Fire as a fundamental ecological process: Research advances and frontiers

Fire is a powerful ecological and evolutionary force that regulates organismal traits, population sizes, species interactions, community composition, carbon and nutrient cycling and ecosystem function. It also presents a rapidly growing societal challenge, due to both increasingly destructive wildfires and fire exclusion in fire‐dependent ecosystems. As an ecological process, fire integrates complex feedbacks among biological, social and geophysical processes, requiring coordination across several fields and scales of study. Here, we describe the diversity of ways in which fire operates as a fundamental ecological and evolutionary process on Earth. We explore research priorities in six categories of fire ecology: (a) characteristics of fire regimes, (b) changing fire regimes, (c) fire effects on above‐ground ecology, (d) fire effects on below‐ground ecology, (e) fire behaviour and (f) fire ecology modelling. We identify three emergent themes: the need to study fire across temporal scales, to assess the mechanisms underlying a variety of ecological feedbacks involving fire and to improve representation of fire in a range of modelling contexts. Synthesis: As fire regimes and our relationships with fire continue to change, prioritizing these research areas will facilitate understanding of the ecological causes and consequences of future fires and rethinking fire management alternatives.


| INTRODUC TI ON
Fire is an Earth system process that has operated for many millions of years.The current context of fire ecology studies (Archibald et al., 2018;Bond, Woodward, & Midgley, 2005;Bowman et al., 2009;Krawchuk, Moritz, Parisien, Dorn, & Hayhoe, 2009;Pausas, Keeley, & Schwilk, 2017;van der Werf et al., 2006) reflects a remarkable paradigm shift-from earlier concepts of fire as a destructive and irreversible force to the current concepts of fire as an inherent and fundamental process influencing most terrestrial ecosystems on Earth (He & Lamont, 2018;Pausas & Bond, 2019).Fire ecologists now view fires as dynamic ecological forces that have evolutionary consequences and are fundamentally shaped by human actions.The goal of fire ecology is to understand the diversity of ways in which fire affects organisms and ecosystems on Earth (Figure 1).
Fire ecologists have constructed an increasingly nuanced, sophisticated and mechanistic understanding of the variable nature of fire as part of the ecological system.Fire is now recognized as a recurrent process, resulting in fire regimes that have direct ecological effects and act as selective forces by shaping species traits throughout the histories of entire lineages (He, Lamont, & Pausas, 2019;Simon et al., 2009).Moreover, fire regimes are important at multiple levels of biological organization, influencing populations, communities and ecosystems.
There is an amazing diversity of fire regimes on Earth.The geographic distribution of fire has been mapped based on current global fire activity (Andela et al., 2019;Krawchuk et al., 2009) and classified into geographically distinct fire regimes called pyromes (Archibald, Lehmann, Gomez-Dans, & Bradstock, 2013).Further, the concept of pyrodiversity-the spatial and temporal heterogeneity of fire regimes-has been examined in ecological contexts such as functional biodiversity and food webs (Bowman et al., 2016).However, a number of questions remain about the causes and ecological consequences of variation in fire regimes.The resulting magnitude and diversity of ongoing fire ecology research is challenging to integrate, given the many different lines of inquiry based on the concept of fire as a central ecological process.
Two notable features of recent fire activity frame fire ecology research.First, many fires are planned by people (see 'prescribed fire' in Box 1) due to the importance of both maintaining fire-adapted systems and the long history of using fire as a management tool (Ryan, Knapp, & Varner, 2013).People also exclude fires from fire-adapted systems, with ecological consequences such as the disappearance of tropical and temperate savannas (Fill, Platt, Welch, Waldron, & Mousseau, 2015;Overbeck et al., 2015).Second, land use and ongoing climate change are altering characteristics of individual fires and changing fire regimes, in some cases pushing them outside the historical range of variability in terms of frequency, size, seasonality or severity (Abatzoglou & Williams, 2016;Balch et al., 2018;Kelly et al., 2013;Miller et al., 2019;Walker et al., 2018).Many recent fires have had negative consequences for natural ecosystems and humans (Balch et al., 2018;Stevens-Rumann et al., 2018).There is a pressing need to project future fire activity under varying scenarios of climate change and land management strategies (Bowman, Murphy, Williamson, & Cochrane, 2014).
Our central goal in this paper is to advance knowledge of fire ecology.As the magnitude of fire ecology research increases, it becomes nutrient cycling and ecosystem function.It also presents a rapidly growing societal challenge, due to both increasingly destructive wildfires and fire exclusion in firedependent ecosystems.As an ecological process, fire integrates complex feedbacks among biological, social and geophysical processes, requiring coordination across several fields and scales of study.
2. Here, we describe the diversity of ways in which fire operates as a fundamental ecological and evolutionary process on Earth.We explore research priorities in six categories of fire ecology: (a) characteristics of fire regimes, (b) changing fire regimes, (c) fire effects on above-ground ecology, (d) fire effects on below-ground ecology, (e) fire behaviour and (f) fire ecology modelling.
3. We identify three emergent themes: the need to study fire across temporal scales, to assess the mechanisms underlying a variety of ecological feedbacks involving fire and to improve representation of fire in a range of modelling contexts.

Synthesis:
As fire regimes and our relationships with fire continue to change, prioritizing these research areas will facilitate understanding of the ecological causes and consequences of future fires and rethinking fire management alternatives.increasingly important to identify priorities in understanding the ecological and evolutionary implications of changing fire activity.Here, we attempt to synthesize the major areas of fire ecology research.
We consider six priority areas, identified through an open online survey distributed broadly to the fire ecology research community but focused in the US (see Supporting Information).We conducted bottom-up coding and qualitative text analysis of responses to the question: 'What are the biggest unmet scientific challenges currently in fire research?'Qualitative analysis grouped common text themes from survey responses into six priority categories.Members of the Future of Fire Consortium further refined the content of these six priority categories during a two-day workshop.We briefly summarize the state of knowledge in each category and propose avenues for progress in our understanding of fire as a fundamental ecological process.Our overall perspective, reflecting the composition of the consortium members, focuses on biophysical and ecological aspects of fire.We also recognize humans as central components of fire regimes, as described effectively in recent work (Maezumi et al., 2018;Roos, Zedeno, Hollenback, & Erlick, 2018).Ultimately, any review based on survey data and expert opinion will produce inevitable gaps in scope and emphasis.However, the value of such a process is to find patterns not evident from any single viewpoint or discipline.

