Land Use Changes in the Teles Pires River Basin’s Amazon and Cerrado Biomes, Brazil, 1986–2020

Round 1

Reviewer 1 Report

Reviewer Report for the Manuscript: Sustainability-1955020

“Land use changes in the Teles Pires River basin’s Amazon and Cerrado biomes, Brazil, 1986-2020”

Journal Name: Journal of Sustainability

REVIEWER REPORT

Summary

The article addresses a very subject of the Amazon rainforest to detect changes occurring in the Teles Pires River basin in Brazil’s center-west over the last 34 years due to agricultural economic development. It uses a combination of satellite spatial data and field validation methods and reveals the transformation path of land cover changes from forest to pasture and from pasture to agriculture. The topic is good and the subject is vital to be addressed for rising awareness and for helping policymakers and environmental activists. The paper is well-structured and reached its research goals. However, the paper needs some improvements in various parts of the article to be publishable.

Abstract and Keywords

1-     The method needs to be more specified in this section

2-     The method/tools names also need to be brought in keywords

Introduction

3-     It seems that this section is supposed to more focus on the cascading effects of changing rainforests to agriculture developments on the Earth system as well as the biomass status in the Amazon. This is important because the main concern of the research here is the damages posed by the transformation have accorded occurred in the study area

4-     This section lacks a comprehensive literature review particularly studies that have been done in the study area and criticize state of the art in the research area in both methodological and results

5-     A compelling argument is needed on the root causes of the investigated transformation, from rainforest to agriculture, as an Amazon countries phenomenon to make this section more contributed to the concurrent literature.

 Theoretical and Literature review

6-   The section should be added to the paper’s structure and give new insights into the  relationships between the transportation of rainforests land and agricultural economic development.

7-   This section also may explain the classification methods and state of the art in this field

Analysis

8- Figure 5. b needs to be more explained as we can see that the percent changes in both agriculture and native vegetation have reached zero in 2021

Presentation

9-     It would be helpful to better understand the transformations that happened in the rainforests if the authors provide a more detailed map to see what exactly happened during the three decades using Figure 4 maps.

Policy implications

10- This section seems to be necessary due to the importance of the results of the present paper for policymakers and global environmental activities.

Author Response

Reviewer 1

Summary

The article addresses a very subject of the Amazon rainforest to detect changes occurring in the Teles Pires River basin in Brazil’s center-west over the last 34 years due to agricultural economic development. It uses a combination of satellite spatial data and field validation methods and reveals the transformation path of land cover changes from forest to pasture and from pasture to agriculture. The topic is good and the subject is vital to be addressed for rising awareness and for helping policymakers and environmental activists. The paper is well-structured and reached its research goals. However, the paper needs some improvements in various parts of the article to be publishable.

Abstract and Keywords

1-     The method needs to be more specified in this section

Answers: We added a sentence to the abstract summarizing the methodology.

2-     The method/tools names also need to be brought in keywords

A: We have added the keyword “supervised classification.”

Introduction

3-     It seems that this section is supposed to more focus on the cascading effects of changing rainforests to agriculture developments on the Earth system as well as the biomass status in the Amazon. This is important because the main concern of the research here is the damages posed by the transformation have accorded occurred in the study area

A: We have added writing (end of first paragraph of Introduction) on the effects of the land-use change from native habitat to commercial agriculture as it relates to water availability which is brought up in the first part of the of the first paragraph of the Introduction section.

4-     This section lacks a comprehensive literature review particularly studies that have been done in the study area and criticize state of the art in the research area in both methodological and results

A: We have added a comprehensive literature review as “2. Using Remotely Sensed Data for Conservation” as its own new section separate from the Introduction section.

5-     A compelling argument is needed on the root causes of the investigated transformation, from rainforest to agriculture, as an Amazon countries phenomenon to make this section more contributed to the concurrent literature.

A: We have added writing on the root causes of this transformation at the end of the first paragraph of the Introduction section.

Theoretical and Literature review

6-   The section should be added to the paper’s structure and give new insights into the relationships between the transportation of rainforests land and agricultural economic development.

A: We have added a comprehensive literature review as “2. Using Remotely Sensed Data for Conservation” where we discuss agricultural economic development as requested.

7-   This section also may explain the classification methods and state of the art in this field

A: We have added a comprehensive literature review as “2. Using Remotely Sensed Data for Conservation” where state-of-the-art classifications methods are covered.

Analysis

8- Figure 5. b needs to be more explained as we can see that the percent changes in both agriculture and native vegetation have reached zero in 2021

A: We have clarified this by adding the following sentence 

Presentation

9-     It would be helpful to better understand the transformations that happened in the rainforests if the authors provide a more detailed map to see what exactly happened during the three decades using Figure 4 maps.

