Integrating plant physiology into simulation of fire behavior and effects

Summary Wildfires are a global crisis, but current fire models fail to capture vegetation response to changing climate. With drought and elevated temperature increasing the importance of vegetation dynamics to fire behavior, and the advent of next generation models capable of capturing increasingly complex physical processes, we provide a renewed focus on representation of woody vegetation in fire models. Currently, the most advanced representations of fire behavior and biophysical fire effects are found in distinct classes of fine‐scale models and do not capture variation in live fuel (i.e. living plant) properties. We demonstrate that plant water and carbon dynamics, which influence combustion and heat transfer into the plant and often dictate plant survival, provide the mechanistic linkage between fire behavior and effects. Our conceptual framework linking remotely sensed estimates of plant water and carbon to fine‐scale models of fire behavior and effects could be a critical first step toward improving the fidelity of the coarse scale models that are now relied upon for global fire forecasting. This process‐based approach will be essential to capturing the influence of physiological responses to drought and warming on live fuel conditions, strengthening the science needed to guide fire managers in an uncertain future.


I. Introduction
Wildland fire burns hundreds of millions of hectares of forests, woodlands, and grasslands annually (Giglio et al., 2013), shaping terrestrial ecosystems (Bond & van Wilgen, 1996;Bond & Keeley, 2005;Bond, 2021), and their impacts on global carbon (Bowman et al., 2009) and water (Li & Lawrence, 2017) cycles.To accurately predict fire behavior and resulting effects, we need to understand the influences of vegetation structure and physiology on combustion (e.g.Byram, 1959;Weise & Wright, 2013) and the microenvironment (e.g.Finnigan, 2000;Banerjee et al., 2020;Atchley et al., 2021).Recent studies demonstrate that burned area shows greater increases in forests and shrublands where vegetation moisture is more sensitive to water limitation (Rao et al., 2022), and that hydraulic traits of woody evergreen species are responsible for up to 3.6-fold variation in live fuel moisture content (LFMC, ratio of biomass water content to oven-dry biomass, Table 1; Scarff et al., 2021).These findings emphasize the importance of accounting for ecophysiological controls on woody vegetation in wildfire forecasting and prescribed fire planning, where fuel effects dominate under less extreme fire danger conditions (Cruz et al., 2022).This will be particularly important in firesusceptible ecosystems which experience significant drought and LFMC declines below thresholds that drive increases in fire behavior (Pimont et al., 2019a).As atmospheric aridity continues to increase with rising temperatures, compounding plant water stress (Grossiord et al., 2020), the number of regions for which live fuels (i.e.living plants) are important in determining the behavior of fires and their ecological outcomes is likely to expand (Resco de Dios et al., 2021).Historically, representation of live fuels in fire behavior models has been limited to static fuel models (Table 2) that generalize vegetation into classes, such as grass, shrub, and timber (Albini, 1976;Scott & Burgan, 2005; but see Rothermel & Philpot, 1973;Hough & Albini, 1978 for seasonally-and agedynamic fuel models).With recent advances in process-based modeling and remote sensing of both fuels and fires, there is now opportunity to capture more realistic fuel heterogeneities, including the physiological dynamics that determine live fuel conditions ('pyro-ecophysiology', Jolly & Johnson, 2018).This will allow exploration of their influence on fire behavior, ecological outcomes (i.e.plant injury, mortality, and recovery; hereafter, 'fire effects') and cascading hazards.Recent work has called for modeling efforts to improve integration of fire behavior and its effects (e.g.Hood et al., 2018;O'Brien et al., 2018;Kleynhans et al., 2021) to better capture fire-vegetation-environment feedbacks.This integration is particularly important in the context of low-intensity and prescribed fire.While fuel heterogeneity has little effect in a highintensity fire environment (Atchley et al., 2021;Cruz et al., 2022), changes in phenology can make the difference between containment and escape for prescribed fires, where combustion dynamics are engineered to achieve desired biological and physical outcomes (e.g.species-specific vegetation survival or mortality, fire risk reduction, soil protection, and smoke management (Hiers et al., 2020)) with implications for future fire and landscape dynamics (Mitchell et al., 2009;Gallagher et al., 2021).Plant water and carbon cycles have been recognized as important to both fire behavior (Nelson, 2001;Macias Fauria et al., 2011;Jolly et al., 2014;Jolly & Johnson, 2018) and ecological fire effects (Michaletz & Johnson, 2007;Hood et al., 2018;B€ ar et al., 2019).Yet, they have not been considered as an integrating framework mechanistically linking the two.
