Volume 11, Issue 11 e2023EF003976
Research Article
Open Access

Escalating Hot-Dry Extremes Amplify Compound Fire Weather Risk

Xuewei Fan

Xuewei Fan

State Key Laboratory of Earth Surface Processes and Resource Ecology, Faculty of Geographical Science, Beijing Normal University, Beijing, China

Contribution: Conceptualization, Methodology, Software, Validation, Formal analysis, ​Investigation, Data curation, Writing - original draft, Writing - review & editing, Visualization

Search for more papers by this author
Chiyuan Miao

Corresponding Author

Chiyuan Miao

State Key Laboratory of Earth Surface Processes and Resource Ecology, Faculty of Geographical Science, Beijing Normal University, Beijing, China

Correspondence to:

C. Miao,

[email protected]

Contribution: Conceptualization, Methodology, Validation, ​Investigation, Data curation, Writing - original draft, Writing - review & editing, Visualization, Supervision, Project administration, Funding acquisition

Search for more papers by this author
Jakob Zscheischler

Jakob Zscheischler

Department of Computational Hydrosystems, Helmholtz Centre for Environmental Research – UFZ, Leipzig, Germany

Technische Universität Dresden, Dresden, Germany

Contribution: Methodology, Validation, Formal analysis, Writing - original draft, Writing - review & editing

Search for more papers by this author
Louise Slater

Louise Slater

School of Geography and the Environment, University of Oxford, Oxford, UK

Contribution: Methodology, ​Investigation, Writing - original draft, Writing - review & editing

Search for more papers by this author
Yi Wu

Yi Wu

State Key Laboratory of Earth Surface Processes and Resource Ecology, Faculty of Geographical Science, Beijing Normal University, Beijing, China

Contribution: Methodology, Formal analysis, Writing - original draft, Writing - review & editing

Search for more papers by this author
Yuanfang Chai

Yuanfang Chai

State Key Laboratory of Earth Surface Processes and Resource Ecology, Faculty of Geographical Science, Beijing Normal University, Beijing, China

Contribution: Methodology, Writing - original draft

Search for more papers by this author
Amir AghaKouchak

Amir AghaKouchak

Center for Hydrometeorology and Remote Sensing, Department of Civil and Environmental Engineering, University of California, Irvine, CA, USA

Contribution: Methodology, ​Investigation, Data curation, Writing - original draft, Writing - review & editing, Visualization

Search for more papers by this author
First published: 06 November 2023
Citations: 1

Abstract

Fire weather compounded by extremely hot and dry conditions often severely impacts society and ecosystems. To mitigate and better adapt to these compound fire weather (CFW) events, a better understanding of recent and future CFW trends is needed. Here we show that in the period 1981–2020, the global average frequency and intensity of CFW events increased by 0.6 days/yr and 0.4%/yr, respectively. Increases in temperature and decreases in relative humidity were responsible for significant trends in the frequency of CFW events in 81.7% and 58.6% of locations, respectively. The same trends contributed to significant increases in CFW intensity in 72.1% and 57.9% of locations. We further demonstrate that anthropogenic climate change (due primarily to greenhouse gas emissions) has aggravated the frequency and intensity of CFW events, particularly in the Amazon region, with over 2-fold and 1.3-fold increases, respectively. Future projections reveal that other (individual) fire weather events are likely to shift toward CFW events accompanied by hot-dry conditions, along with an expected rise in CFW intensity. Furthermore, the increased occurrence of CFW events is likely to substantially augment future population exposure to CFW conditions. Under the SSP5-8.5 scenario, climate change is estimated to contribute 62.6% of the projected increase in population exposure to CFW events by the end of this century. Our findings underscore the urgent need for strong climate action to reduce population exposure to the growing threat of future fire weather events compounded with hot and dry conditions.

Key Points

  • The risk of fire weather events combined with hot-dry conditions increased faster than other fire weather (OFW) events from 1981 to 2020

  • Temperature and relative humidity play a prominent role in the trends of compound fire weather (CFW) events

  • Future projections indicate a shift from OFW events to CFW events combined with hot-dry conditions

Plain Language Summary

In recent years, the combined impact of wildfires, heatwaves, and droughts has inflicted significant global damage. However, existing wildfire risk assessments often neglect the specific hazards arising from the co-occurrence of heatwaves and drought, which exacerbate the consequences of wildfires. To bridge this research gap, we systematically quantified the historical changes, meteorological drivers, and future projections of compound fire weather risk (CFW) resulting from the combination of heatwaves and drought. Our findings reveal accelerated increases in the occurrence of CFW events compared with other fire weather (OFW) events, with a rise of 0.6 days per year in frequency and 0.4% per year in intensity. The predominant climatic factors driving the trend of CFW events are changes in temperature and relative humidity. Furthermore, anthropogenic climate change has further intensified the frequency and intensity of CFW events. Projections indicate a substantial rise in fire weather risk associated with heatwaves and drought throughout the 21st century, while OFW risks (such as individual fire weather events) are expected to decline. This escalation in CFW events poses a grave threat of increased population exposure, particularly under high-emission scenarios.

1 Introduction

Widespread increases in wildfire activity have been documented across multiple regions of the globe in recent years, including several high-profile wildfires in the western United States in 2018 and 2020 (Higuera & Abatzoglou, 2021; D. Wang et al., 2021; J. Wang et al., 2021), Brazil's Amazon rainforest in 2019 and 2020 (Yuan et al., 2022), and Australia in 2019–2020 (Lindenmayer et al., 2020). The escalation of destructive fires has had severe social, economic, and ecological impacts, such as the loss of life, damage to homes and other property, and widespread pollution of air and water (Fann et al., 2018; Robinne et al., 2018). In addition, increasing fire activity can exacerbate climate change by contributing significant amounts of greenhouse gases, including carbon dioxide, methane, and nitrous oxide, to the atmosphere (Larkin et al., 2014). These adverse effects highlight the need for better understanding of fire risks to support wildfire adaptation and mitigation efforts.

