Limiting global warming to 1.5 &deg;C minimises projected global increases in fire weather days, but adaptation to new fire regimes is still needed

Taylor, Inika; Kelley, Douglas I.; Mathison, Camilla; Williams, Karina E.; Hartley, Andrew J.; Betts, Richard A.; Burton, Chantelle

doi:https://doi.org/10.5194/egusphere-2025-720

Preprints

https://doi.org/10.5194/egusphere-2025-720

Preprints

24 Mar 2025

| 24 Mar 2025

Limiting global warming to 1.5 °C minimises projected global increases in fire weather days, but adaptation to new fire regimes is still needed

Inika Taylor, Douglas I. Kelley, Camilla Mathison, Karina E. Williams, Andrew J. Hartley, Richard A. Betts, and Chantelle Burton

Abstract. Understanding future shifts in fire weather risk, including peak season, transitional and off-season, will be crucial for reshaping fire preparation and management in order to adapt to climate change. This study explores future climate-driven projections of fire weather using the McArthur Forest Fire Danger Index (FFDI) across three Global Warming Levels (GWLs) with two future emissions scenarios – 1.5 °C, 2.0 °C under both RCP2.6 and RCP8.5, and 4.0 °C under RCP8.5. Using a large, perturbed physics ensemble, we assess uncertainty in fire weather projections globally and for three regions: Australia, Brazil, and the USA. In addition to season length and peak FFDI, we evaluate transitions in meteorological fire danger periods and shifts in low-fire weather windows to inform fire management throughout the annual cycle. We project a global rise in fire weather days and severity at all GWLs, with the largest increases in Australia, followed by Brazil and the USA. At 1.5 °C, the area exposed to Very High fire weather (FFDI ≥ 24) expands by 31 % (25 %–36 %) relative to a baseline of 1986–2005. Higher GWLs drive further increases, with more than a threefold rise in Very High fire weather days from 2.0 °C to 4.0 °C, emphasising the mitigation benefits of limiting global warming to well below 2.0 °C as intended by the Paris Agreement. The transition from High to Very High, a proxy for the start of the fire season, advances, by 9–12 days in Australia, 16–22 days in Brazil, and 8–24 days in the USA. Despite these changes, low-fire windows persist, providing crucial opportunities for out-of-season preparation such as controlled burns. Our findings highlight the need for both emissions reductions and adaptive strategies, including accounting for changes in out-of-season fire risks when employing management techniques that rely on pre-fire season preparations.

Received: 16 Feb 2025 – Discussion started: 24 Mar 2025

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Inika Taylor, Douglas I. Kelley, Camilla Mathison, Karina E. Williams, Andrew J. Hartley, Richard A. Betts, and Chantelle Burton

Status: closed (peer review stopped)

RC1:
'Comment on egusphere-2025-720', Anonymous Referee #1, 17 Apr 2025
Referee comments

General comments
This is another global and regional assessment study of changes in fire weather (FW) carried out based on: (i) a meteorological fire danger index calculated with values of meteorological variables/climate elements, such as precipitation, temperature and air humidity; and (ii) climate model simulations for different GWLs and scenarios (in this case, the surpassed RCP). Unsurprisingly, the study concludes that the values of the chosen index and metrics will increase and that the increases are proportional to the increasing severity of the GWLs and scenarios. This is one of the main problems with the manuscript: what’s new?
Additionally, this study suffers from other serious problems, which I will detail in the more specific comments. Together, these issues prevent you from recommending this manuscript for publication.

Specific comments
As mentioned in the general comments, the first question that authors have to clarify and answer is what is new about their results/conclusions.

The second question is how much confidence the reader can have in the results and conclusions of the study. This question has to do with the methodological approach and methodology adopted. Studies on the impacts of climate change caused by global warming necessarily include two parts: (i) assessment of changes (e.g., through anomalies or ratios); and (ii) assessment of the statistical significance of these changes. This study only performs part of the assessment of changes, namely in terms of anomalies. Thus, the assessment of statistical significance, which is as fundamental as the first, is missing.

The confidence in the findings is also related to the methodology adopted. In this sense, the authors need to answer the following questions:
Why do you think an index developed for Australia, where wildfires are usually named bushfires, because of the main type of vegetation affected by the wildfires, is suitable for assessing fire danger anywhere in the world?

