Mapping Grasslands in Mixed Grassland Ecoregion of Saskatchewan Using Big Remote Sensing Data and Machine Learning
Round 1
Reviewer 1 Report
Thank you for the opportunity to review the manuscript 'Mapping Grasslands in Mixed Grassland Ecoregion of Saskatchewan using Big Remote Sensing Data and Machine Learning'. The research deals with an interesting topic, and I believe that the manuscript is worth publishing in Remote Sensing. I would, however, recommend some edits for making the results clearer to readers.
Major comments:
The user and producer accuracies of distinguishing tame grasslands are very low. Hence, I can't entirely agree with statements as 'This research demonstrated how powerful machine learning could be when big data is used to generate valuable and accurate information on the spatial distribution of the grassland types (native, tame, and mixed)' (588-589). We can see great results with native and mixed grasslands but not with the tame ones. Simultaneously, I would hypothesise that the problem will be with the mixed category, not with a relatively clearly defined one. Could you discuss this issue more to make readers sure about the appropriateness of your method and conclusions?
In general, I recommend splitting the chapter Results and discussion into two: (i) Results and (ii) Discussion.
Minor comments:
You combine citations by numbers with citations by authors' names. It is unusual for me.
Redundancies like 'two (2)' are not necessary.
Author Response
Dear Reviewer#1,
Thanks for your comments and suggestions!
Comment #1:
The user and producer accuracies of distinguishing tame grasslands are very low. Hence, I can't entirely agree with statements as 'This research demonstrated how powerful machine learning could be when big data is used to generate valuable and accurate information on the spatial distribution of the grassland types (native, tame, and mixed)' (588-589). We can see great results with native and mixed grasslands but not with the tame ones. Simultaneously, I would hypothesise that the problem will be with the mixed category, not with a relatively clearly defined one. Could you discuss this issue more to make readers sure about the appropriateness of your method and conclusions?
We agree with you, this statement has been modified. Actually, the tame results were high, 81.4% and 93.8% for the User’s Accuracy and Producer’s Accuracy, respectively. As shown in figure 5, our results in Fig.5a have higher accuracies than the AAFCs in b, c, and d.
Comment #2:
In general, I recommend splitting the chapter Results and discussion into two: (i) Results and (ii) Discussion.
Thank you for your comment, we split the results and discussion.
Comment #3:
You combine citations by numbers with citations by authors' names. It is unusual for me.
The combination between names and numbers in citations happens when we start a sentence, indicate a definition, or refer to certain findings. We checked with the author's guidelines. Thank you!
Comment #4:
Redundancies like 'two (2)' are not necessary.
We agree with the reviewer; it was unnecessary.
Reviewer 2 Report
Dear authors, the presented research's main topic is an assessment of grassland spatial distribution using remote sensing technologies, machine learning, and big data analysis. The main topic of the paper is interesting and it has an important contribution to environmental and agricultural research. In my opinion, the manuscript is well written, the information given is enough to understand and replicate this research. I think the methodology and the explanation given in it are the most important contribution of this manuscript. There are some minor errors that should be corrected. In my opinion, this manuscript should be accepted for publication after minor revision.
Minor errors:
There is a figure, I guess figure 1 without caption line 139 resolution improvement needed.
Line 223 will used= will be used
Line 246 typo=getSpatialData
Equations and data symbols have low resolution
Figure 5 x-axis has no units
Lines 420-421 I think there is a wording problem
Author Response
Dear Reviewer#2,
Thanks for your comments and suggestions!
Comment #1:
There is a figure, I guess figure 1 without caption line 139 resolution improvement needed.
Thank you, we improved all figures' resolution.
Comment #2:
Line 223 will used= will be used
Done, thank you.
Comment #3:
Line 246 typo=getSpatialData
This is an R package, we added the word "R-package" beside the name of the packages.
Comment #4:
Equations and data symbols have low resolution
Improved, thanks for noticing this issue.
Comment #5:
Figure 5 x-axis has no units
We added the x-axis label.
Comment #6:
Lines 420-421 I think there is a wording problem
Thank you, we revised it!
Reviewer 3 Report
The objective of this work is the evaluation of grasslands using information from optical multispectral sensors MODIS, Sentinel 1 and 2. Without a doubt, the spatial distribution of native mixed and domesticated grasslands is a key task for the monitoring and conservation of ecosystems. The Article is generally well presented with a lot of information, sometimes difficult to read.
The classification approach seems coherent and methodologically correct, using very classical machine learning and dimensionality reduction techniques. Techniques, although classic and widely used, do not detract from the validity of the work.
The conclusions are validated by the results obtained, although the statement of "This research demonstrated how powerful machine learning could be when big data is used ..." could be very enthusiastic, as other classification methods were not evaluated.
Author Response
Dear Reviewer#3,
Thanks for your comments and suggestions!
Comment #1:
The conclusions are validated by the results obtained, although the statement of "This research demonstrated how powerful machine learning could be when big data is used ..." could be very enthusiastic, as other classification methods were not evaluated.
Thank you for your comment. We agree with you; it was a bit enthusiastic. We have changed it
