Wetland Fire Assessment and Monitoring in the Paraná River Delta, Using Radar and Optical Data for Burnt Area Mapping

Round 1

Reviewer 1 Report

After reading the title of this manuscript, "Wetland fire assessment and monitoring using radar and optical data integration for burnt area mapping", we would expect to find a paper where the information from optical and radar sensors is used in a fused way. Where both complement each other so that the final burnt area product obtained would be better than the one generated by each of them independently. And this is not the case.

The authors have also been very concerned about masking the water sheets present in the study area because of their possible spectral signatures similar to that of the burned areas. But, they do not use radar data to improve their detection. They only suggest in conclusions and add, without foundation, in the abstract, that the radar data could help to generate these water masks. Furthermore, in my view, they have relied on the fact that the Sentinel-2 optical products used are cloud-free. But what guarantees do they have of this, especially in humid regions, where a lot of residual clouds are bound to persist? The interesting thing would have been to use the radar data precisely to "see" the surface when there are clouds and, to complete this microwave data with the optical data to obtain the burnt area. In reality, what they do is a comparison (not integration) of two different burnt area products with different sensors (optical versus radar). The only thing they do is to co-locate the images from both sensors in the same file or georeferenced dataset through a SNAP tool.

The chosen study area is, without a doubt, of great interest, as is the idea of mapping the burned area in it due to the problems of flooded and wet zones found there. Merging optical and radar sensors, in this case, could undoubtedly provide interesting results for experts in this field and even generate value-added products for the competent authorities, but this is not the case in this manuscript.

Another separate mention refers to the way the manuscript is written, its structure, and the English. In many cases, I had to reread up to 3 times the same paragraph trying to understand what they meant. This aspect should be corrected in a possible resubmission of the manuscript.

I would also recommend authors read carefully the publisher's suggestions as to what the sections of the manuscript should include. For example, the introduction certainly lacks the state of the art on this topic and a review of previous similar studies, especially those that refer to the fusion of optical and radar data for mapping burned areas. Also, the objective(s) should be right at the end of the Introduction and not in the middle.

Regarding the Materials and Methods section, it is acceptable to start with the study area, but from there, I would recommend the authors briefly present to the readers (with the help of a flowchart) a summary of the whole methodology, from the data used to the assessment of the accuracy of the resulting maps. The following sub-sections would be then detailed and explained.

I could recount numerous fails that I came across in this review. I would point out, at least, that the training and validation sets they have used are not shown in any place, number, total area, or location ..... and have always been done according to the human eye criterion. This point is another major weakness of the present study.

For all these reasons, I cannot recommend this manuscript for publication in Fire.

Author Response

REVIEWER 1

R1 1- After reading the title of this manuscript, "Wetland fire assessment and monitoring using radar and optical data integration for burnt area mapping", we would expect to find a paper where the information from optical and radar sensors is used in a fused way. Where both complement each other so that the final burnt area product obtained would be better than the one generated by each of them independently. And this is not the case.

The topic of the work and the procedural contents follow the protocols of similar works in the literature. There is no mention of data fusion or hybridization at any time, it is only a question of comparing the physical sensitivity of the radar data with the biological sensitivity of the optical data. However, for a better understanding, we changed the title of the paper to “Wetland fire assessment and monitoring in the Parana River Delta using radar and optical data for burnt area mapping”. The originality of the manuscript is based on establishing the advantages and disadvantages of multisensors in the application on wetland fires. The work shows similar cartographic evaluations of the digital classifications made with radar (Sentinel 1) and optical (Sentinel 2), with the same spatial resolution, taken individually.

R1 2- The authors have also been very concerned about masking the water sheets present in the study area because of their possible spectral signatures similar to that of the burned areas. But, they do not use radar data to improve their detection. They only suggest in conclusions and add, without foundation, in the abstract, that the radar data could help to generate these water masks. Furthermore, in my view, they have relied on the fact that the Sentinel-2 optical products used are cloud-free. But what guarantees do they have of this, especially in humid regions, where a lot of residual clouds are bound to persist? The interesting thing would have been to use the radar data precisely to "see" the surface when there are clouds and, to complete this microwave data with the optical data to obtain the burnt area. In reality, what they do is a comparison (not integration) of two different burnt area products with different sensors (optical versus radar). The only thing they do is to co-locate the images from both sensors in the same file or georeferenced dataset through a SNAP tool.

