Effects of fire danger indexes and land cover on fire growth in Peru
- 1US Forest Service, Rocky Mountain Research Station, Fire Sciences Laboratory, 5775 Highway 10 West, Missoula, MT, USA
- 2School of Environmental and Forest Sciences, University of Washington, Seattle, USA
- 3US Forest Service, International Programs, 1 Thomas Circle, NW Suite 400 Washington D.C., U.S.A.
- 1US Forest Service, Rocky Mountain Research Station, Fire Sciences Laboratory, 5775 Highway 10 West, Missoula, MT, USA
- 2School of Environmental and Forest Sciences, University of Washington, Seattle, USA
- 3US Forest Service, International Programs, 1 Thomas Circle, NW Suite 400 Washington D.C., U.S.A.
Abstract. Statistical analyses of wildfire growth are rarely undertaken, particularly in South America. In this study, we describe a simple and intuitive difference equation model of wildfire growth that uses a spread parameter to control the radial speed of the modeled fire and an extinguish parameter to control the rate at which the burning perimeter becomes inactive. Using data from the GlobFire project, we estimate these two parameters for 1003 large, multi-day fires in Peru between 2001 and 2020. For four fire-prone ecoregions within Peru, a set of 18 generalized linear models are fit for each parameter that use fire danger indexes and land cover covariates. Akaike weights are used to identify the best-approximating model and quantify model uncertainty. We find that, in most cases, increased spread rates and extinguish rates are positively associated with fire danger indexes. When fire danger indexes are included in the models, the spread component is usually the best choice. We also find that forest cover is negatively associated with spread rates and extinguish rates in tropical forests, and that anthropogenic cover is negatively associated with spread rates in xeric ecoregions. We explore potential applications of this model to wildfire risk assessment and burned area forecasting.
Harry Podschwit et al.
Status: open (until 12 Feb 2023)
-
RC1: 'Comment on egusphere-2022-742', Anonymous Referee #1, 05 Oct 2022
reply
Review of the article Effects of fire danger indexes and land cover on fire growth in Peru by Podschwit et al.
General comments
This article by Podschwit et al. introduces a novel and simple method to model wildfire growth. Specifically, a difference equation model estimating a spread and an extinguish parameter was described and generalized linear models were fit for each parameter which use fire danger indexes and land cover predictors. The method was tested using fire perimeter data from recent wildfires in four ecoregions in Peru. The approach is certainly interesting and the methodology is mostly comprehensible, with a few key points needing to be further clarified. The overall presentation of the results is sound and the common thread can be followed throughout the paper while language and readability are almost flawless. Here and there, rework is needed to clarify certain points which currently might confuse readers. Therefore I propose that the article can be accepted for publication in this journal after addressing some minor revisions.
Specific comments
While I understand the intention and the setup of the study, I think the methodology currently lacks some clarity. Firstly, Figure 1 makes it seem like direction matters to the approach but if I understood correctly, the extinguish parameter merely decreases the perimeter as a whole and not on a specific side of the circle. Speaking of the sector arc makes this further confusing but I understand that it needs to be calculated for solving the difference equation. I think the best way to avoid this confusion is to adjust Figure 1 and have the “fire” spread in all directions instead of just one. Overall, I think it would still be good if there is some information about the direction of spread (e.g. that you are not modeling it here and why).
At the end of the model description, it would be nice to also show the final difference equation (after Line 116). This might of course be trivial to some but I think others would appreciate to see the final equation directly and it would serve as a good end point of section 2.
The data selection and description is okay. Maybe it is a good idea to state which year the Nature Conservancy data for the ecoregion definition is from. Besides that, I wonder if it is possible to consider land cover changes throughout the study period. Linking a land cover map from 2009 to a fire event from 2019 could be problematic. I understand that the latest GlobCover dataset is only available for 2009 but other global datasets with higher spatial and temporal resolution might increase the validity of the analysis. I was also asking myself why the land cover data was only reclassified into two categorical values. I understand that it is a simple approach but this again raises the question why GlobCover was chosen specifically. At the very least, some more information about the land cover data should be provided (around Line 130).
One more parameter which can potentially also influence fire spread is forest structure (or forest composition). This seems to not be considered here. Is there a specific reason for that?