Fire weather refers to conditions conducive to the occurrence and sustaining of fires (Pausas & Keeley, 2021) and is known to be sensitive to climate change. Several fire weather indices are commonly used to represent the combined influence of different meteorological factors (e.g., wind speed [WS], precipitation [Pr], relative humidity [RH], and temperature [Temp]) and fuel information relevant to the risk of wildfires (Khorshidi et al., 2020; Sharples, 2022). The McArthur Forest Fire Danger Index (FFDI) (Noble et al., 1980), for example, incorporates a measure of vegetation dryness together with the temperature, humidity, and WS on a given day, and is widely applied as a basis for issuing fire weather warnings (Dowdy et al., 2010).

Compound climate extremes are receiving growing attention because the co-occurrence of multiple hazards can amplify their impacts on society and ecosystems (AghaKouchak et al., 2020; Alizadeh et al., 2020; Miao et al., 2022). Wildfires, heatwaves, and droughts frequently arise concurrently due to the intricate interplay of various physical processes spanning multiple spatial and temporal scales (Zscheischler et al., 2018). Due to climate warming, it is anticipated that more severe drought conditions will lead to a rise in both the frequency and intensity of wildfires (M. W. Jones et al., 2022). In recent years, simultaneous occurrences of wildfires and heatwaves, wildfires and droughts, and wildfires and both heatwaves and droughts have been reported to inflict significant damage across various regions of the world (Ribeiro et al., 2022). For example, in the fire season of 2019–2020, Australia experienced wildfires driven by the effects of extremely dry weather, which led to the loss of 33 lives and the destruction of over 3,000 houses (Abram et al., 2021). In the Brazilian Pantanal, unprecedented wildfires in 2020 during combined heatwave and drought conditions burned more than 3.9 million hectares and caused economic losses of about 3.6 billion dollars (Libonati et al., 2022; Silva et al., 2022). The 2022 fire season in southwest Europe—which affected Portugal, Spain, and France—coincided with several heatwaves and has drawn significant international attention due to the scale of destruction. It is believed to be the worst fire season on record since 2006, with at least 469,464 ha burned (Rodrigues et al., 2023). Currently, most existing fire risk assessments tend to focus on the general risk of fires, while overlooking the identification and categorization of specific fire hazards, and particularly those that are compounded by the co-occurrence of heatwaves and/or drought. To address this research gap, there is a need to systematically quantify the global characteristics of compound fire risk resulting from the combination of fire weather, heatwaves, and drought, and to assess historical changes, potential drivers, and future projections of these events.

Previous studies have emphasized the influence of climate change on fire weather (Abatzoglou & Williams, 2016; Dupuy et al., 2020), heatwaves (Kang & Eltahir, 2018), and droughts (Van Loon et al., 2016). Human-induced climate change, caused predominantly by the emission of greenhouse gases, plays a crucial role in the increases in each of these three individual hazards (wildfire, heatwaves, and drought) (Chiang, Mazdiyasni, & AghaKouchak, 2021; Touma et al., 2021; D. Wang et al., 2021; J. Wang et al., 2021). Nevertheless, the influence of anthropogenic forcing on the characteristics (e.g., frequency and intensity) of concurrent fire weather events under hot-dry conditions has not been explicitly quantified. As global temperatures continue to rise, we can expect an increase in the frequency and severity of wildfires, heatwaves, and droughts in many parts of the world (Lehner et al., 2017; Perkins-Kirkpatrick & Lewis, 2020; Perry et al., 2022). As these three hazards become more frequent and severe, they are expected to occur together more often (Bevacqua et al., 2022; Zscheischler & Seneviratne, 2017); however, the extent of this co-occurrence and the specific areas that will be affected are not yet fully understood.

In this study, we conduct a systematic and quantitative analysis of compound fire weather (CFW) events under hot-dry conditions (i.e., CFW–heatwave, CFW–drought, and CFW–heatwave–drought) using climate reanalysis data and evidence from climate models. For simplicity, here we use the term “CFW” to refer to CFW events, and we use the term “OFW” (other fire weather) to describe fire weather events that occur without simultaneous heatwave or drought (see Figure S1 in Supporting Information S1 for detailed classification diagram). We assess patterns and trends associated with recent fire weather, with special emphasis on CFW events. We then discuss the relationship between meteorological drivers and CFW event features to identify critical factors affecting the frequency and intensity of fire weather. Additionally, we quantify the impact of anthropogenic climate forcing on changes in CFW events. Given the significant impacts of CFW events on populations, our study examines future changes in these events and the resulting population exposure across the global land surfaces. The findings from our investigation could inform disaster response measures for wildfires and aid decision-makers in planning for future risks.

2 Materials and Methods

2.1 Data

ERA5 is the latest climate reanalysis produced by the European Centre for Medium-Range Weather Forecasts (ECMWF), providing hourly estimates of a large number of atmospheric, land, and oceanic climate variables (Hersbach et al., 2020). In this study, the ERA5 hourly Temp, 2 m maximum temperature (Tmax), 2 m dewpoint temperature (Td), Pr, and 10 m U and V component wind data for 1981–2020 were converted to a daily scale to calculate the fire weather index. To match the spatial resolution of the climate model data, we interpolated the raw ERA5 data from a 0.25° grid to a 2.5° grid through bilinear interpolation.

