Development of Machine Learning Models to Predict Compressed Sward Height in Walloon Pastures Based on Sentinel-1, Sentinel-2 and Meteorological Data Using Multiple Data Transformations

Round 1

Reviewer 1 Report

Lines 18 and 19: “The aggregation was made either before or after the prediction and similar performances were observed.” This sentence is not clear.

Since Remote Sensing is an international journal, the introduction should focus on the region to which the authors refer. Some of the claims about dairy production are not suitable for some other regions.

As a general comment, I think it might be useful to include a description of the grasslands analyzed in this study. Are they natural grasslands? Multispecies or monospecies?

Line 68: “..Walloon scale..” I am not from Belgium so I am not sure what Walloon is. Please clarify this assuming readers are unfamiliar with this region.

Line 99: Some geolocation coordinates should be included for reference

Line 101: “.,one for each farming area,..”. This sentence is not clear.

2.4. Split the dataset Why those farms were selected for each group?

The Discussion should be condensed, referring only to the most relevant aspect of the results.

Finally, although the models with the best performance are mentioned, it is not clear which is the minimum set of the most relevant variables that must be monitored to feed the model.

Author Response

Please see attachment for the reply and the whole document is available under the following link : https://filesender.belnet.be/?s=download&token=7a80e668-8aa6-430a-a51e-d405dd04234f

Author Response File: Author Response.pdf

Reviewer 2 Report

Overall, I fund the study an interesting work, especially the results ad the discussion sections. However, the Introduction and Method section should be improved before acceptance. See below for my comments:

Major comments:

• Th introduction should be comprehensively revised. Although the study is mainly about the applications of remote sensing data/methods for CSH estimation, there is a little information about different RS methods for this application. Authors should provide more information about different methods and explain several state-of-the-art studies that b have been conducted in this field. The authors described the characteristics of S1 and S2 in the Introduction section (which should be in the dataset section not in Introduction), but they haven’t discussed the studies which used these data for CSH investigation. Also, the authors have utilized machine learning methods for this application, but there is not enough information about different machine learning methods/studies which have been already employed for CSH estimation.
• Grammar should be improved. The authors are incorrectly using past and present forms of verbs in a sentence of a paragraph. There are many informal and awkward words/statements in the text. Please avoid these. For example, “line: 200: Indeed, Ali et al. [85] insisted”, Line: 223: “So” is not a d=formal word to be used in a scientific paper, Line: 246: “including the 12 used” in the variable selection process, Line 253, and many more, some of which are also indicated below. There are several extra “:” in the text (e.g., Line 256 and Line 262 “:” should be removed).
• I think the method should be reorganized again, maybe dividing the information into a couple of meaningful subsections, which explains the main steps (e.g, “data processing”, machine learning (training, validation, etc,)….

 

Minor comment

• Line 63-66: This should be included in the last paragraph of the Introduction sentence where you explain the current study and its objectives
• Please do not bold the abbreviations, such as DSS, PGSUS, etc..
• Line 40: need a reference
• Line 41: Several alternatives for what?
• Line 67: have been widely … or are being widely….
• Line 84: However, ….
• The flowchart is too busy. The authors can remove some of the boxes and keep the main box which show the main steps
• Figure 1 caption: developed for what?
• Line 98: Site description should be actually a separate subsection (study area)
• Line 105-117: I think the CSH data and meteorological date should be provided under two different subsection
• Line 118: The authors did not explain why they only used VV/VH polarizations
• Line 173: I think by “samples”, the authors mean “resampled”. This should eb also revised in the flowchart
• Line 220: “a variables selection method”
• Please rotate 90 degree the Table 1 and Table 2
• Subsection 3.1: most information included in this subsection (including Table 3 and Figure 3) should be in the datsets section. These are not part of results. Table 4 would be in the datasets or method sections.
• Figure 5: The fonts are too large. Why not using black color for the variables
• Concussion can be improved a bit

Author Response

Please see attachment for the reply and the whole document is available under the following link : https://filesender.belnet.be/?s=download&token=7a80e668-8aa6-430a-a51e-d405dd04234f

Author Response File: Author Response.pdf

Reviewer 3 Report

Dear Authors,

I like the topic and your idea, but the manuscript needs to be revised significantly. First, the text is too long, there is a lot of data presented in details. So, the paper is difficult to read, it is necessary to return to selected parts of the text few times due to amount of details. I suggest to focus on a few problems, significantly reducing the text (some may be moved to the supplement/annex).
The article requires detailed editorial verification because there are too many, so-called citation trains, each reference must have an effect on the idea presented in the manuscript, interline and the list of references must be corrected.

More details you can find in the attached manuscript.

best regards

Reviewer

Comments for author File: Comments.pdf

Author Response

Please see attachment for the reply and the whole document is available under the following link : https://filesender.belnet.be/?s=download&token=7a80e668-8aa6-430a-a51e-d405dd04234f

Author Response File: Author Response.pdf

Reviewer 4 Report

This article evaluates the performance of multiple machine learning models that use inputs of meteorological and Earth observations data (optical and radar) for estimating compressed sward height. In the context of grassland management decision making, this article addresses a relevant issue in being able to rapidly asses grazing biomass across large spatial extent. I did, however, generally find this article difficult to follow in places, particularly in the Introduction and methods sections. I would, thus, recommend revisions principally in these sections before the article is accepted. I have highlighted some examples of improvements in the text below.

I would recommend that several sentences should be checked for clarity and restructured throughout the Introduction section. The Introduction would also benefit from more clearly defined objectives and/or key questions.

In the Introduction the term “… transformation of signals …” and Methodology section details the data transformations, however, background and justification on what is meant by this term is required. It is not clear on why the transformation is applied and what the objectives are. Generally, more details should be added to figure and tables captions throughout the manuscript.

