Examining the Potential of Sentinel Imagery and Ensemble Algorithms for Estimating Aboveground Biomass in a Tropical Dry Forest

Round 1

Reviewer 1 Report

Line 47, 55,58 - Please provide some ref. 

recommendations, please provide references for each method used in past for instance, line 47, no ref. provided. 

line 55 among other factors - no ref. provided. what are among others?

Line 58 optical RS to map AGB - no ref. provided. Please adhere these standards and uniformity in case of citations. Also, try to avoid using generalized terms such as accurate than other - better approach is to mention "others" with their names to sound more informative.

The rest of the introduction part, please try to figure out similar issues, I cannot comment on similar problems over and over again. 

Line 78 provides more references when plural terms are used. e.g., attempts, limitations, sensors.

Line 80. I am disagreeing with the statement that not sufficiently employed. Here are some related refs.

https://doi.org/10.1016/j.rse.2022.113232

https://doi.org/10.1016/j.rsase.2022.100897

My question is how different authors are compared with above-listed approaches. 

Line 88- please check the location of ref.

Line 102 - 106, at least one ref. for each method mentioned in the context of AGB should be provided.

tropical dry forests (TDFs), is that acronym is used somewhere else? if so, provide a ref. what I see in the literature is dryland forests. https://doi.org/10.1016/j.rse.2022.113232

Lines 108 - 125 should be a new but single paragraph to make the content arrangement more articulated and well-written.

Line 122, I am again disagreeing with authors claiming that its novel here is ref. where a study previously did the same thing for dryland forests in Africa using S-1, S-2 data for AGB. 

https://doi.org/10.1016/j.rse.2022.113232

 The authors lack a comprehensive literature review and many related studies have not been considered and reviewed before this study.  Therefore, claimed novelty is questionable or the authors were unable to address it. 

While analyzing and comparing existing novel research articles published in remote sensing of environment RSE. titled. 

"Improving above ground biomass estimates of Southern Africa dryland forests by combining Sentinel-1 SAR and Sentinel-2 multispectral imagery".

The above article is a novel contribution, and the present research is a replica of methods and techniques used in that article, therefore, I suggest authors come up with better approaches or novel methods.

 

 

 

Author Response

Dear Reviewer,

Please, see the attachment!

Thanks

Author Response File: Author Response.docx

Reviewer 2 Report

This research paper investigates the efficiency of Sentinel-1 SAR texture-type data and Sentinel-2 multispectral-type data for mapping above-ground biomass (AGB) in Tropical Dry Forests (TDF). It also aims to compare the performance of two machine learning regression techniques, random forest (RF) and extreme gradient boosting (XGBoosting), in estimating AGB using various predictor sets. Additionally, the study aims to create a forest AGB map for the study area using the most appropriate model. By conducting this analysis and employing these methods, the paper contributes to understanding AGB mapping in TDFs and provides valuable insights for effective forest management and conservation.

I have the following comments and suggestions for including information and improvements:

 

Abstract

I suggest the authors add more information about the methods and main results values. The current abstract version must contain information on this aspect, making it a less fragile abstract. Please remove the space between lines 22-23.

 

Introduction

The authors should illustrate more about state of the art in analysis and methods to estimate forest AGB maps in tropical and other regions worldwide.

A more explicit justification must be included regarding the Importance of testing the model in different counties or locations with similar characteristics. This information is essential to give a more general aspect to the study, helping it not appear that the results are only of local Importance.

 

Methods

More details must be included in the sections:

1. Please describe each of the six steps shown in Figure 1.

2. Also include the criteria for choosing the variables associated with choosing images, periods, polarizations, and GLCM metrics.

3. Please replace figures 1 and 2 with better ones. Specifically, in Figure 2, please include cartographic elements such as north, spatial scale, and coordinate grid.

 

Results

1. All results derived from the Variable selection, Performance analysis of RF and XGBoost, Evaluation of AGB estimation models, Importance of predictor variables, and Forest AGB map must be detailed (More details are necessary in all sections)

2. Please replace figures 3 and 4 with better ones.

 

Discussion

Citations that support the results found in this section must be included. It is also imperative that the authors have a topic that compares the results found from the methods and analyses proposed in the article with the literature results.

I suggest that Figure 5 be deleted from the discussion section and must be included in the results section.

Please include a paragraph with possible limitations of the method exposed in your study and the extrapolation of these methods to other drylands worldwide.

 

Conclusions

In this section, information should be added about future advances in this field of research based on the results found in this article.

Comments for author File: Comments.pdf

Author Response

Dear Reviewer,

Please, see the attachment

Thanks

Author Response File: Author Response.docx

Reviewer 3 Report

This study explored the potential of Sentinel 1 SAR texture and Sentinel 2 MSI sensors and the performance of two ensemble modeling algorithms - RF and XGBoost - using different predictor sets for Above Ground Biomass estimation, conducted in a tropical dry forest in Colombia. Although similar work has been conducted, as the authors point out, this research is of interest because nothing similar has been done in the study area.

The article is well presented, and I have only a few minor comments:

Keywords: There are a couple of words that are repeated in the title; I suggest using different ones.

Regarding the format of the references (lines 68, 306, 340, 526, 530, 559, and others), please check that it matches the requirements of the journal.

Line 135: I suggest changing square kilometers to hectares.

Figure 4: Improve the resolution of the figure.

Figure 5C: "Number of pixels" is underlined in red, please remove the underline.

Conclusions: They need improvement; they are actually a summary of the results, not conclusions.

 

I hope these comments are helpful in improving the work.

Author Response

Dear Reviewer,

Please see the attachment

Thanks

Author Response File: Author Response.docx

Reviewer 4 Report

1. Line 125, “the lack of use XGBoosting to estimate AGB” can not be called as a novel.

2. Section 2.3.1, the texture information was extracted from the S1 images with 20 m spatial resolution, how to set the domain size when calculate the GLCM?

3. Line 249, lack a space between the “extraction” and “from”.

4. Section 2.6, the parameters of RF and XGBoost should be clarified.

5. Line 443, the RF is generally outperformed than XGBoost, but a significantly underestimation was found in the RF algorithm, it should be explained and discussed.

6. Line 468, does the importance calculation of RF and XGBoost are same? It should be clarified.

7. Discussion, the texture information describes the spatial correlation of pixel value, how did the AGB determined by the texture with the coarse resolution (20m), which cannot even truly characterize the texture information of an individual tree. It should be explained and discussed.

Author Response

Dear Reviewer,

Please see the attachment,

Thanks

 

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Authors point by point responses are very satisfying. They did not include the changes in the author's response which are difficult to trace down in the revised version.  I mentioned that this study is not a novel contribution and similar studies already exists which turned out that authors are agreed in the response. Therefore, the present study is a replication of an existing research and should not be published unless authors provide a very detail response in the rebuttal that satisfies that authors indeed provided a unique study.