Regarding the problem of water masks. The reviewer has probably misunderstood the use of optical masks, via the NDWI algorithm (Sentinel 2), with radar masks (Sentinel 1). The latter is considered an original contribution of this paper, which makes it possible to perfectly visualize (in the figures) the hydrological dynamics in the Paraná River; this occurs due to the penetrability of the C band in the wetland herbaceous grasslands. Its use obviates the problem of cloudiness which, as we know, is not solvable in optical systems. We value, as an important part of the work, the results obtained with the use of masks. On the other hand, regarding the use of the SNAP tool (collocate), it is precisely a very useful tool that allows us to visualize the optical and radar data as shown in figures 8A-B; this methodology is correct, and is validated by international bibliography.

R1 3- The chosen study area is, without a doubt, of great interest, as is the idea of mapping the burned area in it due to the problems of flooded and wet zones found there. Merging optical and radar sensors, in this case, could undoubtedly provide interesting results for experts in this field and even generate value-added products for the competent authorities, but this is not the case in this manuscript.

With all due respect, we consider this observation to be unfair since never before, as in this manuscript, has the cartography of the Paraná River downspout been shown as one of the causes of extensive burning to obtain revenues from extensive cattle ranching in the area. This work contributes to a (methodological) clarification of the issue of the flood/fire relationship in the Paraná wetland, revaluing the use of radar as shown in figures 10A-D and 11A-F.

 

R1 4- Another separate mention refers to the way the manuscript is written, its structure, and the English. In many cases, I had to reread up to 3 times the same paragraph trying to understand what they meant. This aspect should be corrected in a possible resubmission of the manuscript.

English will always be a weak point for those who, like us, are not native English speakers. Even so, the manuscript has been checked again, and an improvement has been made through an expert translator. We hope that, in its current form, it is better written.

R1 5- I would also recommend authors read carefully the publisher's suggestions as to what the sections of the manuscript should include. For example, the introduction certainly lacks the state of the art on this topic and a review of previous similar studies, especially those that refer to the fusion of optical and radar data for mapping burned areas. Also, the objective(s) should be right at the end of the Introduction and not in the middle.

The estate of the art on this paper includes different subjects, some of them are geographical and other are methodological. We know that, for both, there are a lot of modern bibliography, so we select part of it, and now with the corrections, we add some new. Objective was moved to the end of the introduction.

R1 6- Regarding the Materials and Methods section, it is acceptable to start with the study area, but from there, I would recommend the authors briefly present to the readers (with the help of a flowchart) a summary of the whole methodology, from the data used to the assessment of the accuracy of the resulting maps. The following sub-sections would be then detailed and explained.

Figure 2 preset the workflow analysis.

R1 7- I could recount numerous fails that I came across in this review. I would point out, at least, that the training and validation sets they have used are not shown in any place, number, total area, or location ..... and have always been done according to the human eye criterion. This point is another major weakness of the present study.

Tables 3 and 4 show the evaluations of the classifications, these are not shown in the figures as they constitute a regular placement matrix of 200 points in 2 AOIs. Changes were made in the methodology to better explain what was done.

R1 8- For all these reasons, I cannot recommend this manuscript for publication in Fire.

 

Author Response File: Author Response.docx

Reviewer 2 Report

Line 32-33 The sentence is not very clear. Try to re-write this part.

Add some numerical results in the abstract.

The line space, instead of introduction part is not in accordance with the journal’s specs.

Line 50. These advantages have been exploited in the post wildfire assessment and management (Stefanidis et al., 2022; Gemitzi and Koutsias, 2022).

Stefanidis, S., Alexandridis, V., & Mallinis, G. (2022). A cloud-based mapping approach for assessing spatiotemporal changes in erosion dynamics due to biotic and abiotic disturbances in a Mediterranean Peri-Urban forest. CATENA, 218, 106564.

Gemitzi, A., & Koutsias, N. (2022). A Google Earth Engine code to estimate properties of vegetation phenology in fire affected areas–A case study in North Evia wildfire event on August 2021. Remote Sensing Applications: Society and Environment, 26, 100720.

 

Line 57. The destruction of forest vegetation highly influence directly influence the footprint of wildfires of erosion regulation services of forests (Stefanidis et al., 2022; Silvestro et al., 2021).

 

Stefanidis, S., Alexandridis, V., Spalevic, V., & Mincato, R. L. (2022). Wildfire Effects on Soil Erosion Dynamics: The Case of 2021 Megafires in Greece. Agriculture & Forestry, 68(2), 49-63.

Silvestro, R., Saulino, L., Cavallo, C., Allevato, E., Pindozzi, S., Cervelli, E., ... & Saracino, A. (2021). The Footprint of Wildfires on Mediterranean Forest Ecosystem Services in Vesuvius National Park. Fire, 4(4), 95.

