, to improve the article. We believe that the changes you have suggested have improved the readability and clarity of the article. We hope that this revision satisfies your suggestions.

- you focus on the EnMAP, it isn't a problem, but why you don't try to validate the results on the Sentinel2?

We have added Sentinel2 to the article, and we agree on that it makes the article more balanced. Thank you for the suggestion, it adds a new dimension to the article.

- your Introduction is significantly too general, you don't present a theoretical background of your issue, please add much more details oriented on similar solutions, which are used by other researchers,

We have added more information in the introduction, starting from the historical background by adding information on VIs (lines 47-50), limitations on the remote sensing approach and statistical approaches to them (lines 55-60) and the use of RS toolboxes with their limitations (lines 74-80). In addition, we have included Sentinel-2 in the introduction, and stated that Sentinel-2 can outperform Landsat 8 in winter wheat.

In the RS toolboxes part, we could not find any reference where specialized toolboxes exist for high latitudes. The information available states that correction parameters for satellite images in high latitudes and areas with air moisture (i.e. close to the coast or water bodies) are not available, and that the standard tools may have biases that can affect the applicability of the satellite images.

- you Methods are described too general, please add much more details.

We have structured the methods to clarify the description and added details on the sampling procedure: spectral acquisition is now better explained  (lines 137-142), as well as data pre-processing (lines 174-178) and the satellite simulation procedure (lines 181-185). We have also updated the model validation part, we think it reflects better the procedure now.

- your results are presented clearly, you add a lot of table data, but please present a graph to show better and worst results. A lot of your graphs aren't interpreted, there are non-linear relationships, please, add R2 on the graphs.

We have re-ordered the results section, and decided to only put the main models in the results section, and rather have the rest of the results in supplementary material to make the article easier to read. We present the descriptive statistics as a plot now (Figure 3), which we believe makes it visually more appealing than a table.

As for the non-linearities, we have decided not to discuss them in depth in this article: the models showing non-linearities aim to describe the challenges of using data captured under good conditions to predict data gathered under bad conditions and vice-versa. Therefore, we have mentioned the non-linearities in the text, but the figures are now part of the supplementary material to avoid distracting the readers from our main study, that deals with the yield predictions under the high-latitude challenging conditions.

Much more detailed comments you can find in the manuscript.

We have also harmonized the chapter formatting, we believe we have followed the guidelines now, as well as the plots. If there are any further issues with formatting or figures, we will gladly correct them. Thank you for pointing out e.g. that the map did not comply with the journal standards: we have now updated all the figures to reflect the author guidelines.

Best regards

Reviewer

 

Reviewer 2 Report

The authors made all the corrections based on the suggestions. 

Author Response

Dear reviewer 2,

Thank you for your revision. We have updated the manuscript following the advice of Reviewer 1 and we believe it has improved in readability. In addition, we have included Sentinel-2 into the simulations, which has made the paper more balanced and gives a deeper possibility on applying a combination of satellites for more efficient management of grasslands in high latitudes.

Thank you for the comments throughout the publication process,

The authors

Round 2

Reviewer 1 Report

Dear Authors, thank you very much for your significant improvements. The manuscript looks much better, but please look at the text, you have different styles in different parts of the manuscript. There is missing one figure or table. Please, look at the guidelines to improve it. I marked some problems in the attached text. Best regards Reviewer

Comments for author File: Comments.pdf

Author Response

Dear reviewer,

Thank you for pointing out the variation in style along the document. We have now harmonized the style, the text should have a single style through it. 

We have corrected the table format, removed a few typos and added some extra words for clarification along the document. 

We appreciate our effort along the whole submission process, and hope that you have enjoyed the process as much as we did, the work has improved greatly thanks to your comments, together with the other reviewers and editors.

We will happily provide with further corrections, if required.

Best regards, 

The authors

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

Dear Authors,

I like your idea, because the growing period of vascular plants is very short, and abiotical conditions very hard, otherwise pressure for agriculture area is increasing. So, your idea is interesting and should be developed, but you have to improve the manuscript first.

The Introduction is too general, please add more details on a theoretical background of your topic. You could present more details of cited papers.

In Methods, you could present a research schema of your investigation. Please add more details of data acquisition and processing.

Discussion should be deeper as well, please prepare a table to show a similarity of your scores and results achieved by other researchers.

Much more details you can find in the attached manuscript.

Best regards

Reviewer

Comments for author File: Comments.pdf

Reviewer 2 Report

Something went wrong with the pdf version of the paper. Please provide a correct version. However, some comment are reported in the present document. The method should be described better. How many channels did you use to reach your results ? The bed behaviour of the model in bed weather conditions could be due to the difficulty to calibrate the ground measurements under changing illumination conditions ?