We argue that plant water and carbon cycles drive live fuel moisture and dry mass dynamics, which influence heat transfer into the plant during a fire and subsequent postfire survival and recovery.This focus on plant water and carbon dynamics provides a mechanistic framework that links plant physiology to fine-scale fire behavior and fire-induced effects on plant tissues, addressing a gap in current approaches to modeling fire and vegetation (Fig. 1).We focus in this review on the interactions of woody plants and fire, due to the availability of literature and to maintain focus on the underlying argument.However, many of the needs and underlying connections that we recognize have importance to nonwoody systems.Understanding how within-plant physiology and fire interact will allow exploration of mechanisms important to ensuing plant, stand, and landscape-level vegetation dynamics, which influence subsequent fire behavior.To fully capture fire-ecosystem feedbacks, we propose exploring new ways to bring together fire behavior and effects models at fine scales where both fire and physiological processes can be represented in sufficient detail to define important biophysical mechanisms, with heat transfer to vegetation providing the mechanistic linkage from physiology to fire behavior and subsequent effects (Varner et al., 2021).A fundamental understanding of how fire and plants interact at the fine scale is needed to constrain simulations used at managementrelevant scales.Under unprecedented future climate conditions, for which we do not have past or present analogs, biophysical process modeling will be essential to understand response to conditions that fall outside the range of variability captured by current empirical models.Linking remotely sensed estimates of plant water and carbon status to models of fire behavior and effects offers a mechanistic approach critical to capturing the influence of

Review
Tansley review New Phytologist physiological responses to drought and warming on live fuel conditions under global change.

II. Existing models of fire behavior and effects
There are many existing fire behavior and effects models under use and continuous development.Rather than focusing on any one model or group of models, here we focus on commonly applied scales and model characteristics.We classify these as global to regional-scale models, landscape to stand-scale models, operational models, computational fluid dynamics models, and fine-scale biophysical fire effects models.On global to regional scales (Fig. 1), fire behavior and effects have been combined within ecosystem process models using various approaches to simulate ecosystem dynamics with fire disturbance (see Rabin et al., 2017 for in-depth review).Simulations commonly span decades to centuries and may include the entire globe.In these models, fire behavior and effects are commonly simulated explicitly at the computational grid level (c.0.1-1°) or implicitly at a subgrid level.Fire spread is a function of fuel characteristics (loading, size distribution, and moisture), ignition patterns, weather, and latent suppression.These models often capture a dimension of vegetation growth and/or demographics which can dynamically inform fuel loads and fuel availability.Fire behavior is usually represented empirically based on experimental and modeling studies by Rothermel (1972) and Van Wagner (1973).Fire effects are determined by total energy release and thresholds for mortality by either individuals, cohorts, or by fractioning existing vegetation.For example, recent advances in vegetation demography modeling include the addition of plant

Models 3D Fuel frameworks
Models that generate synthetic 3D fuels as inputs for CFD fire models Computational fluid dynamics (CFD) Physics-based models that simulate the interaction of liquids and gases based on fluid mechanics principles; used in simulation of fine-scale fire behavior and fire-atmosphere interactions within individual stands on timescales of minutes to hours Ecosystem process Mechanistic models based on theoretical understanding of ecological processes; for fire applications, can represent both fire behavior and effects on regional to global scales Empirical fire effects Statistical models that predict tree status (live, dead) as a function of tree characteristics and observed fire injury, often estimated based on flame length Fire danger rating systems Broad-scale assessments of fire ignition, spread, and hazard potential based on current and antecedent weather, fuels, and topography Functional structural plant models (FSPMs) Mechanistic models of 3D plant structure, environment, and physiological response, from gene to community scales  hydraulics in FATES (Christoffersen et al., 2016), allowing decoupling of water and carbon dynamics for more direct determination of live fuel vulnerability to fire at multiple timescales.Live fuel moisture content dynamics are captured by simulating water-and carbon-cycle processes directly at subdaily time scales, allowing for assessment of climate impacts (e.g.change in CO 2 , temperature, and precipitation) on future LFMC trends under different climate scenarios (Ma et al., 2021).However, simplified representation of fire precludes mechanistic understanding of fire-physiology interactions.
