Combining Multitemporal Sentinel-2A Spectral Imaging and Random Forest to Improve the Accuracy of Soil Organic Matter Estimates in the Plough Layer for Cultivated Land

Round 1

Reviewer 1 Report

The authors have estimated soil organic matter from Sentinel-2 images using Random Forest algorithm. The detailed suggestions are as follows:

1.     The title should highlight the importance of the study.

2.     The abstract should be a short and clear summary of the aims, key methods, important findings, and conclusions. Aims and key methods are missing in the current form.

3.     Explanations about experiments and results are too much in the Abstract.

4.     The abstract should clearly bring out the novelty of the study.

5.     Newly published references should be quoted in the paper.

6.     The research question of the study should be clear and appropriate.

7.     The contributions of the study are not clearly expressed in the Introduction section. The authors should provide this with the limitations of current knowledge in this field.

8.     Tables should be given cleaner by reducing line spacing.

9.     Figure-1 Part A should be provided in the Study Area section. Also, it should include location names and legend.

10.  A workflow chart can be added to better summarize the methods used in the study.

11.  Accuracy of the handled GPS should be expressed.

12.  In the Study Area subsection, more details such as geology, soil type, and geography should be provided.

13.  The soil sampling and analysis subsection should be detailed.

14.  In the image resample phase, did the authors use a pan-sharpening or super-resolution algorithm? If not, just using resample with interpolation is not necessary.

15.  Exact dates for the Sentinel-2 images and cloudiness status should be provided.

16.  Theoretical background about the relationship between spectral indices and soil organic matter should be provided with references.

17.  In the conclusion section, the significance of the findings and contribution of the study should be explained. Also, recommendations for further research work can be given.

18.  The language of the paper can be improved.

Author Response

Dear reviewer,

         Thank you for taking your time provide us so many valuable advices. We have modified the manuscript followed by your advices. Our short replies for your comments can be seen in the  attachment. Also, all the specific changes please read the new version of the manuscript marked yellow.

Best regards.

Author Response File: Author Response.docx

Reviewer 2 Report

The manuscript presents a study of some techniques (Partial Least Squares [PLS], Geographically Weighted Regression [GWR], and Random Forest[RF]) for estimating topsoil organic matter (SOM) from satellite images. The paper is well-written and well-organized; in addition, the presented experimental results seem to be robust. However, the main weakness of the paper is its lack of novelty. There exist a lot of papers that use these techniques for the same purpose (some of them are referred to by the authors) or similar applications. These techniques are very established and there are a lot of tools (e.g., WEKA for RF) that allow us to apply them easily. In my opinion, justifying properly the novelty with respect to state-of-the-art techniques is essential to publish the paper.

I think that it would be interesting to include a background section in order to introduce PLS, GWR, and RF. In addition, I believe some details about the used software are required. The abstract is unclear because it includes conclusions that require previous explanations (e.g., the RPIQ acronym).

Author Response

Dear reviewer,

        Thank you for taking your time offer us so many valuable suggestions. We have revised the manuscript followed by your suggestions. Our short replies for your suggestions can be seen in the  attachment. Also, all the specific changes please read the new version of the manuscript marked yellow.

Best regards.

Author Response File: Author Response.docx

Reviewer 3 Report

The authors present the article entitled “Estimating topsoil organic matter in cultivated land through multitemporal satellite images: feature selection combined with random forest”. The contribution describes an integrative procedure to determine the content of Soil Organic Matter (SOM) using multitemporal Sentinel-2A images, spectral information and evaluated three models performance.

The manuscript is very clear, optimized and well planned. It considers field information that provides support to models predictions and geographic SOM potential distribution. The results are clear and convincing for their application.

Some minor observations:

1Figure 1a is not mentioned in the corresponding paragraph (Line 140).

2The word “method” is mentioned two times (line: 156)

3The graphs scale must be homogenized (Fig. 2, Fig. 3, Fig. 4 and Fig.S3), since visually the values seem similar: ej. yellow-brown-earths (maximum scale value=1800) vs paddy soils (maximum scale value=2400).

4Fig. S3, Spectral reflectance data for different soil types with different corresponding to 2018 is lack.

5The fig.5 have to be relocated (from page 11) to their corresponding paragraph (page 13).

6 Line 510: from 16.17 (space between words)

Author Response

Dear reviewer,

        Thank you for taking your time given us so many valuable recommendations. We have changed the manuscript followed by your recommendations. Our short responses for your recommendations can be viewed in the attachment. Also, all the specific changes please read the new version of the manuscript marked yellow.

Best regards.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The author made the corrections in line with all the referee's comments. It is appropriate for the study to be published in the journal.

Reviewer 2 Report

The main weakness of the paper is its lack of novelty. Many papers use these techniques for the same purpose (some of them are referred to by the authors) or similar applications. These techniques are very established and many tools (e.g., WEKA for RF) allow us to apply them quickly. As a consequence of my comments, the authors have reduced the list of contributions they said they had made. The paper should be rejected because it cannot be considered a research paper.