Increasing global wildfire activity is associated with a growing occurrence of extreme wildfire events that pose substantial threats to human communities and natural systems due to their rapid and uncontrollable growth (Duane et al., 2021). Extreme fire events often exhibit exceptional fire behavior, including “blow-ups,” pyrocumulonimbus formation, high net fire radiative power (FRP), and even unusual phenomena such as fire whirls (Castellnou et al., 2022; Cova et al., 2023; Darwish Ahmad et al., 2023). Extreme events exert overwhelming demands on suppression resources and operational capacity, and can result in significant costs, loss of life, as well as unique and enduring ecosystem effects (Bowman et al., 2017; Tedim et al., 2020). Defining extreme wildfire events can be challenging due to inconsistencies between different metrics of fire activity and the socio-ecological impacts of wildfire (Tedim et al., 2018). Further, patterns of fire activity vary across gradients of climate and human influence, and what constitutes an extreme event in one location might represent normal fire activity elsewhere (Hantson et al., 2015). However, regardless of how they are defined or measured, a focus on extreme events casts light on the consequential fact that a relatively small subset of short-duration events is disproportionately responsible for overall wildfire impacts (Wang et al., 2021).

The daily rate of fire spread is an important aspect of extreme wildfire behavior that has received little attention until recently but is of growing interest. In the western US, many recent fire years have been marked by conspicuously large, single-day fire spread events. For example, during the record-setting 2020 fire season, 21 fires in California, Oregon, and Colorado achieved 46 single-day blow-ups that exceeded 10,000 ha/d. Such extreme single-day spread events can account for a very large proportion of cumulative area burned; for example, across fire-prone regions of the western US between 2002 and 2020, the largest 10% of single-day fire spread events accounted for 58% of overall area burned, and the top 1% of events accounted for 20% (Coop et al., 2022). Given the positive relationship between fire size and duration (Wang et al., 2020), large, single-day fire spread events can also disproportionately lengthen fire duration and attendant demands on incident management and affected communities. Therefore, using daily fire progression alongside occurrence of single-day extremes to understand patterns of fire duration could reveal valuable insights towards effective incident management.

Recent advances in our understanding of daily wildfire progression across large landscapes are supported by the widespread availability of satellite-based fire detections coupled with methodological advances for their utilization (Parks et al., 2014; Andela et al., 2019). For example, remote sensing and geospatial approaches have linked various aspects of daily fire progression to climate, weather conditions, and fuel (vegetation) types (Veraverbeke et al., 2014; Holsinger et al., 2016; Hart and Preston, 2020; Wang et al., 2023). However, the extent to which these relationships hold across broader continental scales or vary across narrower geographic regions remains uncertain.

Ecoregions provide a useful framework to assess geographic patterns of variation in ecosystems and biophysical processes such as daily fire progression at the subcontinental scale. By integrating a broad suite of climate and vegetation factors that influence fire, ecoregions offer a useful grouping scheme for evaluation of different components of fire regimes and extreme behaviors (Syphard and Keeley, 2020; Park et al., 2021; Pausas, 2022). For instance, climate variation along two axes, average actual evapotranspiration (AET) and climate water deficit (CWD), shapes vegetation distributions and segregates ecoregions across North America (Stephenson, 1998). AET, which measures both energy and water availability, predicts net primary productivity (Rosenzweig, 1968). Ecoregions with short-statured and sparse vegetation display lower AET on average compared to more productive conifer and broadleaf forests. Similarly, plant species distributions are influenced by CWD, which measures drought or evaporative demand exceeding moisture availability (Lutz et al., 2010). Warm desert ecoregions have high annual deficits, whereas cool or mesic forests and tundra have lower deficits. Because of these relationships to productivity and fuel moisture, gradients of AET and CWD also predict components of fire regimes across ecoregions, such as annual area burned and burn severity (Parks et al., 2014). For example, fuel availability increases and fuel type shifts from herbaceous to woody and from conifer to broadleaf as a function of AET. Similarly, fuel moisture declines and flammability increases with CWD. Given these mechanisms, we would also expect ecoregional AET and CWD climate normals to influence aspects of daily fire progression. For example, low fuel availability could limit daily fire spread in ecoregions with high CWD, whereas herbaceous and woody fuel types can promote very high rates of spread in regions with lower CWD. Conversely, where AET is high and fuels are plentiful, high fuel moisture and less flammable fuel types (e.g., broadleaf forests) may generally constrain daily fire spread.