| FIRE REG IME A S AN ECOLOG I C AL FAC TO R
Fire regimes are important because they help characterize and classify the diversity of fire and its ecological impacts into a simplified set of categories (Agee, 1993).This variation in fire activity and fire effects, over space and time, fundamentally shapes the structure, composition and dynamics of biotic communities across most of Earth's terrestrial ecosystems.Simplifying the diversity of fire activity into fire regimes is as fundamental to fire ecology as simplifying the diversity of plant assemblages into communities is to plant ecology.

| Characterizing past fire regimes
Fire regimes are described with a number of metrics, including fire frequency (point-specific mean return intervals or area-based fire rotation periods), size, seasonality, intensity (the rate of energy release), severity (the direct impacts of fire), type (ground, surface, crown, mixed) and mode of combustion (flaming or smoldering; Keeley, Bond, Bradstock, Pausas, & Rundel, 2012).Past fire activity is interpreted from historical sources such as observational records and maps of area burned (Morgan, Losey, & Trout, 2014), and from palaeoecological proxy archives such as fire scars in tree-rings, tree age structures, charcoal particles preserved in sediments and organic compounds preserved in ice cores (Conedera et al., 2009; Figure 2).Many datasets derived from these sources are publicly available through international databases such as the Global Charcoal Database and NOAA's International Multiproxy Paleofire Database (Gross, Morrill, & Wahl, 2018;Marlon et al., 2016;Power et al., 2008).A key remaining challenge to fire ecology is characterizing and interpreting the high spatial and temporal variability in the characteristics of fire regimes within and among biomes (Figure 1; Table 1).
Few observational datasets or palaeoecological archives of past fire are spatially contiguous.In remote and non-forested regions, palaeoecological data are particularly sparse.In cold tundra and arid grassland ecosystems (with low above-ground productivity), little charcoal is produced from fires and few, if any, trees exist.
Tree-ring fire histories are predominantly from temperate forested ecosystems with surface fire regimes.Where high intensity fires kill most trees or where trees do not form annual growth rings

Available fuel:
The portion of the total fuel that would actually burn under various environmental conditions.Fuel availability is largely influenced by fuel moisture content.
Burn severity: Synonym of fire severity.Burn severity and fire severity refer to the magnitude of effect that fire has on the environment.This can be measured in many ways, using ground-based metrics or remotely sensed data (see Table S1).'Soil burn severity' refers primarily to fire effects on soil properties (e.g.soil structure, soil and soil surface organics, changes in erodibility) and belowground plant parts.'Vegetation burn severity' refers to injury to and mortality of above-ground vegetation.
Crown fire: Fire that occurs in the vegetation canopy.The occurrence of crown fire usually depends on heat released from burning in surface fuels, but under some combinations of conditions (e.g.extreme wind, high connectivity between surface and canopy fuels, high canopy density and/or continuity, extremely dry live fuels) fire can propagate itself through the vegetation canopy independently of surface fuels.

Fire behaviour:
The manner in which a fire reacts to the influences of fuel, weather and topography; quantified via rate of spread (m/s), residence time (s), intensity (kW/m), flame lengths (m) and combustion phase (e.g.smoldering, flaming).
Fire climate: Composite pattern of weather elements over time that affect fire behaviour in a given region.Like climate, fire climate is defined over multiple decades and over regional spatial scales.

Fire ecology:
The study of the relationship between fire and populations, communities and ecosystems.
Fire intensity: the energy released by a fire per unit time per unit area.
Fire regime: A generalized description of the typical characteristics determined from multiple fires (e.g.intensity, frequency, size, extent, type, seasonality) and their ecological impacts (e.g.severity); fire regimes are often defined for a dominant vegetation or ecosystem type, and inherently include variability over space and time.
Fire severity: See burn severity.
Fire weather: Weather conditions that influence fire ignition, fuel conditions and fire behaviour, including wind, atmospheric temperature, stability, humidity, and ignition factors like lightning.
Fuel: Any combustible material including live and dead biomass.Petroleum products and buildings can be additional fuels that influence fire behaviour and effects in the wildland-urban interface.

Fuel load:
The amount of fuel present expressed quantitatively in terms of dry weight of fuel per unit area (kg/m 2 ).See also available fuel.
Prescribed fire: Any wildland fire intentionally ignited by management actions in accordance with applicable laws, policies and regulations to meet specific objectives.
Wildfire: An unplanned wildland fire including those caused by lightning, unauthorized human causes, escaped prescribed fires and all other wildland fires where the management objective is fire containment.
(i.e.tropical forests and savannas), tree-ring records are very sparse or absent (but see Baker & Bunyavejchewin, 2017).In addition to the need for increased spatial coverage, key time periods in the past, such as the Medieval Climate Anomaly (950-1250 CE), may provide important analogues to modern and predicted scenarios of changing fire activity (Kelly et al., 2013;Pierce, Meyer, & Jull, 2004).
Moreover, although fire history records provide valuable data on fire occurrence and frequency, new proxies need to be developed to reconstruct additional characteristics of fire regimes, such as fire severity, fire size and fire temperature (Dunnette et al., 2014;Gosling, Cornelissen, & McMichael, 2019;Leys, Commerford, & McLauchlan, 2017).The detection and quantification of variation in fire regimes is the first step to understanding and predicting their sensitivity to environmental change and the associated ecological consequences.

| Characterizing current fire regimes
Advances in the temporal and spatial resolution of imagery from Earth observation satellites have led to unprecedented descriptions of recent fire activity at the global scale (Figure 2).These methods utilize satellite observations of either fire activity or preand post-fire imagery, with varying levels of field observations for ground verification.Daily fire detection has provided new insights into the seasonality of burning (Giglio, Csiszar, & Justice, 2006;Roy, Boschetti, Justice, & Ju, 2008), and the influence of fireseason length on fire size and total burned area (Andela et al., 2019).Severity metrics derived from MODIS data have been compared within and among regions to define essential fire regime characteristics (Rogers, Soja, Goulden, & Randerson, 2015;Singleton, Thode, Meador, & Iniguez, 2019).Other metrics have been used to model fire impacts on key variables such as carbon emissions (Rogers et al., 2014;Walker et al., 2018).Together, these datasets have revealed much greater heterogeneity in fire effects than previously characterized (Cansler & McKenzie, 2012;Collins et al., 2017).Accurately applying these methods across regionally and globally diverse fire regimes remains a substantial challenge, because existing satellite-based products such as MODIS detect actively burning areas, and assessment of individual fires or fire behaviour is still in development.Overcoming this challenge is a key research need, because the underlying heterogeneity in fire activity is a critical source of ecological diversity within and across landscapes, regions and biomes.

