Mapping Burned Areas of Mato Grosso State Brazilian Amazon Using Multisensor Datasets

Round 1

Reviewer 1 Report

General comments:

 

The manuscript entitled “Mapping burned areas of Mato Grosso State, Brazilian Amazon, using multisensors datasets” by Shimabukuro et al. aims to develop a new method for detecting burned area across the fire-prone southern Amazon and Cerrado biomes using multi-source satellite data. The authors first derive a shade fraction image from reflectance data based on the LSMM method, then classify the shade fraction image using an unsupervised classifier and manually label the burned area (BA) class, and finally assess the BA results using the sentinel-2 based BA, three existing MODIS BA products, and MODIS active fire data. The authors conclude that combination of multi-sensor data improves BA accuracy.

            The study has a very good goal of mapping BA with multi-source data as a single remote sensing data source seems always insufficient to detect BA in tropical regions due to frequent obscuration of clouds and forest canopies. It also chooses a good hot spot, where burnings are widespread during the dry season, for mapping BA and has a good experiment design in terms of assessing BA maps. However, the introduction misses state of art of BA detection algorithms, the method is lack of clarity, and feasibility of the proposed method and accuracy of the derived BA results are not justified reasonably. To make this manuscript a solid one, the following main points should be addressed:

(1)  Provide a brief review of the BA detection algorithms. For instance, the differential normalized burn ratio (dNBR) based on the pre-fire and post-fire remote sensing images has been commonly used to map BA. There is a need to explain why LSMM and the segmentation-based classification was chosen.

(2)  For the method, first it is not clear how the three endmembers of the LSMM are chosen and how other potential confusion surface types (i.e., water bodies) are treated. Second, it would be reasonable to clearly describe the criteria used to label a classified area as burned area because the labeling process is conducted manually, which is very subjective. Because the key part of identifying burned areas in the proposed method is not working automatically, it seems the application potential of the proposed method is very limited, especially for BA at large scales.

(3)  The metrics - omission and commission errors rather than the coefficient of determination (R-square) should be used to assess the overall accuracy of the derived BA results. In the absence of ground-truth BA data, it is a common practice to validation of a remote sensing product using a higher spatial-resolution product of the same kind. First, the sentinel-2 data is a good choice. If the sentinel-based BA is derived at a resolution of 20m or higher, the validation would make sense but the manuscript does not specify it. Second, the omission and commissions errors are used all through the discussion section but the authors infer the commission and omissions errors based on the gridded MODIS active fire data and do not state how errors are calculated. The commission and omission errors should be calculated based on the validation BA data.

(4)  The very large difference between the OLI+VIIRS+PROBA-V based BA and three MODIS BA products needs explanations. In Table 4, the OLI+VIIRS+PROBA-V based BA is more than two times as much as any of three MODIS BA. If the accuracy of the OLI+VIIRS+PROBA-V based BA is very reliable and high, then MODIS BA products are largely underestimated by doing a poor job in detection BA in the study region; then it is the strong point of this paper and deserves emphasis and discussion. Otherwise, the OLI+VIIRS+PROBA-V based BA could be largely overestimated, please explain possible reasons.

 

Additionally, the writing could be improved.

 

Specific comments (hard to comment on it, as line numbers are not added):

Title: consider remove commas. For instance, “Mapping burned areas of Mato Grosso State in Brazilian Amazon using multi-sensor datasets”

 

Abstract (please use consistent tense):

            Would change “still configure as” to “is still configured as”

            Would change “an annual map of the burned area of” to “an annual burned area map of”

            Add “by” in the front of “applying the Linear Spectral Mixture Model…”

           

Introduction

            Please add a short review of BA algorithms and explain a bit why LSMM and a segmentation method were chosen.

            Paragraphs 3-4 could be merged as one paragraph and make it concise.

            P1. May rephrase “Farmers burn their land to convert forests to cropland or pasture and to control the spread of weeds, pests or diseases, as well as to stimulate pasture regrowth” as “Farmers burn their land to convert forests to cropland or pasture, to control the spread of weeds, pests and/or diseases, and  to stimulate pasture regrowth as well.”

            P2. May rephrase “Burns occur every year, generally during the dry season, with a higher incidence at the end of this period” as “Fires generally burn during the dry season every year, peaking at the end of the dry season.”

            P2. “an ignition source”. Which ignition source?

P3. “when a fire regime occurs”. The term fire regime is not proper as it refers to the long-term fire characteristics (i.e., return interval, frequency, severity…).