In addition to ecoregional differences in climate and vegetation distributions, ecoregions vary in a suite of other ecological and human factors that could also lead to differing patterns of extreme single-day events and daily fire progression. For example, ecoregional differences in standing fuel biomass may influence occurrence of extreme single-day events and fire durations moreso than annual productivity inferred from AET climate normals. Ecoregions also face different pressures of environmental change that are relevant to daily fire progression, including patterns of invasion by highly flammable annual grasses (Fusco et al., 2019), vegetation responses to climate change (Hossain and Li, 2021), and climate change velocity (e.g., rate of climatic warming; Loarie et al., 2009). Such climatic changes are broadly expected to promote extreme single-day events and modulate fire seasons (Wang et al., 2015; Kirchmeier-Young et al., 2019; Halofsky et al., 2020; Coop et al., 2022). Consequently, occurrence of single-day blow-ups and their realized impacts are likely to differ among ecoregions, along with associated differences in characteristic fire duration and seasonality.

The increasing availability of gridded climate data (Abatzoglou et al., 2018) and improved consistency of high-resolution fire perimeter maps between the United States and Canada provide new opportunities to gain insights into extreme single-day events and daily fire progression at continental and sub-continental scales. The purpose of this study was to quantify extreme single-day spread events and assess their relative influence in shaping patterns of net fire effects across northern North America, defined here as Canada and the United States. Our primary objective was to characterize North American daily fire progression and extreme single-day events at the continental and ecoregional scales. Our secondary objective was to explore how ecoregions differ in other aspects of daily fire progression, namely total fire duration (number of days of burning), duration of consecutive daily extreme spread, and fire season (timing of fire spread and extreme events within the year), as well as trends in annual occurrence of extreme single-day events over time. Our third objective was to contextualize average and extreme single-day spread along ecoregional annual AET and CWD climate normals, which may directly influence daily fire spread rates and contribute to any differences observed among ecoregions. By exploring the geography and climate space of extreme single-day wildfire spread events, we aim to help identify where and when blow-ups occur most frequently and are most consequential in terms of proportion of area burned. Our findings are particularly relevant to fire management agencies and communities that are increasingly threatened by rapidly spreading fires and changing fire seasonality, as well as land managers tasked with managing novel fire activity and landscape-scale impacts.

As human and ecological communities are increasingly threatened by fires burning under novel fuels and weather conditions (Parisien et al., 2020; Higuera et al., 2023), it is imperative to understand how patterns of daily fire spread vary across North America to provide context for these events and to monitor outcomes and trends. Here, we characterized daily fire progression and extreme single-day fire spread or blow-up events across northern North America and at the ecoregional scale. In doing so, we found substantial ecoregional variation in the typical size of large and extreme single-day blow-up events (i.e., daily area burned thresholds), the prevalence of these single-day events, and how they contribute to overall area burned. This ecoregional variation is likely a result of the spatially and temporally varying factors that control fire spread, such as fuel type, fuel amount and contiguity, and prevalence of fire-conducive weather (Krawchuk and Moritz, 2011; Parisien et al., 2014). We also show that fire duration and fire seasons vary across ecoregions, which is likely a reflection of the specific fire regime characteristics and their controlling factors (Archibald et al., 2013; Jain et al., 2017; Cattau et al., 2020).

Single-day extreme fire spread events are disproportionately shaping North American landscapes (Wang et al., 2021). For example, daily fire spread exceeding 1,704 ha/day (continental large single-day events) burned an average of 2.3 million ha/year in the US between 2002–2021, or 2.2x more area than US public lands treated through mechanical means and prescribed fire annually on average between fiscal years 2002–2021 (USDA Forest Service FACTS, 2023). Furthermore, the low proportions of single-day spread events classified as large or extreme under our ecoregional and continental definitions emphasize that relatively few spread days are responsible for substantial proportions of total area burned, with their cumulative area burned exceeding 80% (large & extreme single-day events) and 50% (extreme single-day events) of the total area burned in some ecoregions. These events have meaningful ecological outcomes. For example, Naficy et al. (2021) found that daily fire spread rates of spatially disjoint day of burning patches was one of the top predictors of fire severity and refugia probability in forests of the Pacific Northwest, with spread rates around the continental or Northwestern Forested Mountains large thresholds (~1,500–2000 ha day⁻¹) associated with higher probabilities of high-severity burn and reduced refugia. Given that the extent and severity of their landscape impacts inherently occur within brief timeframes, extreme single-day events pose substantial challenges to land managers and considerable risks to the safety of affected communities and firefighters (Coen et al., 2020). Our analysis revealing their differential prevalence, extent, and outcomes among ecoregions points to the need to better understand the confluence of top-down and bottom-up factors that lead to extreme single and multi-day spread events, as well as their consequences.