| Characterizing fire regime changes
Detecting changes in fire regimes remains a pressing challenge for ecologists.The ability to characterize fire regimes has been based on construction of probabilistic estimates of fire regime characteristics (McCarthy, Gill, & Bradstock, 2001).These probabilistic distributions are then compared to recent fire events to assess the likelihood of a shift to a different fire regime (Bigio, Swetnam, & Baisan, 2016;Chipman et al., 2015;Kelly et al., 2013).Changes in fire regimes can be detected given a relatively long time period or spatially dense networks of samples (Taylor, Trouet, Skinner, & Stephens, 2016), with some of the clearest examples of change reflecting shifts in humandominated fire regimes (McWethy et al., 2010).Beyond changes in fire frequency and area burned, other changing components of fire regimes have been identified in recent years, including changing spatial patterns of fire severity (Harvey, Donato, & Turner, 2016b;Miller, Skinner, Safford, Knapp, & Ramirez, 2012;Steel, Koontz, & Safford, 2018) and the lengthening of wildfire seasons (Jolly et al., 2015).These different aspects of fire regime can have unique 1. Regimes Increasing the spatial and temporal coverage of fire history records 1. Regimes Linking satellite-derived products of actively burning areas to the diversity of fire regimes

Changing regimes
Integrating fire-adapted plant traits into global fire models

Changing regimes
Predicting fire probability in both fuel-and climate-limited ecosystems under future climate conditions

Changing regimes
Including the many influences of humans in global assessments and projections of fire activity

Aboveground
Understanding post-fire community assembly processes and how they interact with fire regime characteristics (especially severity)

Aboveground
Documenting the plasticity of fire-related traits at the community level

Aboveground
Accounting for fire-induced vegetation change feedbacks with the global climate system through albedo, ash, carbon cycle and smoke

Belowground
Separating the direct effects of fire from the indirect effects of fire on soil properties and microbial composition and function

Belowground
Quantifying the interactive effects of compound disturbances on soil properties

Belowground
Incorporating variation in soil responses (at surface and sub-surface) to fire behaviour and fire severity

Fire behaviour
Linking measurements of fuels to resultant fire behaviour and effects across spatial scales

Fire behaviour
Measuring and characterizing below-ground fuel sources and fire behaviour 6. Models Understanding the impact of spatial and temporal patterns of human-caused ignitions and management 6.Models Studying interactions and feedbacks among multiple disturbances (including multiple fire events) ecological impacts on plant populations and communities, by affecting mortality, establishment, survival and reproduction.

| INTER AC TIONS AMONG ECOLOGY-, CLIMATE-AND HUMAN -DRIVEN CHANG E S IN FIRE REG IME S
What causes fire regimes to change over space and time?
Understanding the multiple ecological interactions in fire regimes is a complex challenge.Nonetheless, three fundamental limits to burning-available fuel, appropriate climate conditions and ignitions-have consistently been identified as important controls of fire activity across multiple spatial and temporal scales (Krawchuk et al., 2009).

| Vegetation as a driver of fire regimes: Co-evolution of fire and biota
Fire has been a dominant ecological and evolutionary force on Earth since plants colonized the land about 400 Mya (Judson, 2017;Pausas & Keeley, 2009).In an evolutionary context, the fire regime is an important force of natural selection in plants and other organisms (Simon et al., 2009).However, strong feedbacks between plants and fire also mean that plants directly influence fire regimes (Beckage, Platt, & Gross, 2009;Nowacki & Abrams, 2008;Platt, Ellair, Huffman, Potts, & Beckage, 2016).While there has been increasing recognition that 'coevolution' of plants and fire regimes drives many ecological processes (Archibald et al., 2018), it is still unclear to what extent such co-evolutionary relationships have influenced landscapes and biomes (Pausas & Bond, 2019).
There is enormous potential to understand when and where fire acts as a macroevolutionary process by studying fire-related plant traits.
Plant communities also alter fire regimes, often through a self-reinforcing cycle that selects for particular traits and species to survive within a given fire regime (Rogers et al., 2015).For example, invasive plant species have the potential to alter fire regimes when they modify the flammability of an ecosystem (Balch, Bradley, D'Antonio, & Gomez-Dans, 2013;Paritsis et al., 2018).Similarly, changes in the spatial distribution of plant traits across a landscape can affect fuel continuity and therefore fire probability and spread.Ultimately, changes in fire regimes resulting from altered plant assemblages have the potential to generate abrupt fire regime shifts and to move ecological systems outside their historical evolutionary arena (Fill et al., 2015;Kane, Varner, Metz, & Mantgem, 2017).One major research challenge is to integrate global databases of fire-adapted traits into global fire and vegetation models for projecting future changes.