Daily climate variables simulated by different experiments from the Coupled Model Intercomparison Project Phase 6 (CMIP6) (Eyring et al., 2016) were used to assess the impacts of anthropogenic activity on CFW event characteristics. The CMIP6 historical model simulations include all-external-forcings simulations (Hist-ALL), greenhouse-gas-only simulations (Hist-GHG), anthropogenic-aerosol-only simulations (Hist-AER), and natural-only simulations (Hist-NAT; solar and volcanic). Comparing the Hist-ALL and Hist-NAT simulations reveals the influence of human-induced climate changes. In addition, the Hist-GHG and Hist-AER simulations were used to isolate the effects of individual human impacts (i.e., greenhouse gases and aerosols) on CFW events. Note that the Hist-ALL experiment ends in 2014, after which the SSP2-4.5 simulations (i.e., a moderate scenario) are employed to extend Hist-ALL simulations until 2020. SSP2-4.5 was selected as it aligns with the current climate goals and efforts of multiple nations to fulfill their existing commitments to mitigate emissions and it has been extensively used in attribution studies (Dong et al., 2022; Ribes et al., 2021). The choice of a finish year of 2020 enables the inclusion of recent observations in the attribution analysis. For future fire weather projections, we chose the SSP2-4.5 and SSP5-8.5 scenarios, representing an intermediate “middle of the road” forcing and a high-emissions “fossil-fueled development” forcing, respectively (O’Neill et al., 2016). The relevant information about available CMIP6 models is listed in Table S1 in Supporting Information S1, and all the model outputs were regridded to a common 2.5° grid using bilinear interpolation. Due to the lack of WS data under the SSP2-4.5 and SSP5-8.5 scenarios in the CNRM-CM6-1 model, the other eight climate models were used for the projections.

To reduce the bias from CMIP6 simulations, a multivariate quantile mapping bias correction method (Cannon, 2018) was used that corrected the marginal distributions and maintained the multivariate dependence structure between the five variables (Temp, Tmax, Pr, RH, and WS) for calculating compound events. The cumulative distribution functions (CDFs) from Hist-ALL simulations were adjusted to the CDFs of ERA5 for the period 1981–2000. Hist-GHG, Hist-NAT, Hist-AER, SSP2-4.5, and SSP5-8.5 simulations were also corrected based on the same CDFs from Hist-ALL simulations to preserve the difference between different experiments and Hist-ALL climates.

Gridded population data at a resolution of 1/8° were obtained from the NASA Socioeconomic Data and Applications Center (SEDAC) (B. Jones & O’Neill, 2016). We used population data sets for four different periods (base period: the 2000s; projection years: the 2030s, 2060s, and 2090s) and under two different SSPs (SSP2 and SSP5). The SSP2-4.5 and SSP5-8.5 scenarios were selected for analyzing future fire weather exposure.

2.2 Methods

2.2.1 Defining CFW Events

We adopted the McArthur FFDI (Noble et al., 1980) to quantify fire danger conditions. This index is intended to provide a quantitative measure of fire weather risk related to the probability of fire occurrence and to the rate of spread. It is computed using the following formula:
FFDI = 2.0 × e ( 0.450 + 0.987 ln D F 0.0345 R H + 0.0338 Temp + 0.0234 WS ) $\text{FFDI}=2.0\times {e}^{(-0.450+0.987\mathrm{ln}\mathrm{D}\mathrm{F}-0.0345RH+0.0338\text{Temp}+0.0234\text{WS})}$ (1)
where DF is the drought factor, Temp is the 2 m air temperature (°C), WS is the wind speed (km/h), and RH is the relative humidity (%). For ERA5, RH is calculated from Temp and Td via the following relation:
RH = 100 112.0 0.1 Temp + T d 112.0 + 0.9 Temp 8 $\text{RH}=100{\left(\frac{112.0-0.1\text{Temp}+{T}_{d}}{112.0+0.9\text{Temp}}\right)}^{8}$ (2)
DF is an empirical estimate of the state of the fuel derived from the Keech-Byram drought index (Fromm et al., 2006), which uses daily temperature and precipitation measurements. To facilitate the comparison of fire weather across space and time, we normalized the daily FFDI values in each grid cell as described by Jolly et al. (2015):
FFDI norm i j = FFDI i j FFDI min FFDI max FFDI i j × 100 ${\text{FFDI}}_{{\text{norm}}_{ij}}=\frac{{\text{FFDI}}_{ij}\,{-\text{FFDI}}_{\min }}{{\text{FFDI}}_{\max }-{\text{FFDI}}_{ij}}\times 100$ (3)

Here, FFDI norm i j ${\text{FFDI}}_{{\text{norm}}_{ij}}$ represents the daily normalized FFDI values, which are bounded by 0 and 100; FFDIij is the daily FFDI for a given grid cell for day i of year j; and FFDImin and FFDImax indicate the historical daily minimum and maximum FFDI for the grid cell. Fire weather days are defined as the number of days when the FFDI norm i j ${\text{FFDI}}_{{\text{norm}}_{ij}}$ was above a threshold value of 50, following Jolly et al. (2015) and Sun et al. (2019). Accordingly, the average difference between the daily normalized FFDI and the threshold of 50 for all fire weather days in a year is taken as the fire weather intensity of that year.

