A Novel Method for Estimating Biomass and Carbon Sequestration in Tropical Rainforest Areas Based on Remote Sensing Imagery: A Case Study in the Kon Ha Nung Plateau, Vietnam
Round 1
Reviewer 1 Report (Previous Reviewer 1)
My main concerns did not be solved. In a tropical rainforest, Sentinel-2 MSI, as an optical sensor, has the saturation problem and weak penetrability of cloud. Why authors chose Sentinel-2 to estimate AGB? And generalized linear models, i.e., log-log, log-lin, lin-log, lin-lin, are not suitable to modeling the complex relationships between AGB and remote sensing indices. I think the study is invalid.
Author Response
Point 1: My main concerns did not be solved. In a tropical rainforest, Sentinel-2 MSI, as an optical sensor, has a saturation problem and weak penetrability of cloud. Why authors chose Sentinel-2 to estimate AGB? And generalized linear models, i.e., log-log, log-lin, lin-log, lin-lin, are not suitable for modelling the complex relationships between AGB and remote sensing indices. I think the study is invalid.
Response 1: In this study, Sentinel-2 satellite imagery was selected as the image that was processed at level 3. Photo data have been minimized by cloud influence due to the time of photo selection in February – the middle of the dry season of the Central Highlands region in Vietnam. Therefore, Sentinel-2 satellite imagery has limited the disadvantages of cloud penetration. Furthermore, Sentinel-2 satellite imagery has a resolution for panchromatic images of 10 m x 10 m, which matches the size of the square forest survey standard cell (the area is 400 m2, side length is 20 m).
The selection of 4 regression models is further explained in section 2.3.1. In this study, the regression model parameters were NDVI values based on remote sensing images, inherited data on forest reserves (constructed from Sentinel-2 satellite imagery), and biomass values from forest survey standard cell data. Furthermore, 20% of the forest survey standard cells were used to assess the model's accuracy based on the R2 coefficient.
Author Response File: 
Author Response.docx
Reviewer 2 Report (New Reviewer)
This study used four regression models to calculate the biomass of tropical rainforest vegetation based on Sentinel-2 satellite image data. It present a medium-resolution free remote sensing imagery-based application for estimating biomass and carbon sequestration in tropical rainforest areas. But, the contribution of the research, especially the novelty, is not obvious. Some major revision may be needed before further processing. The detailed comments are shown below.
DETAILED COMMENTS:
(1) The objectives is not clear, too much irrelevant contents were presented in the introduction. Comprehensive analysis and summary of previous studies is needed, instead of stating them simply.
(2) Too much preprocessing of the image are presented in the flow chart, more detailed descriptions of how to estimate the biomass from image should be added.
(3) A reference should be added for the data source of forest reserve and type.
(4) NDVI is a commonly used index for vegetation analysis, it is unnecessary to spend too much time on it.
(5) Four types of formulas were used while fitting regression models, what is the difference of these formulas? Are the basic assumptions of the linear regression model satisfied?
(6) In this study, the methods of biomass and carbon estimation are the same, it is meaningless, only biomass estimation is recommended.
(7) More detailed description of the field data should be added, such as the number of filed plot? What is the size of sample plot? How to calculate the biomass of sample plot?
(8) The discussion is a little bit messy, few subtitle should be added.
Author Response
Response to Reviewer 2 Comments
Point 1: The objectives is not clear, too much irrelevant contents were presented in the introduction. Comprehensive analysis and summary of previous studies is needed, instead of stating them simply.
Response 1: We have shortened the Introduction section to be more appropriate, focusing on an overview of the research that is relevant to the goal of the article. At the same time, the paragraphs in this section have been more compacted, clearly highlighting the need for this study.
Point 2: Too much preprocessing of the image are presented in the flow chart, more detailed descriptions of how to estimate the biomass from image should be added.
Response 2: We have revised Fig. 3 to make it easier to understand and more in line with the process of this study. Accordingly, it has better explained the process of steps from shaping satellite images, calibrating radiation, establishing NDVI images based on satellite images to building biomass estimation models.
Point 3: A reference should be added for the data source of forest reserve and type.
Response 3: The forest reserve map was founded on Sentinel-2 satellite imagery that we used to estimate biomass in this study. This result has been implemented by the Join Vietnam–Russia Tropical Science and Technology Research Center based on circular 34/2009/TT-BNNPTNT on forest criteria and classification regulations of the Government of Vietnam and accepted by the Scientific Council in May 2022.