| The role of climate and climate-fuel interactions
Climate affects fire regimes across temporal scales ranging from short-term fire weather to millennial-scale climate conditions.In many arid and semi-arid ecosystems, climatic conditions during the fire season are usually favourable for burning, but low fuel loads or a lack of fuel continuity limit fire occurrence and spread.In these fuel-limited systems, widespread fire activity requires periods of increased antecedent precipitation that increase productivity and connectivity of fine fuels (Grau & Veblen, 2000;Swetnam & Betancourt, 1998).In contrast, flammability-limited systems, such as closed canopy forests, typically have high fuel loads, but require an extended period of drought to create conditions favourable to burning.Notably, many ecosystems represent some mixture of fuelor flammability-limited fire regimes, termed 'hybrid' systems by McKenzie and Littell (2017).A fourth category is 'ignition-limited' systems, where dry fuels are abundant but lacking non-human ignition sources, such as lowland sclerophyllous shrublands in California (Steel, Safford, & Viers, 2015) and Mediterranean shrublands in Chile (Keeley et al., 2012).
A key research frontier is understanding how changing climate variables (temperature, precipitation, changes in sequences of extreme wet and dry conditions and the likelihood of ignition) should alter controlling aspects of fire regimes.There is high variability in fire-climate relationships within and across ecosystems, and fire-climate relationships have shifted over time (Kitzberger, Veblen, & Villalba, 1997;Sherriff & Veblen, 2008).Some of this variability is due to differences in species composition, vegetation structure and climate within a region (Gartner, Veblen, Sherriff, & Schoennagel, 2012;Heyerdahl, Brubaker, & Agee, 2001;Taylor & Skinner, 2003) (Jolly et al., 2015).
An example is in the western United States, where the fire-season length has increased by 34% (from 166 to 222 days), ultimately leading to increased annual area burned (Westerling, 2016).Globally, anthropogenic climate change has emerged as a significant driver of increased fire danger, independent of natural climate variability (Abatzoglou, Williams, & Barbero, 2019).These factors have relevance to ecological interactions as humans introduce and remove fire from landscapes (Moritz et al., 2014).A major research challenge is to incorporate all the influences of humans into global assessments and projections, given the inherent complexity and differing regional socio-economic contexts (Table 1).

| EFFEC TS OF FIRE ON ABOVE-G ROUND ECOLOG I C AL PRO CE SS E S
Studies of above-ground fire effects traditionally emphasized plant mortality and regeneration.Yet, fire affects a broader range of aboveground ecological processes, with consequences at spatial scales ranging from establishment of individual plants to global climate.Additional research is needed to understand the feedback mechanisms involved in increasing fire activity, as well as the timing and persistence of plant population traits and community processes after fires.

| Fire-vegetation feedbacks
Post-fire vegetation successional trajectories depend on numerous factors, including fire size and effects, pre-and post-fire climate, recent fire history and plant life-history traits related to survival and recolonization (Davis, Higuera, & Sala, 2018;Johnstone et al., 2016).
Plant traits interact with fire severity to influence post-fire regrowth, reproduction, dispersal, germination and establishment.
Fires in grasslands and in many Mediterranean shrublands, for example, tend to perpetuate the existing plant community because the dominant species resprout from protected basal meristems (Keeley & Rundel, 2005).Forest fires, by contrast, can lead to a variety of post-fire successional pathways, depending not only on fire severity but also on the adaptations (Pausas, 2015) and spatial configurations of surviving trees, and canopy and soil seed banks (Figure 3).Predicting future post-fire vegetation composition is difficult because landscape fragmentation, changing climate, non-native species, herbivory, variable propagule availability and interacting disturbances can alter successional trajectories from those in the past (Batllori et al., 2018;Blackhall et al., 2017).Models can be effective for investigating these dynamics (Crandall & Knight, 2015;Scheller & Swanson, 2015) but many of the underlying mechanisms remain poorly understood.Thus, caution is needed when considering modelling results.
The Some ecosystems demonstrate negative fire-vegetation feedbacks, whereby fire temporarily reduces fuel loading or flammability, which decreases the probability of repeat burning for a period of years to several decades (Figure 3).Such feedbacks are evident in boreal forests and several types of temperate coniferous forest (Heon, Arseneault, & Parisien, 2014;Parks et al., 2015).Differences in the direction and strength of fire-vegetation feedbacks, and their interactions with the trajectories and rates of post-fire vegetation recovery, strongly influence how different landscapes will respond to increasing fire activity as the climate warms (Figure 3).Further, in many systems, feedbacks can switch from negative to positive due to post-fire changes in vegetation or fuel accumulation over unusually long fire-free intervals (Harvey, Donato, & Turner, 2016a).

| Climate-vegetation feedbacks
Interactions between above-ground vegetation and climate at local to global scales are key to understanding the ecosystem consequences of changing fire activity (Archibald et al., 2018).have large impacts on regional and global climate as well (Chakrabarty et al., 2016).A comprehensive understanding of how changing fire activity may influence global and regional climate is still lacking.

| Fire and above-ground biodiversity
Fire is known to affect plant species diversity (He et al., 2019), and the effects can be grouped into a few general patterns.First, many ecosystems exhibit declining plant species diversity with increasing time since fire (Swanson et al., 2011).Second, there may be an optimum range of fire regime characteristics that sustains the highest diversity.The optimum appears to depend on the fire-evolutionary history of the ecosystem and species traits (Kelly & Brotons, 2017).
Thus, an absence of fire can reduce plant diversity in certain ecosystems (Abreu et al., 2017;Parr, Lehmann, Bond, Hoffmann, & Andersen, 2014).Third, the spatial and temporal heterogeneity of and biodiversity is the spatial arrangement of fire refugia-locations that experience longer fire-free intervals or tend to burn with lower severity than the broader landscape in which they are embedded (Crandall & Platt, 2012;Landesmann & Morales, 2018;Meddens et al., 2018) Ponisio et al., 2016).For example, fire indirectly influences ungulate populations and avian communities through its effects on habitat quality, including forage and nesting opportunities (Hutto, 2008;Rupp et al., 2006;Smucker, Hutto, & Steele, 2005).It is also clear that herbivory interacts with fire activity (Blackhall et al., 2017;Fuhlendorf, Engle, Kerby, & Hamilton, 2009).However, the effect of fire on other plant-animal interactions (pollination, dispersal, seed predation) is less known (Lazarina et al., 2017).There is also an increasing recognition that fire may have evolutionary consequences in animals (Pausas & Parr, 2018).Future biodiversity studies could more explicitly examine the effects of fire across a broader range of biotic interactions and trophic levels.Further, there is a need to expand the focus from fire frequency and sometimes severity, to address how a broader range of fire regime attributes affect biodiversity (Miller et al., 2019).
However, even where fire regimes have long remained similar across sites, other contextual factors such as soil type and landscape patterns can exert considerable influence on beta diversity, nestedness and community assemblages (Freeman, Kobziar, Leone, & Williges, 2019).More locally, biotic interactions within and among plant species can further influence changes in deterministic assembly with time since fire and produce changes in fire regimes (Landesmann, Gowda, & Kitzberger, 2016;Tepley, Thompson, Epstein, & Anderson-Teixeira, 2017).In contrast, fire may contribute to stochastic community assembly by influencing random colonization or extinction processes that increase ecological drift.For example, when fire decreases the total number of individuals in communities (community size), it may increase local extinctions due to demographic stochasticity, and increase variation in species composition among sites (Myers, Chase, Crandall, & Jimenez, 2015).Post-fire community assembly will also depend on dispersal abilities of species within the regional species pool as well as rates of dispersal among local communities.Understanding this balance of deterministic and stochastic processes is likely to be as important in fire ecology as it has been in other subfields of ecology (Vellend, 2010).
Fire-related plant traits are important both for basic ecological and evolutionary understanding of the role of fire in plant communities, and for improving models that use simplified plant functional types (see Section 6).For example, whereas dynamic global vegetation models (DGVMs) conventionally used static plant traits and emphasized resource competition as the primary driver of species composition, these models are now being modified to make plant traits an emergent evolutionary process (Scheiter, Langan, & Higgins, 2013).Caution must be taken when using plant traits related to fire for modelling at the global scale, as correlations with other dynamic traits are contingent on biogeographic history, and thus may be ecosystem-dependent.The plasticity of fire-related traits at the community level is virtually unknown, and it is important to identify potential limits to adaptation.