P4. “Moreover, the release…the climate” reads awkward, please rephrase it.

P4. Change “Thus” in the sentence “Thus, burning in tropical regions…” to “For example/instance”.

P6. Change “fire study” to “fire-related study’

P6. Change “between 3 and 11 um” to “between 4 and 11 um”.

P7. The first sentence reads awkward, consider rephrase it.

2. Materials and Methods

2.2.2 VIIRS. Please move (VNP09GA) to the second sentence and specify the data source where the data were obtained.

2.2.3 OLI.

P1. May consider rephrase the first sentence “The Operational …bands (…)” as “The Operational Land Imager (OLI) on Landsat-8 satellite has 11 spectral bands (…)”.

P1. Change “a 16-day of temporal resolution” to “a temporal resolution of 16 days” or remove “of”.

            Please specify the data source where the data were obtained.

2.2.4. Sentinel-2.

            P1. please use a similar description way of L8-OLI for the MSI.

2.3.      Global BA products.

            Table 1. Please add the version/collection number as a column.

            Also please specify data sources where the data sets are available.

P1. I don’t think the term “validation” is proper here for comparison of BA with three MODIS BA product because the MODIS BA has a much coarser spatial resolution than L8 OLI.

2.4. Is this active fire data the same as the NASA official MODIS active fire data (MYD14/MOD14)?

2.5.2. Cropland map. What is 5.3 million? Missing units?

 

2.6. Methodological overview.

            P1. The first paragraph repeats parts of sections 2.2-2.4.

            P2. How are other surface types like water bodies, city/urban, etc, treated? And how endmembers are selected in the LSMM process? Please clarify.

            Fig.3. Does RGB image represent a true- or false-color composite? Please specify the RGB bands for OLI and VIIRS.

            Page8. P2. What kinds of criteria are used to label segments to burned area classes? Please clarify it.

 

2.7. Validation.

            Calculate omission and commission errors here using sentinel-2 based BA (and add discussions of uncertainties in the Sentinel-2 based BA in discussion section). Please also specify the spatial resolution of sentinel-2 BA

            P3. What are “2 groups”?

            Fig.6. A comparison of spatial patterns of the Sentinel-2 based BA and the OLI, VIIRS, PROBA-V based BA (similar to Fig.5) would provide additional information for visual interpretations.

            Fig.7. Please consider revise this figure because it is difficult to tell grids with BA from grids without BA as grids could also be gray color if all BA products have no valid BA in these grids.

            Page 14. P2. It is incorrect to infer omission and commission errors based active fire data, rigorously speaking.

            Table 4. The same as the fourth main point raised in general comments.

Page14. P2. What is “[Erro! Fonte de referência não encontrada.]”? Probably a citation? It also appears at page 16.

 

4. Discussions.

            Please update it with the sentinel-2 based commission and omission errors.

Consider adding discussion of the advantages and potential uncertainties/limitations of the proposed method.

 

5. Conclusions

            Please add findings related the proposed BA method.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The burnt area mapping in the regions which suffer important losses due to fires is needed for the adequate fire management. The presented paper intends to propose a solution for this problem. However I think that the paper has serious flaws, especially in the way of the description of the method. 

In my opinin it is impossible to reproduce the process according to the current version of the method description:

p.7 equation 1 - the explanation of the symbols does not correspond to the symbols used in the equation

p.2, paragraph 2 - What were vales of teo threshold parameters used in the segmentation?

p.8.paragraph2: "...After the classification process, the analyst labels the burnt area segments to the predefined legend class by visual inspection..." Does this mean that visual interpretation was use to assign individual segments as burnt or not burnt. Or all specific class from unsupervised classification were visually reclassified into classes with real meaning? What classes had the "predefined legend"? This sentense also suggests that after all the process of classification is manual.

p9 Validation.  A more clear description of the reference dataset is needed. It is said that 6 Sentinel-2 granules were used for the manual burnt area mapping. How many dates per granule were used? It is also said the 381 points were distributed in MSI/S2 granules. It means that there were 381 points per granule or totally were 381 points. Taking into account the size of the study area I think that this number of test points is too small to give a represenative results.

another issue is how the points were really used. All presented results refer to square kilometers. So I find the current description inconsinstent.