The suitability of FFDI for studying the fire weather globally can be measured based on the number of countries adopting FFDI as a fire danger indicator. So, it is important to know, how many countries use FFDI?

Why do you think an index developed for grasslands is good for assessing meteorological fire danger in other types of vegetation?

How much confidence can one have in results obtained as FFDI for regions as extensive and with such diverse vegetation as Australia, Brazil and the USA?

(Lines 102-107) In general, it is expected that any meteorological danger index will be well correlated with fire activity; This does not demonstrate the usefulness of FFDI.

The authors need to clarify many other aspects of the methodology, namely:
(Lines 214-215) “Throughout this study, we consider days above the “Very High” threshold, corresponding to days above a threshold of 24, as an indicator of fire season length”. Why? This relationship must be demonstrated or there must be a citation.

(Lines 254-266) the authors state that they computed 5 metrics, but in the following lines (267-270) they also state they computed additional metrics. So, how many metrics were computed?

The metrics studied must be well-defined. However, it is not at all clear how the "burnable land surface with an increase in fire weather" is calculated nor what the relationship is between the increase in "fire weather" and "burnable land surface".

To understand the last metric (The number of days in the High FFDI category), it is essential to know the relationship between the FFDI categories and the incidence of fire, on a global scale (the same scale as the study)

The results section also needs a lot of additional work.
Although the exact reproduction of the observed patterns was not expected, the evaluation of the FFDI calculated with simulations and observations (section 3.1 and Figure 1) is not convincing, for several reasons: first, only one metric is evaluated; second, even for this metric what the results demonstrate is that the simulated FFDI is similar only for Australia, the region for which FFDI was developed. The inadequacy of FFDI to assess meteorological fire danger globally is recognized by the authors, at least for boreal regions, and these results do not increase confidence in the methodology adopted and, consequently, in the results and conclusions obtained.

About Figure 1, It is necessary to explain how the "Consensus" is objectively assessed; what do the values indicated in the legend mean? The colour scale is inadequate; it is impossible to perceive the different shades in the figure.

Section 3.2, Figure 2. Why the results for RCP8. 5 to 1.5ºC appear to be less serious than for the RCP2 scenario. 6 to 1.5ºC? This apparent discrepancy is also illustrated in Table 3.

Distributions in Figure 3 and Figure 4 are not normal. What are the consequences for your methods, results and conclusions?

Section 3.3. What is the Mitigation anomaly? Not clear.

The caption of Figure 4 is not clear. What is the " mitigation anomaly of the change in the annual average number of days above the FFDI threshold"? What FFDI threshold?

Section 3.4. Very confusing description of the results. It is necessary to present numerical values that demonstrate what is written. Visual analysis is not enough. Are the reported differences statistically significant?

Line 414. Why the citation?

Line 423. Is the Loess fit described in the methodology?

Table 4. What is the meaning of the colour palette?

In the Discussion, the authors interpret the results but do not validate them. Comparison with the findings of other similar studies is fundamental to increase confidence in the findings of this study. Authors have to focus on what distinguishes and adds value to their study and its results. How much confidence can the reader have in the results/conclusions of this study? Why should we have more confidence in their results than in the results of other studies of this type?

The manuscript is not well written. The title does not adequately describe the study described in the manuscript. The Introduction is very long and confusing. The text is somewhat disorganized, there is some repetition which makes the manuscript difficult and boring to read. Several parts of the manuscript must be rewritten to resolve this issue.

However, perhaps the most serious issue with the written style is the lack of clarity and rigour. Some examples to illustrate this criticism are as follows:
Fire and wildfire are not the same. Wildfires are not “defined as unusual or extraordinary free-burning vegetation events” as claimed by the authors in line 30. Fire is a wanted and controlled biomass combustion that can be used as a tool, while a wildfire is an unwanted, unauthorized and uncontrolled biomass combustion. The scientific community knows the difference and it is time to start using the correct concepts, like wildfire regime, wildfire severity, and so on.

Fire danger and fire risk are also not the same, but the authors seem to use these concepts as synonymous. See, for example, the abstract and the sentences in lines 38-39. The FFDI (McArthur Forest Fire Danger Index) is a fire danger or fire risk index?