| FIRE S E TS ECOLOG I C AL G ROUND RULE S THROUG H SOIL S
The effect of fire on soils is inherently coupled with changes aboveground (Wardle, Jonsson, Mayor, & Metcalfe, 2016).However, responses to fire below-ground may differ than those above-ground because soils contain a relatively large pool of carbon, nutrients and organisms that are at least partially buffered from combustion and mortality during fire events.While plant traits for extensive belowground biomass and resprouting are important features of fireadapted communities (Maurin et al., 2014), we focus this section on soil and soil biota.

| Fire effects on soil physical, chemical and biological properties
Fire alters multiple physical, chemical and biological properties of soil, such as texture, aggregation, pH, nutrient content and microbial community composition.The magnitude of fire effects on soils depends on above-ground fire behaviour (Massman, 2012), organic horizon depth and moisture content (Hartford & Frandsen, 1992) and the physical properties of mineral soil (Giovannini, Lucchesi, & Giachetti, 1988).Fire directly impacts the upper organic horizons via pyrolysis and combustion reactions, including physical loss of the organic horizon, and the underlying mineral soil by conductive and advective heating during a fire (Araya, Fogel, & Berhe, 2017;Certini, 2005;Neary, Ryan, & DeBano, 2008).Fire may also contribute to soil formation (Certini, 2014).
Similar to above-ground fire severity characteristics described earlier, fire severity in soils can be characterized based on change to or loss of the organic horizon, and change to mineral soil physical characteristics.Fire can thus change soil structure (reducing porosity), carbon and nutrient pools and fluxes (reducing concentrations and turnover), soil organic matter composition and biochemistry (increasing pyrogenic carbon and polyphenols) and the composition (decreasing richness, altering community structure) and activity rates (decreasing decomposition) of soil biota (Adkins, Sanderman, & Miesel, 2019;Gutknecht, Henry, & Balser, 2010;Miesel, Hockaday, Kolka, & Townsend, 2015).
Because combustion results in losses of carbon and nitrogen, but not other elements such as phosphorus (Bodi et al., 2014;Butler, Elser, Lewis, Mackey, & Chen, 2018), intense or repeated fires can alter nutrient concentrations and stoichiometry when compared with unburned sites (Ludwig et al., 2018;Pellegrini et al., 2018).These changes to soils have profound impacts on plant growth, community composition and ecosystem processes (da Silva & Batalha, 2008).
Fire effects on the soil microbial community (heterotrophs and symbionts) increase with fire severity (Whitman et al., 2019) which can alter plant recovery and microbially mediated ecosystem processes after fire events.Overall, fire decreases bacterial biomass and diversity, fungal species richness and mycorrhizal colonization, although responses of fungal species can be ephemeral (Dove & Hart, 2017).Responses of microbial biomass and diversity can persist for a decade or more, depending on the ecosystem, fire severity and organic horizon loss (Dooley & Treseder, 2012;Pressler, Moore, & Cotrufo, 2018).Despite these general trends, the limited information on key covariates such as soil pH makes it difficult to determine the mechanisms regulating the responses of soil microbial communities after fire (Pingree & Kobziar, 2019) or effects on subsequent microbial function.

| Effects of compound disturbances on soil properties
Interactions between fire and other disturbances such as drought, wind damage, beetle outbreaks and landslides can have important effects on soil structure and chemistry and subsequent ecological interactions.For example, in ecosystems underlain with permafrost, wildfires can accelerate permafrost thaw and ground subsidence (Gibson et al., 2018), resulting in altered soil hydrology and subsequent likelihood of fire.Additionally, high rates of erosion following post-disturbance vegetation mortality can increase losses of soil organic matter after a fire (Pierson, Robichaud, Rhoades, & Brown, 2019).Thus, interaction effects between fires and other dis-

| Ecological consequences of temporal and spatial variation in soil properties influenced by fire
The temporal responses of soil to fire depend on the variable considered.For example, shifts in microbial biomass and composition occur on sub-annual to decadal time-scales (Dooley & Treseder, 2012) whereas the effect of charcoal formation on soil carbon storage occurs over centennial to millennial time-scales (He et al., 2016).Moreover, the effects of fire on soil carbon dynamics vary across biomes, with soil carbon stocks commonly recovering within a year following fire in temperate grasslands, over decades to centuries in boreal forests and potentially never recovering in some peatlands (Harden et al., 2000).In tropical grasslands, savannas and forests, repeated burning can also deplete total soil carbon and nutrients, but usually only at high frequencies and over decadal time-scales (Liu, Chen, Wang, Hughes, & Lewis, 2015;Pellegrini, Hedin, Staver, & Govender, 2015).These declines may or may not have deleterious impacts on ecosystem productivity (Tierney, 2019).
Fire effects on mineral soils usually emerge over the course of decades due to the mechanism of changes in plant biomass inputs balanced with the turnover time of soil, in contrast to the immediate and direct effects of a single fire on the organic horizon.
Consequently, the coupling between above-and below-ground processes may be altered under changing fire regimes.Furthermore, interactions among pre-fire soil properties, fire severity and post-fire vegetation recovery influence the magnitude of change in soil nutrient pools over time after fire, relative to pre-fire conditions (Godwin et al., 2017;Kranabetter et al., 2016).This post-fire variation in soil properties can have important consequences for microbial communities and plant-microbe interactions (Kardol, Deyn, Laliberte, Mariotte, & Hawkes, 2013).
Soils are heterogeneous across both the soil surface and with soil depth.Combined with the inherent heterogeneity of fire, spatial variability in soil properties can be important for determining post-fire changes in nutrients at the landscape scale that can alter whole ecosystems (Homann, Bormann, & Boyle, 2001).However, it is unclear how heterogeneity in fire behaviour determines heterogeneity in soil responses over time, in large part because of the limited studies that investigate fire severity gradients (Adkins et al., 2019;Garcia-Oliva et al., 2018;Hewitt, Hollingsworth, Chapin, & Taylor, 2016;Kolka et al., 2017;Whitman et al., 2019).Most of our understanding of how fire severity influences soil responses is based on the comparison between wildfire and prescribed fire with the assumption that wildfires are more severe than prescribed fires (Nave, Vance, Swanston, & Curtis, 2011).Moreover, 90% of studies focus only on the upper 30 cm of the soil profile (Richter & Billings, 2015), thus the influences of fire on soil processes throughout the profile remain poorly characterized.

