Mapping Irrigated Rice in Brazil Using Sentinel-2 Spectral–Temporal Metrics and Random Forest Algorithm
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThe manuscript "Mapping Irrigated Rice in Brazil using Sentinel-2 spectral-temporal metrics and Random Forest Algorithm" (remotesensing-3015758). The use of moderate spatial resolution satellites such as Sentinel-2 combined with machine learning such as Random Forest are essential tools in agricultural management today. Irrigated rice cultivation has great agricultural and economic importance for Brazil, given its impact on Brazilian food. Therefore, this study provides important and substantial tools for managing culture in different regions of the country. However, before recommending the manuscript for publication, the authors must improve several aspects of the present study. Therefore, I am recommending this work for major revisions
As small observations, which must be attended to, I highlight:
1 – I noticed some grammatical errors in writing, therefore, I suggest the revision of English by a native speaker.
2 – In the abstract the phrase “Rice is one of the most widely consumed foods in the world, with Brazil being one of the top ten producers. In the country, the majority of rice production is irrigated. Knowledge of where and when rice is planted is essential for the formulation of water use policies, as well as for the planning of stocks, monitoring of harvests and prices. However, the continental size of Brazil, variations in the planting calendar and the diversity of production systems present challenges to the implementation of systematic and automated mapping in the country.”, needs to be summarized in just two lines. Finally, authors must further explore the Results and discussions in the abstract, limiting it to 200 words, in accordance with Remote Sensing standards.
3 – Lines 87-89: The authors must present the hypotheses of this study before the objective.
4 – In Figure 1, authors must present a legend to better identify what the green and red dots mean.
5 – Why did the authors use Mcfeeters' (1996) NDWI and not add Gao's (1996) (https://doi.org/10.1016/S0034-4257(96)00067-3); I believe it could represent the water from the rice leaf well and served as the basis for the Random Forest.
6 – I did not identify the accuracy of Random Forest in the work. Why did the authors not present the Kappa index for the Random Forest used in this study?
7 – In the Appendix All images must show their latitude and longitude.
As a minor and main note, I highlight:
1 – Use the Mendeley Reference Manager for references as well as citations, as both Remote Sensing standards are not standardized in the body of every manuscript.
Comments on the Quality of English LanguageExtensive editing of English language required.
Author Response
Dear Reviewer 1,
Thank you for your valuable suggestions and questions regarding our article. We believe your feedback has been essential in refining the work to the quality expected of a Remote Sensing article.
Responses to your comments are organized in the PDF document submitted along with this form. We are pleased with your suggestions and hope that our revised manuscript meets your expectations.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsReviewer’s Comments
This study aims to investigate the effectiveness of utilizing Sentinel-2 spectral-temporal metrics combined with a Random Forest classifier to develop a potentially scalable method for mapping irrigated rice areas across Brazil. The result shows that the regional classifications often overestimated rice areas due to the influence of cultivation system diversity and confusion with other crops, while the global model effectively mitigated these issues, resulting in reduced overestimations.
However, what is the innovation of this study? The study used spectral-temporal metrics (STM) and RF algorithm to extract rice extent. But I didn’t understand why the authors can use STM data to extract rice extent. Is this the innovation of this study? In addition, I think the training samples is not enough, especially for the regional model. The different sampling date for the training samples is also an major issue for this study. (The National Water and Sanitation Agency (ANA) and the National Supply Company (CONAB) have conducted visual interpretation of Sentinel-2 imagery and field inspections to map irrigated rice in Brazil [1]. However, these mappings correspond to different crop years across the studied states (2017/2018 for TO; 2018/2019 for SC; 2019/2020 for RS)”)
Comments
Point 1: The Abstract is too simple. The authors can describe more about the method, the result.
Point 2: The main contribution or the innovation of this study should be point out in the Introduction.
Point 3: In the Study Region part, the phenological characteristics of rice in each study area should be pointed out.
Point 4: Figure 1, I think it is nonsense to draw the location of samples. It looks like many samples in the study area, but I think it is still not enough for training a good RF algorithm.
Point 5: The description for Figure 2 is not clear. I can get little information only from the Figure 2.
Point 6: Discussion and conclusions should be improved. What is the main contribution of the paper? The limitation of this study?
Author Response
Dear Reviewer 2,
Thank you for your valuable suggestions and questions regarding our article. We believe your feedback has been essential in refining the work to the quality expected of a Remote Sensing article.
Responses to your comments are organized in the PDF document submitted along with this form. We are pleased with your suggestions and hope that our revised manuscript meets your expectations.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for Authors1. Line 145: Table 1. Quantitative of samples for Training and Validation by region per class. In Table 1, There weren’t training samples in some classes, including Forestry, Shrub Vegetation, and Bare Soil, respectively. Why? Please explain this situation. What were the effects of no training samples on this classification?
2. Line 206: In Tocantins, between December and January, the NDVI curve values declined; while between January and February, NDVI values increased, reaching its peak in February. It is necessary to explain this difference.
3. Line 213: In Santa Catarina (Fig. 2.d-f), the NDVI values declined between November and December, then increased up to peak in January. This change was different from that in Tocantins. It is interesting. Please explain more detail.
4. Line 304: “In Sant'Ana do Livramento and São Gabriel, the percentage difference was close to 0. The omission of rice regions was due to confusion with other use classes, especially other agriculture”. Please explain how the other agriculture use class affects the confusion?
5. Line 318: “The confusion matrices indicate that the rice class was most frequently confused with the "agriculture" class.” This is an interesting finding. Explain more detail.
6. Line 460: The use of STM to classify irrigated rice in Brazil is recommended and has the potential to provide classification at a national level. However, it is better to provide some limitations for this study, and to point out some research topics in the future.
Author Response
Dear Reviewer 3,
Thank you for your valuable suggestions and questions regarding our article. We believe your feedback has been essential in refining the work to the quality expected of a Remote Sensing article.
Responses to your comments are organized in the PDF document submitted along with this form. We are pleased with your suggestions and hope that our revised manuscript meets your expectations.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsBased on the corrections provided by the authors, I am considering the present study for publication.
Comments on the Quality of English LanguageMinor editing of English language required.