(lines 85-87) the long-term climate conditions that a fire needs to burn if ignited with available fuel are not “fire weather” but “fire climate”

(Lines 93-94) It is not the “fire indices” that “can forecast changes in the number of fire weather days and season lengths,…”

In lines 15-16, the authors state that they “assess uncertainty in fire weather projections globally and for three regions: Australia, Brazil, and the USA”. This adequately describes the study described in the manuscript?

In line 18, the authors state that they project the fire weather severity. How was this severity assessed?

The authors repeatedly use slang, such as "fire weather days" when they should write “the number of days with…”

(Lines 180-181) Figure 1 does not show the indicated correlation. Show the average number of days per year above the Very High FFDI threshold in 1986-2005 for HadCM3C PPE historically and with the Copernicus FFDI for the same metric and period.

Technical corrections
The manuscript requires a large number of corrections, as a result of an excessive number of typos, errors in writing units (e.g., unit “stuck” to the numerical value, e.g., “1.5m” and “10m” in lines 199-200, but also other errors such as “km/hr1” in line 200) and even citations, for example, “by (Noble et al., 1980)” (line 102), “by (Noble et al., 1980)” (line 116), etc. Please use an n-dash instead of a minus sign to define a period.
Authors should review the entire manuscript to identify and correct any existing errors.
Citation: https://doi.org/10.5194/egusphere-2025-720-RC1
RC2: 'Comment on egusphere-2025-720', Anonymous Referee #2, 06 May 2025

This study investigates projected changes in fire weather, focusing on the Forest Fire Danger Index (FFDI) across three distinct regions: Australia, Brazil, and the USA. While the authors conduct several types of analyses, the overall novelty of the work is somewhat limited. Nevertheless, I am not necessarily opposed to this; the methods and analyses are sound, data sets are public, which I appreciate. However, the study needs to present more specific insights. As the title suggests, the main focus lies in understanding future fire weather changes and potential mitigation strategies. In my opinion, there is potential to deepen this focus. Below are some of the main issues and suggestions on how to address them.
Introduction:

As the first reviewer already pointed out, the study lacks a clearly defined research gap. Given that it compares three climatically and geographically diverse regions, there is an inherent strength in highlighting region-specific mitigation strategies and identifying similarities and differences based on local conditions. However, the current text remains too broad and reads more like a general overview. I recommend restructuring the introduction to clearly define the research gap and then justify the choice of focusing on Australia, Brazil, and the USA. In addition, existing other FFDI products and the state of the art should be summarized, with a clear explanation of how the Perturbed Physics Ensemble approach can yield new insights.
Methods:

There is considerable repetition between the justification provided in the introduction and the methods section. This could be streamlined. The methods section should focus on technical clarity. Important technical details—such as the spatial resolution of the datasets, software used to estimate the FFDI, and other relevant aspects should be added.
Results and Figures:

While the study claims to focus on three specific regions, only global-scale maps are shown. This is a missed opportunity. To align with the paper’s stated goals, I strongly recommend including detailed regional breakdowns, for instance zoom-ins on Australia, Brazil, and the USA, particularly where changes in FFDI are most significant. Instead of clipping results only to non-burnable land, including hotspots of burnable land (e.g., using a proxy for fuel loads) could make the recommendations more specific.
Figure 5 stands out as the most promising in terms of offering new insight into mitigation potentials. It would be valuable to expand this part of the discussion.
Figure 6 shows that several extreme years are expected in the future. This deserves a closer look: what do these extremes look like regionally? Which areas are most at risk? It would be useful to analyze a few “what-if” scenarios more deeply—especially if the PPE approach allows a better understanding of tail risks than traditional GCM ensembles.
Discussion:

The discussion begins to touch on relevant points but currently lacks depth. The section on the different regions holds the most promise in my opinion, but the latter part feels somewhat half-hearted, especially where it lists potential future research questions without any explaining how this study contributes to solving them.
General:

All in all, I share many of the concerns raised by reviewer 1, particularly regarding consistency in terminology and the lack of depth in the analysis. However, I still believe the study could make a valuable contribution once the text is improved and the novel aspects are clarified.