Even though average fire spread rates are greater in some ecoregions (e.g., Great Plains, North American Deserts), it is important to contextualize this finding with the duration of fires (Figure 4). For example, those ecoregions with the highest daily spread rates also tend to experience lower duration fires, as previously reported by Andela et al. (2019), resulting in a smaller mean final fire size relative to other ecoregions with lower fire spread rates and longer duration fires. Longer duration fires in the Northern Forests, Northwestern Forested Mountains, and Taiga ecoregions are likely a result of larger woody fuels and thick surface fuels (e.g., pine duff, peat) that can retain burning embers even under variable weather conditions, which then contribute to fire spread when weather becomes more fire conducive (Krawchuk and Moritz, 2011; Prichard et al., 2017). Furthermore, these regions’ rugged terrain can complicate suppression efforts, and their lower population densities may lead to patterns of fire management that allow fires to continue burning when less infrastructure and fewer lives are at risk (Park et al., 2021).

Our analyses also indicate that longer fires are generally associated with greater maximum daily spread rates. This suggests that large or extreme spread days may be more likely to occur the longer a fire burns, or that their occurrence extends fire duration by rapidly expanding the area of active burning. Alternatively, the same top-down and/or bottom-up conditions that promote rapid fire spread may also be conducive to longer fire durations. Regardless of the directions of causality and mechanisms linking maximum daily spread rate and duration, these patterns may be expected to produce the exponential relationship between final fire size and duration in many ecoregions (Xi et al., 2019; Wang et al., 2020). Furthermore, even if relatively few fires achieve consecutive extreme spread days, longer duration fires inherently have greater opportunity for consecutive extreme behavior, particularly given the potential for greater daily spread with each additional day of burning. Such feedbacks between daily spread and fire duration raise the possibility of a self-reinforcing loop wherein extended fire duration amplifies the potential for large or extreme spread days, further extending the lifespan of a fire and likelihood of subsequent extreme behavior, drastically expanding the final fire size.

In addition to differential spread and fire duration, fire seasons and seasonality of extreme single-day events also varied among ecoregions. Although some ecoregions have more confined fire seasons (e.g., Tundra, Taiga, Northern Forests, Northwestern Forested Mountains, Marine West Coast Forest), fire occurs nearly year-round in some ecoregions (e.g., Great Plains, Eastern Temperate Forests), with extreme single-day events typically distributed across nine of the twelve months of the year. Interestingly, a bimodal fire season was observed in the Mediterranean California, Marine West Coast Forests, Eastern Temperate Forests, and Great Plains ecoregions, a phenomenon also reported by Benali et al. (2017) in the central and southwest regions of the continent. In some locations like Mediterranean California, the November peak is likely associated with Santa Ana winds (Westerling et al., 2004). Santa Ana winds are also likely responsible for the later average seasonal timing of extreme single-day events in the region. Elsewhere, particularly in broadleaf forests, the phenological stage of vegetation may influence fire season timing (Parisien et al., 2023) or that of extreme single-day spread. For example, in some ecoregions with bimodal fire seasons such as the Eastern Temperate Forests, Great Plains, and Marine West Coast Forests, early season fires and extreme single-day events may occur before vegetation green-up; then fire activity diminishes until late summer and early fall when fuels again become sufficiently dry to carry fire. The abrupt truncation of the fire season in the Northern Forests, Taiga, and Tundra ecoregions is likely due to rapid onset of weather that is prohibitive to fire spread (i.e., snow) compared to more southerly ecoregions. Similar to our spread rate and fire duration findings, ecoregional differences in vegetation, fuel loads, and climate likely contribute to the observed variability in fire season and seasonality of extreme single-day events.