Better state the research gap in this research.

Fig 1. Add a base map with label in the figure 1.

Line 115. Do not use capital letters for the sub-chapter heading.

Line 202. Please justify the selection of the used vegetation index (a small literature review).

 

 

 

 

 

 

 

Author Response

REVIEWER 2

 

-Line 32-33 The sentence is not very clear. Try to re-write this part.

 

Done. Also English grammar was checked for all the manuscript.

 

-Add some numerical results in the abstract.

 

Numerical data about the burned surface was added in the abstract.

 

-The line space, instead of introduction part is not in accordance with the journal’s specs.

 

Thanks for the comment. Formal specifications issues of submitting guidelines were checked and specific problems found were corrected. Sometimes the use of templates are not as easy as one want, and final corrections have to be done at the end of the editing process.

 

-Line 50. These advantages have been exploited in the post wildfire assessment and management (Stefanidis et al., 2022; Gemitzi and Koutsias, 2022).

 

We appreciate too much the bibliographic contribution which is specific and new; we include the citations in the new version.  

 

-Line 57. The destruction of forest vegetation highly influence directly influence the footprint of wildfires of erosion regulation services of forests (Stefanidis et al., 2022; Silvestro et al., 2021).

 

Same than previous comment. 

 

-Better state the research gap in this research.

 

Fig 1. Add a base map with label in the figure 1.

 

Unfortunately, we were unable to make the requested change in figure 1 at this time, but we think that the figure present the base map (Argentina), and in the detailed portion coordinates are present.

 

-Line 115. Do not use capital letters for the sub-chapter heading.

 

The request was corrected and the manuscript was controlled from the author's rules.

 

-Line 202. Please justify the selection of the used vegetation index (a small literature review).

 

The indices were justified in section 2.3.1 using bibliographical references. New references to NDVI were added as it was not previously justified. Some of the references are our own, and show our previous experience using indexes in the area. We use those that best fit; In general, all the indices used are widely accepted in the academic literature.

Author Response File: Author Response.docx

Reviewer 3 Report

The authors did an excellent job motivating their research objective asserting the importance of using digital land degradation mapping and other initiatives that enhance the tracking of land degradation. Overall, the results seem to be reasonable. The authors were able to justify the low overall accuracy of detecting burned and not-burned spots. The complexities and heterogeneity imposed post-fire events affect the accuracy of the spectral indices. However, the following suggestions are recommended to be resolved.

The sentence that starts at line 32 is redundant. 

Some of the references cited in the paper are written in Spanish. 

The authors stated that data was obtained at different orbits. I think it is worth noting to mention that these satellites are in a sun-synchronous orbit. 

The use of Google Earth and Bin images need to be justified. The two frameworks are known to have a delay in updating their imagery. Some imagery dates could have a delay of years before it gets updates. The authors provided a date for retrieving the information form Google Earth but they did not state the latest imagery update. The authors need to state that it was used solely for polygon representation testing and not for comparative purposes if that is the case.  

The heading 2.3 methodological issues does not include any issues associated with the methodological framework. The heading should be renamed appropriately to reflect the content.

In line 184, what does the " them" refer to? Does it refer to the clouds?

What is the justification for using the normalized difference vegetation index and not using other kinds of vegetation indexes?

All the indexes used in the analysis have mathematical formalism. The authors only included the formula for the separability index. What ranges of wavelengths are used in these indexes?

In line 180, The same information was already mentioned in the previous lines. 

The authors frequently introduce acronyms without stating what they refer to. For example, Q3 and Q4 mentioned in line 210 are defined until later in the document. 

AOI was mentioned early in the document but what it refers to was not mentioned until Figure 9.

The authors mentioned that the thresholds for the spectral indices were determined based on visual inspection of fire scars. How was the visual inspection carried out?

In line 53, there is a grammatical error with the word “monitoring”. 

Author Response

REVIEWER 3

R3 1- The authors did an excellent job motivating their research objective asserting the importance of using digital land degradation mapping and other initiatives that enhance the tracking of land degradation. Overall, the results seem to be reasonable. The authors were able to justify the low overall accuracy of detecting burned and not-burned spots. The complexities and heterogeneity imposed post-fire events affect the accuracy of the spectral indices. However, the following suggestions are recommended to be resolved.

R3 2- The sentence that starts at line 32 is redundant. 

Many different sentences all along the manuscript were checked also erasing, or modifying, redundant comments. English grammar was controlled.

R3 3- Some of the references cited in the paper are written in Spanish. 

