The Impact of Stratospheric Aerosol Injection on Extreme Fire Weather Risk
Abstract
Stratospheric aerosol injection (SAI) would potentially be effective in limiting global warming and preserving large-scale temperature patterns; however, there are still gaps in understanding the impact of SAI on wildfire risk. In this study, extreme fire weather is assessed in an Earth system model experiment that deploys SAI beginning in 2035, targeting a global temperature increase of 1.5°C above pre-industrial levels under a moderate warming scenario. After SAI deployment, increases in extreme fire weather event frequency from climate change are dampened over much of the globe, including the Mediterranean, northeast Brazil, and eastern Europe. However, SAI has little impact over the western Amazon and northern Australia and causes larger increases in extreme fire weather frequency in west central Africa relative to the moderate emissions scenario. Variations in the impacts of warming and SAI on moisture conditions on different time scales determine the spatiotemporal differences in extreme fire weather frequency changes, and are plausibly linked to changes in synoptic-scale circulation. This study highlights that regional and spatial heterogeneities of SAI climate effects simulated in a model are amplified when assessing wildfire risk, and that these differences must be accounted for when quantifying the possible benefit of SAI.
Key Points
-
Impacts of stratospheric aerosol injection (SAI) on extreme fire weather frequency varies spatially and regionally
-
SAI limits projected 21st century increases in extreme fire weather risk in many global regions
-
Anomalous drying under SAI simulations leads to increases in extreme fire weather frequency in some regions
Plain Language Summary
Under human-caused climate change, wildfire risk is expected to increase in many parts of the globe as temperatures increase and precipitation, humidity, and wind patterns shift. This study investigates how one form of solar climate intervention—injecting sunlight-reflecting aerosols high into the atmosphere, or Stratospheric Aerosol Injection (SAI)—could slow down this trend. The resulting global cooling, when SAI is simulated in a climate model, limits the increases in meteorological conditions that can lead to wildfire spread. However, in some locations there are reductions in precipitation and humidity and increases in wind speed, which leads to regional increases in wildfire risk under SAI. This study highlights both the benefit and unintended consequences of SAI on global wildfire risk and the need to continue understanding the role of climate intervention in limiting increases in extreme climate events.
1 Introduction
Larger and more severe fires in western North America, Siberia, Australia and the Amazon have been linked to anthropogenic warming through enhanced long-term drying and more frequent extreme fire weather conditions (Abatzoglou et al., 2021; Balch et al., 2018; Goss et al., 2020; Juang et al., 2022). As global temperatures continue to warm due to increases in human-driven greenhouse gas emissions, drier conditions and a higher likelihood of extreme fire weather conditions are expected over many regions (Abatzoglou et al., 2019; Park et al., 2022; Touma et al., 2021). Under a moderate anthropogenic emissions scenario, approximately 3-fold increases in extreme fire weather conditions are expected by 2100 over large parts of the globe (Park et al., 2022), and under a high emission scenario, extreme fire weather frequency is expected to emerge above the historic variability in 70% of the globe by 2080, including western North America, the Amazon, and Siberia, potentially leading to a “new normal” of wildfire occurrence and severity (Abatzoglou et al., 2019; Touma et al., 2021).
Recent studies have explored the role of climate intervention in limiting the impacts of projected human-driven emissions on the climate. Stratospheric aerosol injection (SAI), one method of solar climate intervention (SCI), has been at the forefront of this effort given its technological and political plausibility (NASEM, 2021). Global Earth system model (ESM) simulations have shown that SAI could potentially limit increases in global mean temperature and preserve hemispheric temperature gradients (Richter et al., 2022; Tilmes et al., 2018). Due to large and numerous non-linearities, however, SAI may also lead to regionally and temporally heterogeneous impacts on extreme heat and precipitation (Dagon & Schrag, 2017; Ji et al., 2018; Lee et al., 2020; MacMartin et al., 2022; Tye et al., 2022) and drought (Abiodun et al., 2021; Alamou et al., 2022; Coughlan de Perez et al., 2022; Odoulami et al., 2020). These impacts of SAI have been thoroughly discussed in Tye et al. (2022). More recently, compound extremes under SAI have been investigated over Africa, where impacts on warm-dry events were also found to vary spatially but cold-dry events generally increased over the continent (Obahoundje et al., 2023). SAI has also been found to decrease evaporative demand but simultaneously reduce precipitation over Africa, thereby complicating drought outlook and management in many regions (Abiodun et al., 2021). Additionally, irreducible uncertainty driven by internal climate variability can highly influence the impact of SAI on extreme climate events, especially for the first few years after climate intervention is first implemented (Barnes et al., 2022; Keys et al., 2022).