CFW events occurring under hot-dry conditions includes three scenarios: (a) fire weather that occurs during heatwave periods but without accompanying drought; (b) fire weather that occurs during drought periods but without accompanying heatwaves; and (c) fire weather that occurs during compound heatwave and drought periods. Here, a heatwave is defined as a period of at least three consecutive hot days with daily Tmax values exceeding the long-term (1981–2000) daily 90th percentiles for each calendar day (in a 15-day moving window) (Perkins & Alexander, 2013). The Standard Precipitation Index (SPI) is applied to identify drought conditions, and a drought event is defined as occurring when the SPI at a 1-month scale is smaller than −1 (Singh et al., 2021). We converted the monthly SPI to daily values; that is, if the SPI of a month is less than −1, then any day of that month is considered a drought day. Next, we calculated the occurrence in time and space of each independent hazard (fire weather, heatwave, and drought) and identified the number of days with fire weather co-occurring with heatwave (not dry) or drought (not hot), as well as the number of days with co-occurrence of these three hazards. In the context of these definitions, two indices are constructed to measure characteristics of CFW events: CFW frequency—the total number of CFW events observed for any given year—and CFW intensity—the average value of fire weather intensity for all CFW events in any given year. We defined fire weather occurring without simultaneous heatwave or drought as OFW events; for example, individual fire weather, as well as fire weather events that occur in conjunction with other factors (such as wind extremes).

2.2.2 Identification of the Dominant Climate Factor

Climate variables including Temp, RH, WS, and Pr are known to have a substantial impact on fire activity (Flannigan et al., 2016). These four variables are used for calculating the FFDI. Thus, we conducted a partial Mann–Kendall test (PMK test) (Jain et al., 2022; Libiseller & Grimvall, 2002) to assess the influence of the covariates (i.e., Temp, RH, WS, and Pr) on the Theil-Sen trends in CFW event characteristics. The PMK assesses whether the trend detected in the response variable (here, the frequency and intensity of CFW events) is statistically significant after considering the correlation with a given covariate. If the p-value of the PMK test becomes statistically insignificant (e.g., p > 0.05), that is, larger than the significance level of 5%, this suggests that the relevant covariate has a significant effect on the trend in CFW event frequency (or intensity) (Jain et al., 2022; Mediero et al., 2014), and the detected trend is removed. Note that since multiple covariates can be drivers of CFW event trends, the sum of the attribution percentages of all climate variables considered can exceed 100%.

2.2.3 Attribution of Anthropogenic Forcings

Anthropogenic impacts on the properties of CFW events are assessed by comparing the changes in frequency and intensity of CFW events from the CMIP6 simulations including and excluding anthropogenic forcings. We calculated the risk ratio (RR) to demonstrate the impact of anthropogenic forcings on the frequency and intensity of CFW events (Chiang, Greve, et al., 2021; Stott et al., 2016; Zscheischler & Lehner, 2022). The RR of each pixel is defined as P1/P0, where P0 is empirically calculated as the number of CFW event occurrences divided by the total number of days or the cumulative intensity of the total historical reference period under the natural-forcing conditions; P1 is the equivalent value estimated under anthropogenically forced conditions (Hist-ALL, Hist-GHG, and Hist-AER). If RR > 1, it can be concluded that anthropogenic climate change increases the risk of seeing a higher frequency and intensity of CFW events, relative to the natural forcings.

2.2.4 Future Risk Exposure

Exposure is defined here as the number of individuals exposed to compound or OFW events. It is computed in each grid cell by multiplying the annual total number of CFW (or OFW) event occurrences and the number of people exposed to each fire weather event for both the baseline and future periods. As B. Jones et al. (2015) proposes, the exposure change (ΔE) can be decomposed into three parts, including the population effect, the climate effect, and the population–climate interaction effect, estimated by the following equation:
E = C 1 × P + P 1 × C + C × P ${\increment}E={C}_{1}\times {\increment}P+{P}_{1}\times {\increment}C+{\increment}C\times {\increment}P$ (4)
Here, C1 is the number of days with fire weather and P1 is the population in the historical period; ΔC and ΔP are the changes in fire weather days and population, respectively, in the future scenarios compared with the baseline period. Thus, we refer to C1 × ΔP, P1 × ΔC, and ΔP × ΔC to represent the roles of population change, climate change, and the change in their interaction, respectively. Then, three types of effects are calculated using the following equations:
CR pop = C 1 × P E × 100 ${\text{CR}}_{\text{pop}}=\frac{{C}_{1}\times {\increment}P}{{\increment}E}\times 100$ (5)
CR clim = P 1 × C E × 100 ${\text{CR}}_{\text{clim}}=\frac{{P}_{1}\times {\increment}C}{{\increment}E}\times 100$ (6)
CR int = C × P E × 100 ${\text{CR}}_{\mathrm{int}}=\frac{{\increment}C\times {\increment}P}{{\increment}E}\times 100$ (7)

Here, CRpop is the percentage contribution of the population effect; CRclim is the percentage contribution of the climate effect; and CRint is the percentage contribution of the population–climate interaction effect.