Point 4: NDVI is a commonly used index for vegetation analysis, it is unnecessary to spend too much time on it.
Response 4: We have removed some unnecessary descriptions of NDVI in section 2.3.2.
Point 5: Four types of formulas were used while fitting regression models, what is the difference of these formulas? Are the basic assumptions of the linear regression model satisfied?
Response 5: Professor Rick Nordheim, an expert in statistical mathematics at the University of Wisconsin, has confirmed that in applied statistics, there is no concept of right or wrong model, but only whether the model matches or does not match reality, whether the model reflects the nature of the phenomenon studied or not. This is an empirical view and an important point that all app makers have in common. Therefore, we always have to consider the suitability of the model every time we solve a regression problem.
The 4 selection models are 4 common forms of linear regression models. Sometimes experimentally, if we take the original value, the difference between the largest value and the smallest value is extremely large. When the log is removed, the deviation in the data decreases, which is more in line with the model assumption (the model usually assumes the data have a standard distribution, and the deviation must also be within a certain limit with the mean). Through the experimental process, depending on the time of image acquisition and the characteristics of the study area, the best model is selected. In this study, the best model was selected by evaluating the R2 coefficient. Twenty percent of the total standard cells were retained to evaluate the accuracy of the models based on the R2 coefficient.
Point 6: In this study, the methods of biomass and carbon estimation are the same, it is meaningless, only biomass estimation is recommended.
Response 6: Since this research also serves forest managers and planners in the Kon Ha Nung Plateau area in the area of paying for carbon services, please allow us to retain the sections on carbon research. This is a very urgent issue for the research sector in particular and Vietnam in general. With the paper published in a reputable journal with the input of leading international experts in this field, we propose future carbon management and payment methods for local managers. Thank you.
Point 7: More detailed description of the field data should be added, such as the number of filed plot? What is the size of sample plot? How to calculate the biomass of sample plot?
Response 7: We have added detailed data on how standard cells are selected and established as well as how to calculate the actual biomass of standard cells in section 2.2.1.
Point 8: The discussion is a little bit messy, few subtitle should be added.
Response 8: We have added 2 subtitles to the Discussion section so that readers can easily follow the sequence of this section:
4.1. Estimation of tropical rainforest biomass based on remote sensing data;
4.2. Status quo and fluctuations of biomass in the Kon Ha Nung Plateau area for the period 2016-2021
Author Response File: Author Response.docx
Reviewer 3 Report (New Reviewer)
The abstract should be more emphasized and should be concise. Especially, when it comes to its begining.
Free satellite images do not seem a good keyword.
is AGB stands for terrestrial biomass estimates?
The Introduction should be consisting of 4 to 6 paragraphs, please reduced the paragraph in the introduction to concise sentences. I think this does not have a good impact on your manuscript from the reader.
In the abstract section, You mentioned "GIS" without writing what it stands for.
Please provide an overview of what describes each section of your paper in the last part of introduction section.
The legend of Figure 1 is not clear to readers.
The justification must be checked in Table. 1. The titles are not justified with descriptions.
Original or Raw satellite imagery in Figure. 2.
What do you mean by control, map documents in Figure. 2?
Are fail and true correct acronyms in Figure. 2?
What does the star sign in "* Weighting for forest types" mean?
Some grammatical mistakes in the passage like ".."
Please check the font throughout the paper (e.g., Table. 2).
Please use the same Pseudo-colored composite for both images in Figure. 3.
Please check Equation. 8. It seems it is incorrect.
Based on which criteria you classified the R or R2 value?
Please use the same symbol of R or R2 (Line 304).
Figures. 5-6 is not clear to read.
Table. 4 must be modified. At first glance, reader might be confused.
In Table. 6, * and ** are not described.
Please check the font of Table. 7.
What do other region(s) and other land(s) mean in Figure. 7?
Why did you use 2016 and 2021?
What does Mg mean?
The discussion section is long and should be concise.
The reference section must be well-acquainted with the template of journal.
Author Response
Response to Reviewer 3 Comments
Point 1: The abstract should be more emphasized and should be concise. Especially, when it comes to its beginning.