The potential benefit of SAI on wildfire risk and emissions has been explored in previous studies. Using a 4-member SAI simulation experiment in an ESM with uniform injection, Burton et al. (2018) found regional differences in the impact of SAI on fire danger. However, the study was not able to capture the impact of time-varying magnitudes of SAI, and by only assessing time slices of the simulations, the temporal variability of these impacts was not investigated. A more recent study that leveraged experiments from the Geoengineering Model Intercomparison Project 6 found that both SAI and solar dimming have regionally varying impacts on burn area and fire emissions when trying to keep temperature changes in a high-emissions scenario at more moderate levels (Tang et al., 2023). By using burn area and fire emissions products from the ESM experiments, the study allows insight into annual changes in wildfire, but does not explicitly quantify changes in daily fire weather conditions that can shape the severity of fires and potential damages to human and natural systems.
The creation of “large ensembles” of SAI ESM experiments has accelerated in recent years to allow a more robust quantification of the role of climate intervention on atmospheric, land, and ocean processes amid internal climate variability. The Geoengineering Large Ensemble (GLENS; Tilmes et al., 2018) is a 20-member experiment using the Community Earth System Model version 1 (CESM1) with the Whole Atmosphere Community Climate Model (WACCM) that deploys SAI in 2020 to prevent further increases in global mean surface temperature and maintain north-to-south and equator-to-pole temperature gradients under the high emissions RCP8.5 scenario. Tye et al. (2022) found that, while SAI prevented some large changes in extreme dry and wet days expected under RCP8.5, SAI in GLENS led to large decreases in heavy precipitation days and large increases in consecutive dry days over parts of South America and western Africa. On the other hand, SAI produced large increases in heavy precipitation days and reductions in consecutive dry days in parts of Southeast Asia and Australia. Using GLENS, Abiodun et al. (2021) also found that SAI could reduce the upper limit of drought stress by reducing temperatures but could increase the lower limit by impeding precipitation over much of Africa. Additionally, using machine learning methods, Barnes et al. (2022) showed that SAI impacts in GLENS could be detected within 1 year of SAI implementation for temperature extremes, and within 15 years for precipitation extremes.
More recently, a new large ensemble SAI experiment has been created. The Assessing Responses and Impacts of SCI on the Earth system with Stratospheric Aerosol Injection (ARISE-SAI) effort (Richter et al., 2022) uses version 2 of CESM and version 6 of WACCM. ARISE-SAI has similar temperature-stabilization goals and employs the same injection protocol as GLENS, but has a more realistic deployment year and baseline warming scenario. Similar to GLENS, heterogeneous impacts of SAI on temperature and precipitation over the globe also emerge in the ARISE-SAI experiment. In this study, the potential benefit of SAI on projected increases in wildfire risk under climate change are investigated in the ARISE-SAI experiment. By using this large-ensemble experiment, the role of SAI in modulating daily fire weather conditions is examined while accounting for internal climate variability. Additionally, the role of SAI in modifying drought conditions on daily to seasonal scales that lead to extreme fire weather conditions is investigated.
2 Data and Methods
2.1 SSP2-4.5 and ARISE-SAI-1.5 Simulations
The climate change experiments utilized here are the 10-member Community Earth System Model version 2 with the Whole Atmosphere Community Climate Model version 6 (CESM2 (WACCM6)) simulations under the Shared Socioeconomic Pathway scenario (SSP2-4.5) forcing scenario from 2015 to 2100. Parallel climate intervention experiments are the 10-member ARISE-SAI-1.5 simulations that also follow the SSP2-4.5 scenarios, but with SAI implemented from 2035 until 2069. The SAI protocol in this experiment, denoted by “1.5”, ensures that global mean temperature remains close to 1.5°C above the pre-industrial temperature, while also preserving both the north-south temperature gradient and the equator-to-pole temperature gradient. The SSP2-4.5 and ARISE-SAI-1.5 simulations are further detailed in Richter et al. (2022), and the data are publicly available at https://doi.org/10.5065/9kcn-9y79 and https://doi.org/10.26024/0cs0-ev98, respectively.
