Impact of Training Set Configurations for Differentiating Plantation Forest Genera with Sentinel-2 Imagery and Machine Learning
Round 1
Reviewer 1 Report
The authors presented an article titled "Impact of training set configurations for differentiating plantation forest genera with Sentinel-2 imagery and machine learning" to be considered for publication in Remote Sensing. The article is within the scope and aim of the journal.
The article is very interesting and will be of great interest to the readers.
First of all, the article should be properly formatted, following the template available at the journal website.
Concerning the abstract, it can be considered well written and containing the necessary information. However, taking into account the journal template, the abstract shall be limited to 200 words. So, I recommend the authors to reduce and adapt the text to this condition.
Figure 1 must be enlarged because smaller areas are not properly identified.
Table 3 must be rearranged. It is very difficult to understand. Try to find a different type of table, or simply explain the content in plain text.
Divide Figure 4 in two figures (a) and (b) and enlarge each one of them.
I do not know if Table 5 and Figure 5 are aceptable in landscape format. Probably must be converted to portrait. In any case, the information presented is difficult to read. A good option could be sending this information to supplementary materials.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
1. There is a worrying and fundamental relationship in the work: forest plantations and water use. It is in my view the root cause that triggers the research. This aspect should be incorporated in the abstract.
2. Throughout the text there are different calls indicating "(Error! Reference source not found.)". They must be checked and corrected.
3. The use of references is up to date and focused on how to solve the problem. It could be clearer if it were synthesised in a matrix linking the authors, with the sample size and its positive and negative aspects. It will help the reader and also to know the path chosen by the authors of this paper.
4. Since the authors focus on modelling to generate predictions using the random forest algorithm, it would be interesting to mention its creator, the statistician Leo Breiman.
5. I recommend paying more and better attention to maps. An article focusing on improving forest plantation genus maps should be more rigorous in graphic semiology (see Jacques Bertin).
6. Some graphic and cartographic considerations:
Figure 1. Map background in grey, why? Grey is already a colour with value What does it represent?
Figure 2. The geo is represented by a shadow model, and the variable in solid colour, to take advantage of the shadow model, better to apply transparency to the variable. A figure with three maps, two different scales with different semiological criteria. This does not allow comparison. Match the scale to see the representation correctly.
Figure 3. While the homogeneity of map a has a good legibility, map b does not. It would require more detail of one or several areas on different maps.
7. I don't quite understand the meaning of the colour gradations in table 5.
8. Some maps of band combinations, approximations to conflict zones to observe colours, textures, shapes of the sample or samples, etc.
9. The presentation of a methodological table would help to better understand the processes applied.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Title: Impact of training set configurations for differentiating plantation forest genera with Sentinel-2 imagery and machine learning
Imagery is very popular among scientists. They try to use these images for classification purposes and forestry, too.
1. Keywords should not be the same as title words and shorter. Please correct it.
2. The Introduction chapter explains the methods consistently. The purpose of the investigation was clearly presented.
3. Table 3. The summary of Experiments A to D should be repair. The information is not readable in Table 3. Also size of information is too much in all tables.
4. May big results table 5 and figure 5 should be in the supplementary file.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The authors addressed all my comments and I recommend now the article to be published.
Author Response
Please see the attached.
Author Response File: 
Author Response.docx