A: In order to better visualize these transformations shown in Figure 3 and Figure 4, Figure 3 was reduced down to maps for 4 years with approximate 15-year intervals (1986, 1991, 2005, 2020) during the study period so that these changes are more visually clear. The remaining maps from the other years (1996, 2000, 2011, 2015) were moved to Appendix A.

Policy implications

10- This section seems to be necessary due to the importance of the results of the present paper for policymakers and global environmental activities.

A: We agree and have organized “5.3. Policy Implications” as its own sub-section.

Reviewer 2 Report

The review file is attached.

Comments for author File: Comments.pdf

Author Response

Reviewer 2

 The manuscript “Land use changes in the Teles Pires River basin’s Amazon and Cerrado biomes, Brazil, 1986-2020” evaluates the changes in land use in the Teles Pires River basin, Brazil, over a 34-year period. The study is interesting as it evaluates land use changes driven by socioeconomic factors. The study is also important as it evaluates the land use changes including deforestation over the significant amazon region. The methodology and results are well explained, also the results are well discussed. It is recommended to accept the manuscript after the following minor corrections.

1. In the first author’s affiliation (Postgraduate Program in Environmental Sciences), is it their job description or the institute’s name? It is better to use only the institute’s name.

A: We have clarified this as just Environmental Sciences and not program.

2. In keywords, three keywords cover the study area, is it important to mention all of them? It is better to add “land use” as a keyword.

A: We have added land use as suggested as a key word.

3. Please correct the page number.

A: We have corrected all page numbers.

4. P2, L49-50 “Prior research analyzed changes in….” is a very confusing statement. Please elaborate clearly.

A: We have clarified the writing on this to be more direct.

5. P2, L52, “such as that conducted by [11],…..” is not the correct form of citation. You can rewrite it as, “such as that conducted by Kar et al. [11]….”

A: We have clarified the citation to more generally state what the study did and cite as [#] at the end of the sentence.

6. P2, L54, Land cover and land use.

A: We have made this correction.

7. P2, L57-60, and 62-65 are too big sentences. Please try to break it into a more logical structure.

A: We have broken up the writing in both of these two places into shorter sentences.

8. Mapping the same result in two different figures (Fig 4 and Fig 5) is not appropriate. Although by doing so, the image size is suitably large to distinguish all land covers, however, in my opinion, this can be compromised for better synchronization of figures and Fig 4 and 5 shall be combined.

A: We have kept the graphs separate from the maps and not combined these. We have also simplified the map presentations in Figure 3 and Figure 4 to only 4 years to more easily see the changes over time. We have then moved the remaining 4 maps to the Appendix A in Figure A1.

9. It is a better option to summarize land cover percentages in marked pie-charts for each observing year i.e eight pie-charts (in single figure) for the entire study period. it would be a visible representation of table 3.

A: We appreciate the suggestion to summarize Table 3 as pie charts so we have added this as a new figure.

10. It is a good practice that the symbology of a parameter remains the same in all figures, however, in Fig 5a and 5b you show the same two parameters with opposite symbology. Also, it is too large figure with very less information. Figure 7 has the same type of information with different parameters. It is good to merge 5a and 7a, similarly, 5b and 7b.

A: We have combined Figure 5 with Figure 7 into a more efficient single figure. The line color and data point marker are the same for agricultural areas and likewise for native vegetation.

11. Apparently, the hyperlink in reference No 17 is not referenced in a proper format. Also, when I checked the link, it gives Error 404. Please carefully check this matter.

A: We have replaced this references with another one whose web-link works.

Reviewer 3 Report

The text aims to examine changes in land use and land cover in the Teles Pires river basin between 1986 and 2020. Although discussing changes in land use in Brazil, a country that has suffered substantial changes due to the process of deforestation, is a good and vital purpose, the manuscript has to be better structured to be published in the Sustainability journal. I suggest rejecting the manuscript's publication and leaving my comments divided into two parts. 1) General comments, 2) Methodology:

1) General comments: the description of the results was well presented, the maps have good resolution and the graphics are clear. About the objectives, I leave my comments below:

Lines 67-70: The goals of our research were to expand knowledge about land-use dynamics in the region in order to better manage water resources and plan economic activities. The objective of our study was to evaluate the changes in land use in the Teles Pires River basin over a 34-year period from 1986 to 2020. 