I would also suggest giving some more information as to how the elements of the GLM were chosen. It seems like an inverse-link and a Gamma density function work here and make sense but when reading the text, this assumption comes out of nowhere and it would increase the understanding of the modeling approach if this part was a bit more elaborated on.
To give the article a better structure, I suggest moving the results section into a separate chapter (e.g. a new chapter 4). Currently, the results are announced in chapter 3.4 but then the actual results follow in chapters 3.5-3.7. It would make much more sense to put them into a new chapter to separate them from the data and modeling section.
On a similar note, I suggest getting rid of chapter 3.7 and work the contents into chapter 3.6 where the same relationships are already discussed. Figure 7 should be kept and the growth curves should be discussed more thoroughly.
I think the discussion part is okay and covers the interpretation of results, potential flaws of the approach and possible future work. However, what I’m most concerned with is the explanation for the counterintuitive results of the relationships between the extinguish rates and the environmental variables. Your explanation sounds like this comes solely from the correlation with the spread rates. This sounds correct but it would mean that the extinguish parameter is not independently modeled and is therefore flawed. It would be great to get some more insights as to why the analysis was still carried through with this approach.
Finally, I’d like to make a general remark about the title and the contents of the article. The title made me expect an application of known methods to a specific study region. However, while reading it felt more like a methodological article which describes and evaluates a novel and simple approach for modeling fire growth. Maybe the title can be adjusted to accurately state what the key aim of this study was.
Technical comments
Line 24: The subordinate clause after the comma sounds a bit strange.
Line 72: The word “areal” should be changed as it can be easily confused with “radial”.
Lines 120/127/130: It’s always better to include a citation to datasets. Posting links is okay but if possible, add a citation to fully acknowledge the source.
Line 148: Maybe include a reference to the appendix already here.
Lines 149/154: There is no author called “Computing”. The author in this case is “R Core Team” – this should also be adjusted in the References section in Line 342.
Figure 6: While I understand how to interpret this figure, it can be confusing to other readers to have the ascending probability values on the y-axis (may seem like ERC for example always has a probability of below 0.1 while Intercept-only always has a probability close to 1.0). Additionally, the colors, especially the shades of blue, are very hard to distinguish. Consider choosing other colors or at least bigger “gaps” between the different shades.
Line 204: Remove the comma before the citation.
Line 213: I think it’s better to change “area burned” to “burned area”.
Figure 7: I had a hard time understanding the legend. Does the dashed line stand for 10% forest cover or for 10% anthropogenic cover? Or even for both scenarios together? Please make it clearer which land cover extreme corresponds to which line style.
Lines 289/290: This sounds like extreme fire events often occur in the conditions that are typical for other normal fires. So are the conditions now unique or not? Please rephrase this sentence.
Appendix A: The structure and syntax of the appendix is confusing. The subchapters are named A0.1 and A0.2 while the figures are then named A1 and B1. Please name the chapters and figures consistently.
Figures 1/3/4/6/A1/B1: It should be considered to add labels to every subplot instead of just one for all. This was already done in Figures 5 and 7 and improves readability as the reader doesn’t have to look at the whole figure to understand a subplot (e.g. if only looking at the lower left plot in Figure 3, it’s hard to infer that frequency is displayed on the y-axis).
Several occasions: It often seems that spaces between words in parenthesis are too big (see e.g. Lines 125/159/168/175/181/207). Please check the whole article for this.
-
RC2: 'Comment on egusphere-2022-742', Anonymous Referee #2, 19 Jan 2023
reply
The manuscript provides a simple wildfire growth model for four ecoregions in Peru. The model relates fire growth to fire danger indices and land cover through difference equation models that are parameterized to best fit 1003 large multi-day fires in terms of radial spread speed and perimeter length extinguish rates. The differences found for Andean (grass), Xeric, Dry Forest and Amazon Forest are calculated and discussed. Potential applications are alluded to.
General comments:
- The manuscript is well written and clearly discusses the materials.
- The method and calculations are well explained.
- As constructed, the validity of the approach is uncertain since no validation was attempted.
- Be careful in discerning where your model is giving you insights into fire behavior as opposed to where results are artifacts of the model constraints or assumptions.