| FIRE B EHAVI OUR D IREC TS ECOLOG IC AL OUTCOME S
Living and dead plant fuels play a prominent role in fire ecology because they help determine fire behaviour and fire effects from below-ground to the upper atmosphere.Fuels are also the most readily manipulated factor influencing fire behaviour, making them a focal point of fire management.The type and amount of fuels consumed during a fire are key links in feedbacks between fire and climate.

| Characterizing fuels
A leaf-to-globe approach is necessary to characterize fuels (See 'fuels' in Box 1 and Figure 4).Moving from small to large scales, fuel characteristics include those of individual plants (e.g.leaf anatomy, chemistry and water content), whole-plant morphology, vegetation composition and structure and ecosystem productivity.Each of these characteristics can be translated into data relevant to relationships between fuels and fire characteristics; these terms include the ratio of fuel surface area to volume, fuel particle size and density, moisture and chemical content, total and available fuel mass and spatial arrangement.These fuel characteristics are used to parameterize fuel models (see below).However, research relating fuel characteristics to determinants of fire behaviour and effects is not as advanced as other trait-centred investigations (see previous sections).
At the global scale, developments in remote-sensing methods have enabled major advances in fuels measurement.For example, measurement of aerial fuels has been revolutionized with LiDAR methods that remotely capture canopy, crown, leaf shape and spatial arrangement, and provide estimates of fuel mass (Andersen, McGaughey, & Reutebuch, 2005).Below the canopy, surface fuels can be characterized with precise 3D ground-based terrestrial laser scanning (TLS; Loudermilk, Hiers, & O'Brien, 2017).Recent efforts to bridge these techniques with dynamic fuel moisture (so-called 4-D fuels) and fire behaviour models significantly advance the ability to understand fuel combustion processes.Because below-ground fuels are driven by finescale topography, hydrology, tree stand age and species composition, they are currently difficult to quantify at large scales.

| Fire behaviour links fuels with fire effects
Two different modes of combustion-flaming and smolderingoccur during a fire but ecological processes are not necessarily related to combustion mode.Rather, they depend on a combination of fire intensity, duration of exposure and the transfer of that energy to objects in the environment.Estimating fire intensity from field-based observations is difficult and requires either infrared sequences (Clark, Radke, Coen, & Middleton, 1999) or direct measurements of heat flux (Butler et al., 2016).Fire radiative power can be detected from the thermal bands of satellite sensors, but these are either limited by temporal frequency (a few detections per day from MODIS and VIIRS) or spatial resolution (at coarser satellite scales).
The relationship between fire behaviour and fire effects has been improved by quantification of fire intensity, at least above-ground (Hiers, O'Brien, Mitchell, Grego, & Loudermilk, 2009).

F I G U R E 4
The relative role of factors driving fire behaviour and effects varies with spatial scale.Several factors (shown as different colours) determine the relationship between fire behaviour and fire's ecological effects.For example, for tree mortality, factors could include leaf moisture content, ecosystem productivity and forest structure and composition as Factors 1, 2 and 3 respectively.The relative importance of these factors (percent of ring with a given colour) changes across spatial and temporal scales.These scales range from fires acting on individual organisms (a), to stands (b), to large landscapes (c).We posit that thresholds of spatial and/or temporal scales exist at which the relative influence of these factors shifts substantially (d, e) New approaches to measure thermal properties, dubbed fire metrology (Kremens, Smith, & Dickinson, 2010), link plant injuries to post-fire recovery of ecosystems.For example, at the level of a leaf, fire behaviour is affected by the micrometeorological conditions of the boundary layer (Dickinson & Johnson, 2001).Heat fluxes can be measured at very fine scales using thermography at sub-centimetre and sub-second scales (Figure 4;O'Brien et al., 2016), which allows detailed knowledge about heat transfer and tree injury (Sparks et al., 2017).Measures of heat flux at scales relevant to plant, soil and microbial processes enable mechanistic understanding of ecological change (Butler & Dickinson, 2010).Below-ground fire behaviour has been advanced by a small number of observational, experimental and model-based approaches (Huang & Rein, 2015;Massman, 2012).However, the field of below-ground fire behaviour is much less developed when compared to the more-readily observed above-ground processes.