3 Results

3.1 Observed Trends in CFW Events

We first compared the characteristics of CFW and OFW events during the period 1981–2020 (Figures 1a and 1b). The probability distribution for these two types of fire weather reveals that, over the past four decades, the total frequency of CFW events was under 3,000 days, whereas the maximum total frequency of OFW was approximately 8,000 days. In terms of the cumulative intensity of fire weather events, the significantly rightward shift (p < 0.05, using a two-sample t-test) of the CFW events distribution with respect to the OFW events distribution indicates that fire weather events compounded with heatwaves, or droughts, or both tend to have higher intensity. Globally, over the past 40 years, OFW events were more frequent than CFW events (54% vs. 46% of the total frequency of fire weather events, respectively). Nevertheless, the cumulative intensity of CFW events accounts for 55% of the total cumulative intensity of fire weather events, exceeding the cumulative intensity of OFW events (45%). The spatial differences between CFW events and OFW events further demonstrate the dominant impact of CFW events. While the frequency of OFW events is spatially more prevalent than that of CFW events, the annual mean intensity of CFW events is approximately 1.5–3 times higher than that of OFW events (Figure S2 in Supporting Information S1). The frequency of CFW risk events has increased over almost all regions of the world, with a global average increase of 0.6 days/yr (Figure 1c). The Amazon region, the western United States, Central Africa, the Mediterranean region, and southeastern Australia have seen the strongest increases in CFW event frequency, with an upward trend of over 2 days/yr. In contrast, trends in OFW event frequency are less significant, and the global average of OFW event frequency exhibits a weak declining trend of 0.04 days/yr (Figure 1e). The frequency of OFW events in the Amazon region and Central Africa even shows a significant decreasing trend of 1.5 days/yr, while significant positive trends are detected for CFW events. The intensities of both CFW events and OFW events show predominantly increasing trends for the period 1981–2020 (Figures 1d and 1f), with stronger increases in global average CFW event intensity (0.4%/yr) than in global average OFW event intensity (0.2%/yr). Positive trends in CFW event intensity are particularly evident in many regions of the northern hemisphere, the Amazon region, and southeastern Australia. These results underscore the higher risk of fire weather events when combined with hot-dry conditions, as compared with OFW events, as reflected in the stronger intensity and greater magnitude of the increasing trends. In addition, the results show that hotspot regions with positive trends in CFW event frequency also exhibit the largest trends in CFW event intensity, implying that regions that have encountered increasing maximum CFW event frequency have also experienced increased excess fire weather severity when CFW events occur.

Details are in the caption following the image

Comparison between metrics associated with compound fire weather (CFW) events and other fire weather (OFW) events. (a, b) show the probability distributions of total frequency and cumulative intensity, respectively, for CFW and OFW events for 1981–2020. The pie chart insets show the proportions of (a) total frequency and (b) cumulative intensity of total fire weather events contributed by CFW and OFW events. Theil-Sen trends in annual (c) CFW event frequency, (d) CFW event intensity, (e) OFW event frequency, and (f) OFW event intensity were calculated using the ERA5 climate reanalysis data set over the period 1981–2020. The trends in global annual average CFW event frequency, CFW event intensity, OFW event frequency, and OFW event intensity are provided at the bottom left of panels (c–f). Diagonal lines indicate significance at the 0.05 level, determined using the Mann–Kendall test.

3.2 Identification of the Dominant Climatic Driver of CFW Events

Figure 2 shows the global percentage contributions of the four climate variables to the significant long-term trends in frequency and intensity seen for the different categories of fire weather (see Methods section “Identification of the dominant climate factor”). In regions where significant trends in CFW event frequency and intensity are observed, Temp is the most dominant factor driving these trends. Specifically, 81.7% (72.1%) of the grid cells in these regions show significant correlations between temperature change and the frequency (intensity) of CFW risk (Figure 2a). These regions are mainly located in the Amazon region, the western United States, Central Africa, the Mediterranean region, southeastern Australia, and northeastern Asia (Figure S3 in Supporting Information S1). RH was found to be the second most important factor influencing the trends of CFW event frequency and intensity in 58.6% and 57.9% of the grid cells in these regions, respectively (Figure 2a). The area of influence for RH in Central Africa and the Amazon region is markedly smaller than that of temperature (Figure S3 in Supporting Information S1). By comparison, Pr and WS were found to contribute relatively little (about 20% of the grid cells) to the significant trends in CFW event characteristics, as they were only detected as drivers in a small part of Africa and the Amazon region (Figure 2a, Figure S3 in Supporting Information S1). The ranking of driving factors affecting trends in OFW event characteristics is consistent with that of CFW event characteristics—Temp and RH are more important drivers than Pr and WS (Figure 2b, Figure S4 in Supporting Information S1). Previous studies have reported that high temperatures can influence fire risk by drying fuels (e.g., grasses, shrubs, and trees), thus making them more flammable and easier to ignite (Gutierrez et al., 2021; Westerling et al., 2006). Consequently, areas that are already dry are often at higher risk of fire hazard when heatwaves occur (Jyoteeshkumar Reddy et al., 2021). RH usually declines as air temperature increases, and low humidity levels lead to further drying of vegetation fuels and therefore lead to fuels becoming increasingly flammable (Trigo et al., 2006).

Details are in the caption following the image

Dominant drivers of trends in fire weather risk. Percentage of grid cells with statistically significant Theil-Sen trends in the frequency and intensity of (a) compound fire weather and (b) other fire weather that can be explained by Theil-Sen trends in four climatic variables.

