Creation of a Walloon Pasture Monitoring Platform Based on Machine Learning Models and Remote Sensing
Round 1
Reviewer 1 Report
The manuscript analyses the suitability of compressed sward height estimation using machine learning models and remote sensing imagery. The approach is relevant and the analysis seems consistent for the proposed aim. However, it lacks relevant description of the Material and Methods sections as well as appropriate presentation of the Results and Discussion sections. Moreover, the Conclusions section provides more future perspectives than the important conclusions according to the aim of the manuscript. The manuscript topic has relevance to be future published, but it requires a consistent review and reorganisation.
Some other comments are presented below:
Despite the authors providing an interesting review about the most relevant model to understand and monitor pastures (Annex 5.1), they do not provide any of the references and, so it is difficult for the readers to follow up the review and critically analyse it.
The manuscript presents a good overview about the importance of grasslands and the necessity of developing Decision Support System (DSS) to monitor pastures. The manuscript focuses on the modelling of the compressed sward height (CSH). However, the Introduction section does not have information about the relevance and suitability of the CSH to monitor grassland systems throughout a DSS. Then, an appropriate citation and description of the CSH importance is necessary.
The authors referenced their previous paper for describing part of the methods and supporting the results. It is interesting that the manuscript is a continuation and improvement of their previous research. There is no problem with this type of auto-citation. However, the manuscript lacks important and clear description about the way the modelling was structured, the reference field data, time stamp of the field data according to the dates from the satellite dataset, the predictor variables, the feature selection algorithm, among others. For an appropriate comprehension of this manuscript is practically mandatory to read the previous paper.
The section 3.2 provides texts that should be better placed in the Material and Methods section because they do not add any results or discussion instead of just justifying some decisions adopted in the methodology.
Page 9, line 333: What is PAA? It needs a full name description.
Page 10, line 382: The numbering of the Annex 9.3 (and the further ones) differs from the previous ones (Annex 5.1 and 5.2).
Page 11, line 417: “(from 104 to 276)”. What is the source of the value 104?
Page 11, Table 3: What is the difference between the row “N dates*” and “N dates acquisition S2”?
Page 12, Figure 4: The x and y labels are not appropriately described. The month names (x-axis) are not in English.
Page 15, Figure 5: The x and y labels are missing. Also, the explanation of this figure in the text is very difficult to follow.
Page 16, lines 531-534: The text misses information about the reasonable explanations on the over and under-estimates between the estimations.
Page 19, Figure 7: Does this figure present the results for only one parcel or multiple parcels? Where are the parcels? Why do some pixels seem to be borders of some fields?
Page 20, Table 7: It lacks an appropriate explanation in the text about the ranking approach used.
Author Response
Dear reviewer,
Thank you for your kind and enriching review and improvements hints. Please find attached the document gathering your remarks and our answers.
Kind regards
Author Response File: Author Response.docx
Reviewer 2 Report
Dear Authors
I find the manuscript very interesting.
The Acronym can generate some doubt. Please make sure that the literature is free from similar acronyms. The title is too long.
But in some parts, I've been lost with the list of packages, the complexity of diagrams, and the reference body not comprehensive.
A package list can be put into a table or converted to supplementary materials.
Diagrams are exemplary (flow diagrams), but histograms like fig 5 are not very easy to contextualize; please figure out how this information can be turned into a more appealing one.
Reference body
one sentence with one reference, the most updated, or two if strictly related; please avoid writing a few words followed by a group of references; in some case, I've seen 5 of them.
Hyperparameters
The community, as well as readers of all kinds, need to know the range of hyperparameters tested in validation.
Knowing the final set of hyperparameters is of little sense.
Kind regards
Author Response
Dear reviewer,
Thank you for your kind and enriching review and improvements hints. Please find attached the document gathering your remarks and our answers.
Kind regards
Author Response File: Author Response.docx