Results

Figure 6. If I understand well it seems that the results obtained with the proposed method are worse that the results iven by global products. in such case a very strong justification is needed for the publication of the proposed method.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Summary

In this study, burned area maps were created through unsupervised classification of shade fraction images derived through linear spectral mixture modeling of Landsat OLI, PROBA-V, and Suomi NPP-VIIRS images. The resulting burned area maps were then compared to several global burned area products and Sentinel-2 reference image. The highest correlation between the burned area maps and Sentinel-2 reference data was with Landsat OLI while the least was with VIIRS-derived maps. The authors attribute the higher agreement between of OLI-derived burned area maps and reference Sentinel-2 data to its superior spatial resolution. Despite the higher spatial resolution of PROBA-V derived maps n(100m) compared to VIIRs derived maps (300 m), the maps had similar agreements to Sentinel reference data. The authors attributed the 'under-performance' of the PROBA-V imagery to their lower temporal resolution compared to VIIRS.

The study is a welcome contribution to the growing literature on techniques to improve the mapping of burned areas. I am convinced many readers will find practical applications of the method they propose of combining high temporal/low-spatial resolution imagery with low temporal/high-spatial resolution imagery to create better burned area maps. The manuscript will, however, need close editing to clarify meaning and correct English grammar. The authors also need to elaborate on some of the methods and analysis. For example, it is not quite clear Sentinel-2 data were used to generate reference data for validation of burn area maps from other sensors. How was this 'reference data' derived? What Sentinel-2 bands were used to generate the data? Was the image classified to map areas burned? Even more confusing is the statement by the authors that Sentinel-2 imagery were available for only half the study area. How did this impact validation? The explanations on the burn area maps from the different sensors were compared to the 1 km MODIS active fire product is unclear. I must also state that reviewing of the manuscript was made more challenging because of the missing line numbers in the manuscript.

Specific Issues

I will now highlight several of the issues that I encountered in various sections of the manuscript. I will not attempt to type all these out here, but this is an indication that the authors will need to edit their manuscript thoroughly.

Title:

Delete 's' in 'multisensors to change to 'multisensor'

Abstract

Replace 'still configure as' with 'remain a'

Line 2: Modify sentence beginning here to begin 'In this paper, we propose……..'

Line 9: Replace 'computed in the' with 'evaluated in terms of land use and land cover classes over the Amazon….'

Lines 11-13: Sentence beginning here does not make sense and 'occasing' is not a recognized English word. It is not clear what the authors are stating regarding burned areas with the forested class. Please elaborate and clarify meaning.

Lines 16-18: Modify to 'the areas estimated were quite different.' The meaning of the next sentence beginning with 'Our results' is not clear. Please elaborate and clarify.

Lines 19-20: modify as 'about the patterns of fire in various biomes of Mato Grosso State, Brazil that are important……….' The rest of the sentence is vague and needs to be more specific.

Introduction

Line 1: delete 'the' before 'burning'

Line 2: delete 'the' before 'cattle'

Line 3: delete 'the' before 'agricultural'

Page 2

Paragraph 1: Suggested 'Burns occur every year, and especially during the dry season, with most of the fires occurring towards end of the dry season. During this time, risk of fire is highest due to factors such as ………..'flammable materials, and increased exposure to ignition sources [2,7-10].

Paragraph 2: Does the Mha refer to 'millions of hectares'? If so, spell it out in full the first time and then you can use the abbreviation subsequently.

Line 2: affected by fire' is vague. Be more specific.

Line 3: '….escaping from nearby agricultural areas into forests may be a…..'

Lines 4-5: '….year when fires occur…'

Line 5: What do you mean by 'will be resistant to this event'?

Line 6: Are burned forest areas resilient that they will recover to former states?

 Lines 7-8: The statement 'where biomass burning accelerates the availability of nutrients for plans [12]' needs to be qualified with conditions that make that possible. Many physical, chemical, and mineralogical, biological soil properties are influenced by severity of the burn, and it is the severity that will determine whether the effects are beneficial, neutral, or deleterious.

Line 11: Delete 'the' before 'vegetation'

Paragraph 3

Begin as follows: 'Forest burning brings numerous…..'

Line 5: The sentence beginning here is very long and needs to be split and restructured. Begin sentence here as 'Biomass burning in tropical regions is estimated to contribute to 32% of global CO₂ emissions……..'

Paragraph 4

The paragraph needs to be re-structured to correct grammar and clarify meaning. Perhaps restructure first sentence in the paragraph as follows 'Analysis of the timing and distribution of forest fires is an important tool in efficient fire management across regions experiencing varied fire regimes.'