Specific comments:
- The model makes two simplifying assumptions, namely that fire spreads at a constant rate from an ignition point, and that a constant length of the fire’s perimeter is extinguished after the first time step. This is an understandable expediency but how and when these assumptions might lead to erroneous results should be discussed. Basically, what sort of fire behavior or conditions would tend to ‘break’ this model?
- Provide more detail on the burned area product being used. The GlobFire data is derived from a 500m MODIS burned area product.
- Many fires over many years were used to derive the specific ecoregion models, which is good, but effectively what is developed are average values for each fire. Since daily spread is not examined, it would seem that basically, the final Area (km2) and Duration (days) are used to solve for an average spread rate (r) and fixed amount of perimeter being extinguished. The lack of daily progressions is clearly discussed in the Discussion section but more could be done to 1) support the approach, and 2) explain why it may not work so well for some ecoregions.
- Why were all the fires used in the parameterization? Generally it would be expected to break into training and validation datasets. This division could be done and analyzed in a number of ways to better understand how well the model(s) are performing. Equally well across fire danger levels? Equally well for each year?
- How was the fire danger that was used calculated? From the text, it states that the value on the date and location of the reported ignition was used? Did you look at how much daily variability there was over at least a selection of the fires in each ecoregion? Given the approach, the median or average value for the duration would seem more likely to prove significant. This may be part of the reason why the model fit drops as the duration (and hence area) grows.
- It should be noted that the model did fabulously in the Amazon (n = 663), reasonably in the Andean (n=252) and marginally/poorly in the Xeric and Dry Forest (n = 38 and 50). Could performance be sample size driven? That may not be the driving issue but it could be tested to bound the matter.
- Given the methods, I understand why single day fires were excluded but where does the 405 hectare limit come from? Is it reasonable to have it be the same in all ecoregions?
- Line 162 - “The spread rate and extinguish rate were highly correlated with one another”: Given that they were both calculated from the total area burned and duration of burning, is this a finding or a necessity from the calculations?
- Do the median RMSE values for the ecoregions have any independent comparative value? Shouldn’t they be normalized by the median spread rate or fire size for each ecoregion? It is unclear if 0.5 km2 is a large or small error, for example.
- Figure 5 (Time – Days) – this looks like the error term is an exponential term. The long term rise in errors would drive the larger error for large fires and ecoregions typified by them. Then again it may be a sample size issue….
- Line 191 – “Changes in fire weather and land cover were predicted to have no effects on fire growth in the dry forest regions.”- Really? I think this is misstated or fundamentally miscomprehended. What was shown here is that neither fire weather nor land cover had statistically significant predictive value for fire growth in these forests. Why is the question? Is it because there were no other land cover types encountered? Was it because fire growth was more strongly determined by a variable that wasn’t included (e.g. topography)? Was it because of the relatively small sample size, especially since the data were skewed by at least one very large fire?
- Line 209 – additional parameters or covariates could be included” – such as what? How would they help?
- Figure 7 – Wettest 2% of day? If that were true no fire spread would be expected. I suspect this is the wettest 2% of days when fire spread was observed.
- Line 228 – How would the spread tool identify when/where fire occurrence is possible since it is not spatial? It might help identify conditions when fire spread is possible that could perhaps be used in this way.
- Lines 244-245 – “relatively sparse human habitation” – perhaps but in the case of the Amazon forests most if not all of those ignitions are tied to those areas of sparse human habitation. The fires rarely if ever happen in remote forests.
- Lines 256-257 “This implies that the spread rate is proportional to the extinguish rate, with a harmonic number as the scaler” – see comment 5 above.
- Line 269 – “additional covariates” – such as?
- Line 274 – “ENSO-related effects on fire spread” – perhaps better stated as – “ENSO-related effects on weather conditions that affect fire spread”.
- Line 305 – “increased fire danger increased extinguish rates” seems nonsensical physically. It could arise from the model structure for fires that either were extinguished because they spread more rapidly to where landcover or features prevented further fire spread or when sudden events (e.g. rain) extinguished a fire that started during high fire danger conditions.
Harry Podschwit et al.
Harry Podschwit et al.
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
288 | 86 | 12 | 386 | 5 | 5 |
- HTML: 288
- PDF: 86
- XML: 12
- Total: 386
- BibTeX: 5
- EndNote: 5
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1