However, there are several key caveats to interpreting our fire seasonality results. First, our analyses only considered wildfires ≥400 ha; the inclusion of spread days from smaller wildfires, as well as those from prescribed and cultural burning practices would likely result in greater fire season heterogeneity and pyrodiversity. Second, we used relatively broad EPA Level 1 Ecoregions to group fires across the continent. This choice of spatial grouping unit could influence our results, with finer-scale spatial groupings likely to reveal greater heterogeneity in regional fire regimes. Because of this finer-scale heterogeneity, grouping fires across broader spatial scales may also contribute to some of the bimodal fire seasons we observed. For example, we merged the Eastern Temperate Forests ecoregion with the relatively small but pyroregionally unique Tropical Wet Forests region, which comprises southern Florida. Fires may occur throughout most of the year in Florida, in contrast to the Appalachians where fire generally occurs in the fall when precipitation and relative humidity are low following the seasonal accumulation of fine fuels in the form of falling leaves. Thus, ecoregions do not necessarily coincide with pyroregions (Galizia et al., 2021), and pyroregional diversity within larger ecoregions creates further nuance in interpreting the patterns of fire seasonality presented here.

Although understanding ecoregional variation in the seasonal timing of large and extreme single-day events throughout a year is valuable for management considerations, it is also crucial to examine how the prevalence of these events has changed over time, particularly given expectations that they will occur more frequently in a warmer and drier future (Wang et al., 2015; Kirchmeier-Young et al., 2019; Halofsky et al., 2020). At the time of our analysis, annual frequencies have not increased since 2002 in most ecoregions or across the continent; however, extreme events are occurring more frequently in the Northwestern Forested Mountains (Supplementary Figure S5). Notably, the Northwestern Forested Mountains are the most fire-prone ecoregion in our dataset in terms of having the greatest number of fires larger than 400 ha and the most fire spread days (e.g., Tables 1, 2). Therefore, although the occurrence of extremes is broadly projected to increase due to climate change, our analysis of a short timeseries demonstrates that outlier extreme events are already occurring more frequently in the most fire-prone ecoregion of North America despite its comparatively low average daily spread rate. However, it is important to recognize that our ability to detect temporal trends in the frequency of extreme single-day events may be limited by the relatively short duration of the time period over which we are able to interpolate daily fire progression.

Anticipating extreme single-day events is critical yet challenging given the wide range of conditions under which fire occurs across the continent (Figure 1B; Parisien et al., 2016). As such, understanding ecoregional prevalence and extent of extreme single-day events will be important to improve resource deployment and community protection, particularly given the substantial differences among ecoregional extreme thresholds (e.g., 3,829 ha to 16,626 ha). This vast difference in extreme thresholds coupled with the variability in human populations and development patterns across the continent result in differential risk to communities and infrastructure (Nagy et al., 2018; Park et al., 2021). For example, in Mediterranean California, extreme single-day events exceeding 13,885 ha/d are particularly concerning given the high human population density and geography of the area. Understanding this risk can help identify communities that should be preemptively evacuated, and strategically plan points of engagement that may be more effective given the likelihood of the large single-day spread events (Thompson et al., 2018). In contrast, the exceptionally high extreme threshold in the North American Deserts (16,626 ha/day) may potentially be mitigated by growing the invasive species management programs within the region and by creating effective fuel breaks. However, response options will inevitably vary elsewhere, such as the Northern Forests ecoregion that is characterized by high fuel loads, low human population density, and extreme event sizes exceeding 12,634 ha/day with the largest mean fire size of any ecoregion. While population density is fairly low, this ecoregion has many rural, remote, and dispersed communities that often include First Nations Indigenous people who are disproportionately impacted by these extreme events (Davies et al., 2018; McGee, 2021). Furthermore, evacuations are complicated by having few or indirect roads to these remote communities (Tepley et al., 2022). Human impacts due to fire duration are also important to consider for those communities that are experiencing long evacuations (Stidham et al., 2011) and the strain on fire fighters who have extended deployments on these long-duration incidents (Vincent et al., 2016; Jeklin et al., 2020). Ecoregional links to climate and ecoregional fire spread rates provide useful information for tailoring management efforts to specific regions and for anticipating future fires and extreme single-day events.