Response 1: We have shortened the summary and further emphasized the significance of this study by proposing a method for estimating biomass using basic linear regression models based on satellite imagery.
Point 2: Free satellite images do not seem a good keyword.
Response 2: We changed the keyword "Free satellite images" to "Sentinel-2" to emphasize the type of satellite images that were used in this study.
Point 3: Is AGB stands for terrestrial biomass estimates?
Response 3: Yes, AGB is just an estimate of the above-ground biomass of vegetation. Since the standard plot data are only a measurement of terrestrial biomass and not subterranean biomass measurements, we chose to estimate terrestrial biomass only in this study.
Point 4: The Introduction should be consisting of 4 to 6 paragraphs, please reduced the paragraph in the introduction to concise sentences. I think this does not have a good impact on your manuscript from the reader.
Response 4: We have shortened the Introduction section to be more appropriate, focusing on an overview of the research that is relevant to the goal of the article. The Introduction section has been shortened to 5 paragraphs. Many thanks for the reviewer's comments on this. Our earlier introduction made the specific goal of the article not stand out.
Point 5: In the abstract section, You mentioned "GIS" without writing what it stands for.
Response 5: We added the phrase "Geographic Information System" to explain the word GIS in the Abstract section.
Point 6: Please provide an overview of what describes each section of your paper in the last part of introduction section.
Response 6: We have added descriptions summarizing the implementation sections of the article in the last paragraph of the Introduction section. Thank you very much for your comments.
Point 7: The legend of Figure 1 is not clear to readers.
Response 7: We have edited and enhanced the resolution for Fig. 1.
Point 8: The justification must be checked in Table. 1. The titles are not justified with descriptions.
Response 8: I have replaced Tab. 2 header with "Technical parameters and acquisition time of Sentinel-2 remote sensing images" to better fit the description of the table.
Point 9: Original or Raw satellite imagery in Figure. 2.
Response 9: The Sentinel-2 image in this study is “Raw satellite imagery”. Satellite imagery was processed at level 3 prior to the start of the process. We have re-edited Fig. 2.
Point 10: What do you mean by control, map documents in Figure. 2?
Response 10: This means using photo control points and original topographic map materials to shape the image of the correct coordinates of the study area. We edited it to "Image control point, map documents" in Fig. 2.
Point 11: Are fail and true correct acronyms in Figure. 2?
Response 11: This is the step of calculating biomass using 4 basic regression models and selecting the appropriate model based on the R2 coefficient to establish a biomass estimate map. We revised it in Fig. 3 to make it easier to follow.
Point 12: What does the star sign in "* Weighting for forest types" mean?
Response 12: The * sign is just a symbol for marking 1 small subsection in section 2.3.2. We have removed it to match the format of the magazine and replaced it with italics for each of these subsections. We apologize for the inconvenience.
Point 13: Some grammatical mistakes in the passage like ".."
Response 13: We apologize for these typos. We rechecked the entire article and removed the undue errors. Thank you very much.
Point 14: Please check the font throughout the paper (e.g., Table. 2).
Response 14: We have re-edited the font of the tables to the required format. Thank you very much.
Point 15: Please use the same Pseudo-colored composite for both images in Figure. 3.
Response 15: We used the same pseudo-plant color combination (Nir-Red-Green) to combine photos from both 2016 and 2021. Due to the contrast mode of colors when using ENVI software, when exporting the image file, the colors of the 2 images are slightly different. We have added a caption (Nir-Red-Green) to the photo so that readers can more easily follow it.
Point 16: Please check Equation. 8. It seems it is incorrect.
Response 16: We double-checked the formulas in the article.
Point 17: Based on which criteria you classified the R or R2 value?
Response 17: We have added more references for the correct division of the levels of the R2 value. Accordingly, 0.3 ≤ R2 < 0.5 indicates low correlation, 0.5 ≤ R2 < 0.7 indicates moderate correlation, and 0.7 ≤ R2 indicates high correlation.
Point 18: Please use the same symbol of R or R2 (Line 304).
Response 18: We have agreed on the usage as "R2" throughout the article.
Point 19: Figures. 5-6 is not clear to read.
Response 19: We have re-edited Figures 5 and 6 in the article to make it easier to follow.
Point 20: Table. 4 must be modified. At first glance, reader might be confused
Response 20: We have re-edited Tab. 4 and Tab. 6. Since the figures are quite long, please allow us to keep it in 2 main columns to show the points.