Line 12: “…most performing models..” highest performing models?

Line 25: “….good public image….” This phrase is subjective. I would suggest re-phrasing.

Line 24-28: The authors mention a “higher difficulty of management of grazed pastures” it is, however, not clear what the alternative is to grazing pasture. It would be good to mention the alternatives and the associated disadvantages.

Line 31: “…by facilitating their work…” is not clear. Please elaborate on this.

Line 38-39: Non-destructive in-situ measurement (e.g., LAI, height, visual analysis) could also be used to infer dry biomass per area. This approach would of course be less accurate and still present the issue of scale (i.e., from point to field) but would be less laborious than the assessment of destructive samples. The authors mention non-destructive measurements in the proceeding paragraph but would be good to mention in the context of model inputs.

Line 45-46: It should be mentioned that a big unknown is how to position the RPMs to best capture the spatial variability in height of given pasture.

Line 54: “…prism of the remote sensing field…” please consider revising this sentence. Please also be specific about remote sensing, e.g. from satellite and airborne platforms.

Line 58: “..bunch..” consider re-phrasing.

Line 58: “..recent technological advancements..” please add details here.

Line 74: Please detail what is meant by “…quality…” of S1 and S2.

Line 75: Should be “..offers..”

Line 78: Why “Likewise …”?

Line 80: It should be mentioned that the S1 spatial resolution is also mode-dependant.

Line 84-85: Please add more detail on this, “…encompass any transformation of those signals…”?

Line 87: Again, please define what is meant by “…encompass multiple transformations…”.

Figure 1: Figures seems untidy (e.g., TRained ML) I would suggest revising the figure to make it clearer for the reader to follow. Some parts of the figure could also be summarised as they are covered in subsequent text.

Figure 2: Please add more detail in the caption, e.g. what does the red box with the cross-hatching represents?

Line 112: Why “degree-day-18”? This is unclear and has to be defined.

Line 116-117: Are eight meteorological variables considered in this study? Please check this.

Line 122: The spatial resolution of the S1 data products should be mentioned here.

Line 168: I guess the sub-division would be determined by the highest resolution of the satellite data, i.e. 10 m is the highest resolution for the S2 data.

Line 173: “…sampled using the same way…” please consider revising this text to read more scientifically.

Line 177: Should be “…packages…” perhaps.

Line 186: Is meteorological data really considered non spatial? Please consider changing this.

Line 195: “1&”? is this is typo?

Line 203: Untidy URL. Please consider including the reference section instead.

Table 1: Text right-aligned is difficult to read. Please consider changing this.

Figure 3: More detail should be provided in this figure caption.

Figure 11: Text too small and a legend label should be added.

 

Author Response

Please see attachment for the reply and the whole document is available under the following link : https://filesender.belnet.be/?s=download&token=7a80e668-8aa6-430a-a51e-d405dd04234f

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Although the authors applied most of my comments, there are still a couple of comments with no answer. Specifically, the authors should add more information to the Introduction section. Here are my two previous major comments on this:

- The introduction should be comprehensively revised. Although the study is mainly about the applications of remote sensing data/methods for CSH estimation, there is a little information about different RS methods for this application. Authors should provide more information about different methods and explain several state-of-the-art studies that b have been conducted in this field.

- Also, the authors have utilized machine learning methods for this application, but there is not enough information about different machine learning methods/studies which have been already employed for CSH estimation.

Author Response

Dear reviewer,

Thank you for your thorough review. Please find attached the updated and highlighted version of the manuscript and the answer to the comments: https://filesender.belnet.be/?s=download&token=4ab21ae4-c991-40f9-9134-cb3f6d9377ae

Kind regards,

Charles

Author Response File: Author Response.docx

Reviewer 3 Report

Dear Authors,

thank you very much for your improvements, the manuscript looks very interesting. You addressed all my comments.

There should be revised some formats, e.g. journal names should be shortened.

I wish you a lot of citations.

best regards

Reviewer

Author Response

Dear reviewer,

Thank you for your thorough review. Please find attached the updated and highlighted version of the manuscript and the answer to the comments: https://filesender.belnet.be/?s=download&token=4ab21ae4-c991-40f9-9134-cb3f6d9377ae

Kind regards,

Charles

Author Response File: Author Response.docx

Reviewer 4 Report

I appreciate the authors' thorough responses to my previous comments. Overall, I think the manuscript has been significantly improved and my suggestions have been largely addressed. Before recommending for publication, I just have some minor comments that should be addressed:

Line 25: For me the phrase “…better acceptance from the public…” is still a bit ambiguous and, I think, is superfluous.

Line 50: Are there any studies where they have explored different sampling patterns. If so, past research might be useful to cite here.

Line 78: I think the sentence “The Sentinel constellations (Sentinel-1 and Sentinel-2) were chosen” is now redundant, consider removing.

Line 84: Consider removing “…many …”

Figure 1: I still think this figure could be more visually appealing. I think, at least, the caption should be expended to explain the figure more to help guide the reader through the flow diagram.

Line 115: I think the degree-day-18 has to be justified in the methods section (under “Meteorological data”). Why was it selected? Any citations for this?

Line 177: Please mention that the 10 m sub-division was based on the highest resolution data (i.e., Sentinel-2).

Figure 3: You could make the caption more stand-alone for the reader. Define the CSH, what do the error bars mean? Mention that these are weekly CSH measurements, Farms or fields?

Author Response

Dear reviewer,

Thank you for your thorough review. Please find attached the updated and highlighted version of the manuscript and the answer to the comments: https://filesender.belnet.be/?s=download&token=4ab21ae4-c991-40f9-9134-cb3f6d9377ae

Kind regards,

Charles

Author Response File: Author Response.docx