Ecoregional variability of fire spread rates (mean daily area burned, large and extreme single-day events) suggests that broad-scale vegetation and climatic factors captured in ecoregional delineations have bottom-up and top-down effects on fire spread rates. For example, the ecoregions with the highest average daily spread rates and most large and extreme single-day events, calculated either by frequency or percent of area burned, were the Great Plains and Mediterranean California, which are generally characterized as drier ecoregions largely comprised of grasslands and shrubland fuel types which tend to drive rapid fire spread (Cheney et al., 1993; Cruz et al., 2022). The high daily fire spread rates in the North American Deserts ecoregion could be linked to invasive annual grasses that promote rapid fire spread [e.g., Bromus tectorum (cheatgrass); Balch et al., 2013]. Conversely, ecoregions with lower daily spread rates including the Marine West Coast Forests, Eastern Temperate Forests, and Northwestern Forested Mountains generally have higher biomass and lower CWD, though the Northern Forests ecoregion does not follow this pattern. Greater average spread in the Northern Forests despite low CWD may be a function of bottom-up factors, such as landscapes of highly contiguous fuels.

Fire spread rates (daily, large, extreme) generally increase with CWD annual climate normals. The Temperate Sierras ecoregion may deviate from this overall relationship between fire spread and climate due to specific landscape patterns unique to this region. Specifically, fires in isolated and relatively small mountain ranges (i.e., sky islands) are likely limited in size and spread rates by the lack of large, contiguous areas of fuel (Wardle et al., 1997). Nonetheless, this general relationship between fire spread and CWD broadly agrees with other studies showing that metrics of fire activity (fire frequency, area burned, fire severity) can vary along climatic gradients (Krawchuk and Moritz, 2011; Parks et al., 2014). However, the lack of a strong relationship is not completely unexpected, as Archibald et al. (2013) demonstrated a weak relationship between fire regime and climate. For example, the lack of correlation between fire spread rates and AET in our dataset may be attributed to the relatively low variability among ecoregions’ 30-year AET annual climate normals. Eastern Temperate Forests are a clear outlier with high AET, but they also had the lowest mean fire size and a low-moderate number of spread events. Consequently, annual normal AET and CWD may help set the stage for fire by influencing vegetation distributions and fuel conditions, but daily spread may be more strongly influenced by other finer-scale vegetation or fire weather drivers (Potter and McEvoy, 2021; Jain et al., 2022) or internal self-reinforcing weather dynamics (i.e., winds) that drive extreme blow-up behavior (Coen et al., 2018).

Here, we define ‘extreme’ based entirely on daily fire spread area. We do not account for other fire characteristics that may be considered extreme in other contexts, such as community impacts (evacuations, smoke, loss of life and property), firefighting resources needed, ecological consequences of fire (fire severity, carbon emissions), or physical and hydrological impacts (debris flows, erosion, landslides, flooding). As such, extreme in this context may not always have negative impacts on human or natural communities, and conversely many events not labeled as “extreme” here may have wide-ranging human and ecological impacts. Consequently, more research is needed to understand how large single-day events correspond to community, ecological, physical, and hydrologic impacts.

Extreme single-day spread clearly poses a threat to communities in fire-prone regions of North America, as experienced by communities such as Paradise, California (2018), Detroit, Oregon (2020), Fort McMurray, Alberta (2016), among many others. As we continue to navigate novel climate conditions that are resulting in more wildfires and shifting fire regimes, understanding extreme single-day fire spread events is critical for planning and conservation of our landscapes and communities. Our approach of studying daily fire progression complements past geographic characterizations of wildfire regimes, which primarily relied on ignition points, dates, and final fire sizes. By providing detailed spatio-temporal fire spread information, mapping daily fire progression can facilitate the fine-scale analyses that are needed to identify the top-down and bottom-up mechanisms that create blow-ups and their consequences to ecosystems and society.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found at: https://doi.org/10.5061/dryad.2jm63xswg.

JB: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Visualization, Writing – original draft, Writing – review & editing. JC: Conceptualization, Funding acquisition, Project administration, Supervision, Writing – original draft, Writing – review & editing. MK: Funding acquisition, Writing – review & editing. CN: Funding acquisition, Writing – review & editing. M-AP: Funding acquisition, Writing – review & editing. SP: Conceptualization, Funding acquisition, Investigation, Writing – original draft, Writing – review & editing. CS-R: Conceptualization, Funding acquisition, Writing – original draft, Writing – review & editing. EW: Conceptualization, Funding acquisition, Writing – review & editing.