2.2 Fire Weather Index
A modified version of the Canadian Forest Fire Weather Index (FWI; Dowdy et al., 2009; Wagner, 1987) is used to quantify meteorological conditions related to wildfire spread. The FWI is based on an empirical model that was developed for conditions observed in Canada, but has been employed for global studies (e.g., Flannigan et al., 2013; Jolly et al., 2015; Park et al., 2022). As with any empirical index, the contributing variables do not interact linearly and so the relationship may evolve or change over time as different variables exacerbate or counteract each other (Barbero et al., 2020). However, FWI is also well understood and used operationally by many countries around the world so is an appropriate tool to examine the potential impacts of SAI.
Figure S1 in Supporting Information S1 shows the workflow of calculating the FWI for the ARISE-SAI-1.5 and SSP2-4.5 experiments. The equations for each step are thoroughly documented in Dowdy et al. (2009). The model variables used to compute FWI are near-surface daily average temperature (TREFHT), precipitation (PRECT), relative humidity (RHREFHT), and wind speed (U10). While daily maximum or noon-time temperature is more commonly used in the calculation of FWI instead of average temperature, five of the 10 SSP2-4.5 ensemble members do not have daily maximum temperature available (Richter et al., 2022). To include all 10 members of the SSP2-4.5 ensemble and better represent internal climate variability, daily average temperature is used. Importantly, changes in extreme fire weather frequency computed from maximum temperature or average temperature under both SSP2-4.5 and ARISE-SAI-1.5 are highly comparable (see Figure S2 in Supporting Information S1).
For all 20 of the combined SSP2-4.5 and ARISE-SAI-1.5 simulations, three moisture codes are first computed for the full simulation periods for each model grid point. While these three moisture codes explicitly account for meteorological variables only, they are empirically formulated to represent relative moisture levels in different parts of the canopy and on different time scales. The Fine Fuel Moisture Code (FFMC) is computed using TREFHT, PRECT, RHREFHT, and U10, and represents moisture conditions for shaded litter fuels and on daily timescales. The Duff Moisture Code (DMC) is computed using TREFHT, PRECT, and RHREFHT, and represents moisture conditions for decomposed organic material and on monthly timescales. Lastly, the Drought Code (DC) is computed using TREFHT and PRECT and represents drying deep into the soil and on monthly time scales. The FFMC and U10 are used to calculate the Initial Spread Index (ISI), which reflects the fire spread rate, and DMC and DC are used to calculate the Build-up Index (BUI), which reflects the potential heat release or severity of fires. The FWI is then calculated using the ISI and BUI. The FWI does not account for variations in vegetation types and does not capture land surface dynamics, but instead quantifies daily to seasonal scale atmospheric conditions that can lead to the buildup of burnable fuels and wildfire spread. All moisture codes and indices are unitless (Dowdy et al., 2009).
2.3 Extreme Fire Weather Events
To identify extreme fire weather events, the 99th percentile threshold of the FWI is first calculated for each grid point across all 10 SSP2-4.5 ensemble members in the “baseline” period, that is, 2015–2035. Similar extreme thresholds using the FWI and other fire weather indices have been used in previous studies to characterize conditions that underlie large and severe wildfires (Burton et al., 2018; Goss et al., 2020; Kirchmeier-Young et al., 2017; Touma et al., 2021, 2022). Moreover, using the full ensemble best accounts for the impact of internal variability on the FWI distribution in the baseline period. Then, extreme fire weather events are identified as days that meet or exceed the 99th percentile for all ensemble members in the SSP2-4.5 and ARISE-SAI-1.5 experiments for the full simulation period, 2015 to 2069.
Changes in extreme fire weather events are evaluated from the baseline period (2015–2035) for the globe over two periods after SAI deployment: 2035–2054 and 2050–2069. Two overlapping 20-year periods are used to match the number of years from the baseline period. By using the grid-box specific 99th percentile of the daily FWI in the baseline period as an extreme threshold, every ice-free land location will experience 3–4 extreme fire weather events per year in the baseline period. Given that not all these locations experience wildfire risk, six fire prone global regions are used for further analysis (Table S1 in Supporting Information S1). Annual and seasonal time series of changes in extreme fire weather events for these regions are assessed. Changes are defined as robust when at least two-thirds of the ensemble agree on the sign of change or difference.
