R-Unet: A Deep Learning Model for Rice Extraction in Rio Grande do Sul, Brazil

Round 1

Reviewer 1 Report

1.     Line 63 ：The author say “The traditional DL model has a simple structure……. “

Line 65: The author say “Recently, DL, as a popular research method……”

So, What is the difference between the “DL” mentioned in sentence 63 and sentence 65, respectively

 

2.     I suggest author redraw the Table 2 and Table 5, it now looks messy and does not clearly express the results of the experiment.

 

3.     Could you please elaborate further the attention mechanism module. As we know, there are lots of attention mechanism structure, such as ECA,CMBA et al.

 

4.     In this paper , author say “the Precision, Compared with U-Net, IOU and MCC of R-Unet were increased by 2.1%, 2.8% and 2.5%, respectively.”  and Moreover，table S1 shows that Dataset03 and Dataset07 have similar accuracy.

So, though in-depth and good work has been done academically, but the solution has been presented in a very complex manner. Do you think, such complex solution will be acceptable to the practical implementers?  Is there any practical point in improving such a little bit of precision?

Minor editing of English language required

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper "R-Unet: A deep learning model for rice extraction in Rio Grande do Sul, Brazil" presents an innovative approach to rice extraction using a deep learning model, R-Unet, based on Sentinel-2 and time-series Sentinel-1 data. The paper is well-structured and the methodology is sound. Here are some comments and suggestions for improvement:

Clarity and Organization: The paper is well-structured and the objectives are clearly stated. However, the authors could improve the flow of the paper by providing a brief overview of the methods and results in the introduction. This would help to orient the reader and provide context for the detailed discussions that follow.

Methodology: The methodology is sound and well-explained. However, the authors could provide more detail on the R-Unet model, including the attention-residual module and the multi-scale feature fusion (MFF) module. This would help readers to understand the technical aspects of the study and to replicate the methods if desired.

Results and Discussion: The results are interesting and the discussion is insightful. However, the authors could do more to interpret their results in the context of previous research. For example, how do their findings compare with those of studies using other deep learning models or traditional machine learning methods? Also, the authors could discuss the implications of their findings for agricultural management and food security in Brazil.

Limitations: The authors have acknowledged some of the limitations of their study, such as the difficulty of obtaining high-quality data and the need for multi-source data to extract the optical and SAR features of rice. However, they could provide more detail on how these limitations might affect their results and how they could be addressed in future research.

Conclusion: The conclusion could be strengthened by summarizing the key findings, discussing their implications, and suggesting directions for future research. The authors could also discuss the potential applications of their research in more detail.

References: The authors have cited a good range of sources, but they could update their literature review to include more recent studies. This would help to situate their research in the current state of knowledge in the field.

Overall, this is a promising study that could make a significant contribution to the field of agricultural remote sensing. With some revisions to improve clarity and detail, this paper has the potential to be a valuable resource for researchers and practitioners in the field.

Minor English language changes are required.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

This is a competent manuscript.  It is well written and presented.  There is a clear statement of intent and a logical flow to the text.  The tables and figures are informative and professional.  There is a very extensive set of relevant references but with some format inconsistencies.

There is a very long history of the use of remote sensing for mapping agriculture.  This study focuses on rice in Brazil and adds to that literature. The authors follow standard research methods comparing data and procedures using varied measures to assess accuracies.  The manuscript is however complex and some reduction or simplification might be considered.  There is some repetition between sections.

As in almost all manuscripts, there are editorial suggestions as follow for consideration by the authors:

1.       Typically define and then utilized acronyms independently in both the abstract and main text.  For example SAR in line 17 and IOU etc.

2.       Line 23, governments?

3.       There is inconsistent use of serial commas.

4.       Lines 26-37, timely and accurate repeated.

5.       Line 45, number is a poor word choice.  Amount perhaps?

6.       There is excessive use of ‘etc’.

7.       Line 84, awkwardly worded.  What is high-score?

8.       Line 108 areas?  Why are the testing and training areas not overlapping.  This is a very unusual approach.   Perhaps this is just poorly worded from later text?

9.       Line 147, perhaps calibration for training.  Common with validation.  The use of training, validation and test separately is somewhat confusing. This should be clarified.

10.   Table 1.  Dataset does not need to be in each line.

11.   Line 173.  Why 4 and two?

12.   Line 201. And (d)?

13.   Line 227 the formulas might be in the same order as the sentence.  Similarly, order in lines 241 and following.

14.   Table 2.  Typically insert table after text reference. Also Figure 7.

15.   Lines 282 and 283, Table S1?

16.   Table 5. OA?

17.   Line 429, verification?  Validation?

18.   There are inconsistencies in the formatting of the references.  Article titles use upper case and not.  Journal titles are abbreviated and not.

As stated earlier, this is a competent manuscript and certainly suitable with minor editorial changes for this journal.

Well written, minor issues.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

I have no questions!

Minor editing of English language required