Assessing Wildfire Hazard in the Wildland–Urban Interfaces (WUIs) of Central Portugal
Round 1
Reviewer 1 Report
Dear authors!
I think your paper is fit to Forests journal, but at present version paper has some disadvantages.
Introduction
I think this is good variant of Introduction.
Materials and Methods
You considered probability in SEction 2.3 Wildfire hazard.
But I did not any information on ignition sources situated in forest covered or rural segment of the WUI. I think you must provide some information and speculations on this topic.
Table 2. How did you select Hazard scores for different hazard classes? Why did you not use regular class definition? What is variable x and y in Success curve trend?
What software used with ArcSDM toolbox? Please, fully describe mathematical content of this toolbox.
Results
Figure 9. Wildfire probability. Please, link this map with some information on ignition sources over considered territory.
Discussion
Please, make some speculations on ignition sources in this section.
Also, you can provide some discussion of WUI fires in context of advocacy programs within home owners like NFPA program in the USA.
Conclusion
Maybe, it will be suitable to mark out 3-4 key finding with corresponding conclusions in this section.
References
I think this section content and references quantity are enough for this paper.
No comments
Author Response
The authors are very grateful for all the comments and suggestions made to our paper, which really contributed to the improvement of the final quality of the article. All the suggestions were considered, and the changes made in the text are underlined in yellow.
Introduction
I think this is good variant of Introduction.
R: Thank you very much for the comment.
Materials and Methods
You considered probability in SEction 2.3 Wildfire hazard.
But I did not any information on ignition sources situated in forest covered or rural segment of the WUI. I think you must provide some information and speculations on this topic.
R: In Portugal, the number of ignitions is very high and occurs mainly in areas of higher population density and, possibly, into or near urban-forest interface areas. However, a small number of ignitions is responsible for the high number of burnt areas in the Central region of Portugal and these large forest fires (with burnt areas of 10,000, 20,000 ha) put people and their property at risk. Moreover, in Portugal, the ignition points are not precisely located. The location of these ignitions corresponds to the centroid of the parish where the fire started. Therefore, the data related to fire ignition to be used may be biased. For these reasons, we considered that the burnt area is the most important variable when analysing the forest fire risk (both hazard and vulnerability) in Portugal.
Table 2. How did you select Hazard scores for different hazard classes? Why did you not use regular class definition? What is variable x and y in Success curve trend?
R: As we mention in the text, the methodology was adopted from Oliveira et al., 2020. In this paper, as well as in ours, the final scores for classes result from the product of the susceptibility combined with probability to wildfire. The 5 classes were defined for the final model, adopting the following assumptions: (i) the ‘very high’ class had to validate 50% of the burnt area, and the ‘very low’ class could not incorporate the burnt area, in the years used for result validation, 2005 and 2020. The remaining classes were defined considering the following assumptions: (i) model data, (ii) validated burnt area, (iii) breaks in the model success curve and (iv) the trend lines of the segments between these breaks (Table 3 and Figure 5). The success curve results from the combination of the cumulative of burnt area (variable y) and hazard (variable x). Please, see the figure 5.
What software used with ArcSDM toolbox? Please, fully describe mathematical content of this toolbox.
R: The SDMtoolbox (v2.5 for ArcMap 10.0), a python-based ArcGIS toolbox, was applied to validate the model by calculating the AUC (Area Under the Curve). ROC toll and calculate ROC Curves and AUC values were used to calculate respectively ROC and AUC values. Additional information was included in the text.
Results
Figure 9. Wildfire probability. Please, link this map with some information on ignition sources over considered territory.
Discussion
Please, make some speculations on ignition sources in this section.
R: As the dependent variable is the burnt area, and not the number of ignitions, we considered that we should not address this issue in the results nor in discussion, as mentioned above.
Also, you can provide some discussion of WUI fires in context of advocacy programs within home owners like NFPA program in the USA.
R: Wildfires in Portugal in the context of political and advocacy programs are discussed between lines 424-458. Please see the discussion section.