2.4 Assessing Differences in FWI Fuel Moisture Codes
The roles of the FWI fuel moisture codes, FFMC, DMC, and DC are assessed in driving the daily variability of the FWI under both the SSP2-4.5 and ARISE-SAI-1.5 scenarios. First, FFMC, DMC, and DC are averaged for each grid point over the baseline period and two analysis periods for each ensemble member. The difference between each ARISE-SAI-1.5 ensemble member and the ensemble-mean SSP2-4.5 fuel moisture code is then calculated. Robust differences are identified when at least two-thirds of the ensemble members agree on the sign of the differences. The spatial and ensemble spread of each of the codes over each region during extreme fire weather days is also shown. The statistical significance of the differences between SSP2-4.5 and ARISE-SAI-1.5 distributions is calculated using the Kolmogorov-Smirnov test.
3 Results
3.1 Global Fire Weather Risk Under Climate Change
Under the SSP2-4.5 scenario, extreme fire weather events are projected to increase in many fire-prone regions across the globe throughout the 21st century (Figures 1, 2, and 3). Robust (>two-thirds ensemble member agreement) increases of up to eight extreme fire weather events per year are seen over parts of the Amazon, Siberia, central and eastern Africa, and Southeast Asia (∼100% increase) and of up to four over the Mediterranean, Canada, and southern Africa (∼50% increase) during 2035–2054 compared to 2015–2035 (Figure 1a and Figure S2 in Supporting Information S1). Increases in these regions result from both an expansion and amplification of extreme fire weather seasons (Figure 2). By 2050–2069, further increases of up to ∼150% in the frequency of extreme fire weather events occur in some of these regions (Figure 1b), as their extreme fire weather seasons continue to amplify (Figure 2).

(a and c) Ensemble-mean change in extreme fire weather event frequency (exceedance of 99th percentile of FWI in 2015–2035) under SSP2-4.5 in (a) 2035–2054 and (c) 2050–2069 compared to SSP2-4.5 in 2015–2035. (b and d) Ensemble-mean difference between ARISE-SAI-1.5 and SSP2-4.5 annual frequency in (b) 2035–2054 and (d) 2050–2069. Areas in white are desert, ocean, or ice, and areas in gray show less than two-thirds ensemble-wide agreement on the sign of change. Boxed regions are selected for further analysis shown in Figures 2, 3, and 5 (see Table S1 in Supporting Information S1 for regional boundaries).

(a, c, e, g, i, and k) Change in 30-day extreme fire weather event frequency per 105 km2 compared to 2015–2035 in 2035–2054 and (b, d, f, h, j, and l) 2050–2069 in the SSP2-4.5 (coral, red) and SAI-1.5 (light purple, dark purple) simulations for (a and b) western North America, (c and d) northeast Brazil, (e and f) western Amazon, (g and h) Mediterranean, (i and j) west Central Africa, and (k and l) northern Australia. The thick lines represent the median change across the ensemble, and the shading around each line shows the interquartile range (IQR) of the ensemble distribution. The gray shading shows the IQR of the 2015–2035 ensemble distribution.

Change in extreme fire weather event annual frequency per 105 km2 compared to 2015–2035 ensemble-mean for the SSP2-4.5 (coral solid line) and ARISE-SAI-1.5 (blue dashed line) simulations for (a) western North America, (b) northeast Brazil, (c) western Amazon, (d) Mediterranean, (e) west central Africa, and (f) northern Australia from 2020 to 2069. The thick lines represent the 10-year moving ensemble median, and the shading represents the ensemble interquartile range.
Changes in extreme fire weather frequency also show large decadal and multi-decadal variability under SSP2-4.5 (Figure 3). In both northeast Brazil and the western Amazon, an additional ∼15 extreme fire weather events per 105 km2 are projected by 2050. However, over the subsequent 20 years, these decline to ∼10 extreme fire weather events per 105 km2 (Figures 3b and 3c). Similarly, a rapid increase of ∼7 extreme fire weather events per 105 km2 is evident over northern Australia between 2040 and 2050 followed by a rapid decline by 2065 compared to 2015–2035 (Figure 3f).