At stand to landscape scales, a large number of models were developed for spatial plant ecology and fire occurrence at the scale relevant to forest management (10-100 km 2 ; e.g.FIRE-BGC (Keane et al., 2004), LANDIS-II (Sturtevant et al., 2009;Scheller et al., 2019)).These models are applied to decadal to century-long simulations.Observed or modeled fire dynamics account for standto-stand fire spread probabilistically or through the estimation of mechanistic processes.These models simulate cohorts or representative sets of individual plants/trees to estimate vegetation development and demography which inform fuel loading and availability.While this class of models can be leveraged to explore fire behavior and effects, they do not mechanistically link the two (Keane et al., 2004).Within the stand (1-10 km 2 ), fire effects are usually homogenous or stochastically based on homogenous traits of the stand (Furniss et al., 2022).For example, FIRE-BGCv2 (Keane et al., 2011) links biogeochemical processes with stand-scale fire behavior and effects based on live and dead fuel loading, leveraging FARSITE (Finney, 1998), but does not address within-stand variability.Keane et al. (2004) and Sturtevant & Fortin (2021) provide extensive reviews of landscape to stand-scale models.
Operational fire models such as BehavePlus (Andrews, 2013), FARSITE, FLAMMAP (Finney, 1998), or FFE-FVS (Reinhardt & Crookston, 2003) also represent fire behavior and effects.However, these models, which need to be computationally inexpensive to be useful, rely on empirical representations of fire behavior (Rothermel, 1972), and use fixed, stylized fuel models (Scott & Burgan, 2005) to determine flame length and resulting mortality.Their linkage between fire behavior and effects is thus very simple, and largely does not account for dynamic plant physiology or demography.
At the substand scale (< 1 m-1 km), the most advanced representations of fire behavior and effects are currently found in distinct classes of models.Fire behavior is best captured with Computational Fluid Dynamics (CFD) models such as FIRETEC (Linn et al., 2002) or the Wildland-Urban Interface Fire Dynamics Simulator (WFDS, Mell et al., 2009).These models simulate individual fire events, or portions thereof, on short timescales of minutes to hours.Fuel parameters are static, with exception of moisture mass and dry mass of the fuel, which may decrease through dehydration and consumption, respectively.These models do not represent fire effects, as they do not explicitly distinguish between live and dead fuels, or between different vegetation species or functional groups.In absence of explicit vegetation response to changing meteorological conditions, they cannot resolve changes in LFMC or the sensitivity of fire behavior to live fuel moisture dynamics (Jolly, 2007).
Prediction of fine-scale ecological fire effects (postfire vegetation injury, mortality, and recovery) has been approached using both empirical and process-based models, with fire intensity provided as input from other models or measurements.Most postfire mortality models rely on empirical correlations (Woolley et al., 2012;Hood et al., 2018) which predict vegetation status (live or dead) as a function of plant characteristics (e.g.species, bark thickness, height, and diameter) and observed fire injury (e.g.crown scorch, and bark char), often estimated based on flame scorch height ( Van Wagner, 1973).A number of process-based biophysical fire effects models predict injury to different woody plant compartments (e.g.stem (FireStem2D, Chatziefstratiou et al., 2013) or crown (Michaletz & Johnson, 2006) heating models).Few, however, have considered injuries to more than one part of the plant (Michaletz & Johnson, 2008), and none address the influence of interacting injuries across multiple woody plant compartments on postfire plant function.
Direct coupling of fine-scale fire behavior and effects models remains challenging due to the different temporal scales that are relevant for both.As an alternative to solve this challenge, we propose exploring nested modeling frameworks (Gettelman et al., 2022;Shuman et al., 2022) which can transfer information across models at scales where both fire and physiological processes can be represented in sufficient detail to define important biophysical mechanisms.Recent work has made strides in mechanistically linking fire behavior to ecological effects through plant traits (Zylstra, 2021), but without mechanistic determination of vegetation density and moisture content resulting from plant water and carbon dynamics.By incorporating fine-scale physiology, we can capture dynamic water and carbon to represent these properties more realistically.