By defining an objective to relate the dynamics of land use with water resources management and planning, the authors should better elaborate the discussion of this issue. Important information such as the installation of hydroelectric power plants (HPPs) is cited to justify the increase in the water area and the soy moratorium and increase in commodity prices for changes in the agricultural area. However, this discussion should be better elaborated.

Regarding the expansion of water due to the HPPs, I advise the authors to highlight how this construction will lead to a loss of biodiversity and a change in the carbon stock, tying the article to the issue of climate change and how this would effect erosion and sediment deposition.

2) Methodology: The delineation of the Teles Pires river basin and the satellite image classification method, two key components of the methodology, were not sufficiently explained by the authors. I recognize the following considerations should be made for an article resubmission in the future:

Lines 88-92: The area of study was delimited using ArcGis 10.1 Software, ArcHydro extension, the Digital Elevation Model (DEM) from the Brazilian Agricultural Research Corporation also known as Embrapa [15]. We used the Shuttle Radar Topography Mission (SRTM) data with spatial resolution of 90 meters and the drainage network from the Brazilian Institute of Geography and Statistics [16].

It is unclear how the study area is delineated in practice. According to the authors, they employed the IBGE drainage network and DEM from Embrapa. The ArcHydro extension's goal is to create the drainage network first, then the watershed. Additionally, there are DEMs with spatial resolutions larger than 90 meters, such as TanDEM-X - Digital Elevation Model (DEM) - Global, 30m and NASA SRTM Digital Elevation 30m (https://developers.google.com/earth-engine/datasets/catalog/USGS SRTMGL1 003).

For the delimitation of the hydrographic basin's bounds to be accurate, the stages of delimitation must be accurately described. What is detailed in the following lines is in direct conflict with how the IBGE drainage network is used:

Lines 101-104: Soon after, the correction of depressions was performed and the flow direction was obtained by the Eight Direction Pour Point Model method, which assumes that the water flows to one of the eight neighboring cells according to the greatest slope, used for generating the accumulated flow, calculation of the upstream cells that drain to each cell of the raster.

In this section, the authors describe how to create a drainage network by using the DEM's depression correction and flow direction procedures. Why utilize an approved IBGE base if a drainage network was created?

—-----------—-----------—-----------—-----------—-----------—-----------—-----------—----------

Lines 121-123: We used data from the dry season in the region between the months of May and September since, when there is low cloudiness, favoring the classification of images. 

I propose including a table with the image's ID, date, and orbit/point details.

—-----------—-----------—-----------—-----------—-----------—-----------—-----------—----------

Lines: 137-139: Supervised classification was then used for identifying the classes present in the area and selecting representative samples, which consisted of the training phase of the classifier. 

In studies involving remote sensing, the supervised classification technique is frequently used. The authors cite the Mapbiomas platform when comparing their classification outcomes with Mapbiomas maps to show that recent developments in this field show a greater availability of more modern techniques, such as the use of machine learning, decision trees, or artificial neural networks that are widely used. Here are a few research examples:

Souza, C.M., Jr.; Z. Shimbo, J.; Rosa, M.R.; Parente, L.L.; A. Alencar, A.; Rudorff, B.F.T.; Hasenack, H.; Matsumoto, M.; G. Ferreira, L.; Souza-Filho, P.W.M.; de Oliveira, S.W.; Rocha, W.F.; Fonseca, A.V.; Marques, C.B.; Diniz, C.G.; Costa, D.; Monteiro, D.; Rosa, E.R.; Vélez-Martin, E.; Weber, E.J.; Lenti, F.E.B.; Paternost, F.F.; Pareyn, F.G.C.; Siqueira, J.V.; Viera, J.L.; Neto, L.C.F.; Saraiva, M.M.; Sales, M.H.; Salgado, M.P.G.; Vasconcelos, R.; Galano, S.; Mesquita, V.V.; Azevedo, T. Reconstructing Three Decades of Land Use and Land Cover Changes in Brazilian Biomes with Landsat Archive and Earth Engine. Remote Sens. 2020, 12, 2735. https://doi.org/10.3390/rs12172735

Y. E. Shimabukuro et al., "Discriminating Land Use and Land Cover Classes in Brazil Based on the Annual PROBA-V 100 m Time Series," in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 13, pp. 3409-3420, 2020, doi: 10.1109/JSTARS.2020.2994893.