Yes, that’s, correct, the reason of this is because the references are cited as they were written, in Spanish, and in the way they can be found in the internet. In all cases we follow the original reference configuration.

R3 4- The authors stated that data was obtained at different orbits. I think it is worth noting to mention that these satellites are in a sun-synchronous orbit. 

The word was removed. Although we want to clarify that the initial allusion was done referring to the polarization of radar products; it was mentioned as “descending right orbit”.

R3 5- The use of Google Earth and Bin images need to be justified. The two frameworks are known to have a delay in updating their imagery. Some imagery dates could have a delay of years before it gets updates. The authors provided a date for retrieving the information form Google Earth but they did not state the latest imagery update. The authors need to state that it was used solely for polygon representation testing and not for comparative purposes if that is the case.  

The entire paragraph was removed, since these images were only used to visually corroborate the different landcovers. Neither the images of Google Earth, nor those of Big are part of the analyzes carried out for burned area mapping.

R3 6- The heading 2.3 methodological issues does not include any issues associated with the methodological framework. The heading should be renamed appropriately to reflect the content.

The heading was changed for “workflow”.

R3 7- In line 184, what does the "them" refer to? Does it refer to the clouds?

Yes, it refers to the clouds. Even so, the sentence was adjusted to make it more readable.

R3 8- What is the justification for using the normalized difference vegetation index and not using other kinds of vegetation indexes?

Various indices were used in the work (NDVI, NDWI, NBR, and NBR 2), but specifically the NDVI was preferred over others such as SAVI since, it is a classic index and, on the other hand, because the group has been working in the area and has tried different indices and the NDVI shows an appropriate setting for the characteristics of this wetland environment.

R3 9- All the indexes used in the analysis have mathematical formalism. The authors only included the formula for the separability index. What ranges of wavelengths are used in these indexes?

The NDVI is the normalized difference of the red and the infrared band, calculated as NDVI = (NIR-RED) / (NIR+RED). Using Sentinel-2 correspond Band 8 (NIR) and Band 4 (Red).

R3 10- In line 180, The same information was already mentioned in the previous lines. 

Erased.

R3 11- The authors frequently introduce acronyms without stating what they refer to. For example, Q3 and Q4 mentioned in line 210 are defined until later in the document. 

It was corrected. Now the acronyms are mentioned with their full name the first time and, subsequently, in a reduced form. The added phrase “the focus was on quarters 3 and 4 (Q3 and Q4)”.

R3 12- AOI was mentioned early in the document but what it refers to was not mentioned until Figure 9.

Problem fixed. The AOI is now firstly mentioned in the methodology in point 2.3.4

R3 13- The authors mentioned that the thresholds for the spectral indices were determined based on visual inspection of fire scars. How was the visual inspection carried out?

Bibliography of the group that has already worked on the determination of this limit has been added. Even so, we want to comment that the threshold on the NBR2 index was determined using a frequency distribution for the different existing burned covers in the area (different types of grasslands and shrublands), considering that the carbon signal may be different for different landcovers. The index varies between -1 and 1, and if it is less than -0.2 (used by us), it rules out anomalous values ​​that are above the 95% percentile.

R3 14- In line 53, there is a grammatical error with the word “monitoring”. 

Done.

 

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The authors have made some changes to the wording and structure of the manuscript that have only slightly improved its current presentation. However, there are still underlying problems that have not been addressed. The authors insist, in different parts of the text (from the abstract to the conclusions) on talking about synergy and/or integration of the Sentinel-2 optical sensor with the Sentinel-1 SAR. See, for example, lines 17 and 73, where they refer to "synergistic use of radar and optical data" and in lines 27, 182, 188, 198, 506, and 522, where they talk about "SAR and optical data integration". It is useful at this point to look up the meaning of these two words to show that this is not what the authors actually do.

Integration: the act of combining or adding parts to make a unified whole.

Synergistic: causing or involving synergy (= the combined power of working together that is greater than the power achieved by working separately).

I also cannot agree with the thresholds defined for the case of the burnt area with the NBR 2 index from "visual inspection". Have you not found any better statistical or physical/mathematical criteria for their definition? In the same line, how can both S1 and S2-derived products are visually assessed using 10m S2 images (lines 254 and 255)? I believe that there is still a lack of methodological rigor in the use of these satellite-derived products, as well as in the accuracy assessment of the burnt area maps.

Despite all the above, and acknowledging all the work and effort put in by the authors, I believe that if they can improve the key points that I discussed in my previous review, as well as the current one, the manuscript could be published in Fire after major revisions.