Evidence for the role of plant water and carbon in fire behavior
Fire behavior is broadly controlled by complex interactions among fuel density and moisture content, topography, and the atmosphere (Countryman, 1966).These factors not only interact, but also can be highly variable in space and time, resulting in dynamic fire environments.Live fuel moisture content is a key fuel property governing fire behavior (Fig. 2a).Declining LFMC is associated with an increase in area burned (Dennison & Moritz, 2009;Nolan et al., 2016;Pimont et al., 2019b), and LFMC below 100% sharply increases fire rate of spread (ROS, Pimont et al., 2019a) and the probability of large fires (Martin-StPaul et al., 2018) (Fig. 2a).Given its impact on fire behavior, LFMC has been incorporated into fire danger rating systems (e.g.Deeming et al., 1977;Stocks et al., 1989) and fire behavior models (e.g.Rothermel, 1972).
The water and dry mass components of LFMC are outcomes of plant water-and carbon-cycle processes, respectively, and are controlled by environmental conditions, soil properties, and plant physiology (Fig. 2b,c;Macias Fauria et al., 2011;Jolly & Johnson, 2018;Ruffault et al., 2018).Plant water and carbon can vary independently in both space and time, with opposing New Phytologist (2023) 238: 952-970 www.newphytologist.comÓ 2023 The Authors New Phytologist Ó 2023 New Phytologist Foundation effects on LFMC.For example, a sharp prebudburst (early June) increase in foliar dry mass of older, live pine needles has been shown to reduce LFMC despite increasing foliar water mass (Fig. 2b; Jolly et al., 2014Jolly et al., , 2016)).Based on these findings, Jolly & Johnson (2018) proposed a framework decomposing LFMC into plant water and carbon cycle processes (Fig. 2c), where the numerator is determined by whole plant hydraulics (e.g.transpiration; relative water content; and xylem embolism), and the denominator by carbon allocation (e.g. chemical composition/density; phenology; and canopy architecture).These processes influence fire behavior both by changing tissue-and canopy-level properties that influence heat transfer, and by altering the proportion of dead biomass in fuels and on the ground surface.Seasonal or drought-induced senescence and shedding are examples of physiological processes that alter leaf, fine branch, and whole plant flammability, transforming live fuels into dead fuels and increasing litter accumulation.Tissue senescence is associated with shifts in osmolytes and water content (Milla et al., 2007), which can increase or decrease LFMC.Tissue mortality also increases the dead to live fuel ratio, causing a strong increase in fire spread and intensity (e.g.Cruz et al., 2015;Sieg et al., 2017;Balaguer-Romano et al., 2020).When senesced tissues are shed, litter accumulates in the surface fuel-bed and may increase the likelihood of surface fire.
The importance of both plant water-and carbon-cycle processes for fire behavior becomes clear when examining their effects on heat transfer and combustion (Michaletz & Johnson, 2007; B€ ar Preheating, the process by which heat transfer evaporates water, dries, and decomposes carbon-containing polymers in unburned fuels, occurs via conduction, convection, and radiation.Thermal conductivity varies with material density, temperature, and water content, while convection depends on geometry and orientation of the fuel surface.The proportion of radiation absorbed by unburned fuel depends on both its thermal absorptance as well as geometry and orientation.Fuel chemistry and water content are therefore important to both conductive and radiative heat transfer through their influence on thermal conductivity and absorptance, while convection and radiative heat transfer to the fuel surface depend on vegetation geometry and orientation, including leaf, branch, crown, or canopy structure (Michaletz & Johnson, 2006;Pausas & Moreira, 2012).The rate at which the temperature increases in response to heat transfer depends on the mass and specific heat of the biomass, which depend on plant tissue water content and composition (Boardman et al., 2021).Water content also determines how much energy is required to heat the fuel to evaporation and combustion temperatures (Yashwanth et al., 2016), while the content of carbon-based polymers determines the amount and rate of thermal decomposition, influencing the rate of preheating and time to combustion (Kilzer & Broido, 1965).Fuel chemistry and water content are also critical in determining whether combustion reactions are self-sustaining and how they will contribute to fire intensity and rate of spread (Quintiere, 2006;Matt et al., 2020).During combustion, the content of organic polymers determines how much energy can be released, and water content can determine whether flaming or smoldering combustion occurs.Higher rates of fire spread and probability of flaming combustion, which occurs at high temperature and/or low moisture, are therefore more likely when live fuels are water stressed and the ratio of water mass to dry mass (LFMC) is lower.