Therefore, I suggest that the authors reformulate their article image classification technique or provide well-founded justifications that demonstrate the advances achieved with the use of the supervised classification technique, within a context where this technique has already been overcome by significant advances in the area of remote sensing. Below are some suggestions for articles dealing with image classification:

Asokan, A.; Anitha, J.; Ciobanu, M.; Gabor, A.; Naaji, A.; Hemanth, D.J. Image Processing Techniques for Analysis of Satellite Images for Historical Maps Classification—An Overview. Appl. Sci. 2020, 10, 4207. https://doi.org/10.3390/app10124207

Mahmon, N. A. and Ya'acob, N. A review on classification of satellite image using Artificial Neural Network (ANN). 2014 IEEE 5th Control and System Graduate Research Colloquium, 2014, pp. 153-157, doi: 10.1109/ICSGRC.2014.6908713

Anil B. Gavade & Vijay S. Rajpurohit (2021) Systematic analysis of satellite image-based land cover classification techniques: literature review and challenges, International Journal of Computers and Applications, 43:6, 514-523, DOI: 10.1080/1206212X.2019.1573946

What new scientific advancement does the manuscript make in the field of image classification if there is already a platform like Mapbiomas that offers land use and land cover data on the Brazilian territory in annual maps?

Author Response

Reviewer 3

Comments and Suggestions for Authors

The text aims to examine changes in land use and land cover in the Teles Pires river basin between 1986 and 2020. Although discussing changes in land use in Brazil, a country that has suffered substantial changes due to the process of deforestation, is a good and vital purpose, the manuscript has to be better structured to be published in the Sustainability journal. I suggest rejecting the manuscript's publication and leaving my comments divided into two parts. 1) General comments, 2) Methodology:

1) General comments: the description of the results was well presented, the maps have good resolution and the graphics are clear. About the objectives, I leave my comments below:

Lines 67-70: The goals of our research were to expand knowledge about land-use dynamics in the region in order to better manage water resources and plan economic activities. The objective of our study was to evaluate the changes in land use in the Teles Pires River basin over a 34-year period from 1986 to 2020. 

By defining an objective to relate the dynamics of land use with water resources management and planning, the authors should better elaborate the discussion of this issue. Important information such as the installation of hydroelectric power plants (HPPs) is cited to justify the increase in the water area and the soy moratorium and increase in commodity prices for changes in the agricultural area. However, this discussion should be better elaborated.

Regarding the expansion of water due to the HPPs, I advise the authors to highlight how this construction will lead to a loss of biodiversity and a change in the carbon stock, tying the article to the issue of climate change and how this would effect erosion and sediment deposition.

A: We thank the reviewer for this suggestion and have added a paragraph on the impacts of dams in this region in the first sub-section 5.1. of the Discussion section.

2) Methodology: The delineation of the Teles Pires river basin and the satellite image classification method, two key components of the methodology, were not sufficiently explained by the authors. I recognize the following considerations should be made for an article resubmission in the future:

Lines 88-92: The area of study was delimited using ArcGis 10.1 Software, ArcHydro extension, the Digital Elevation Model (DEM) from the Brazilian Agricultural Research Corporation also known as Embrapa [15]. We used the Shuttle Radar Topography Mission (SRTM) data with spatial resolution of 90 meters and the drainage network from the Brazilian Institute of Geography and Statistics [16].

It is unclear how the study area is delineated in practice. According to the authors, they employed the IBGE drainage network and DEM from Embrapa. The ArcHydro extension's goal is to create the drainage network first, then the watershed. Additionally, there are DEMs with spatial resolutions larger than 90 meters, such as TanDEM-X - Digital Elevation Model (DEM) - Global, 30m and NASA SRTM Digital Elevation 30m (https://developers.google.com/earth-engine/datasets/catalog/USGS SRTMGL1 003).

For the delimitation of the hydrographic basin's bounds to be accurate, the stages of delimitation must be accurately described. What is detailed in the following lines is in direct conflict with how the IBGE drainage network is used:

Lines 101-104: Soon after, the correction of depressions was performed and the flow direction was obtained by the Eight Direction Pour Point Model method, which assumes that the water flows to one of the eight neighboring cells according to the greatest slope, used for generating the accumulated flow, calculation of the upstream cells that drain to each cell of the raster.

A: The use of the IBGE drainage network in the basin boundary definition stage, in the authors' perspective, did not configure a conflict with the use of the ArcHydro extension, since the data served to recognize the basin extension and was also used to carry out the reconditioning of the DEM, the first step performed in the delimitation of the basin (Ettritch et al., 2014). The reconditioning of the DEM consists of imposing the drainage pattern on the DEM, forcing it to correspond to the drainage vector through adjustments in the increase values, which helps in the hydrological analyzes in areas with topographical differences. Previous tests carried out for the Teles Pires river basin without this step resulted in inconsistencies in the boundaries of the area, especially in the portion close to the mouth of the basin, justifying the need to carry out such a procedure. As described in the methodology, in the subsequent step, the raster drainage network was generated using the accumulated flow data as a base (standard procedure in automated methods of delimitation of watersheds). Finally, the drainage raster was applied in the basin delimitation.