The decadal variability in extreme fire weather events under SSP2-4.5 is tied to the large annual variability in the FWI moisture codes in some regions (see Figure S5 in Supporting Information S1). For example, in the western Amazon, the FFMC and DC increase continuously under warming, however, the DMC stabilizes and slightly decreases at the end of the simulation period, causing extreme fire weather event frequency to drop. In northern Australia, all three moisture codes begin to decrease at the end of the period, leading to reductions in extreme fire weather frequency. Interestingly, these decadal variations are not present when assessing global averages of extreme fire weather conditions across the CMIP6 ensemble under SSP2-4.5 (Park et al., 2022). This may be due to model uncertainty across the CMIP6 ensemble in the representation of decadal variability over different regions, and differences in the baseline period used in the analysis (Park et al. (2022) used 1950–1979 while this study uses 2015–2035).
Note that by using a baseline period of 2015–2035, increases in extreme fire weather event frequency are beyond those already observed in recent years. Both observational and model-based studies have concluded that in some regions, statistically significant changes in the frequency of extreme fire weather and wildfire occurrence, as well as wildfire size, have already emerged above pre-industrial and historic levels of natural variability (Abatzoglou et al., 2019; Goss et al., 2020; Touma et al., 2021). Additionally, our findings of increased extreme fire weather frequency are congruent with previous studies, confirming that global warming, even under a moderate emissions scenario, will continue to amplify wildfire risk in many vulnerable regions.
3.2 Global Fire Weather Risk Under SAI
The impact of SAI as simulated in the ARISE-SAI-1.5 experiment is to limit increases in extreme fire weather frequency over much of the globe, though the magnitude varies regionally and through time. In northern South America, central Asia, the Mediterranean, and parts of Southeast Asia, there are approximately two fewer extreme fire weather events per year than in the SSP2-4.5 simulations within the first two decades after deployment (2035–2054; Figure 1b), and approximately four fewer within the last two decades of the simulation period (Figure 1d). For northeast Brazil and the Mediterranean, this means that extreme fire weather frequency remains similar to that in the baseline period (Figures 3b and 3d), limiting the expansion and amplification of the fire weather season seen under the SSP2-4.5 scenario (Figures 2c, 2d, 2g, and 2h).
There are, however, several regions where SAI does not prevent increases in extreme fire weather events or, instead, drives higher increases compared to SSP2-4.5. Over western Amazon and northern Australia, there are no robust differences in the annual frequency of extreme fire weather events between ARISE-SAI-1.5 and SSP2-4.5 during the first two decades after SAI deployment (Figures 1, 3, and 3f). Two and a half decades after SAI deployment, extreme fire weather frequency in the ARISE-SAI-1.5 and the SSP2-4.5 simulations begin to differ, with SAI leading to large and robust increases in extreme fire weather events (Figures 1, 3, and 3f). In west central Africa, SAI leads to increases in extreme fire weather frequency, beyond those observed in SSP2-4.5 throughout the intervention period (2035–2069) (Figure 3e), and leads to further amplification and expansion of the fire season (Figures 2i and 2j). These results show that there are strong temporal and spatial variations in the impacts of SAI on extreme fire weather events. These variations are driven by the non-uniform impact of SAI on daily to seasonal drought conditions and are further discussed in the following section.
The impact of SAI on extreme fire weather also varies seasonally. For example, in the western Amazon, SAI leads to four more extreme fire weather events per month and per 105 km2 in the early baseline fire season compared to SSP2-4.5, but two fewer events in the late fire season during the first two decades after SAI deployment (Figure 2e). On the other hand, SAI leads to less frequent extreme fire weather events (∼two events/month/105 km2) over western North America in the early fire season compared to SSP2-4.5, and slightly more frequent events in the late fire season (∼one event/month/105 km2; Figure 2a). By mid-century (2050–2069), the impact of SAI is more consistent throughout the respective fire seasons; increases in extreme fire weather events in the western Amazon and decreases in western North America occur continuously between June and November under the ARISE-SAI-1.5 experiment (Figures 2b and 2f).