Citation: https://doi.org/10.5194/egusphere-2025-720-RC2

Status: closed (peer review stopped)

RC1:
'Comment on egusphere-2025-720', Anonymous Referee #1, 17 Apr 2025
Referee comments

General comments
This is another global and regional assessment study of changes in fire weather (FW) carried out based on: (i) a meteorological fire danger index calculated with values of meteorological variables/climate elements, such as precipitation, temperature and air humidity; and (ii) climate model simulations for different GWLs and scenarios (in this case, the surpassed RCP). Unsurprisingly, the study concludes that the values of the chosen index and metrics will increase and that the increases are proportional to the increasing severity of the GWLs and scenarios. This is one of the main problems with the manuscript: what’s new?
Additionally, this study suffers from other serious problems, which I will detail in the more specific comments. Together, these issues prevent you from recommending this manuscript for publication.

Specific comments
As mentioned in the general comments, the first question that authors have to clarify and answer is what is new about their results/conclusions.

The second question is how much confidence the reader can have in the results and conclusions of the study. This question has to do with the methodological approach and methodology adopted. Studies on the impacts of climate change caused by global warming necessarily include two parts: (i) assessment of changes (e.g., through anomalies or ratios); and (ii) assessment of the statistical significance of these changes. This study only performs part of the assessment of changes, namely in terms of anomalies. Thus, the assessment of statistical significance, which is as fundamental as the first, is missing.

The confidence in the findings is also related to the methodology adopted. In this sense, the authors need to answer the following questions:
Why do you think an index developed for Australia, where wildfires are usually named bushfires, because of the main type of vegetation affected by the wildfires, is suitable for assessing fire danger anywhere in the world?

The suitability of FFDI for studying the fire weather globally can be measured based on the number of countries adopting FFDI as a fire danger indicator. So, it is important to know, how many countries use FFDI?

Why do you think an index developed for grasslands is good for assessing meteorological fire danger in other types of vegetation?

How much confidence can one have in results obtained as FFDI for regions as extensive and with such diverse vegetation as Australia, Brazil and the USA?

(Lines 102-107) In general, it is expected that any meteorological danger index will be well correlated with fire activity; This does not demonstrate the usefulness of FFDI.

The authors need to clarify many other aspects of the methodology, namely:
(Lines 214-215) “Throughout this study, we consider days above the “Very High” threshold, corresponding to days above a threshold of 24, as an indicator of fire season length”. Why? This relationship must be demonstrated or there must be a citation.

(Lines 254-266) the authors state that they computed 5 metrics, but in the following lines (267-270) they also state they computed additional metrics. So, how many metrics were computed?

The metrics studied must be well-defined. However, it is not at all clear how the "burnable land surface with an increase in fire weather" is calculated nor what the relationship is between the increase in "fire weather" and "burnable land surface".

To understand the last metric (The number of days in the High FFDI category), it is essential to know the relationship between the FFDI categories and the incidence of fire, on a global scale (the same scale as the study)

The results section also needs a lot of additional work.
Although the exact reproduction of the observed patterns was not expected, the evaluation of the FFDI calculated with simulations and observations (section 3.1 and Figure 1) is not convincing, for several reasons: first, only one metric is evaluated; second, even for this metric what the results demonstrate is that the simulated FFDI is similar only for Australia, the region for which FFDI was developed. The inadequacy of FFDI to assess meteorological fire danger globally is recognized by the authors, at least for boreal regions, and these results do not increase confidence in the methodology adopted and, consequently, in the results and conclusions obtained.

About Figure 1, It is necessary to explain how the "Consensus" is objectively assessed; what do the values indicated in the legend mean? The colour scale is inadequate; it is impossible to perceive the different shades in the figure.

Section 3.2, Figure 2. Why the results for RCP8. 5 to 1.5ºC appear to be less serious than for the RCP2 scenario. 6 to 1.5ºC? This apparent discrepancy is also illustrated in Table 3.

Distributions in Figure 3 and Figure 4 are not normal. What are the consequences for your methods, results and conclusions?

Section 3.3. What is the Mitigation anomaly? Not clear.

The caption of Figure 4 is not clear. What is the " mitigation anomaly of the change in the annual average number of days above the FFDI threshold"? What FFDI threshold?