Results from leaf-level burn experiments support the influence of plant water and carbon processes on preheating.Time to ignition (Fig. 2d) has been shown to increase with moisture (Grootemaat et al., 2015;Bianchi et al., 2019;but see Fletcher et al., 2007) and leaf mass per area (LMA, inverse of specific leaf area, Grootemaat et al., 2015), which is associated with accumulation of total structural carbohydrates and lignin (Poorter et al., 2009).Flame and smolder duration (Fig. 2e) also increase with dry mass (Grootemaat et al., 2015), but decrease with moisture (Bianchi et al., 2019).These contrasting effects of leaf-level water and carbon on different fire behavior characteristics underscore the importance of capturing LFMC components independently.Despite evidence from Alam et al. (2020) that certain leaf and shoot flammability metrics (e.g.ignitibility and combustibility) are decoupled, the same study showed a positive relationship of leaf dry matter content with shoot and species-level flammability, suggesting that effects of leaf water and carbon on fire behavior are scalable.At canopy scales, Jolly et al. (2016) found that increased leaf density, which accounts for changes in dry mass associated with phenological change in foliar chemistry and carbon allocation, increased modeled crown fire propagation and area burned.Additionally, canopy bulk density, a measure of canopy architecture influenced by water-and carbon-cycle processes via growth rate, environment, and phenology, has long been recognized as an important driver of spread rate in active crown fires ( Van Wagner, 1977;Resco de Dios, 2020).The independent relationships of water-and carboncycle processes with fire behavior, and their complex interdependence, highlight the need for a more mechanistic representation of LFMC that accounts for underlying physiology and allows for dynamic change.

Evidence for the role of plant water and carbon in fire effects
First-order effects of fire on vegetation are the direct result of combustion and heat transfer to plant tissues (Michaletz & Johnson, 2007), and can be modulated by plant physiological characteristics.The same traits that influence fire behavior through controls on heat transfer also influence fire effects by modulating tissue temperatures and exposure times, particularly in low and mixed-severity fire regimes where vegetation is not combusted and immediately killed (O'Brien et al., 2018;Varner et al., 2021).Percent mortality has been demonstrated to increase with fire radiative energy density (Smith et al., 2017;Steady et al., 2019), or time-integrated radiative flux density from ignition to cessation of the fire.Since absorbed fire radiative energy is tied to flame and smolder duration (O'Brien et al., 2016), which increases with dry mass and decreases with moisture (Fig. 2e), we expect first-order mortality to increase with dry mass and decrease with moisture content (Fig. 3a).This is consistent with significant increases in simulated canopy fuel consumption with reduced canopy fuel moisture, particularly under low-wind scenarios (Sieg et al., 2017).While the physical relationship between moisture and flame and smolder duration should hold across species and ecosystems, further research is required to validate the relationship between moisture content and first-order mortality, as tissues with higher moisture contents can also be more heat sensitive (Wright & Bailey, 1982).
When vegetation is not consumed, fire-induced injuries, including hydraulic dysfunction, cambium and apical meristem necrosis, leaf and fine root necrosis, are proposed to influence whole-plant carbon and water budgets resulting in hydraulic failure and/or carbon starvation (Fig. 3b; Midgley et al., 2011;Michaletz et al., 2012;Hood et al., 2018;Michaletz, 2018;Silva et al., 2018;B€ ar et al., 2019;Berenguer et al., 2021).This framework suggests that second-order effects on live vegetation are driven, in part, by postfire water and carbon availability, along with integrity of the water and carbon uptake and transport systems (Fig. 3b,c; Michaletz & Johnson, 2007;Hood et al., 2018;Michaletz, 2018;B€ ar et al., 2019).Postfire environmental conditions and speciesspecific traits determine whether resulting functional and growth limitations will lead to vegetation recovery or mortality (B€ ar et al., 2019;Sayer et al., 2020;Hood, 2021;Ruswick et al., 2021).For instance, traits such as water stress resistance are observed to differ between disturbance-dependent and obligate sprouters in Mediterranean-type climate regions (Pratt et al., 2012), with implications for postfire community composition.Franco et al. (2014)

Review
Tansley review New Phytologist are expected under climate change, the interplay between temperature optima and CO 2 fertilization effects on photosynthesis, photorespiration, and respiration will define how much carbon is available for postfire plant growth and resprouting, determining the acclimation potential of a given species, plant community or ecosystem.Indeed, from Mediterranean to Tropical ecosystems, the balance between forest expansion and retreat, between seedling recruitment and mortality, has been attributed to species differences in drought sensitivity (e.g.embolism resistance, rooting depth, and stomatal regulation) and carbon balance (e.g.stomatal regulation, osmotic adjustment, and carbon allocation) (Keeley, 1998;Franco et al., 2014).