| Fire behaviour integrates climate and fuels
The longer-term ecological outcomes of fire often depend on the relationship between above-ground and below-ground fuels.
Below-ground properties and processes critically influence aboveground fuel recovery (via vegetation regeneration, seed germination and soil nutrients and symbionts), physiological status, moisture content and decomposition rates.Under climate warming, below-ground fuels (thermokarst soils, drought-exposed organic horizons and peat soils) are increasingly available for combustion during fires.Roots act as one of the many connections between above-ground and below-ground fire behaviour and its subsequent impacts (Hood, Smith, & Cluck, 2010).Recent work has advanced from observations of fire effects on above-ground indicators such as tree mortality to quantifying below-ground fire behaviour (Varner et al., 2009).Below-ground fire intensity and heating duration affects soil physical, chemical and biological properties and processes (see Section 4), yet the conditions that determine thermal severity remain poorly understood.
The relationship between fire behaviour and fuels dictates the quantity and quality of combustion products.The physical and chemical products of fuel consumption -including smoke, aerosols and volatilized gases-influence the translocation of carbon, nutrients and living organisms, and affect human health (Bowman & Johnston, 2014;Kobziar et al., 2018).Several aspects of smoke and particulate emissions are understudied from an ecological standpoint, including smoke transport of living microorganisms (Kobziar et al., 2018) and the promotion of germination and plant growth by smoke.Smoke alters UV profiles, obscures sunlight, provides substrates for water and ice condensation, and deposits particulates on plants and the environment.These processes occur both within burn perimeters (Bell, Stephens, & Moritz, 2013) and beyond via atmospheric transport.Aerosol radiative forcing is also a major uncertainty in our understanding of the net effects of fire on Earth's climate (Landry, Matthews, & Ramankutty, 2015).The ecological effects of smoke over multiple spatial and temporal scales should be more fully characterized (Table 1).

| MODELLING FIRE AC TIVIT Y
Fire models accomplish several unique goals essential to answering ecological questions, including (a) characterizing fire and its effects at large spatial extents, (b) testing scenarios and hypotheses regarding the interactions described in previous sections and (c) projecting fire behaviour and effects into the future.Numerous fire models are currently in use and under development (Table S2).

| Fire behaviour models
Modelling fire behaviour began with the work of (Rothermel, 1972) and (Van Wagner, 1973) and has improved over time (Table S2), including coupling with dynamic vegetation and/or Earth System models spanning a wide range of spatial and temporal scales.Further, coupled weather-fire models now capture the unfolding of complex landscape-scale events over a period of days (Coen, Stavros, & Fites-Kaufman, 2018), while global models continue to improve their representation of fire frequency, burned area and seasonality across years to decades (Forkel et al., 2019;Hantson et al., 2016).
The representation of dynamic policy and management actions, and the evolving wildland-urban interface, requires concerted interdisciplinary efforts with social scientists (Kline et al., 2017).An ongoing challenge for global fire modelling is scaling up existing fire behaviour models from point scales to grid cells tens to hundreds of kilometres in size, without downplaying the importance of spatial heterogeneity and spatially interactive processes occurring within those large grid cells.

| Fire effects models
Fire effects models represent combustion, intensity and severity via physically based models and vegetation characteristics (e.g.bark thickness, resprouting, serotiny).As models improve their representation of vegetation such as with the incorporation of plant physiology, growth, and size-structure (Fisher et al., 2018) and traitbased modelling (Fisher et al., 2015), simulation of fire mortality also improves.
Dependent on the duration, magnitude and depth of heat transfer, fire can alter soil in ways that influence vegetation recovery (Brando, Oliveria-Santos, Rocha, Cury, & Coe, 2016;Johnstone, Hollingsworth, Chapin, & Mack, 2010), the structure of invertebrate and microbial communities (Hewitt et al., 2016) and availability of nutrients (Karam, Weisberg, Scheller, Johnson, & Miller, 2013).The representation of interactions between fire and below-ground processes must be improved within models to properly project landscape and biome changes (Foster et al., 2019;Pausas et al., 2018).
The forefront of fire effects model development is to include interactions with other disturbances (drought, harvesting, insect outbreaks, wind-throw; Kane et al., 2017;Scheller et al., 2018), with human behaviour (suppression, human ignitions), and more refined representation of fire behaviour and effects, including the feedbacks among fire, vegetation and climate described in Section 3.

| EMERG ENT THEME S FOR FIRE ECO LO GY
Across the six identified research areas in fire ecology from the research community-characteristics of fire regimes, changing fire regimes, fire effects on above-ground ecology, fire effects on below-ground ecology, fire behaviour and fire ecology modellingthree common themes emerge.

| Understanding the ecological consequences of fire through time
A wide array of time-scales is necessary for understanding fire processes, from measures of fire behaviour during seconds to minutes, to multi-millennial records of fire history.However, most research tends to study fire processes on a single time-scale, exposing a large untapped potential to link across time-scales.There is a particular opportunity to understand how fire regimes change over multi-decadal to centennial and millennial time-scales.Fundamental knowledge of ecosystem ecology could then be increased through mechanistic links between fire regimes and fire effects.Our current understanding of fire regimes is based on well-developed approaches for reconstructing fire history-particularly fire return interval-through dendrochronology and palaeoecology.Expanding on these methods to produce metrics used to study contemporary fire ecology, including area burned, fire severity and seasonality, would further advance our understanding of fire regimes.Better integration of fire histories into models would also lead to insights regarding the ecological and evolutionary consequences of changing fire regimes.Finally, archaeological and palaeoecological research can provide important information about the human dimension of fire regimes over the long-term, and this integration is likely to provide key insights for living with fire in the future.

| Characterizing feedbacks and nonlinearities
Despite the long recognition of fire as an important ecological process, the mechanisms underlying the interactions and feedbacks of fire with other ecological processes are only beginning to be explored in a systematic way.Determining the generality of such feedbacks, especially in systems with a long evolutionary relationship with fire, would greatly deepen our understanding of fire's role on Earth while enabling better projections of future fire effects.
To better project where positive or negative feedbacks between vegetation and fire will occur, we suggest four research priorities.
First, continue to study the individual components of the primary drivers of fire activity and the relatively simple pairwise feedbacks between components, including human influences.Second, diversify the types of fire regimes studied, especially those where fire behaviour is highly variable in space and time (grasslands, savannas, Mediterranean systems).Third, expand the study of feedbacks to include not only vegetation, but also higher trophic levels and below-ground components of fire-frequented ecosystems, especially those that influence fuel composition and dynamics.Fourth, use the power of models to explore feedbacks that are difficult to observe because they either develop slowly or are difficult to interpret without controlling for other influencing factors in the field.