3.3 SAI Impacts on Daily to Seasonal Drying
To understand spatial and temporal variations in the impact of SAI on extreme fire weather risk, the roles of the FWI fuel moisture codes—that is, FFMC, DMC, and DC—are examined. These quantities represent the variability of moisture conditions from daily to seasonal timescales. Robust reductions in FFMC, DMC, and DC in the ARISE-SAI-1.5 experiment are prevalent where extreme fire weather frequency increases are prevented, such as over Central America, the Mediterranean, the Sahel, and South Asia (Figure 4). Conversely, regions with larger increases in extreme fire weather frequency in ARISE-SAI-1.5 compared to SSP2-4.5, such as west central Africa and the western Amazon, show robust increases in all three drought codes (Figure 4), but only increases in west central Africa are statistically significant (p-value <0.01; Figures 5c and 5e). However, robust and statistically significant (p-value <0.05) increases in drought codes in ARISE-SAI-1.5 compared to SSP2-4.5 do not necessarily lead to increases in extreme fire weather event frequency, which is the case over northeast Brazil and northern Australia (Figures 4, 5, and 5f).

(a and b) Ensemble-mean difference in Fine Fuel Moisture Code (FFMC), (c and d) Duff Moisture Code (DMC), and (e and f) Drought Code (DC) between SSP2-4.5 and ARISE-SAI-1.5 in 2035–2054 (a, c, and e) and 2050–2069 (b, d, and f). Areas in white are desert, ocean, or ice, and areas in gray show less than two-thirds ensemble-wide agreement on the sign of the difference between SSP2-4.5 and ARISE-SAI-1.5. FFMC, DMC, and DC are unitless, with higher values representing higher likelihood of fire weather conditions.

Ensemble distribution of Fine Fuel Moisture Code (FFMC), Duff Moisture Code (DMC), and Drought Code (DC) in SSP2-4.5 (orange) and ARISE-SAI-1.5 (purple) in 2015–2034 (left), 2035–2054 (middle) and 2050–2069 (right) for (a) western North America, (b) northeast Brazil, (c) western Amazon, (d) Mediterranean, (e) west central Africa, and (f) northern Australia. The box represents the interquartile range of the ensemble and over the region, and the horizontal line represents the median. The whiskers represent the rest of the distribution and outliers are shown with points above and below the whiskers. Statistically significant differences between moisture code distributions under SSP2-4.5 and ARISE-SAI-1.5 at 0.01 (black borders) and 0.05 (gray shading) levels calculated using the Kolmogorov-Smirnov test are shown for the respective boxplot pairs. FFMC, DMC, and DC are unitless, with higher values representing higher likelihood of fire weather conditions.
This discrepancy in changes in the FWI moisture codes and extreme fire weather frequency is due to differences in the seasonality of changes among moisture codes (see Figure S4 in Supporting Information S1). For example, increases in the FFMC in west central Africa in ARISE-SAI-1.5 are widespread throughout the fire season, and coincides with large increases in the DMC and DC, leading to large increases in extreme fire weather event frequency (Figures 1, 2, and 3e, and Figure S4e in Supporting Information S1). However, in northeast Brazil, increases in the FFMC in ARISE-SAI-1.5 are widespread but relatively smaller throughout the season, and do not coincide with the large increases in the DMC and DC, leading to small changes in extreme fire weather frequency compared to the baseline period (Figures 1, 2, and 3b, and Figure S4b in Supporting Information S1). This leads to the lack of synchrony among drier levels of moisture codes needed to reach extreme thresholds of FWI and is tied to the varied impact of SAI on temperature, precipitation, relative humidity and wind speed on different time scales.
The impacts of SAI on DMC and DC in the period immediately following deployment (2035–2054) are weaker than those in the second half of the analysis period (2050–2069) for some regions including northeast Brazil (Figures 4c, 4e, and 5b, and Figures S5 and S6 in Supporting Information S1). However, west central Africa, the Mediterranean, and Northern Australia show large differences between ARISE-SAI-1.5 and SSP2-4.5 in DMC and DC immediately after deployment (the Mediterranean also has large increases in the FFMC) (Figures 4, 5, and 5f, and Figures S5 and S6 in Supporting Information S1). The spatiotemporal differences in the changes in DMC and DC in both the SSP2-4.5 and ARISE-SAI-1.5 scenarios might be driven by variations in the enhancement of land-atmosphere feedbacks under increases in CO2 (Berg et al., 2016). The temporal variations in changes in the moisture codes under SAI could also be linked to the differences in the timing of direct cooling, as well as the timing of rapid adjustment and feedbacks from cloud and water vapor changes (Kashimura et al., 2017).