In this section, the authors describe how to create a drainage network by using the DEM's depression correction and flow direction procedures. Why utilize an approved IBGE base if a drainage network was created?

A: The use of the EMBRAPA SRTM DEM (90m) was considered compatible with the planned detailing in the study area, since it is a large basin, with a large N-S extension and a large area (approximately 142,000 km²) and also suitable for availability of hydrographic data, complete for an area at 1:250,000 scale, as mentioned. Although the existence of more detailed DEM is known, it was decided to work with the original SRTM product, considered with the scale adopted for the analysis and maintaining all the corrections planned in the processing of this nature.

—-----------—-----------—-----------—-----------—-----------—-----------—-----

Lines 121-123: We used data from the dry season in the region between the months of May and September since, when there is low cloudiness, favoring the classification of images. 

I propose including a table with the image's ID, date, and orbit/point details.

A: We thank the reviewer for this suggestion and have added this as a new Table 1 in the Materials and Methods section.

—-----------—-----------—-----------—-----------—-----------—-----------—-----

Lines: 137-139: Supervised classification was then used for identifying the classes present in the area and selecting representative samples, which consisted of the training phase of the classifier. 

In studies involving remote sensing, the supervised classification technique is frequently used. The authors cite the Mapbiomas platform when comparing their classification outcomes with Mapbiomas maps to show that recent developments in this field show a greater availability of more modern techniques, such as the use of machine learning, decision trees, or artificial neural networks that are widely used. Here are a few research examples:

Souza, C.M., Jr.; Z. Shimbo, J.; Rosa, M.R.; Parente, L.L.; A. Alencar, A.; Rudorff, B.F.T.; Hasenack, H.; Matsumoto, M.; G. Ferreira, L.; Souza-Filho, P.W.M.; de Oliveira, S.W.; Rocha, W.F.; Fonseca, A.V.; Marques, C.B.; Diniz, C.G.; Costa, D.; Monteiro, D.; Rosa, E.R.; Vélez-Martin, E.; Weber, E.J.; Lenti, F.E.B.; Paternost, F.F.; Pareyn, F.G.C.; Siqueira, J.V.; Viera, J.L.; Neto, L.C.F.; Saraiva, M.M.; Sales, M.H.; Salgado, M.P.G.; Vasconcelos, R.; Galano, S.; Mesquita, V.V.; Azevedo, T. Reconstructing Three Decades of Land Use and Land Cover Changes in Brazilian Biomes with Landsat Archive and Earth Engine. Remote Sens. 2020, 12, 2735. https://doi.org/10.3390/rs12172735

10. E. Shimabukuro et al., "Discriminating Land Use and Land Cover Classes in Brazil Based on the Annual PROBA-V 100 m Time Series," in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 13, pp. 3409-3420, 2020, doi: 10.1109/JSTARS.2020.2994893.

Therefore, I suggest that the authors reformulate their article image classification technique or provide well-founded justifications that demonstrate the advances achieved with the use of the supervised classification technique, within a context where this technique has already been overcome by significant advances in the area of remote sensing. Below are some suggestions for articles dealing with image classification:

Asokan, A.; Anitha, J.; Ciobanu, M.; Gabor, A.; Naaji, A.; Hemanth, D.J. Image Processing Techniques for Analysis of Satellite Images for Historical Maps Classification—An Overview. Appl. Sci. 2020, 10, 4207. https://doi.org/10.3390/app10124207

Mahmon, N. A. and Ya'acob, N. A review on classification of satellite image using Artificial Neural Network (ANN). 2014 IEEE 5th Control and System Graduate Research Colloquium, 2014, pp. 153-157, doi: 10.1109/ICSGRC.2014.6908713

Anil B. Gavade & Vijay S. Rajpurohit (2021) Systematic analysis of satellite image-based land cover classification techniques: literature review and challenges, International Journal of Computers and Applications, 43:6, 514-523, DOI: 10.1080/1206212X.2019.1573946

What new scientific advancement does the manuscript make in the field of image classification if there is already a platform like Mapbiomas that offers land use and land cover data on the Brazilian territory in annual maps?