Paragraph 5

'Remote sensing represents a particularly useful tool for mapping and quantifying areas burned in large and difficult to reach areas. These data allow for i0 active fire detection……'

Line 8: Sentence beginning here: replace 'orbital sensors' with 'remote sensing'

Line 9: '…dynamics as they allow the observation of large areas of the surface at high temporal resolutions [28,28].'

Paragraph 6

Line 1: Delete 'However,' and begin sentence with 'Remote sensing'

Line 2: 'Significant sources of error include….'

Line 7: Data are always plural, therefore modify sentence here as follows: 'by using sensor data that have geometric, radiometric, temporal resolutions appropriate…..'

Page 3

Paragraph 1

Line 2: 'and Fire CCI but these provide different burned area estimates……..' the rest of the sentence is vague. What are these data characteristics and applied methods you are referring to?

Line 6: '….but their limited temporal resolution (16 days) reduce their utility in the savanna and pastureland areas which regenerate rapidly following fires. In contrast, the MODIS sensor…..'

Line 10: Modify sentence beginning here as follows: 'Other sensors with similar high temporal resolution as MODIS include VIIRS and PROBA-V [39].' You should mention the temporal and spatial resolution of these sensors.

Paragraph 2

Line 1: Combining information from different sensors can….'

Line 2: Explain what you mean by 'systematic burned area product'.

Lines 4-6: Delete the first six words in the sentence beginning here. Begin sentence as follows 'An alternative approach to reduce the dimensionality of image data and enhance specific information for digital interpretation is the use of fraction images [40,41].'

Lines 7-10: '…are generated using Linear Spectral Mixing Model (LSMM).' Next sentence needs to be elaborated and meaning clarified.

Paragraph 3

Line 1: Begin sentence as 'The main objective of this research to propose a new method to map burned areas using multisensor fraction images by taking advantage….'

Line 3: '……..Landsat OLI images for the year 2015.'

Lines 4-5: '…..area mappingobtained using Sentinel-2 data, and compare.

The 'Introduction' needs to be expanded to include a review of PROBA-V, VIIRS, and Sentinel-2 sensors, as they are used in mapping wildfires. This will help justify their use in your study.

Material and Methods

Study Area

Paragraph 1

Line 1: The study area is Mato Grosso State (Figure 1), the third largest state in Brazil located in the central-west region and covering an area approximately 904,984 km2. According to the Brazilian National Classification System [42], Mato Grosso State consists of three biomes:………'

Lines 5-6 : You list 'climate, terrain relief, precipitation systems, and length of annual seasons' as explanations for the complex biodiversity of Mato Grosso State. But 'climate' will encompass both precipitation systems, and length of annual seasons'.

Line 6: Delete 'For instance,' and begin sentence starting here as 'The amazon consists of …….'

I have noticed you are frequently and incorrectly using the terms 'presents' or 'presented' when describing characteristics of a dataset or feature.

Paragraph 2

Line 1: Delete 'Furthermore,' and begin sentence with 'Mato Grosso Sate stands out as a center of Brazilian agribusiness tied to illegal deforestation. The state falls partly within the 'arc of deforestation' in the southern part of the Brazilian Legal Amazon, an area with one of the highest annual…..'

Lines 4-9: Not clear. Needs to be restructured and elaborated.

Page 4

Paragraph 1

Mato Grasso State is among the states that lead in the number of active fires detected [48]. For example, in the 1988-2020 period, 40,000 fires were detection in the state, accounting for approximately 27% of the total active fires observed in the Legal Amazon [48]'. Most of the active fires occurred between June and November, a period with the least precipitation received'.

Lines 5-6: Sentence here beginning with 'In the study area, fire is used….' Needs to be deleted as it was previously stated at the top of page 2.

This is as far as I will go in specific comments as there many more to continue listing in this already long review. I will point you to the last paragraph of page 16 and ask you elaborate on how you arrived at the conclusion that 'a shift towards more intensive agriculture may result in a decrease in fires'. Also, you refer to relationship between burned area and economic development. I believe this should be omitted from the paper because this is not something you investigated in your research or even mentioned in your literature review. Would you also address what you considered limitations of your study and make recommendations for future studies in the conclusions section?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

My comments have been well addressed. I recommend the publication of the revised version. 

Author Response

REVIEWER 1

Comments and Suggestions for Authors

My comments have been well addressed. I recommend the publication of the revised version. 

Response: Thanks for the recommendation to publish our paper. Your comments and suggestions included in the revised version really improved our manuscript.

