Effects of Climate Change on Burn Probability of Forests in Daxing’anling

Round 1

Reviewer 1 Report

Please, find below the itemized list of the corrections brought to Manuscript ID: forests- 527202 :  “Impact of Climate Change on the Flammability of 2 Forests in Daxing’anling, China »

 

General Remarks:

This title has too many unnecessary words :

« Burn probability model … »

Some important considerations are highlighted.

Abstract :

 

-Does the abstract reflect the essentials of the methodology and findings

-What problem are they trying to solve? provide a scientific basis for the implementation of forest fire adaptation technology..it should be included in the abstract?.

Introduction :

 

-The research problem and research questions must be clear. What make this paper relevant?

Materials and Method:

- The study area design?

 

- Parameters estimation (model calibration)?

Discussion.

- How the results compare to another study which was published very recently?

 

Conclusion

 

Forest fire adaptation technology?

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

GENERAL COMMENTS

 

Dear authors,

 

Please find enclosed my report of manuscript forest-527202 entitled “Impact of Climate Change on the Flammability of 2 Forests in Daxing’anling, China”. I found the article interesting, especially for the very detailed fire-weather information and because annual fire number and burned areas emerge as a result of the interaction between daily weather conditions, fire process and fuel configuration. However, I found important drawbacks and aspects that should be addressed by the authors before recommending for publication.

 

The specific objectives of the present work should be more explicit at the end of the introduction. The authors list a number of studies for the same region predicting areas burned under climate change scenario. However, it is not clear where is the novelty or contribution of this work to the previous research conducted in the study area. There is also a list/description of papers using fire models in the introduction but apparently unrelated with the contribution of this paper to the field of fire modelling/management. 

 

The methods section is not well organized. The authors should start describing the modeling approach and framework (sections 3.4 and 3.5) and continue explaining the data sources and how fire weather indices and codes are computed, including references to previous versions to the model (if exist) or the formulation behind the fire weather indices computation. It is also unclear to me: 1) how fire intensity can be modeled if fuel (vegetation) structure is not considered, and 2) how fire ignition is simulated, especially those caused by human activities, according to the model description, apparently simulated from a fire ignition density layer. 

 

My main concern is an apparently large difference between the simulated and observed burned areas (from +12.4% to -59%), that is much larger than the difference between burned area simulated under present and future conditions (from 10.5% to 20.4%). This is translated into a high level of uncertainty that can invalidate the approach. Authors should better justify model reliability. In addition, this validation exercise (observed vs simulated) needs to be supported by some statistical analysis to really inform about the level of model accuracy and predictions confidence.

 

I do not understand why the authors only represent average values instead of box-plots with the variability around mean or median values if they run 10,000 simulations for each fire season (i.e. they are losing the opportunity to represent the inherent uncertainty associated with the stochastic nature of fires). Please consider this alternative way of showing the results or justify the average values. 

 

Finally, I think that the authors should try to better explain the advantages of their modelling approach. They are considering fire suppression, that it was found to be a key factor affecting fire regime. That is very interesting and, unfortunately, rarely considered in fire modelling (see e.g. [1–5]). Besides, number of fires is an emerging fire regime property derived from their model approach, that is not easy to see in fire modeling exercises. The authors aim to predict the impact of climate change on fire regime, and for doing so, they isolate the effect of this important driver on fire regime descriptors. However, the effect of changes in land use or fire management (e.g. [6–8]) and, especially, the spatial interaction between fire and vegetation dynamics (including natural succession and post-fire regeneration) can definitively affect fire regime [9], as well aspossible feedbacks on climate change effects on vegetation [10]. They authors should better discuss these issues, including model limitations, to avoid misleading conclusions.

 

I hope the authors find my comments and suggestions useful to improve the manuscript. 

 

All the best. 