Point 21: In Table. 6, * and ** are not described.
Response 21: We replaced the * and ** symbols in table 6 with the paraphrase just before table 6 for RMSE and MAE symbols.
Point 22: Please check the font of Table. 7.
Response 22: We have moved the font for table 7 to the same format as the entire article.
Point 21: What do other region(s) and other land(s) mean in Figure. 7?
Response 23: They are all common 1 meaning that other types of mantle are not forests and are not biomass. We have revised the commentary of Fig. 7 to make it easier to follow.
Point 21: Why did you use 2016 and 2021?
Response 23: According to the forest survey cycle in Vietnam, there will be a basic census of forest reserves every 5 years. These are the years we inherited data on forest inspection standards in the study area. In addition, since 2016, policies on forest conservation and management in the Kon Ha Nuong Plateau area have been promoted, especially policies on forest protection and forestry development. We want to compare forest biomass results during this period so that we can assess the effectiveness of forest protection policies here.
Point 24: What does Mg mean?
Response 24: Mg/ha is equivalent to tons/ha. We have further explained and agreed on all the symbols in the article.
Point 25: The discussion section is long and should be concise.
Response 25: We have added 2 subtitles to the Discussion section so that readers can easily follow the sequence of this section:
4.1. Estimation of tropical rainforest biomass based on remote sensing data;
4.2. Status quo and fluctuations of biomass in the Kon Ha Nung Plateau area for the period 2016-2021
Point 26: The reference section must be well-acquainted with the template of journal.
Response 26: We have rechecked and edited the reference system according to the journal's template.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report (Previous Reviewer 1)
Accepted
Author Response
Sincerely thank
Reviewer 2 Report (New Reviewer)
The authors have incorporated most of the modifications proposed, to improve their manuscript, which is highly appreciated. But, the contribution of the research, especially the objectives, is not obvious. Some major revision may be needed before further processing. The detailed comments are shown below.
DETAILED COMMENTS:
(1) The objectives is still not clear, please summarize them to (1)(2)(3).
(2) Generally, the NDVI is not sensitive to the AGB, especially the areas with high AGB, which called "saturation effect". So I am a little doubt about the feasibility of this study. Some statements should be added to support the methodology in this paper.
(3) Instead of using only NDVI, sets of the metrics, such as vegetation indices, texture, should be extracted and used for biomass estimation. It is not convincing to use NDVI individually.
(4) The saturation effect of NDVI to the AGB estimation should be comprehensively assessed. The uncertainty analysis of the maps of AGB or carbon stock must be conducted.
(5) There are some problems with tenses in the paper. The present and past tenses are mixed.
Author Response
Point 1: The objectives is still not clear, please summarize them to (1)(2)(3).
Response 1: We have added 2 specific objectives of the article in the last paragraph of the Introduction section including (1) selecting an appropriate biomass estimation model for the rainforest eco-system in the Kon Ha Nung Plateau area and (2) developing biomass, carbon stock, and CO2 equivalent values for the period 2016 – 2021 for forest vegetation in the Kon Ha Nung Plateau area, Vietnam.
Point 2: Generally, the NDVI is not sensitive to the AGB, especially the areas with high AGB, which called "saturation effect". So I am a little doubt about the feasibility of this study. Some statements should be added to support the methodology in this paper.
Instead of using only NDVI, sets of the metrics, such as vegetation indices, texture, should be extracted and used for biomass estimation. It is not convincing to use NDVI individually.
The saturation effect of NDVI to the AGB estimation should be comprehensively assessed. The uncertainty analysis of the maps of AGB or carbon stock must be conducted.
Response 2: After learning more about plant indicators, we decided to use EVI instead of NDVI to estimate forest vegetation biomass.
Enhanced Vegetation Index (EVI) is the best alternative. EVI is designed for high or dense vegetation cover. EVI has benefits over NDVI through de-coupling of the canopy background signal (using soil line) and a reduction in atmosphere influences (using blue band). EVI is calculated similarly to NDVI but uses additional wavelengths of light to correct for the inaccuracies of NDVI. Variations in solar incidence angle, atmospheric conditions like distortions in the reflected light caused by the particles in the air, and signals from the ground cover below the vegetation are corrected for using EVI.