Drought and herbivory research, where tree mortality has a long history of study, has defined mortality thresholds for hydraulic failure in terms of percent loss of hydraulic conductivity (PLC; Hammond et al., 2019), and for carbon starvation in terms of nonstructural carbohydrate (NSC) depletion (Barker Plotkin et al., 2021).Evidence of increased PLC from laboratory experiments (Michaletz et al., 2012;West et al., 2016;Partelli-Feltrin et al., 2021), as well as from forest fires (B€ ar et al., 2018), indicates that fire increases vulnerability to cavitation (Midgley et al., 2011;Hood, 2021).In addition, fire can decrease root NSC (Varner et al., 2009), reducing stores available to support metabolism, growth, and subsequent stress response, resulting in mortality independent of canopy damage.Together, this research suggests similar relationships for second-order fire mortality (Fig. 3c) as for mortality from drought and herbivory, where sufficiently high PLC or low NSC results in mortality.This is consistent with recent work linking mortality and recovery from low-intensity fire to prefire water status (van Mantgem et al., 2018;Partelli-Feltrin et al., 2020) and concentrations of NSC (Zhu et al., 2012;Sayer et al., 2020).In addition, smoldering consumption can impact water uptake directly through root loss (O'Brien et al., 2010).In conifers, trade-offs have also been documented between resistance to cavitation and to fire, likely mediated by allocation of carbon to either building thick bark or dense xylem, but not both (Resco de Dios et al., 2018).These tradeoffs have implications for woody vegetation recovery from fire under climate change, where drought conditions can predispose woody vegetation to hydraulic (Partelli-Feltrin et al., 2020) and carbon (Sayer et al., 2020) limitations, impacting capacity to recover from fire-induced injury and/or resprout.Indeed, intense postfire drought can also cause significant resprout mortality resulting from simultaneous loss of hydraulic conductivity and depletion of root starch (Pratt et al., 2014).

Importance of plant water and carbon dynamics
Given the demonstrated role of plant water and carbon status in both fire behavior (Fig. 2) and effects (Fig. 3), accurately capturing their dynamics (seasonal variation, differences across species or functional types, with topography, and natural disturbance) will lead to improvements in fire behavior and effects models.Carbon and water status of live fuels change dynamically through time with environmental conditions and seasonal phenology (Baffoin et al., 2021).Variation in plant dry matter is driven by changes in organic molecules, largely NSC and lipids synthesized in the leaf and allocated to different plant organs (i.e.roots, stems, leaves, and reproductive structures) as needed for respiration and storage (Chapin et al., 1990).Diurnal variation results from daily synthesis and export patterns, while seasonal variation is largely driven by reallocation to meet seasonal demands associated with phenology or source limitation (e.g.drought-induced reduction in photosynthesis; Mart ınez-Vilalta et al., 2016).Changes in water content are driven by the interplay between soil water availability, physiological and morphological regulation, and atmospheric vapor pressure deficit through the soil-plant-atmosphere continuum (Nelson, 2001;Macias Fauria et al., 2011).Water and carbon status differ across organs and over time as water moves from root to leaf via transpiration and as carbon is assimilated and transported across organs via the phloem (Fig. 4a, Zhou et al., 2020), and these dynamics differ by species and/or plant functional type.For instance, while evergreen conifers show strong seasonal variation in leaf NSC and dampened variation in stem NSC and water deficit (ΔW), the opposite is true for evergreen angiosperms (Fig. 4b, S anchez-Costa et al., 2015;Mart ınez-Vilalta et al., 2016).Topographic variation (microclimates, water availability, and soil resources available for growth) can also modify physiology, and fuel availability and loading, and has been shown to govern fire behavior and effects (Knapp et al., 1993;Krawchuk et al., 2016;Swann et al., 2022).Different functional strategies in relation to seasonal water deficit and disturbance, such as those defined by leaf lifespan (deciduous vs evergreen) and minimum water potential (Ackerly, 2004), will determine how these water and carbon dynamics manifest across species.Even within an individual species, plants can modulate the effects of variability in water resources by modifying leaf area, rooting depth, and/or stomatal conductance (Rambal, 1993).Among co-occurring chaparral shrubs, differences in rooting depth can drive differential responses of leaf water potential to seasonal drought (Davis & Mooney, 1986;Smith & Richardson, 1990), and those with similar responses may use different modes of seasonal osmotic adjustment to accomplish changes in osmotic potential (i.e.shift in water volume vs solute concentration; Bowman & Roberts, 1985), with strong implications for LFMC.Species differences in drought response and associated leaf senescence and shedding have implications for litter accumulation and fuel-bed flammability as well due to species variation in litter flammability and nonadditive effects in multispecies litter mixtures (de Magalhães & Schwilk, 2012, 2021).