 

References

 

1         Keeley, J.E. et al.(1999) Reexamining fire suppression impacts on brushland fire regimes. Science (80-. ).284, 1829–1832

2         Turco, M. et al.(2013) Brief communication decreasing fires in a mediterranean region (1970-2010, NE Spain). Nat. Hazards Earth Syst. Sci.13, 649–652

3         Cumming, S.G. (2005) Effective fire suppression in boreal forests. Can. J. For. Res.35, 772–786

4         Ruffault, J. and Mouillot, F. (2015) How a new fire-suppression policy can abruptly reshape the fire-weather relationship. Ecospherein press, 1–19

5         Brotons, L. et al.(2013) How fire history, fire suppression practices and climate change affect wildfire regimes in Mediterranean landscapes. PLoS One8, e62392

6         Regos, A. et al.(2016) Synergies between forest biomass extraction for bioenergy and fire suppression in Mediterranean ecosystems: insights from a storyline-and-simulation approach. Ecosystems19, 786–802

7         Duane, A. et al.(2019) Adapting prescribed burns to future climate change in Mediterranean landscapes. Sci. Total Environ.677, 68–83

8         Aquilué, N. et al.(2019) The Potential of Agricultural Conversion to Shape Forest Fire Regimes in Mediterranean Landscapes. EcosystemsDOI: 10.1007/s10021-019-00385-7

9         Pausas, J.G. and Paula, S. (2012) Fuel shapes the fire-climate relationship: evidence from Mediterranean ecosystems. Glob. Ecol. Biogeogr.21, 1074–1082.

10      Brotons, L. and Duane, A. (2019) Correspondence : Uncertainty in Climate-Vegetation Feedbacks on Fire Regimes Challenges Reliable Long-Term Projections of Burnt Area from Correlative Models. FireDOI: 10.3390/fire2010008

 

 

SPECIFIC COMMENTS

 

ABSTRACT 

 

Line 13-14. Not sure what the authors mean in this first sentence. Why the implementation of adaptation measures for fire management would be helpful to project burn burnability? or they refer to the incorporation into the modelling framework?

 

Lines 25. What does “the simulated areas …. similar changes to the actual burned areas” mean? Changes in relation to what? Not sure if the authors refer to the difference between simulated and observed burned areas (as validation measure) or to the interannual fluctuations between observed and simulated burned areas. 

 

Line 27-28. To highlight the need of proactive management, and to refer more explicitly to the spatial outputs of your fire simulations, I would conclude that “the implementation of proactive fire management in area with high predicted BP values will be key for an effective mitigation of future wildfire impacts”. 

 

INTRODUCTION

 

Lines 38-39. Total burned area in 2017, but where?  US, China? 

 

Line 44. The authors can also mention northern Europe (Sweden or Norway).   

 

Lines 51-57. The authors go back to large forest fires in US. I suggest merging this paragraph with the previous one (Lines 38-42). 

 

Line 62. Please replace disaster-prone by fire-prone. I would be more neutral in the terminology.  

 

If the study is developed for a China’s region, I do not understand why the authors make so much emphasis on North America. I encourage the authors referring to other fire-prone regions such as Australia, South Africa or Southern Europe.

 

For instance: Turco,M., Rosa-Cánovas, J. J., Bedia, J., Jerez, S., Montávez, J. P., Llasat, M. C., & Provenzale, A. (2018). Exacerbated fires in Mediterranean Europe due to anthropogenic warming projected with non-stationary climate-fire models. Nature communications, 9(1), 3821. 

 

Lines 87-101. After reading this paragraph, I am wondering what is the contribution of this paper to previous research conducted in the same study area? 

 

Lines 102-117. What it is the objective of listing all these different applications of fire models and simulation in different regions.  

 

Line 122. What model? It is an updated version of a previous one? Why the authors mention fire suppression here? This is a new implementation of a previous model? What are the main goal and specific objectives of the study, and how this study contributes to the previous research conducted in the same area?  

 

STUDY AREA

 

Study area description is usually included in material and method section. 

 

Lines 140-141. The authors state that the study area is an important fire management region because is fire prone. Please rephrase. If the authors want to highlight the fire management policy currently implemented in that region, please briefly explain it. 

 

What is the source of fire statistics? Fire reports, remote sensing? 1 million hectares burned in a single year is a huge proportion for an area of 8 million hectares, especially if there is a policy aimed at stopping all fire as soon as possible (lines 119-120), and the low number of fires per year. I would like to see more information here regarding the data source, and the fire suppression policies or resources of the region, to better understand these figures.  