In addition, in this study, weights of forest vegetation types were also added for study in the biomass estimation model (according to table 2). According to the results of the new model using the EVI index, the R2 value has increased for both times selected to calculate AGB (0.76 and 0.765).
Point 3: There are some problems with tenses in the paper. The present and past tenses are mixed
Response 3: We rechecked the grammar for the entire article.
Author Response File: Author Response.docx
Round 3
Reviewer 2 Report (New Reviewer)
Thanks a lot for your efforts to improve the quality of the paper.
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
Suggestion for sustainability-1845628
Authors adopted 4 regression models and spatial analysis of GIS to estimate forest aboveground biomass (AGB) as well as its temporal dynamics based on Sentinel-2 images. Overall, the manuscript is lack of innovation with poor writing. I suggest this manuscript should be directly rejected or under a major revision. In a tropical rainforest, Sentinel-2 MSI, as an optical sensor, has the saturation problem and weak penetrability of cloud. Why authors chose Sentinel-2 to estimate AGB? I think it is invalid. The specific suggestions are follows:
1. Abstract: the last sentence should be the explicit contribution of this study, which is also the solution degree of the research gap that you reported in the second sentence. There are too many long sentences, which is quite difficult for reading. The content should be more concise and informative. The innovation and explicit contribution of this study, not the detailed workflow and all results, should be emphasized in the Abstract. The remote sensing data used in this study should be reported. Authors said the research gap is the lack of the calculation of biomass and carbon absorption capacity of forest ecosystems, especially tropical rainforests, which is quite substantial in previous studies. I think authors should find out the special innovation and contribution of this study, and then rewrite the content of Abstract.
2. Keywords: the keywords should also emphasize the innovation and explicit contribution. Authors just selected some phrases from the title. Besides, the title should also rewrite, which should be the explicit innovation and explicit contribution of this study. The existing title is vague and lack of information.
3. Main body: language needs polish by native speakers.
1) Introduction: the reasons authors used regression models and Sentinel-2 data in this study should be introduced.
2) Materials and Methods: how authors dealt with this problem of the temporal inconsistency between field data and Sentinel-2 image? Field campaign should be reported in details. Why authors used single-date images? The topography and climate should be introduced.
3) Discussion: the uncertainty should be discussed. The comparison to relevant studies should be discussed.
4) Conclusions: this part should be concise. Do not repeat the content that reported in above parts. Pleas highlight the contribution of this study, and cautions about the application this methodology in other areas.
4. References: references should be revised according to the requirement of the journal.
Comments for author File: 
Comments.pdf
Reviewer 2 Report
This paper presents a methodological framework for predicting biomass and C stock of tropical rainforest vegetation in the Kon Ha Nung Plateau, Vietnam. This methodology combines field surveys data and remote sensing approaches for C stock estimation.
I found the paper to be easy to follow. The introduction did a nice job reviewing the topic of why it’s necessary to use remote sensing techniques to estimate AGB. More previous studies about how AGB are estimated in tropical forests might be added.
The methods behind the approach and theoretical model need more detail descriptions, like missing mention the field survey time, the scale of stand cell investigation, explanation why choosing the satellite imagery data sources at February.
The empirical model and simulation exercise are straightforward, the results are logical and well explained in the text, although many of the figures could be better described and organized to improve readability and interpretation.
I have some general comments that could improve the paper:
· While the results from the empirical example are logical and well discussed, I am still left wondering the ‘so what’ in terms of how these findings can be broadly applied by policymakers, especially if there are multiple sources of uncertainty. In response, the paper could be expanded in a few ways:
o Apply the empirical model for seasonal NDVI time-series besides a single time NDVI to improve accurate and less saturation problem.
o Is there’s a way to estimate below ground biomass?
o Summarize the literatuies of data used to estimate AGB in empirical methods (etc, LiDAR data, medium/high-resolution optical images). Add also citations about the theoretic studies using NDVI to model AGB.
o Except generalized linear models, does the nonparametric methods has better performance?
Specific comments
· Line 20, should be “log-log” instead of “log,log”.
· Page 8, Line 270, using consistent font size and subscript. Figure 5 and Figure 6 are difficult to interpret due to small font size.
· Page 6, Tables 2. Forest types were the same for STT=6 and STT=7
· Page 7, Line 243, need to rewrite the sentence “Excel is used to analyse the relationship between …”