Climate and phenology drive seasonal change in these firerelevant traits, regulating water and NSC balances.Under future climate conditions, drought and phenological shifts can affect greenup, senescence, and susceptibility to mortality, all of which impact fuel load and flammability (Jolly et al., 2016).For instance, prefire drought increases the likelihood of mortality (van Mantgem et al., 2018;Partelli-Feltrin et al., 2020) and of resprouting failure (Karavani et al., 2018;Resco de Dios et al., 2020).Topographic variation can be an important control on drought induced mortality and resulting fuel loads.For example, increased water availability in concave areas can mediate the increase in hydraulic stress typical at lower elevations (Tai et al., 2017).Prefire drought and heat stress may also diminish flowering and seed production, negatively impacting species with fire-cued recruitment (Nolan et al., 2021).Heat susceptibility of tissues (B€ ar et al., 2021) and probability of vegetation survival can also be influenced by plant phenological stage during a burn (Trabaud, 1991;Knapp et al., 2009;Ruckman et al., 2012;Pratt et al., 2014), highlighting the importance of capturing seasonal variation in prediction of mortality (Bond & van Wilgen, 1996).Beyond drought, other natural disturbances can cause physical damage to vegetation (e.g.pathogens and herbivory), change in fuel loads and availability due to rapid mortality (e.g.bark beetles and tropical cyclones), or increased resource availability for plant growth (e.g.windfall), all of which can be incorporated into plant physiology models to understand their feedbacks on fire behavior and effects (Karp et al., 2021;Rouet-Leduc et al., 2021;Fettig et al., 2022;Ibanez et al., 2022;Lee et al., 2022).Acknowledging dynamic physiology recognizes that fire behavior and its effects will vary, within and across species, with plant water status (Nolan et al., 2018), carbon dynamics (photosynthesis, respiration, allocation), phenology (B€ ar et al., 2021), species (McAllister & Weise, 2017;Nolan et al., 2018; Resco de Dios, 2020), and time (Fig. 4).

IV. Integrating fine-scale physiology with fire models
The wildland fire research community has called for the use of process-based models to explore the potential mechanisms and interactions driving fire dynamics and effects and to conduct virtual experiments that allow for consideration of no-analog future climate conditions (Michaletz et al., 2013;Hoffman et al., 2018;O'Brien et al., 2018).The demonstrated importance of plant water and carbon to fire behavior and effects, and their variability across species and time, underscore the value in applying plant carbon and water process models to explore the biophysical mechanisms linking vegetation to fire behavior and effects under present and future conditions.Recent advances in whole-plant modeling offer a promising framework for linking simulations of plant water-and carbonrelated processes to finer-scale models of fire behavior and effects.Nolan et al. (2018Nolan et al. ( , 2020) ) suggested a physiological basis to model LFMC for fire behavior applications using leaf water potential as a proxy.This approach allows for species-specific variation with environmental drivers, while acknowledging the importance of capturing variation in leaf mass to account for decoupling of seasonal changes in the water and carbon contents that comprise LFMC.Mechanistic functional structural plant models (FSPMs) (Vos et al., 2010;Louarn & Song, 2020;de Vries, 2021), of 3D plant structure, environment, and physiological response, from gene to community scales, offer a potential solution to capture both water and carbon cycle processes.Functional structural plant models, which can simulate water and carbon flows across 3D plant compartments (root, stem, and leaf) in response to environmental conditions, have been demonstrated for use in simulation of both small plants and trees and include open-source platforms that enable coupling with external modeling tools (e.g.Zhou et al., 2020).
