Review Reports

Round 1

Reviewer 1 Report

General comments

The authors of the article, have made some improvements to the text that I have reviewed before. However, before it is published, the sections newly added (in red) need to be proofed for English language and in some instances rewritten so they make sense in English (even though I appreciate the effort of using a certified textcheck.com they did not do a good job for the added paragraphs). Please make sure they are reviewed and corrected.

Abstract: The first sentence in the abstract should be improved to something like: Human induced deforestation has a major impact of forest ecosystems and therefore its detection and analysis methods should be improved.

Introduction: It contains much better description of the study (L86-100) than other versions of the article

Results: section added (L268-278) needs to be checked for English language and to be rewritten so it becomes more fluid.

 

Specific comments:

L164: soft/hard wood not capital letter

L164: Why do you mean are distributed?

L207: what do you mean algorithms were used to precisely construct database?

L270: misclassified by (English mistake)

L271: results for buildings (no comma)

L291: should be written: performance in classes with small number of pixels

L295: review of … confirmed that

L311-312: does not make sense in English

L314: from high

L325: First talk about overall then about specific classes

L336: which may improve

Author Response

Thanks for the reviewer's comments. Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

From my consideration, I think the quality of this paper has been improved a lot. It can be accepted after the minor revisions.

 

The caption of each figure is not holistic. You should improve the title of each figure.

Author Response

Thanks for the reviewer's comments. Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

The authors have made a reasonable attempt to address my comments and suggestions. However, there is still no analysis or quantification of disturbance or deforestation, and I stand by my original comments. 

The issue is not the use of the term 'disturbance', as opposed to 'deforestation'. Deforestation is in fact a form of disturbance. The issue is that both these terms imply a change in state from a 'undisturbed' or 'intact' state to one that is considerably altered by internal or external drivers, such as human activities. Although the presence of human made structures implies a change in state from an intact, unaltered ecosystem, i.e. a forest, to a disturbed one, the authors would still need to quantify that change between two points in time. 

Therefore, the authors need to either 1) perform classification of a pre-disturbed scene and quantify the amount of change post-disturbance, or 2) change the scope of the paper to focus on the use of deep learning for land cover classification in more general terms. 

Author Response

Thanks for the reviewer's comments. Please see the attachment.

Author Response File: Author Response.docx