A: As for image classification methods, we thank you for the suggested articles. However, it was not the intention of our work to present innovation or scientific advance in image classification protocol. A classification method was adopted which, although it has already been surpassed by more recent techniques, as mentioned, was sufficient to achieve the proposed objectives of quantifying the areas and enabling analysis of the temporal dynamics. The execution of supervised classification possible, from the point of view of human resource training, the application of knowledge acquired in theory in practice applicable to a relevant regional issue. The advancement of technology in the area of remote sensing occurs quickly, new approaches such as the use of neural networks, artificial intelligence, machine learning, etc., become more common, but the theoretical foundation remains and decision-making based on the results obtained is still carried out by the analyst's interpretation. Once the technique was applied and the results obtained, the achieved accuracy was considered satisfactory for the research.

The mention and use of data from Mapbiomas in the context took place for the purpose of verifying and comparing the result found by the supervised classification, aiming to highlight the analysis of the result obtained with the product made available by INPE.

The authors have used the data distributed by Mapbiomas in other research and recognize the very important contribution of this important database in providing accurate and free data for application by the scientific community.

However, in this study we sought to create our own land use and occupation database, where we qualified the land uses that best suited the needs of our line of research. Our data are available for other studies that aim to estimate the impact of changes in land occupation on water availability, the occurrence of extreme events and the production of sediments in the watershed of the Teles Pires river, surveys that are part of the project “Rede de Pesquisas no rio Teles Pires: disponibilidade hídrica e de sedimentos em diferentes cenários ambientais” (Research Network in the Teles Pires river: water availability and sediments in different environmental scenarios”).

References

Ettritch, G.; Hardy, A.; Bojang, L.; Cruz, D.; Bunting, P.; Brewer, P. Enhancing Digital Elevation Models for Hydraulic Modeling Using Flood Frequency Detection. Remote Sensor. Environment. 2018, 217, 506–522. https://doi.org/10.1016/j.rse.2018.08.029.

Round 2

Reviewer 1 Report

Most of my concerns have been met in the revised versions. However, the following comment has yet to be clearly understand and addressed as I meant that why percent changes in both agriculture and native vegetation have reached zero in 2021? And it is somehow vague to readers that how they have reached to a similar point in 2021? Please revise or explain

8- Figure 5. b needs to be more explained as we can see that the percent changes in both agriculture and native vegetation have reached zero in 2021

A: We have clarified this by adding the following sentence 

Author Response

8- Figure 5. b needs to be more explained as we can see that the percent changes in both agriculture and native vegetation have reached zero in 2021

A: We have clarified this by adding the following two sentences to the end of the 4^th paragraph of the Results section on L376-379: “Between 2015 and 2020, both declines in the increase in agricultural areas and the decrease in native vegetation to near 0% (Figure 5b) do not mean that there is no expansion of agriculture nor no deforestation. Rather this indicates a relative stabilization of recent rates of agricultural expansion and deforestation.”

Reviewer 2 Report

The authors have revised the manuscript upto my satisfaction. I recommend acceptance of this manuscript to be published in its present form.

Reviewer 3 Report

I continue to stand by my first decision to reject the paper. The authors' responses to my questions do not reflect the developments in remote sensing and watershed planning. The article is not original and does not bring relevant scientific contributions.

Author Response

Comments and Suggestions for Authors

The text aims to examine changes in land use and land cover in the Teles Pires river basin between 1986 and 2020. Although discussing changes in land use in Brazil, a country that has suffered substantial changes due to the process of deforestation, is a good and vital purpose, the manuscript has to be better structured to be published in the Sustainability journal. I suggest rejecting the manuscript's publication and leaving my comments divided into two parts. 1) General Comments, 2) Methodology:

1) General Comments: What new scientific advancement does the manuscript make in the field of image classification if there is already a platform like Mapbiomas that offers land use and land cover data on the Brazilian territory in annual maps?

We applied our methodology outlined in section 3.3. to classify land use for 1986, 1991, 1996, 2000, 2005, 2011, 2015, and 2020 using Landsat imagery. We have then validated our land use classifications using methods outlined in section 3.3. with ground truthed data that our research team collected in 2020 from 1477 geo-locations. We then proceeded to use methods outlined in section 3.3. to classify land use for 1986, 1991, 1996, 2000, 2005, 2011, and 2015 using  Landsat images and then we proceeded to validate these land-use classifications we made against MapBiomas data. Our contribution to scientific advancement is in the methods we used for land classification and our validation with real world data surveyed in the field in 2020.

We based our validation analysis on Shimabukuro et al. 2020 that you provided:

10. E. Shimabukuro et al., "Discriminating Land Use and Land Cover Classes in Brazil Based on the Annual PROBA-V 100 m Time Series," in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 13, pp. 3409-3420, 2020, doi: 10.1109/JSTARS.2020.2994893.