Author Response File: Author Response.pdf

Reviewer 2 Report

Dear Authors. Thank you for the clarifications and for the changes in the manuscript. Now the method is much clearer. 

I have only one more suggestion. In the 2.2. there are links to the access to all data except Sentinel-2. I would suggest to add it. 

Author Response

REVIEWER 2

Comments and Suggestions for Authors

Dear Authors. Thank you for the clarifications and for the changes in the manuscript. Now the method is much clearer.

Response: Thanks for your comments and suggestions that really improved our manuscript.

I have only one more suggestion. In the 2.2. there are links to the access to all data except Sentinel-2. I would suggest to add it.

Response: Thanks for the suggestion. The link for Sentinel-2 was included now in the revised manuscript. (https://jeodpp.jrc.ec.europa.eu/forobs/sentinel.py).

Author Response File: Author Response.pdf

Reviewer 3 Report

I commend the authors for their hard work. The manuscript is much improved. I have some minor issues that I paste below.

Line 173: The 'N' in 'nadir' not be capitalized.

Lines 174-175: Suggested rephrasing: In this study we used 5-day Top-Of-Canopy PROBA-V image composite acquired at a spatial resolution of 100 m for the period between May and October 2015.

Lines 176-177: Link given is broken

Line 178: What is this 'program developed'? Is it open-source and available for other interested users?

Lines 183-185: We downloaded 153 VIIRS daily reflectance (VNP09GA) images acquired between May and October 2015 from the US Geological survey at https://e4ftl01.cr.usgs.gov/VIIRS/VNP09GA.001/. The downloaded images consisted of three spectral bands, I1 (red, 600-680 nm), I2 (NIR, 846-885 nm), and I3 (MIR, 1580-1640 nm), and had a spatial resolution of 500 m. The 153 daily reflectance images were later composited into a cloud-free monthly VIIRS surface reflectance images using a script in Google Earth Engine and image quality flags 1 (QF1) layer.

Line 193: 30 m spatial resolution does not apply to the panchromatic band.

Lines 205-207: But isn't August-November the same period you obtained your Landsat OLI images? How does unavailability of Landsat OLI and Sentinel-2 images for the May-July period impact your results?

Line 209: Delete 'problem'

Lines 228-229: Please elaborate as it is not clear what you mean.

Lines 231-232: Meaning not clear.

Lines 233-234: '…event by relating fire scars to thermal emissions…'

Lines 239-241: Please rephrase and correct grammar.

Line 247: Delete 'the' before '2015' and 'an' before 'auxillary'.

Line 248: '..data were…'

Line 250: Delete 's' in 'millions'.

Please explain product by Arai et al. Was this a thematic crop map or was it a multispectral image?

Line 266: Delete 'Following' and begin sentence ' Using these fraction images……'. Please clarify the rest of the sentence.

Line 227: Delete 'Then' and begin sentence as 'Burned areas were obtained through digital classification…..'

Line 311: Rephrase and clarify meaning. Begin ' In order to delineate burned areas in this study, 381 randomly distributed points in Sentinel-2 ….' I am not clear what the granules you are referring to are.

Line 249: Not clear what you mean by 'mapped by the final composite'.

Line 374: 'cells had values…'

Lines 396-399: Please elaborate and clarify meaning.

Line 407: '…burned area obtained using the proposed…'

Lines 408-410: Not clear what you mean by 'the achievement percentage'

Lines 410-412: '…where there is a burned area detected but no active fire observed, PROBA-V produced the highest commission error (51.27%) while the Fire CCI data had the lowest (8.79%).

Lines 414-415: What does 'final classification results' refer to?

Lines 419-421: Begin 'There were substantial differences in mapped areal extent of burned areas across the various datasets…'

Lines 427-428: Fix grammar and clarify meaning.

Lines 453-455: Fix grammar and clarify meaning.

Lines 477-478: 'However, the areal extents of burned areas were very different….'

Line 479: '…maps in this study…'

Lines 490-491: Delete the parenthesis and words in it.

Line 481: Lines 505-506: Fix grammar.

Lines 521-522: 'Cloud shadows, topographic shadowing, and water bodies have similar spectral characteristics as burned areas.'

Line 523: 'other features that can be confused with burned areas include clearcuts, ….' 'land conversion' is vague. What do you have in mind here?

Line 539: Replace 'still presented' with 'resulted in'

Line 591: Not clear what you mean by 'confident estimates of uncertainties'.

Line 602: 'OLI sensor resulted in the …'

Author Response

Please see the attachment.

Author Response File: Author Response.pdf