Conclusion
Maybe, it will be suitable to mark out 3-4 key finding with corresponding conclusions in this section.
R: The conclusion has been revised in order to identify the main results obtained, as well as the most innovative aspect of the research. Please see underlined text.
References
I think this section content and references quantity are enough for this paper.
R: Thank you for your comment.
Author Response File: 
Author Response.docx
Reviewer 2 Report
I have published on WUI and wildfires and I find this study to be comprehensive and a very useful and scientifically sound contribution. I am also a geographer and I can see this is a geographically solid analysis definitely worthy of publication.
I have a few suggestions for possible improvement.
First, I note that this study builds on the methodology used in Oliveira et al. (2020). For the reader’s benefit, it would be best to explain this in the last paragraph of the Introduction. Tell us in a summary sentence or two what is the new contribution of this study relative to Oliveira et al. How is this study distinguished and what is the new contribution? It could be useful to explain the new information relative to Oliveira et al. also in the Discussion. I seoe that you mention agreement with Oliveira et al. on lines 370-371, but it would be nice to hear more about what you learned beyond what Oliveira et al. found in a sentence or two.
Second, I note that Geographers make excellent maps, and I like nearly all of them. But, I would like to see every one of them enlarged, because the printed names on the maps are now difficult to read. Please enlarge them to fill or nearly fill the entire available width which is the right 2/3 of each page. Of course, it would be convenient to have two maps on one page, so perhaps each map cannot quite fill the right 2/3 of each page in order to fit two of them on a page along with their Figure Heading. There are also a few things about particular maps:
Figure 3. Lots of classes make this map unlikely to successfully show the reader everything. You might remedy this by choosing to highlight the fewer classes that are most important to the story, perhaps with brighter or more prominent colors.
Figure 7, 12. These maps both look like areas (polygons) are shaded, but the Legend shows lines. The lines in the Legend are difficult to match up with the shaded areas on the map. I don’t know quite how to fix this. You could show shaded polygons (e.g., circles, squares) in the Legend, but if that is not possible, you maybe could thicken the lines in the Legend so more color is showing. Maybe if this is enlarged, it will not need any fixing....?
Figure 9. Perhaps change the gradient in the shading so it goes from green to red, so there is more distinction evident in the map. More distinction is needed so it is easier to see the gradient in probability.
Figure 15. The colors/shading used are just not bright and clear. Contrast this bar chart with Figure 16, which is excellent (maybe high and very high are a little hard to distinguish).
Minor fixes:
Line 32. “vicinity to settlements” should be “vicinity of settlements”
Line 122. I don’t know, I am used to seeing Latin names in italics, so Pinus pinaster, but I don’t know how the Forests journal does these.
Line 123. It might be good to explain what “scrubland” is for foreign readers. Just a sentence should be enough. Is it native shrubs?
Line 125. Can you tell us what are some of the species that dominate the “invasive species forests?” Isn’t Eucalyptus an invasive species too?
Line 128. I notice throughout the paper you have it as “km2" not km2 which I am used to seeing. Do you want to change this or is it OK as km2?
Table 1. “Other groforestry” maybe should be “Other agroforestry”?
Line 166+ on P. 6 at the bottom. Types of wildland urban Interface. Please move this so it is all on one page. Should this be Table 2 not Table 1? Check the references to this table in the text.
Line 171. Please say what the probability refers to. Probability of what? Please do the same on Lines 208 and 210. Could also repeat the meaning of Susceptibility in the same way. These are just to make it easier for the reader to recall what you are doing.
Line 176. Did you explain what the source was for the “areas burned between 1975 and 2020"?
Line 182. Lri should be LRi. Also please explain what i is. If it is an index from i to the number of classes, maybe it should be LR with a subscript i and the same for Si and Ni? I am just guessing.
Line 207-208. Here, too, I am guessing you are adding up across all the i = 1 to n classes? Please explain how the “sum of scores” uses the nomenclature you are using.
Line 212. You might tell us the pixel resolution for the raster (e.g., 30 m).