We have added both a new Table 5 for validating land use classifications using methods outlined in section 3.3. against the 1477 locations of ground-truthed data. We have also added Appendix Table A1 and Table A2 which validates land use classifications we made using Landsat images (using methods outlined in section 3.3.) against MapBiomas for the other years we evaluated. The burned class is excluded since there is no corresponding class in MapBiomas and keeping those points would only increase errors. All new Kappa indices remained within the range of 0.8 to 1.0, considered Excellent.

Our results suggest that there is more need for ground truthing especially for the crop classification in Brazil since for the crop category using Landsat data that validating with ground-truthed data that we collected in the field that this crop category was only 73.35% accurate. This is compared to validating the Landsat data with MapBiomas which has 94.72% accuracy for crops. The ground truthing we discuss in the Teles Pires River basin has not been done before and thus contributes to the literature. More importantly, if this is not more widely known that this discrepancy exists for crops then future studies would not be able to improve the accuracy of land-use classification relative to ground-truthed data.

2) Methodology: The delineation of the Teles Pires river basin and the satellite image classification method, two key components of the methodology, were not sufficiently explained by the authors. I recognize the following considerations should be made for an article resubmission in the future:

Lines 88-92: The area of study was delimited using ArcGis 10.1 Software, ArcHydro extension, the Digital Elevation Model (DEM) from the Brazilian Agricultural Research Corporation also known as Embrapa [15]. We used the Shuttle Radar Topography Mission (SRTM) data with spatial resolution of 90 meters and the drainage network from the Brazilian Institute of Geography and Statistics [16].

It is unclear how the study area is delineated in practice. According to the authors, they employed the IBGE drainage network and DEM from Embrapa. The ArcHydro extension's goal is to create the drainage network first, then the watershed. Additionally, there are DEMs with spatial resolutions larger than 90 meters, such as TanDEM-X - Digital Elevation Model (DEM) - Global, 30m and NASA SRTM Digital Elevation 30m (https://developers.google.com/earth-engine/datasets/catalog/USGS SRTMGL1 003).

For the delimitation of the hydrographic basin's bounds to be accurate, the stages of delimitation must be accurately described. What is detailed in the following lines is in direct conflict with how the IBGE drainage network is used:

Lines 101-104: Soon after, the correction of depressions was performed and the flow direction was obtained by the Eight Direction Pour Point Model method, which assumes that the water flows to one of the eight neighboring cells according to the greatest slope, used for generating the accumulated flow, calculation of the upstream cells that drain to each cell of the raster.

A: The use of the IBGE drainage network in the basin boundary definition stage, in the authors' perspective, did not configure a conflict with the use of the ArcHydro extension, since the data served to recognize the basin extension and was also used to carry out the reconditioning of the DEM, the first step performed in the delimitation of the basin (Ettritch et al., 2014). The reconditioning of the DEM consists of imposing the drainage pattern on the DEM, forcing it to correspond to the drainage vector through adjustments in the increase values, which helps in the hydrological analyzes in areas with topographical differences. Previous tests carried out for the Teles Pires river basin without this step resulted in inconsistencies in the boundaries of the area, especially in the portion close to the mouth of the basin, justifying the need to carry out such a procedure. As described in the methodology, in the subsequent step, the raster drainage network was generated using the accumulated flow data as a base (standard procedure in automated methods of delimitation of watersheds). Finally, the drainage raster was applied in the basin delimitation.

In this section, the authors describe how to create a drainage network by using the DEM's depression correction and flow direction procedures. Why utilize an approved IBGE base if a drainage network was created?

A: The use of the EMBRAPA SRTM DEM (90m) was considered compatible with the planned detailing in the study area, since it is a large basin, with a large N-S extension and a large area (approximately 142,000 km²) and also suitable for availability of hydrographic data, complete for an area at 1:250,000 scale, as mentioned. Although the existence of more detailed DEM is known, it was decided to work with the original SRTM product, considered with the scale adopted for the analysis and maintaining all the corrections planned in the processing of this nature.

—-----------—-----------—-----------—-----------—-----------—-----------—-----

Lines: 137-139: Supervised classification was then used for identifying the classes present in the area and selecting representative samples, which consisted of the training phase of the classifier. 