3.3 Anthropogenic Contributions to CFW

To assess whether the human influence on climate has contributed to a greater frequency of CFW events, we calculated the RR using the different forcing simulations from the CMIP6. According to the historical all-forcing conditions, anthropogenic climate change has increased the frequency and severity of CFW events over most land areas of the world (with RR > 1, Figures 3a and 3b). The most pronounced RR values appear in the Amazon region, where anthropogenic climate change has caused more than 2-fold and 1.3-fold increases in the frequency and intensity of CFW events, respectively. A further increase in RR under greenhouse-gas-only forcing reveals that both the frequency and intensity of CFW events exhibit a distinct response to the presence of greenhouse-gas-driven climate changes, and this is true almost everywhere, with good inter-model agreement (stippling; Figures 3c and 3d). Most notably, the greenhouse gas emissions increase the occurrence frequency of CFW events by approximately 2.5 times relative to natural-only forcing in much of Europe, Africa, the Amazon region, Indonesia, and the United States. Likewise, we discovered that the intensity of CFW events with greenhouse-gas-only forcing is approximately 1.5 times greater than with natural forcing over the Amazon region, the Mediterranean region, and parts of the United States. Conversely, when we focus on the influence of anthropogenic aerosols on CFW event properties, we see the calculated RR for most regions of the world is less than unity, which suggests that anthropogenic aerosols have reduced CFW events in terms of frequency and intensity (Figures 3e and 3f). The spatial distribution of climatological values (1981–2020 mean) for the frequency and intensity of CFW events in the historical all-forcing and greenhouse-gas-only forcing experiments has better consistency with the results of the ERA5 data set than with the natural-only and anthropogenic-aerosol-only experiments, which supports the robustness of our attribution result (Figure S5 in Supporting Information S1). Our findings are consistent with previous results on fire weather attribution (Abatzoglou et al., 2019; Li et al., 2021; Touma et al., 2021), supporting the important role of anthropogenic climate change (mainly greenhouse gases) in aggravating CFW event properties. Conversely, we also find that anthropogenic aerosols may offset the greenhouse-gas-induced increase in the frequency and intensity of fire weather events compounded with hot-dry conditions, which has been proven by previous research from the perspective of general fire weather events (Touma et al., 2021).

Details are in the caption following the image

The risk ratio (RR) under different anthropogenic forcing conditions. RR for frequency (left column) and intensity (right column) of compound fire weather risk over the time period 1981–2020 under historical all forcings (Hist-ALL; (a, b)), historical greenhouse-gas-only forcing (Hist-GHG; (c, d)), and historical anthropogenic-aerosol-only forcing (Hist-AER; (e, f)). Diagonal lines indicate areas where at least 75% of models show RR > 1.

3.4 Future Projection of CFW Events

Future projections for fire weather based on the multi-model ensemble mean of CMIP6 under its worst-case Shared Socioeconomic Pathway climate scenario (i.e., SSP5-8.5) for the end of the current century are shown in Figure 4. Significant increases in CFW event frequency appear worldwide under the SSP5-8.5 scenario (Figure 4a). The most remarkable increases (more than 75 days) are expected to occur in Africa and the Amazon region. The largest intensification of CFW event intensity is predicted in the middle and high latitudes of the northern hemisphere and the Amazon region, with an increase of more than 100% compared with the historical reference period (Figure 4b). The long-term projection under the SSP5-8.5 scenario indicates a relatively small increase (about 50%) in CFW event intensity in areas of the tropics and the southern hemisphere, excluding the Amazon region. Looking at future projections for OFW event frequency at the end of this century, the results are contrary to the changes projected for CFW event frequency. OFW event frequency in most regions is projected to decrease or maintain its historical level into the far future, with the most notable decreases (about 75–150 days) occurring in Africa and the Amazon region (Figure 4c). We note in particular that these two regions are also the regions with the largest increases in CFW event frequency, which indicates that the future increasing emissions will lead to a shift in fire weather from individual fire weather events to more frequent occurrence of conditions compounded with heatwave, drought, or both in Africa and the Amazon region. The rise in frequency and intensity of CFW events is directly linked to the increasing occurrence and duration of warm and dry periods, which are closely linked to climate change. This has created an environment that is highly susceptible to wildfires, unlike any encountered in the past (Mukherjee & Mishra, 2021; Pechony & Shindell, 2010). Globally, OFW event intensity changes only slightly compared with that of the historical reference period (Figure 4d).

Details are in the caption following the image

Projected changes in future fire weather under the SSP5-8.5 scenario. Spatial distributions of future changes in annual mean (a) compound fire weather (CFW) event frequency, (b) CFW event intensity, (c) other fire weather (OFW) event frequency, and (d) OFW event intensity for a future time period (2081–2100) compared with the reference period (1981–2000). Diagonal lines represent the areas that show significant changes between the two periods at the 0.05 significance level, based on the two-sample Student's t-test.

We analyzed the uncertainty of future estimates by comparing these projections with the fire weather under the SSP2-4.5 scenario. Overall, the spatial patterns for future changes in fire risk metrics in the medium-emissions scenario (SSP2-4.5) (Figure S6 in Supporting Information S1) are consistent with those in the highest-emissions SSP5-8.5 scenario. However, the changes are smaller in magnitude under the SSP2-4.5 scenario. By the end of the 21st century, annual CFW event frequency in Africa and the Amazon region is projected to increase by about 50 days relative to the historical climatology under the SSP2-4.5 scenario. For CFW event intensity, the magnitude of increase in the middle and high latitudes of the northern hemisphere is about 50% compared with the 1981–2000 climatological mean. These lower projections in the SSP2-4.5 scenario (compared with the SSP5-8.5 scenario) suggest that some of the future changes in CFW risk could be mitigated through lower greenhouse gas emissions. The change amplitudes for frequency and intensity of OFW events are also lower under the SSP2-4.5 scenario than under the SSP5-8.5 scenario.

Further analysis of future changes in fire weather reveals that the CMIP6 models project continuous increases in the global mean frequency and intensity of CFW events during the 21st century (Figures 5a and 5b). The increasing trends in CFW event frequency are projected to be 0.4 and 0.9 days/yr under SSP2-4.5 and SSP5-8.5, respectively. Estimates of future CFW event intensity show an increasing trend of 1.0%/yr under SSP5-8.5, versus only 0.3%/yr under SSP2-4.5. Conversely, a substantial reduction in OFW event frequency is projected during the 21st century under both future emissions scenarios as a consequence of transformation to CFW events (Figure 5c). The associated decreasing trends in OFW event frequency are 0.1 and 0.2 days/yr under the SSP2-4.5 and SSP5-8.5 scenarios, respectively. The global mean OFW event intensity decreases slightly under both the SSP2-4.5 and SSP5-8.5 scenarios, and there are no obvious differences in the future variation between the two SSPs over time (Figure 5d). The model spread for changes in CFW event frequency, CFW event intensity, and OFW event frequency is larger for the higher-emissions scenario (SSP5-8.5), which indicates a greater uncertainty in the projections under that scenario. Moreover, the two scenarios start to diverge in about 2040.