Line 221. I am not sure I understand what you mean by “had to validate 50% of burnt area” here.
Lines 223-225. This “interactive process” is not something that the reader could repeat with just this description. I think it is better to make this clearer so someone else could do the same thing.
Figure 5. I see “Modelo de Hazard” do you want this to be in English?
Table 4. I am unsure why the entries for LRi are given as 0,365 etc., not 0.365?? Same question for Table 5 too.
Line 282. Should cordilheira cental have the first letters capitalized?
Lines 286 and 287. Why is “high” not capitalized, but Low is?
Lines 410, 414. Should “defendable” be “defensible”?
Line 465. Could remove “a” before “huge”
No comments. English is pretty good, minor little things not worth fixing...
Author Response
The authors are very grateful for all the comments and suggestions made to our paper, which really contributed to the improvement of the final quality of the article. All the suggestions were considered, and the changes made in the text are underlined in yellow.
I have published on WUI and wildfires and I find this study to be comprehensive and a very useful and scientifically sound contribution. I am also a geographer and I can see this is a geographically solid analysis definitely worthy of publication.
R: Thank you very much for the kind and constructive comment.
I have a few suggestions for possible improvement.
First, I note that this study builds on the methodology used in Oliveira et al. (2020). For the reader’s benefit, it would be best to explain this in the last paragraph of the Introduction. Tell us in a summary sentence or two what is the new contribution of this study relative to Oliveira et al. How is this study distinguished and what is the new contribution? It could be useful to explain the new information relative to Oliveira et al. also in the Discussion. I seoe that you mention agreement with Oliveira et al. on lines 370-371, but it would be nice to hear more about what you learned beyond what Oliveira et al. found in a sentence or two.
R: The reference to the use of the Oliveira et al. methodology is now mentioned in the conclusion. In addition, the conclusion has been revised in order to identify the main results obtained, as well as the most innovative aspect of our research in relation to the work carried out by Oliveira et al. Please see underlined text.
Second, I note that Geographers make excellent maps, and I like nearly all of them. But, I would like to see every one of them enlarged, because the printed names on the maps are now difficult to read. Please enlarge them to fill or nearly fill the entire available width which is the right 2/3 of each page. Of course, it would be convenient to have two maps on one page, so perhaps each map cannot quite fill the right 2/3 of each page in order to fit two of them on a page along with their Figure Heading. There are also a few things about particular maps:
Figure 3. Lots of classes make this map unlikely to successfully show the reader everything. You might remedy this by choosing to highlight the fewer classes that are most important to the story, perhaps with brighter or more prominent colors.
Figure 7, 12. These maps both look like areas (polygons) are shaded, but the Legend shows lines. The lines in the Legend are difficult to match up with the shaded areas on the map. I don’t know quite how to fix this. You could show shaded polygons (e.g., circles, squares) in the Legend, but if that is not possible, you maybe could thicken the lines in the Legend so more color is showing. Maybe if this is enlarged, it will not need any fixing....?
Figure 9. Perhaps change the gradient in the shading so it goes from green to red, so there is more distinction evident in the map. More distinction is needed so it is easier to see the gradient in probability.
R: All maps were redone with high resolution. The legends have been enlarged and suggestions regarding colour gradation have also been taken on board. We think that the overall quality of the maps has improved significantly.
Figure 15. The colors/shading used are just not bright and clear. Contrast this bar chart with Figure 16, which is excellent (maybe high and very high are a little hard to distinguish).
R: Based on your suggestion, the bar chart of figure 15 was modified.
Minor fixes:
Line 32. “vicinity to settlements” should be “vicinity of settlements”
R:The text was corrected accordingly.
Line 122. I don’t know, I am used to seeing Latin names in italics, so Pinus pinaster, but I don’t know how the Forests journal does these.
R: Pinus pinaster is the correct form.
Line 123. It might be good to explain what “scrubland” is for foreign readers. Just a sentence should be enough. Is it native shrubs?