In studies involving remote sensing, the supervised classification technique is frequently used. The authors cite the Mapbiomas platform when comparing their classification outcomes with Mapbiomas maps to show that recent developments in this field show a greater availability of more modern techniques, such as the use of machine learning, decision trees, or artificial neural networks that are widely used. Here are a few research examples:

Souza, C.M., Jr.; Z. Shimbo, J.; Rosa, M.R.; Parente, L.L.; A. Alencar, A.; Rudorff, B.F.T.; Hasenack, H.; Matsumoto, M.; G. Ferreira, L.; Souza-Filho, P.W.M.; de Oliveira, S.W.; Rocha, W.F.; Fonseca, A.V.; Marques, C.B.; Diniz, C.G.; Costa, D.; Monteiro, D.; Rosa, E.R.; Vélez-Martin, E.; Weber, E.J.; Lenti, F.E.B.; Paternost, F.F.; Pareyn, F.G.C.; Siqueira, J.V.; Viera, J.L.; Neto, L.C.F.; Saraiva, M.M.; Sales, M.H.; Salgado, M.P.G.; Vasconcelos, R.; Galano, S.; Mesquita, V.V.; Azevedo, T. Reconstructing Three Decades of Land Use and Land Cover Changes in Brazilian Biomes with Landsat Archive and Earth Engine. Remote Sens. 2020, 12, 2735. https://doi.org/10.3390/rs12172735

10. E. Shimabukuro et al., "Discriminating Land Use and Land Cover Classes in Brazil Based on the Annual PROBA-V 100 m Time Series," in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 13, pp. 3409-3420, 2020, doi: 10.1109/JSTARS.2020.2994893.

Therefore, I suggest that the authors reformulate their article image classification technique or provide well-founded justifications that demonstrate the advances achieved with the use of the supervised classification technique, within a context where this technique has already been overcome by significant advances in the area of remote sensing. Below are some suggestions for articles dealing with image classification:

Asokan, A.; Anitha, J.; Ciobanu, M.; Gabor, A.; Naaji, A.; Hemanth, D.J. Image Processing Techniques for Analysis of Satellite Images for Historical Maps Classification—An Overview. Appl. Sci. 2020, 10, 4207. https://doi.org/10.3390/app10124207

Mahmon, N. A. and Ya'acob, N. A review on classification of satellite image using Artificial Neural Network (ANN). 2014 IEEE 5th Control and System Graduate Research Colloquium, 2014, pp. 153-157, doi: 10.1109/ICSGRC.2014.6908713

Anil B. Gavade & Vijay S. Rajpurohit (2021) Systematic analysis of satellite image-based land cover classification techniques: literature review and challenges, International Journal of Computers and Applications, 43:6, 514-523, DOI: 10.1080/1206212X.2019.1573946

A: As for image classification methods, we thank you for the suggested articles. However, it was not the intention of our work to present innovation or scientific advance in image classification protocol. A classification method was adopted which, although it has already been surpassed by more recent techniques, as mentioned, was sufficient to achieve the proposed objectives of quantifying the areas and enabling analysis of the temporal dynamics. The execution of supervised classification possible, from the point of view of human resource training, the application of knowledge acquired in theory in practice applicable to a relevant regional issue. The advancement of technology in the area of remote sensing occurs quickly, new approaches such as the use of neural networks, artificial intelligence, machine learning, etc., become more common, but the theoretical foundation remains and decision-making based on the results obtained is still carried out by the analyst's interpretation. Once the technique was applied and the results obtained, the achieved accuracy was considered satisfactory for the research.

The mention and use of data from Mapbiomas in the context took place for the purpose of verifying and comparing the result found by the supervised classification, aiming to highlight the analysis of the result obtained with the product made available by INPE. The authors have used the data distributed by Mapbiomas in other research and recognize the very important contribution of this important database in providing accurate and free data for application by the scientific community.

However, in this study we sought to create our own land use and occupation database, where we qualified the land uses that best suited the needs of our line of research. Our data are available for other studies that aim to estimate the impact of changes in land occupation on water availability, the occurrence of extreme events and the production of sediments in the watershed of the Teles Pires river, surveys that are part of the project “Rede de Pesquisas no rio Teles Pires: disponibilidade hídrica e de sedimentos em diferentes cenários ambientais” (Research Network in the Teles Pires river: water availability and sediments in different environmental scenarios”).

References

Ettritch, G.; Hardy, A.; Bojang, L.; Cruz, D.; Bunting, P.; Brewer, P. Enhancing Digital Elevation Models for Hydraulic Modeling Using Flood Frequency Detection. Remote Sensor. Environment. 2018, 217, 506–522. https://doi.org/10.1016/j.rse.2018.08.029.

Round 3

Reviewer 1 Report

Regarding the revised version, the paper is now publishable. 

Reviewer 3 Report

I maintain my first and second evaluations and reject the article's publication.