Section 3.4. Very confusing description of the results. It is necessary to present numerical values that demonstrate what is written. Visual analysis is not enough. Are the reported differences statistically significant?

Line 414. Why the citation?

Line 423. Is the Loess fit described in the methodology?

Table 4. What is the meaning of the colour palette?

In the Discussion, the authors interpret the results but do not validate them. Comparison with the findings of other similar studies is fundamental to increase confidence in the findings of this study. Authors have to focus on what distinguishes and adds value to their study and its results. How much confidence can the reader have in the results/conclusions of this study? Why should we have more confidence in their results than in the results of other studies of this type?

The manuscript is not well written. The title does not adequately describe the study described in the manuscript. The Introduction is very long and confusing. The text is somewhat disorganized, there is some repetition which makes the manuscript difficult and boring to read. Several parts of the manuscript must be rewritten to resolve this issue.

However, perhaps the most serious issue with the written style is the lack of clarity and rigour. Some examples to illustrate this criticism are as follows:
Fire and wildfire are not the same. Wildfires are not “defined as unusual or extraordinary free-burning vegetation events” as claimed by the authors in line 30. Fire is a wanted and controlled biomass combustion that can be used as a tool, while a wildfire is an unwanted, unauthorized and uncontrolled biomass combustion. The scientific community knows the difference and it is time to start using the correct concepts, like wildfire regime, wildfire severity, and so on.

Fire danger and fire risk are also not the same, but the authors seem to use these concepts as synonymous. See, for example, the abstract and the sentences in lines 38-39. The FFDI (McArthur Forest Fire Danger Index) is a fire danger or fire risk index?

(lines 85-87) the long-term climate conditions that a fire needs to burn if ignited with available fuel are not “fire weather” but “fire climate”

(Lines 93-94) It is not the “fire indices” that “can forecast changes in the number of fire weather days and season lengths,…”

In lines 15-16, the authors state that they “assess uncertainty in fire weather projections globally and for three regions: Australia, Brazil, and the USA”. This adequately describes the study described in the manuscript?

In line 18, the authors state that they project the fire weather severity. How was this severity assessed?

The authors repeatedly use slang, such as "fire weather days" when they should write “the number of days with…”

(Lines 180-181) Figure 1 does not show the indicated correlation. Show the average number of days per year above the Very High FFDI threshold in 1986-2005 for HadCM3C PPE historically and with the Copernicus FFDI for the same metric and period.

Technical corrections
The manuscript requires a large number of corrections, as a result of an excessive number of typos, errors in writing units (e.g., unit “stuck” to the numerical value, e.g., “1.5m” and “10m” in lines 199-200, but also other errors such as “km/hr1” in line 200) and even citations, for example, “by (Noble et al., 1980)” (line 102), “by (Noble et al., 1980)” (line 116), etc. Please use an n-dash instead of a minus sign to define a period.
Authors should review the entire manuscript to identify and correct any existing errors.
Citation: https://doi.org/10.5194/egusphere-2025-720-RC1
RC2: 'Comment on egusphere-2025-720', Anonymous Referee #2, 06 May 2025

This study investigates projected changes in fire weather, focusing on the Forest Fire Danger Index (FFDI) across three distinct regions: Australia, Brazil, and the USA. While the authors conduct several types of analyses, the overall novelty of the work is somewhat limited. Nevertheless, I am not necessarily opposed to this; the methods and analyses are sound, data sets are public, which I appreciate. However, the study needs to present more specific insights. As the title suggests, the main focus lies in understanding future fire weather changes and potential mitigation strategies. In my opinion, there is potential to deepen this focus. Below are some of the main issues and suggestions on how to address them.
Introduction:

As the first reviewer already pointed out, the study lacks a clearly defined research gap. Given that it compares three climatically and geographically diverse regions, there is an inherent strength in highlighting region-specific mitigation strategies and identifying similarities and differences based on local conditions. However, the current text remains too broad and reads more like a general overview. I recommend restructuring the introduction to clearly define the research gap and then justify the choice of focusing on Australia, Brazil, and the USA. In addition, existing other FFDI products and the state of the art should be summarized, with a clear explanation of how the Perturbed Physics Ensemble approach can yield new insights.
Methods:

There is considerable repetition between the justification provided in the introduction and the methods section. This could be streamlined. The methods section should focus on technical clarity. Important technical details—such as the spatial resolution of the datasets, software used to estimate the FFDI, and other relevant aspects should be added.
Results and Figures:

While the study claims to focus on three specific regions, only global-scale maps are shown. This is a missed opportunity. To align with the paper’s stated goals, I strongly recommend including detailed regional breakdowns, for instance zoom-ins on Australia, Brazil, and the USA, particularly where changes in FFDI are most significant. Instead of clipping results only to non-burnable land, including hotspots of burnable land (e.g., using a proxy for fuel loads) could make the recommendations more specific.
Figure 5 stands out as the most promising in terms of offering new insight into mitigation potentials. It would be valuable to expand this part of the discussion.
Figure 6 shows that several extreme years are expected in the future. This deserves a closer look: what do these extremes look like regionally? Which areas are most at risk? It would be useful to analyze a few “what-if” scenarios more deeply—especially if the PPE approach allows a better understanding of tail risks than traditional GCM ensembles.
Discussion:

The discussion begins to touch on relevant points but currently lacks depth. The section on the different regions holds the most promise in my opinion, but the latter part feels somewhat half-hearted, especially where it lists potential future research questions without any explaining how this study contributes to solving them.
General:

All in all, I share many of the concerns raised by reviewer 1, particularly regarding consistency in terminology and the lack of depth in the analysis. However, I still believe the study could make a valuable contribution once the text is improved and the novel aspects are clarified.

Citation: https://doi.org/10.5194/egusphere-2025-720-RC2

Inika Taylor, Douglas I. Kelley, Camilla Mathison, Karina E. Williams, Andrew J. Hartley, Richard A. Betts, and Chantelle Burton

Data sets

Global Daily FFDI Projections from HadCM3C Perturbed Physics Ensemble - RCP2.6 I. Taylor et al. https://doi.org/10.5281/zenodo.14860331

Global Daily FFDI Projections from HadCM3C Perturbed Physics Ensemble - RCP8.5 I. Taylor et al. https://doi.org/10.5281/zenodo.14859064

Model code and software

Analysis code for Limiting global warming to 1.5°C minimises projected global increases in fire weather days, but adaptation to new fire regimes is still needed I. Taylor et al. https://doi.org/10.5281/zenodo.14871362

Inika Taylor, Douglas I. Kelley, Camilla Mathison, Karina E. Williams, Andrew J. Hartley, Richard A. Betts, and Chantelle Burton

Viewed

Total article views: 847 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
702	125	20	847	12	34

HTML: 702
PDF: 125
XML: 20
Total: 847
BibTeX: 12
EndNote: 34

Views and downloads (calculated since 24 Mar 2025)

Month	HTML	PDF	XML	Total
Mar 2025	63	18	2	83
Apr 2025	39	29	4	72
May 2025	48	15	2	65
Jun 2025	40	5	2	47
Jul 2025	23	14	0	37
Aug 2025	90	32	4	126
Sep 2025	383	5	5	393
Oct 2025	16	7	1	24

Cumulative views and downloads (calculated since 24 Mar 2025)

Month	HTML	PDF	XML	Total
Mar 2025	63	18	2	83
Apr 2025	39	29	4	72
May 2025	48	15	2	65
Jun 2025	40	5	2	47
Jul 2025	23	14	0	37
Aug 2025	90	32	4	126
Sep 2025	383	5	5	393
Oct 2025	16	7	1	24

Viewed (geographical distribution)

Total article views: 881 (including HTML, PDF, and XML) Thereof 881 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 21 Oct 2025

Short summary

Climate change is reshaping fire seasons worldwide and, in many places, increasing fire weather risk. We use climate model simulations to project future changes in fire danger at different levels of global warming, focusing on Australia, Brazil, and the USA. Keeping warming below 2 °C significantly limits the increase in fire risk, but even at 1.5 °C, fire seasons lengthen, with more extreme conditions. However, low-fire weather periods remain, offering critical windows for fire management.


Total:	0
HTML:	0
PDF:	0
XML:	0