R: Yes, it is native shrubs as a consequence of land abandonment or wildfire recurrence. In the text native shrubs was included.
Line 125. Can you tell us what are some of the species that dominate the “invasive species forests?” Isn’t Eucalyptus an invasive species too?
R: In Portugal, eucalyptus trees are not considered an invasive species, as they are mostly planted by man. Among the most important invasive species, Acacia stands out. This information was included in the text.
Line 128. I notice throughout the paper you have it as “km2" not km2 which I am used to seeing. Do you want to change this or is it OK as km2?
R: km2 is the correct form. Changes was made in the text.
Table 1. “Other groforestry” maybe should be “Other agroforestry”?
R: Word corrected.
Line 166+ on P. 6 at the bottom. Types of wildland urban Interface. Please move this so it is all on one page. Should this be Table 2 not Table 1? Check the references to this table in the text.
R: Tables 2 and 3 have been correctly numbered.
Line 171. Please say what the probability refers to. Probability of what? Please do the same on Lines 208 and 210. Could also repeat the meaning of Susceptibility in the same way. These are just to make it easier for the reader to recall what you are doing.
R: The text has been clarified to mention what probability and susceptibility refer to.
Line 176. Did you explain what the source was for the “areas burned between 1975 and 2020"?
R: The source of burned areas are included in the text. See the text highlighted in yellow.
Line 182. Lri should be LRi. Also please explain what i is. If it is an index from i to the number of classes, maybe it should be LR with a subscript i and the same for Si and Ni? I am just guessing.
R: Please, see the text were i is clarified.
Line 207-208. Here, too, I am guessing you are adding up across all the i = 1 to n classes? Please explain how the “sum of scores” uses the nomenclature you are using.
R: We believe that your doubt about the sum of scores is clarified with follow phase, which was included in the paper: “The final product of the susceptibility, for each pixel, results from the sum likelihood ratios associated to its classes in the variables landcover, slope, and elevation.”
Line 212. You might tell us the pixel resolution for the raster (e.g., 30 m).
R: All of the maps have a spatial resolution of 10 by 10 m. Please, see the text.
Line 221. I am not sure I understand what you mean by “had to validate 50% of burnt area” here.
R: In fact, one of the model's criteria is that 50% of the pixels burnt in the years used to validate the results (2005 and 2020) obtained by the model coincide with the High hazard class, otherwise the model (through the variables used) does not fulfil the necessary requirements for its validation.
Lines 223-225. This “interactive process” is not something that the reader could repeat with just this description. I think it is better to make this clearer so someone else could do the same thing.
R: We agree that ““interactive process” is not adequate, thus the text was changed.
Figure 5. I see “Modelo de Hazard” do you want this to be in English?
R: Modelo of hazard was changed in the figure by Hazard model.
Table 4. I am unsure why the entries for LRi are given as 0,365 etc., not 0.365?? Same question for Table 5 too.
R: The LRi values were given in x.xx, thus we change the comma by point in the table 4 and 5.
Line 282. Should cordilheira cental have the first letters capitalized?
We agree and proceed accordingly.
Lines 286 and 287. Why is “high” not capitalized, but Low is?
R: In the text, we have chosen to put all hazard classes in lower-case letters.
Lines 410, 414. Should “defendable” be “defensible”?
R: We agree and proceed accordingly. So defendable was changed by defensible.
Line 465. Could remove “a” before “huge”.
R: We proceed accordingly.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
Dear authors!
Thank you for your revision, but your paper is still needed the improvements.
Materials and Methods
Put your answer in paper text (I did not find this inclusion):
In Portugal, the number of ignitions is very high and occurs mainly in areas of higher population density and, possibly, into or near urban-forest interface areas. However, a small number of ignitions is responsible for the high number of burnt areas in the Central region of Portugal and are these large forest fires (with burnt areas of 10,000, 20,000 ha) that put people and their property at risk. Moreover, in Portugal, the ignition points are not precisely located. The location of these ignitions corresponds to the centroid of the parish where the fire started. Therefore, the data related to fire ignition to be used may be biased. For these reasons, we considered that the burnt area is the most important variable when analysing the forest fire risk (both hazard and vulnerability) in Portugal.