Details are in the caption following the image

Projected temporal changes in future fire weather. Time series of annual mean anomalies for (a) compound fire weather (CFW) event frequency, (b) CFW event intensity, (c) other fire weather (OFW) event frequency, and (d) OFW event intensity relative to the historical average (1981–2000) under the SSP2-4.5 and SSP5-8.5 scenario. The shading shows the range of the multi-model ensemble between the 25th and 75th percentiles, and the solid lines show the multi-model means.

3.5 Future Population Exposure to CFW Risk

As fire weather events are expected to become more frequent and severe worldwide due to increasing heatwaves and droughts, it is crucial to quantitatively assess the population's exposure to these events for effective risk management. For this purpose, we computed the population exposure to fire weather events under the SSP2-4.5 and SSP5-8.5 scenarios for three future periods, referred to for brevity as the 2030s (2021–2040), the 2060s (2051–2070), and the 2090s (2081–2100). The spatial distribution of future changes in CFW event exposure shows an almost global increase relative to the historical reference period (Figures 6a, 6c, and 6e). The highest changes in population exposure to CFW events are mainly concentrated over India and some parts of Central Africa, reaching more than 10,000 million person-days (with one person-day being one person exposed to a wildfire in one day) in each of the three periods. In addition, the CFW event exposure over regions of Europe, the Amazon region, the United States, Indonesia, and eastern China is projected to increase more than 100 million person-days under the future scenarios. In contrast, future changes in OFW event exposure are more variable at the global scale. Increases of more than 10 million person-days are found in most of Africa, some parts of the Europe, the United States, and India in the 2030s and 2060s; but for the 2090s, OFW event exposure decreases in India and Central Africa with respect to the historical state. In South America and most of Asia, the exposure to OFW events is projected to decrease in the three future periods compared with the historical period due to a decreasing trend in OFW event frequency (Figures 6b, 6d, and 6f). The projected changes in CFW event and OFW event exposure exhibit similar patterns under the SSP2-4.5 scenario but feature a relatively smaller increase across the world, relative to SSP5-8.5 (Figure S7 in Supporting Information S1).

Details are in the caption following the image

Projected changes in population exposure to fire weather under scenario SSP5-8.5. The left panels show the population exposure changes for compound fire weather events in the (a) 2030s (i.e., 2021–2040), (c) 2060s (i.e., 2051–2070), and (e) 2090s (i.e., 2081–2100); the right panels show population exposure changes for other fire weather events in the (b) 2030s, (d) 2060s, and (f) 2090s. The projected changes are calculated from the CMIP6 multi-model means with respect to the historical reference state (1981–2000).

We then decomposed the total change in exposure into three effects: the climate effect, the population effect, and the population–climate interaction effect. At the global level, the exposure to CFW events is dominated by the climate effect. Both the climate effect and the population–climate interaction effect show progressive increases in the three future periods under the SSP2-4.5 and SSP5-8.5 scenarios (Figures 7a and 7b). In comparison, CFW risk exposure caused by population change slowly increases under the SSP2-4.5 scenario and first increases and then decreases under the SSP5-8.5 scenario. As a result of the above changes, the climate effect is found to have the largest contribution to changing population exposure to CFW events for both scenarios, and its role grows over time under the SSP5-8.5 scenario. By the end of this century, under the SSP5-8.5 scenario, the CFW risk exposure caused by the climate effect is roughly 62.6% of total projected exposure, whereas exposure caused by the population effect is only 9.5%, leaving 27.9% due to the interaction effect (Figure 7b). In contrast, the change in CFW risk exposure due to the climate effect is 43.0% of the total for the SSP2-4.5 scenario by the end of this century, whereas the changes in exposure caused by the population effect and the interaction effect are 21.7% and 35.3% of the total, respectively (Figure 7a). For OFW events, exposure is primarily driven by the population effects under both the SSP2-4.5 and SSP5-8.5 scenarios, with continuously increasing contributions over the three future periods (Figures 7c and 7d). The decreasing trend in OFW events in the future leads to a reduction of future population exposure caused by the climate change effect and the interaction effect under the future scenarios. However, this decrease in OFW risk exposure cannot offset the increase in exposure led by population growth effects. Thus, the contribution ratios of both the climate and interaction effects are usually negative, while the population effect exhibits positive contributions. An opposite case that needs to be pointed out is that by the end of the 21st century, population change shows a negative effect, but the contribution of climate and interaction effects is positive under the SSP5-8.5 scenario. This is likely due to the combination of two factors: the population structure under SSP5 characterized by a peak-then-fall trend and a rapid decline in OFW risk occurrence. These factors result in a lower exposure of the population to OFW risk by the end of the century compared with the historical reference level.

Details are in the caption following the image

Effects of changes in climate, population, and their interaction on exposure to fire weather events in the future. The top panels show the decomposition of aggregate global population exposure changes for compound fire weather events under the (a) SSP2-4.5 and (b) SSP5-8.5 scenarios, and the bottom panels show the decomposition of aggregate global population exposure changes for other fire weather events under the (c) SSP2-4.5 and (d) SSP5-8.5 scenarios. In each subplot, the lines in the left section represent the population exposure changes caused by population change (green), climate change (purple), and their interaction effect (blue), while the bars in the right section represent the percentage contributions of the same variables to population exposure changes in three future periods.

