Mapping Agricultural Intensification in the Brazilian Savanna: A Machine Learning Approach Using Harmonized Data from Landsat Sentinel-2

Round 1

Reviewer 1 Report

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

This study is based on HLS data and utilizes machine learning and hierarchical classification methods to classify crops in the Brazilian Savanna region in detail. The article has a clear logical structure, feasible methods, reliable data, and achieves good classification accuracy. I agree to publish this article after minor revisions.

Below are a few minor issues that require the author's explanation and clarification.

1.Please provide a detailed explanation of the quality control process for HLS products. As the study area is located in a tropical region, has the optical remote sensing data been affected by cloud cover and precipitation in the long term? The data description of HLS contains a lot of information, which may give the impression that the study directly used HLS.s\HLS.l data. However, it is actually utilizing three vegetation indices derived from HLS data. I suggest that the vegetation indices part should be included in the datasets section.

2.The color differentiation in the legend of Figure 5 is not very distinct. It is recommended to use colors with higher color contrast or different shapes to represent the data points.

3.Can the study period be extended? Currently, the research only covers a complete growth cycle. Is it possible to continuously use this classification work?

The quality of the English writing is acceptable, but there are minor revisions needed in the results section for better expression.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

This study used Landsat and Sentinel-2 data to map agricultural intensification.  The researchers compared tRF, ANN, XGBoost in LCLU classification. The structure and organization of the paper is coherent. The methods are adequately described. The results and discussion are welly presented. Overall, I would accept the manuscript with minor revisions. 

Some details of comments are here: 

Abstract: remove subtitle of (1) to (4). 

Fig.1 legend ≥ 302 should be ≤ 302; 

Fig. 7 RF, XGBoost, and ANN should in the same order across all sub figs. 

Table 6: double check ‘Omission error’ or ‘Commission error’

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

Édson Luis Bolfe et al. provide a method for Mapping Agricultural Intensification in the Brazilian Savanna.

Although the topic is novel and useful, due to the quality of presentation, I have to reject this paper. But a further invitation for resubmission is welcome. The authors should carefully revise their manuscript especially writing the paper in scientific writing form, such as refer to the papers published in top journals. The abstract section is totally not acceptable in the journal articles! The authors must improve this.

Since the paper organization is weak, I only can provide a few points for the authors to improve their manuscript.

The introduction section is too long, please write it in brief way to show the scientific questions and the limitation in the area of mapping.

The method section is fine but it should have more detail at the method description such as how to calculate the SAVI and other abnormal Vis.

Figure 5 should be more clear, the color is hard to read.

Figure 7 it seems that different ML methods have different accuracy at different stages of mapping. Please clarify the difference.

Figure 8 do the authors just map the savanna from NDVI as data source?

Figure 9 after getting different components of vegetation covers, it needs to be validate with the reliable statistics such as ground surveying. This point must be improved before resubmission! This is the most important point that whether this study is reliable or not.

Although there are not much grammar mistakes in the manuscript, the quality of presentation is not in a scientific writing form. I do not think it could be improved in a brief time. The author must know what scientific writing is before resubmitting this manucript .

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Author Response

Dear Reviewer,

We appreciate all observations and suggestions, especially in round 1. All the indicated points were very important for the improvement of the Article: “Mapping Agricultural Intensification in the Brazilian Savanna: A Machine Learning Approach using Harmonized Data from Landsat Sentinel-2”. Adjustments and changes are marked in blue and red in the text.

Authors,

Reviewer 4 Report

The current manuscript is much better than the previous version.

I just have a few suggestions to improve the paper.

Figure 6, you may put the color in the right axis in red.

Figure 8 the longitude and latitude axis could be put in each subfigure and each level.

Besides, if the code is available to the public will make this paper have high impact to the research community.

The Quality of English Language is much better than the previous version.

Author Response

Dear Reviewer,

We appreciate all observations and suggestions.

Figure 6, you may put the color in the right axis in red.

Figure 6 adjusted as requested.

Figure 8 the longitude and latitude axis could be put in each subfigure and each level.

Figure 8 adjusted as requested.

Besides, if the code is available to the public will make this paper have high impact to the research community.

Information adjusted as requested.

Data Availability Statement: The data, maps and codes generated are available online, which can be accessed: https://doi.org/10.48432/1YYF9Y

Authors,