| Harnessing the data revolution and using models to explore the diversity of fire on Earth
The many data sources described in this manuscript provide unique and complementary views of fire on Earth, offering ecologists an incredible opportunity to draw new insights about how fire is changing.
Over a dozen satellites and space-borne sensors are collecting information about fire events and their effects in real time (e.g.Landsat, MODIS, Sentinel, VIIRS, Planet, DigitalGlobe's Worldview Collection, GEDI, ECOSTRESS, and others), climate and Earth System models are operating on a variety of spatial scales and databases are providing unprecedented detail about fuels and ignitions.Ongoing data efforts include further development and better integration of palaeorecords and government incident reports addressing spatial heterogeneity, and incorporation of new sources of information about fire-from social media to drones.However, there is still a long way to go to develop better data infrastructure, to improve curation of data in a standardized format that includes controlled vocabularies that describe what was measured (Gross et al., 2018), to increase transparency, reusability and interoperability of data products across political and disciplinary boundaries and to apply new data analysis techniques such as machine learning to discover important large-scale patterns about fire behaviour, fire regimes and ecological outcomes.
There is significant intellectual momentum building in the area of fire models, with several parallel but largely disconnected efforts under way.The diversity of modelling approaches reflects the spatio-temporal issues described earlier, the diversity of fire regimes on Earth, and the individual goals of the disparate modelling communities, which range from detailed landscape models to Earth System Models.One urgent fire modelling need is to better understand the fuel-fire-atmospheric conditions that generate extreme fire behaviours that exceed the predictive ability of our current fire models.An additional modelling need is the explicit incorporation of human interactions with fire, which will be required to project the consequences of future fires.To advance this area, we need a more accurate parameterization of how people create and alter fire regimes, from changes in spatial and temporal patterns of ignitions to alteration of fuel type and continuity, coupled with past and future land use and land cover scenarios, and global data layers that provide a mechanistic understanding of human decision-making processes.

| CON CLUS IONS
Fire is a fundamental component of most terrestrial ecosystems on Earth.In many cases fire is key to understanding population, community and ecosystem ecology.Theories of how fire interacts with evolutionary processes are just beginning to develop and, although currently focused largely on plants, these theories are poised to advance rapidly in the animal and microbial realms.The fire ecology research community has made immense progress in recent years to understand the many aspects of fire and fire regimes, including the number, timing, changes and ultimate effect of fires on the Earth system.Here, we offer guidance to continue the important mission of understanding the fundamental role of fire in a diverse array of ecological systems.

ACK N OWLED G EM ENTS
This manuscript is a product of discussions at the Future of Fire workshop held in November 2017 in Boulder, Colorado, USA.
Support was provided by NSF-DEB-1743681 to K.K.M. and A.J.T.
We thank Shalin Hai-Jew for helpful discussion of the survey and qualitative methods.
At a local scale, recent investigations into the interactions between climate variables (especially drought) and seedling recruitment have highlighted decreases, or failures, in post-fire tree regeneration (Stevens-Rumann & Morgan, 2019).Local mechanisms, such as reductions in tree seedling recruitment, growth and survival under warmer conditions may drive fundamental changes in ecosystems (Enright, Fontaine, Bowman, Bradstock, & Williams, 2015) but the timing and persistence of these transitions are largely unknown.Understanding of the mechanisms and degree to which fireinduced vegetation change can alter regional climate also needs to be further advanced(Beringer et al., 2015;Liu, Ballantyne, & Cooper, 2019).At the global scale, increasing fire activity can reduce ecosystem carbon storage(De Faria et al., 2017).The resulting increase in atmospheric carbon dioxide concentrations accelerates climate warming, potentially driving further increases in wildfire activity.In addition to carbon cycle feedbacks, changes in surface energy partitioning, including changes in surface albedo, have important influences on regional climate(Rogers, Randerson, & Bonan, 2013), and aerosols emitted during fire combustion could fire regimes can either increase or decrease biodiversity, but existing examples are ecosystem-dependent(Martin & Sapsis, 1992;Parr & Andersen, 2006), and the influence of fires on biodiversity may depend on trophic levels(Davies, Eggleton, Rensburg, & Parr, 2012;Maravalhas & Vasconcelos, 2014).Another important factor for fire F I G U R E 3 Feedbacks between fire and vegetation interact with post-fire forest recovery rates to mediate responses to altered fire regimes in landscapes that currently support forest cover.High-severity fire alters fuel and microclimate, making sites either more (positive feedback) or less (negative feedback) likely to reburn soon after a severe fire.The direction and strength of the feedbacks (f − and f + ) interact with the time to forest recovery (r) after high-severity fire to determine how the proportion of a landscape that supports forest cover (FC) varies as fire activity increases.Negative feedbacks (a) reduce the probability of reburning, limiting the amount of forest lost as the climate become more conducive to fire.Positive feedbacks (b) produce a threshold, where small increases in either fire activity or the time to forest recovery after severe fire can shift the system from forest to non-forest cover.White dots on the left response surfaces represent hypothetical baseline conditions in different forest regions.Climate change and increased wildfire activity alter the shape of the response surfaces.At the same time, the x-axis positions of the regions shift (white arrows) as forest recovery slows under a more arid environment with larger patches of high-severity fire.Short and long arrows represent regional differences in sensitivity to the slowing of forest recovery (afterTepley et al., 2018) turbances should be considered as these interactions may amplify or dampen changes to soil properties after fire.Given that global change projections indicate increased frequency of fire and other disturbances, additional work is needed to classify the additive or interactive effects of compound disturbance(Bradford et al., 2012).The existence of numerous experiments manipulating fire frequency(Godwin, Kobziar, & Robertson, 2017;Guinto, Xu, House, & Saffigna, 2001;Holdo, Mack, & Arnold, 2012) provides the opportunity to impose additional disturbances-such as reduced precipitation-to study the combined responses of soil properties such as soil aggregation, pH, nutrient pools, respiration rates and microbial community composition.

Humans contribute to and are affected by changing fire regimes
(Bowman et al., 2011;Guarinello de Oliveira Portes, Safford, & Behling, 2018;Kobziar, Godwin, Taylor, & Watts, 2015;McWethy et al., 2013)ra Portes, Safford, & Behling, 2018;Kobziar, Godwin, Taylor, & Watts, 2015;McWethy et al., 2013).The mechanisms by which humans alter fire regimes include: (a) changing the frequency, timing Finally, humans alter fire regimes through anthropogenic climate change.Across much of the globe, climate conditions are becoming increasingly conducive for fire.The length of the fire-weather season (days with fire danger metrics above their median values) has increased by nearly 20% from 1970 to 2013 . Refugia may sustain local populations of fire-sensitive species within large burned areas, increasing post-fire habitat diver-