 

METHODS

 

153-154. What is the MODIS product used for vegetation? What are the vegetation (fuel) types represented? What is the resolution of the DEM?

 

Line 182. What threshold?

 

Lines 188-193. Are these formulas extracted from literature? Please include references. What are these “a” and “b” values? 

 

It is not clear at all the role of these variables in the modelling framework because the BP model was not described yet. Authors should introduce the modelling framework earlier because at this stage is not clear to me (or other potential readers) what type of modelling approach is used and the role of each variable on it. 

 

Lines 199-206. Several codes and indices are listed but more information about how they are computed and integrated in the BP model is needed. 

 

Lines 205. Please clarify that data from stations is used for present conditions and from grids for future weather conditions. 

 

Please move section 3.4 and 3.5 to the beginning of method section. 

 

Line 209. Please describe the fire growth model here.

 

Lines 216-223. I understand that the number of fire events predicted in the model is an emerging property derived from the relationships between fire weather indices and fire occurrence. However, I do not understand how fire occurrence is modeled: How fire ignitions are simulated? Human-caused ignitions (but not lightning-caused fires) often depend on accessibility (e.g. distance to roads), wildland-urban interface or human density, or are related to socioeconomic factors. I do not see any of these factors considered in the model. There is a fire ignition probability layer or fire ignitions are randomly simulated within the density layers? 

 

Lines 250. What type of information provide the raster layer “study area”? Fuel types are not described, nor the source (only referred as a MODIS product). 

 

Please move fire growth model to line 210. 

 

Figure 2. Please include the meaning of the abbreviatures in the map of fuel types in the figure legend.   

 

Lines 267. Here the authors should briefly explain why they use 10,000 iterations (simulations) for each fire season and refer to supplementary material for further details (see my comment in the discussion related to this issue).  

 

RESULTs

Line 285. 1.1? Please correct. 

 

290-294. This section shows results derived from models that were not explained in the method section (e.g. annual fire numbers and lighting fire occurrence). I would describe the objectives and methods used for modelling these two variables and then I would simply show R2 and p-values in the results section. 

 

Lines 307-308. This bellows to method section. 

 

Lines 322-324. This should be moved to study area description. 

 

Lines 327-328. This result needs to be supported by some statistical analysis or a simple correlation.

 

Line 331. I would use inter-annual fluctuations or variability instead of “changes”.

 

Lines 346. Replace “increase” by “changes”. An increase of -46% does not make sense. 

 

Figure 7. Please add a new line showing the historical burned areas per year (baseline) as indicated in Figure 9. 

 

Lines 363. Remove one “in”.

 

Lines 266-367. Please move this sentence to discussion section. 

 

DISCUSSION

 

This section describes a bit better the novelty of the present work in relation to the previous research, but this should clearly state in the introduction. Sensitivity analysis should be move to an appendix. The advances and limitations of the modelling framework should be more explicitly and better discussed, and supported with relevant studies such as those I previously suggested (see general comments above).  

 

Lines 406. An ensemble of models (e.g. average) allow accounting for uncertainty regarding the different models or scenarios, but do not eliminate uncertainty.  

 

Lines 407-419. The authors surprisingly show a new analysis in the discussion, not previously described neither in methods nor in results sections. This part could easily move to supplementary material (and thus address one of my comments).  

 

 

Author Response

Please see the attachment

Round 2

Reviewer 2 Report

The authors have addressed all my main concerns, followed my suggestions and clarify all issues. The manucript is now improved and it is ready for being accepted for publication. 

Please correct line 318. "Predicted".

All the best. 

Author Response

1. The specific objectives of the present work should be more explicit at the end of the introduction.
Specific objectives continued not being clear. The objectives are explicit in discussion section (lines 414-416). Please, move them at the end of the introduction.
—Revised. We moved it into the introduction. “We used a BP model to evaluate the effects of climate change on burn probability in the region. The model considered the effects of fire suppression, which is suitable for such intensively managed areas. We predicted the annual fires in 2030s (2021-2050) and simulated daily fire occurrences, spread, and suppression based on the information of fuel type, terrain, weather, fire weather indices, fire behavior, and suppression ability in the integrated model. The results will provide a scientific basis for the implementation of forest fire adaptation technology”.