Comments for author File: Comments.pdf

Reviewer 3 Report

The research article “Yield Estimates by Hyperspectral Measurements in Grasslands at High Latitudes” is a very interesting article that study the traditional grassland yield with remotely sensed data but in a high latitude areas. The study is scientifically sound, time-series results make sense, the PLS method for analyzing hyperspectral data is reasonable. There are some editorial corrections need to be done but overall it’s good quality. It should be acceptable by the journal.

Reviewer 4 Report

In spite of not being an absolute novelty the use of hyperspectral data to estimate biomass in grasslands, the northern location and the use of a PLS to retrieve robust models enhance the importance of the data, that deserve to be published after a thoroulgy review.

I suggest authors  reflect in the first place on the objectives and possible outputs of their work and review the whole study.

 

Introduction

A better revision of the impact of climate change on vegetation dynamics at high latitudes, on the use of remote sensing for assessing vegetation cover and other properties is necessary. It should focus on the vegetation indices selcted for the analysis, on the potential use of the handheld spectroradiometer (Fieldspec), and on the opportunities provided by remote platforms for vegetation monitoring. The objctves of the study are also not clear.

The sentence “explore potential and limitations of the handheld” does not clarify for which uses the handheld is proposed (i.e. only for investigating which wavelengths better represent temporal dynamic in vegetation changes? Or for practical uses? ), equally the choice of the simulation of EnMap  in comparison with other satellites deserves a better explanation.

Material and methods

The material and methods section is also poor. The criteria for sampling points distribution is not clear.

Page5 line 3. Which are the directions of cross transects and of the grid transect? How long they were? How many measurements were taken every transect?

Page5 line 6. The 0.25m2 plot refer to the field of view of the FieldSpec.

In addiction the height of measurements and the field of view angle need to be specified.

Page5 line 6. The white reference is used to normalize measurements for variations in the atmospheric conditions not for calibration, which is usually performed in the factory.

Page5 line 7. What authors means when they mention to have developed a model based on the 350-900nm? They simulate reflectance from the Fieldspec in that range?

Hyperspectral measurements with he fielddspec should be collected only under sunny conditions. all point collected under cloudy weather should be deleted.

Results

Vegetation indices

No trends were visible in the correlation observed between vegetation indices and dry biomass. Data are presented in the figure with different trend according to the year. Did authors observe any differences in the regressions across years? These differences could be evidenced and commented.

Hyperspectral

Results from the hyperspectral measurements describe only the results of the PLS. Parameters of the PLS shown in table 3 are not mentioned in the text. Parameters should include adjusted R2 instead of R2 and RMSE for the calibration and validation data set.

The most useful information possibly retrieved from this study, the wavelengths with major influence in the PLS models are not shown.

Table 3. It is generally recommended to use the handheld spectroradiometer under clear sky conditions. Indeed, data modelling under cloudy conditions need

Figures and tables

Figures frequently don’t show units, in composite figures different frames need to identified (a,b,..).

Figure 2 shows the relationship between 3 different vegetation indices and above ground biomass (named “measured yield”). The term “measured yield” is not proper, authors refer to the amount of dry biomass collected in the same field of view of Fieldspec measurements. Units are missing.

Figure 3. Units are missing on both axes. All pictures should have the same scale for a better interpretation of results.

Table1. The legend of the last 3 columns is not clear. What N refer to? The number of handheld replicates in the 1ha area?

Table 2. The table was supposed to show the amount of dry biomass collected after each field measurement. The purpose of the table is to show the range of the biomass collected at each measurement data? Maybe it would have been more useful to show this information in a figure, showing the temporal trend of biomass along time.

Table 3 the parameters of the PLS RMSECV e RMSEP are not specified

Detailed comments

Throughout the whole manuscript the term yield is used to define the amount of dry biomass collected above the soil. I would rather adopt: above ground biomass since yield is usually referred to production along a time period.

Reviewer 5 Report

Partial least squares regression (PLS) is already a very popular regression model in remote sensing. Therefore, the current submission is lack of novelty. The advantages of your study were not sufficiently appealed and a thorough rewriting and restructuring are required in order to improve the scientific level.

1. Introduction

Something is missing about why vegetation indices are special in remote sensing applications and why they are even more important.

You said that more comprehensive models including MLR and ML have been developed as well as PLS. Why did you ignore MLR and ML?

2. Materials and Methods

Vegetation indices

Why did you select only three indices? There are a lot of VIs. You should describe the reasons.

Hyperspectral analyses and modelling

You should clarify the details.

Did you repeat the sampling procedure (L.133)? It should be better for more robust conclusions.

Did you conduct any variable selection?

3. Results

Vegetation indices

You denoted some R2 values.

Were the relationships significant?