4 Discussion and Conclusions

Climate change has led to a rise in wildfire incidence, evidenced by many studies (Abram et al., 2021; Goss et al., 2020). Typically, fire weather events occur during hot and dry weather conditions, which can exacerbate soil water deficits and atmospheric heat and contribute to more-intense and longer-burning fires (Libonati et al., 2022; Sutanto et al., 2020). However, our current understanding of CFW events is limited, which hinders the development of effective adaptation strategies. Therefore, our study aimed to improve understanding of historical fire weather risk and conduct a comprehensive quantitative assessment of various categories of fire weather risk in the future. Our results demonstrate that although fire weather conditions without simultaneous heatwave, drought (OFW events) occurred more frequently during the period 1981–2020, their cumulative intensity was less than that of CFW events, which highlights the more destructive impacts of CFW when it occurs simultaneously with hot-dry conditions. Our results support the argument that fire weather, heatwaves, and droughts influence and promote one another. High temperatures exacerbate the drying out of soils and later warm the atmosphere as it subsequently gains less water from evaporation (Humphrey et al., 2021; Sutanto et al., 2020). The increased atmospheric demand for evaporation enhances sensible heat flux, which can trigger a heatwave event (Aminzadeh et al., 2021). High temperatures and soil moisture deficits are important drivers of fire weather, and as a reinforcing feedback, wildfires emit carbon dioxide and other greenhouse gases that warm the planet and may subsequently increase the likelihood and intensity of drought (Kim & Sarkar, 2017). The frequency and intensity of CFW events have significantly increased over most parts of the global land surface in recent decades, most notably in the Amazon region, the western United States, Central Africa, the Mediterranean region, and southeastern Australia.

Trends in the frequency and intensity of CFW events are increasing, and here we show this can be attributed to rising Temp and decreasing RH. While Pr and WS changes have played a role, their effects have been more limited. We find Pr changes are inadequate in offsetting the warming effect, and fire weather occurrence depends more on Pr frequency than on the amount of Pr (Flannigan et al., 2016; Jain et al., 2022). Our conclusion supports previous work deconstructing the primary drivers of global and regional fire risk (Jain et al., 2022; Richardson et al., 2022; Ruffault et al., 2020). Moreover, anthropogenic climate change has increased the occurrence frequency and intensified the magnitude of CFW events for many regions. We found the most pronounced anthropogenic impacts to be in the Amazon region, Africa, Indonesia, and the Mediterranean region. Further decomposing the contributions from greenhouse gases and anthropogenic aerosol forcings suggests that the changes in frequency and intensity of CFW events are robustly associated with greenhouse gases and partially offset by anthropogenic aerosols.

The global increase in heatwave and drought frequency and intensity under a future warmer climate (Domeisen et al., 2022) is likely to reinforce the occurrence frequency and intensity of CFW events in most parts of the world. Projections of fewer CFW event days and lower intensity under the moderate-emissions SSP2-4.5 scenario compared with the high-emissions SSP5-8.5 scenario emphasize the significance of reducing greenhouse gas emissions to mitigate future CFW events. The most remarkable increases in the frequency of CFW events are expected to occur in Africa and the Amazon region, while the largest increases in CFW event intensity are projected in the middle and high latitudes of the northern hemisphere. Spatial patterns of exposure under varying climate and socioeconomic scenarios show increases in population exposure for CFW events relative to the historical base period. The primary driver of this exposure is the climate change effect, which features rapid increases in the number of CFW event days. As a result, there is an urgent need for climate action plans to reduce overall exposure to the growing conditions of future fire weather events compounded by hot-dry events. A limitation of this study is that it used a simple proxy (FFDI) for fire potential days. FFDI is based solely on weather and ignores the influence of fuel load and human (natural) ignition (M. W. Jones et al., 2022). The role of changes in fire drivers in different seasons is also not assessed in this study. More comprehensive analyses will therefore be needed to reduce uncertainty in fire risk projections and raise awareness of fire management among decision makers.

Acknowledgments

This study was supported by the National Natural Science Foundation of China (42041006), the State Key Laboratory of Earth Surface Processes and Resource Ecology (2022-ZD-03), and the Fundamental Research Funds for the Central Universities. We acknowledge the World Climate Research Programme's Working Group on Coupled Modelling, which is responsible for the Coupled Model Intercomparison Project Phase 6 (CMIP6), and we thank the climate modeling groups listed in Table S1 in Supporting Information S1 for producing and making available their model outputs. We also thank the European Centre for Medium-Range Weather Forecasts (ECMWF) for providing the ERA5 reanalysis data and NASA Socioeconomic Data and Applications Center (SEDAC) for providing the population data.

    Conflict of Interest

    The authors declare no conflicts of interest relevant to this study.

    Data Availability Statement

    CMIP6 model simulations are introduced in Table S1 in Supporting Information S1. The World Climate Research Programme's Working Group on Coupled Modelling is responsible for CMIP6 (Eyring et al., 2016), and relevant model outputs are accessible online through https://esgf-node.llnl.gov/search/cmip6/. The ERA5 hourly temperature, maximum temperature, dewpoint temperature, precipitation, and wind speed data were obtained from the Copernicus Climate Change Service Climate Data Store (Hersbach et al., 2023). Population data were originally derived by NASA Socioeconomic Data and Applications Center (SEDAC) and can be accessed in B. Jones and O’Neill (2020).