Put your answer in paper text (I did not find this inclusion):
The 5 classes were defined for the final model, adopting the following assumptions: (i) the ‘very high’ class had to validate 50% of burnt area, and the ‘very low’ class could not incorporate the burnt area, in the years used for result validation, 2005 and 2020. The remaining classes were defined considering the following assumptions: (i) model data, (ii) validated burnt area, (iii) breaks in the model success curve and (iv) the trend lines of the segments between these breaks (Table 3 and Figure 5).
Results:
I insist on inclusion of some ignition information to the paper.
The easiest way is to provide typical map with hotspots obtained from MODIS Terra/Aqua space device for some fire dangerouos season.
Discussion
You must extensevely describe limitations and disadvantages of your research in context of ignition sources. Moreover, you must suggest future researches to eliminate this disadvantage of your research.
Conclusion
I suggest to reorganize part of this section to numbered paragrahs with key findings and corresponding conclusions instead of solid text.
No comments
Author Response
Thank you for your revision, but your paper is still needed the improvements.
The authors are grateful for all the comments and suggestions made to our paper, which really contributed to the improvement of the final quality of the article. All the suggestions were considered, and the changes made in the text are underlined in green.
Materials and Methods
Put your answer in paper text (I did not find this inclusion):
In Portugal, the number of ignitions is very high and occurs mainly in areas of higher population density and, possibly, into or near urban-forest interface areas. However, a small number of ignitions is responsible for the high number of burnt areas in the Central region of Portugal and are these large forest fires (with burnt areas of 10,000, 20,000 ha) that put people and their property at risk. Moreover, in Portugal, the ignition points are not precisely located. The location of these ignitions corresponds to the centroid of the parish where the fire started. Therefore, the data related to fire ignition to be used may be biased. For these reasons, we considered that the burnt area is the most important variable when analysing the forest fire risk (both hazard and vulnerability) in Portugal.
Put your answer in paper text (I did not find this inclusion):
R: Please, see lines 214-222.
The 5 classes were defined for the final model, adopting the following assumptions: (i) the ‘very high’ class had to validate 50% of burnt area, and the ‘very low’ class could not incorporate the burnt area, in the years used for result validation, 2005 and 2020. The remaining classes were defined considering the following assumptions: (i) model data, (ii) validated burnt area, (iii) breaks in the model success curve and (iv) the trend lines of the segments between these breaks (Table 3 and Figure 5).
R: This information is included in the methodology, between lines 233-238
Results:
I insist on inclusion of some ignition information to the paper.
The easiest way is to provide typical map with hotspots obtained from MODIS Terra/Aqua space device for some fire dangerouos season.
R: We understand the insistence on the integration of ignitions in this analysis, but as we mentioned (now also in the text) because most ignitions do not originate fires and a small percentage are at the origin of large fires (and these are the ones that put people and property at risk) we considered not including them. On the other hand, the burnt area allows us to have longer data series. The use of data obtained from MODIS Terra/Aqua space device seems an excellent option to locate ignition hotspots, in a particular season, but makes it difficult to obtain long series, which allows a more accurate characterization of the ignition risk.
Discussion
You must extensevely describe limitations and disadvantages of your research in context of ignition sources. Moreover, you must suggest future researches to eliminate this disadvantage of your research.
R: Please see the last paragraph of the discussion. Although the limitations of this paper regarding the non-inclusion of ignitions in this work are not discussed, the problem of ignition risk appears as a working hypothesis in future research on forest fire risk. Moreover, this need is framed within a ministry resolution through a "Program to Reduce the Number of Fires".
Conclusion
I suggest to reorganize part of this section to numbered paragrahs with key findings and corresponding conclusions instead of solid text.
R: The conclusion has been reorganized and now identifies, at several points, the key findings of this research. Please see the text.
Author Response File: Author Response.docx