2. how fire intensity can be modeled if fuel (vegetation) structure is not considered? This question continues being without a clear answer. Please provide more information about how this variable was derived and the data sources.
—The fuel type data were inputted to the model. The fuel structure is considered in FBP system. The fire behavior curves (spread speed, fire intensity, flame flight, ect.) were made for different fuel types under different weather and FWI indices conditions. The fuel characters were considered in fuel type classification, such as tree species, fuel loadings, and structure. “The model simulates the spread of each fire based on the fire growth model (Wildfire) [Todd 1999]. It is an eight-point cellular fire growth model that uses the Fire Behavior Prediction (FBP) System.”

My main concern is an apparently large difference between the simulated and observed burned areas (from +12.4% to -59%), that is much larger than the difference between burned area simulated under present and future conditions (from 10.5% to 20.4%).
—Revised. Originally, we want to show the difference between the simulated burned areas and observed. Actually, it doesn't have meaning to compare the simulated areas for 1988-2000. So we delete these sentences. For the simulation in baseline and 2030s, we used the fire suppression ability in 2010, which is different with that before 1987. We compared the results between the 2030s and baseline, which indicted the impacts of climate change on burn probability. The model validation was used with result of simulated the period 1991-2010. "Since there were a clearly change on fire management policy and fire suppression ability in the baseline period, we take fire simulation results in the period 1991-2010 to validate BP model for the region".

This is a problem, which continues, in the revised paper, and authors should comment this issue in the discussion including model limitations. Moreover, authors assume that there are not changes in ignition distributions and fuel types in future 30 years, and they don´t consider this assumption as a limitation, in spite of reviewer recommendations.
—Revised. This is a limitation of the research. “However, the simulations described in this paper do not consider the changes in fire sources and fuel types in the future 30 years and the model assumed that the ignition distribution and fuel type would be basically the same as in the baseline period. This is a limitation of this research. If the fire management and fuel types will not have fundamental changes in the next 30 years, a small change in these factors can be considered unlikely to affect the reliability of the conclusion overall”.

Lines 87-101. After reading this paragraph, I am wondering what is the contribution of this paper to previous research conducted in the same study area?
Authors answered: —The previous research mainly focused on the impacts of climate change on fire weather and fire occurrences. They didn’t simulate fire occurrence and spread on daily for different scenarios. The paper will help us to better understand fire regime changes in landscapes. The results can help the local fire agency to carry out management measures in the parts with high BP values. We also showed the changes of fires and burned areas in the next 30 years.
This answer should be included in the main text of the paper.
—Revised. “The previous research mainly focused on the impacts of climate change on fire weather and fire occurrences. They didn’t simulate fire occurrence and spread on daily for different scenarios. The paper will help us to better understand fire regime changes in landscapes. We predicted the annual fires in 2030s (2021-2050) and simulated daily fire occurrences, spread, and suppression based on the information of fuel type, terrain, weather, fire weather indices, fire behavior, and suppression ability in the integrated model. The results will provide a scientific basis for the implementation of forest fire adaptation technology”.

Line 257. What is the MODIS product used for vegetation? This question continues being without answer in the revised manuscript.
—Revised. It is MODIS-L1B data. We added the download site. “The vegetation map was derived from Modis satellite remote sensing data MODIS-L1B in 2010 (http://www.gscloud.cn/)”.

It would be better if authors listed in a single paragraph all the variables that are needed to run the models, and afterwards, explain each of them in its corresponding section. In this current form, it is very difficult to follow the logic of the model framework.
—Revised. We added figure 3 to show the BP framework.

Finally, I have not found any supplementary material which is cited in the main text.

—Revised. “We analyzed the effects of iterations on the accuracy of BP simulation and found that 10,000 iterations are appropriate for the study (Appendix A for further details)”.

Author Response File: Author Response.pdf