Mapping Secondary Vegetation of a Region of Deforestation Hotspot in the Brazilian Amazon: Performance Analysis of C- and L-Band SAR Data Acquired in the Rainy Season

Round 1

Reviewer 1 Report

In this paper, C- and L-band SAR data were acquired to discriminate secondary vegetation in the rainy season with the advantage of penetration cloud. Two classification methods，RF and SVM, were employed to extract the secondary vegetation using backscattering coefficients and texture features with difference sizes. The accuracy of classification using L band ALSO-2 SAR images are higher than that from C band sentinel SAR images. The experiment of this paper is relatively complete, and the selected data and methods are relatively common. Further exploit and innovations should be extracted from the results. However, there still has some problems as follows:

(1) The part of introduction should add some previous results about the classification using optical images and SAR images, specially in secondary vegetation regions;

(2) for classification, there are many methods to select features. In this paper, just PCA was applied to get the optimal feature set, it is necessary to clearly explain the reasons of choosing the PCA method.

(3) In this paper, the number of samples is 122, and a set of 2844 points were selected for PF, 1782 points for SV, 2296 points 223 for CP, 834 points for SP, and 168 points for BS, the relationship between samples and points should be clarified. Normally, accuracy of classification is affected by the purity of sample.

(4) Table 3 and table 4 indicated that the accuracy of classification is related with the size of extracting texture features. The reasons should be explained in the paper.

(5) In this paper, the classification results were extracted from L band and C band SAR images, respectively. Confused features set extracted from L band and C band SAR images should be also applied to retrieve the information of secondary vegetation.

Author Response

Dear reviewer,
Please check the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The aim of this manuscript was to evaluate the potential of combined C- and L-band SAR data to map secondary vegetation. The research topic is of significant and the contents are appropriate foe the forest journal. Unfortunately, this paper failed to achieve its objective. Detailed comments and suggestions are listed below:

1. The structure of the article in the second part is a bit confusing. i.e., section 2.4, 2.5, 2.6, 2.8 and 2.9 are all description of the methods while section 2.7 is more biased data.

2. The article devoted a large part of the introduction about the importance, status, government's policy and some protects of re-deforestation of secondary vegetation in Brazilian. And the article did not provide a good review of the current state of this research and problems, which could lead to the content and objective of this research.

3. The flowchart did not show the technical flow and methodological point of the article well. I suggest redrawing the flowchart.

4. I think a period punctuation mark is missing between “9° 45’ 26.5” The study area” in line 107.

5. As demonstrated in section 2.7, the samples of each land cover were severely unbalanced, which may affect the prediction accuracy of the model.

6. I suggested changing “kappa index” to “kappa coefficient”.

7. Random Forest classifiers has the property of feature importance ranking, why did this article use PCA to select features?

8. The conclusion of the article needs to be further condensed.

Author Response

Dear reviewer,
Please check the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

In this paper, C- and L-band SAR data were acquired to discriminate secondary vegetation in the rainy season with the advantage of penetration cloud. Two classification methods，RF and SVM, were employed to extract the secondary vegetation using backscattering coefficients and texture features with difference sizes. The reviewed manuscript has been greatly improved, there is two comments for the manuscript :

1） In the part of Keywords, Secondary Vegetation and Rainy Season should be added. Moreover, Machine learning should be deleted, because the RF and SVM classification referred in manuscript are two of machine learning methods.

2)  The description of the results in the abstract is too simple, and it is necessary to further summarize the innovation of this manuscript.

Author Response

Please see the attachment. 

Dear editor,
Enclosed, please, find the revised version of the manuscript (second revision) and the response to the comments to the first reviewer.
We would like to pay the publication fees using the voucher from the second author (Edson E Sano), if it is still possible.
Best Regards
The authors

Author Response File: Author Response.docx

Reviewer 2 Report

The authors have resonably addressed my questions.

Author Response

No additional comments.