The changes in the FFMC, DMC, and DC under ARISE-SAI-1.5 could be linked to changes in synoptic systems in some regions. Specifically, decreases in precipitation under ARISE-SAI-1.5 over west central Africa and the western Amazon could reflect changes in monsoonal circulation (see Figure S3 in Supporting Information S1). Over west central Africa, increases in DMC and DC occur during the West African monsoon and post-monsoon seasons (July to October; see Figure S4 in Supporting Information S1), which has been shown to weaken and produce less precipitation under SAI in previous studies (e.g., Da-Allada et al., 2020; Sun et al., 2020). However, increases in the FFMC occur throughout the year over this region and could be linked to changes in mesoscale weather systems. Similarly, SAI leads to the largest increases in DMC and DC moisture codes in the western Amazon during September to November, which could be related to the weakening of the South American monsoon under SAI (Sun et al., 2020). Here, however, the largest changes in FFMC coincide with changes in DMC and DC, and therefore could also be linked to changes in the monsoonal circulation.
4 Discussion and Conclusions
Based on the ARISE-SAI-1.5 experiment, which uses one SAI strategy to limit warming to 1.5 degrees Celsius above pre-industrial levels, the increases in the frequency of extreme fire weather events over much of the globe that are projected under a moderate warming scenario are dampened. Impeding further increases in extreme fire weather frequency brings relief to regions that have already experienced anthropogenically driven increases in drought and extreme fire weather conditions such as the Mediterranean region (Curt & Frejaville, 2018; Jain et al., 2022). However, the ability of SAI to mitigate increases in extreme fire weather events is not regionally uniform. In west central Africa and the western Amazon there are increases in extreme fire weather event frequency evident in the ARISE-SAI-1.5 simulations that are greater than those expected under the SSP2-4.5 scenario. Similar to previous studies that have assessed SAI potential in mitigating changes in dry and wet extremes in a warming climate (Lee et al., 2020; MacMartin et al., 2022; Sovacool et al., 2022), these results point to the need for the evaluation of SAI strategies to account for impacts beyond changes in global surface temperatures. Additionally, limiting increases in fire weather risk could be used as a target for SAI strategies—however, the spatial heterogeneity in the FWI, as well as the daily temporal resolution and dependence of multiple variables of the FWI could make this an ineffective or unrealistic target (Lee et al., 2020).
The increased global 21st century extreme fire weather risk under the SSP2-4.5 warming scenarios evident in this study is in general agreement with recent literature that assess future wildfire risk (Abatzoglou et al., 2019; Park et al., 2022; Touma et al., 2021). Similarly, the reductions in extreme fire weather events in the boreal regions of North America, Europe, and Asia, as well as northeast Brazil in the ARISE-SAI-1.5 simulations shown in this study support previous studies that investigate the impact of SCI on burn area, fire emissions, and fire danger (Burton et al., 2018; Tang et al., 2023). However, the increases in wildfire risk under SAI over west central Africa, northern Australia, and the western Amazon are not evident in previous work. This discrepancy could be attributed to differences in SAI strategies, baseline scenarios, and variables used for wildfire assessment (e.g., FWI vs. burn area). Therefore, this study widens our understanding of wildfire risk by accounting for daily to seasonal wildfire-relevant conditions and employing simulations with a more realistic baseline scenario and deployment year.
While previous studies have assessed the role of SAI on extreme events (Barnes et al., 2022; Ji et al., 2018; Obahoundje et al., 2023; Tye et al., 2022) and drought conditions (Abiodun et al., 2021; Alamou et al., 2022; Coughlan de Perez et al., 2022; Odoulami et al., 2020), the use of a fire weather index allows for an understanding of the impact of SAI on all fire-relevant variables and therefore wildfire risk. While fire weather indices are highly sensitive to changes in temperature under future climate change scenarios (Abatzoglou et al., 2019; Flannigan et al., 2016; Touma et al., 2021), this study shows that reduction in temperature alone under SAI does not always lead to fewer extreme fire weather events. A limitation, however, is that the impact of warming or SAI on fuel availability for wildfires has not been assessed, which have shown to play a role when examining SAI impacts on annual fire emissions (Tang et al., 2023). While previous studies have shown large SAI-driven reductions in gross primary production and leaf area index in the GLENS simulations compared to RCP8.5 (e.g., Tye et al., 2022), it is unclear how these reductions relate to fuel availability or limitations for wildfire. Additionally, studies have shown that model parameterizations can drive large uncertainties in the response of vegetation to SCI strategies (Dagon & Schrag, 2019). More nuanced understanding of vegetation and wildfire dynamics under SAI for different regions is necessary to tease out this relationship (Zarnetske et al., 2021).
The role of changes in natural or human ignition in both the SSP2-4.5 and ARISE-SAI-1.5 experiments is also not assessed in our study. While studies have shown that lightning events that occur with little-to-no rainfall (“dry lightning”) have historically ignited large fires (Kalashnikov et al., 2022), the relatively coarse resolution of the CESM2(WACCM6) model does not allow for relevant processes to be captured sufficiently. Understanding the role of SAI in potentially changing dry lightning events through both changes in weather systems and atmospheric chemistry could lead to a more holistic understanding of wildfire risk under SAI. However, the probability of human ignition would be comparable between the ARISE-SAI-1.5 and SSP2-4.5 experiments given that there are no differences in the assumptions around the spatial and temporal evolution of population densities and land-use land-cover change.
The large ensemble approach facilitates an understanding of the role of internal climate variability in modulating forced responses in both the climate change and SAI simulations. In some regions, the impact of SAI on extreme fire weather frequency is consistent across the ensemble (e.g., the Mediterranean and west central Africa), but in others, the impact is clouded by internal variability (e.g., western North America and northern Australia), and may require simulations with additional ensemble members for a clearer signal to emerge, or with a stronger SAI forcing scenario. Additionally, decadal variability in extreme fire weather frequency in both the SSP2-4.5 and ARISE-SAI-1.5 experiments in some regions may require longer simulations to quantify the full impact of SAI on extreme fire weather throughout the 21st century. Given that global temperatures continue to increase under continued increases in CO2 concentrations (even as CO2 emissions start to decrease) in the second half of the century under the SSP2-4.5 scenario (Meinshausen et al., 2020; Visioni et al., 2021), longer simulations and a larger ensemble could enable the isolation of decadal variability from the forced response during this period (Deser et al., 2020).
This study is the first to shed light on the potential impact of SAI on global wildfire risk by assessing daily fire weather conditions in a plausible, large ensemble, SAI experiment, and paves the way for further assessments of the impact of SAI on extreme fire weather. Future studies should assess the roles of SAI scenario choice(s), as well as the ESM(s) used for SAI experiments, in shaping the impact on wildfire risks. Different SAI scenario choices, which includes the choice of baseline scenario, time of SAI deployment, temperature or other meteorological goals, injection strategy, and SAI interruptions or phase-outs, could lead to a multitude of outcomes on wildfire risk (MacMartin et al., 2022). For example, a lower temperature target (e.g., 1° above pre-industrial temperature) may lead to increased dampening of fire risk in the Mediterranean or Western Amazon but could also lead to a further amplification in west central Africa. Additionally, model uncertainties can impact the amount of SAI needed, the stratospheric climate response, and the subsequent impact on surface climate variables, including those important for wildfire risk (Visioni et al., 2021). This study broadens our understanding of the potential benefits and unintended consequences of SAI that is needed to make informed decisions about the role of climate intervention in future climates.
Acknowledgments
We would like to acknowledge high-performance computing support from Cheyenne (https://doi.org/10.5065/D6RX99HX) provided by NCAR's Computational and Information Systems Laboratory, sponsored by the National Science Foundation. This work was supported by the National Center for Atmospheric Research, which is a major facility sponsored by the National Science Foundation under Cooperative Agreement No. 1852977. D.T. and J.W.H. were supported by the Walter Scott, Jr. College of Engineering at Colorado State University and the LAD climate group. M.T. and K.D. were supported by SilverLining through its Safe Climate Research Initiative.
Open Research
Data Availability Statement
SSP2-4.5 and ARISE-SAI-1.5 data are publicly available at DOIs https://doi.org/10.5065/9kcn-9y79 and https://doi.org/10.26024/0cs0-ev98, respectively. Analysis and visualization scripts are available at https://doi.org/10.5281/zenodo.7976073.