Such a model could be coupled to fine-scale fire behavior models using fuel modeling frameworks such as Fuel3D (Parsons et al., 2011), FuelManager (Pimont et al., 2016), STANDFIRE

Review
Tansley review New Phytologist (Parsons et al., 2018), or FastFuels (Parsons et al., 2020).These frameworks can generate the 3D structure of canopy, mid-, and under-story vegetation for input to physics-based fire models, such as FIRETEC or WFDS, allowing examination of within-stand 3D fuel-fire interactions.Currently, meteorological or remotely sensed data (see the Remote sensing of plant water and carbon to inform process modeling section) can be used to set the moisture properties in these fuel models.Moisture is most often assigned a fixed value by the user for the course of the simulation across minutes to hours, and only FuelManager allows for assignment of alive vs dead status.
As an advance to enable mechanistic evaluation of fine-scale interactions between physiological and heat transfer processes and their influence on fire behavior and effects, we propose incorporation of a physiological process model to determine fuel condition based on physiological descriptions (plant water and carbon-related processes) of individual plants (e.g.Cochard et al., 2021;Ruffault et al., 2022a,b).This could be achieved through use of a FSPM with remotely sensed parameterization (Fig. 5a,b) to inform 3D fuel properties (Fig. 5c).These properties would provide the necessary inputs to parameterize fire behavior models (Fig. 5d), the outputs of which could inform predictions of subsequent plant water and carbon dynamics (Fig. 5a).Model representations of these postfire physiological processes could then inform second-order fire effects (Fig. 5e) through a process-based mortality model (e.g.Michaletz & Johnson, 2008;Butler & Dickinson, 2010) and subsequent 3D fuel properties (Fig. 5c).This approach will require parallel advances in mechanistic modeling of whole-plant fire injury processes to determine whether interacting injuries result in ultimate recovery, mortality, or functional limitations (Hood et al., 2018;B€ ar et al., 2019), which will be essential to determining subsequent fuel properties.
The individual plant (or within-plant) detail of this approach will enable further exploration of the plant traits most critical to the process of, and sensitivity to, heat transfer, allowing for determination of scale-relevant parameters.A model sensitivity approach varying plant physiological input parameters could be used to define functional groupings, based on species traits (response of tissue-level water and carbon to environmental drivers) that influence fire behavior and effects, which may allow for more generalized response functions to reduce computational resource requirements for use at larger scales.
By directly linking combustion, physiological process, and fireinjury plant-mortality models to hydrologic and plant succession models in a spatially explicit way, this framework could potentially be extended to simulate landscape-scale ecohydrology, and its response to fire disturbance under present and future climate conditions.By calculating transient heterogenous fuel moistures and temperatures based on incident radiation, heat fluxes, and topographic variation, this model extension could capture species-(or functional type-) dependent responses to soil moisture at different depths.Landscape-scale models of ecohydrology could use available downscaled climate reconstructions and models and spatially explicit maps of soil characteristics, topography, and aspect (e.g.Thornton et al., 2014;Soil Survey Staff, 2022) to simulate plant and fire response to landscape heterogeneity.This would allow plant water and carbon to vary spatially, influencing both fire behavior and vegetation mortality, which feedback to alter hydrology and plant succession, influencing ecosystem trajectory and response to subsequent fire disturbance.Such a framework could also allow for exploration of the role of other natural disturbance, alongside species composition and competitive interactions, in determining fuel load and the likelihood of future fire.
Another way to link fine-scale mechanistic models to coarser scale outcomes is through metamodels and nested model design.These techniques aim to find the crucial aspects of the mechanism that would be propagated at a higher level of spatial or temporal coarseness.Traditional or machine-learning metamodels can emulate nonlinear processes at scales where simulation would be unfeasible (Sparks et al., 2011;Huang et al., 2016).These can be integrated into existing coarse-scale models to propagate fine-scale understanding in a more computationally efficient manner (Lu & Ricciuto, 2019).A nested model design can allow for experimentation at relevant scales of physiology and mechanism to constrain coarser models or metamodels.For example, species trait-based functional types of fire behavior and effects can be created and used at coarser scales; while separately simulating component species at scales relevant to physiology, combustion, or fire injury (Neilson et al., 2005).Bayesian methods can further allow for integrating multiple nested models with observational data to constrain the performance of coarse-scale models along with the propagation of uncertainty (Talluto et al., 2016).

V. Remote sensing of plant water and carbon to inform process modeling
Remote sensing of vegetation exploits the biophysical links between spectral information contained in the electromagnetic radiation reflected or emitted from the earth's surface and vegetation state, function, and dynamics.Optical data have been used to map LFMC using multispectral sensors at local (Marino et al., 2020), continental (Yebra et al., 2018a), and global (Quan et al., 2021) scales, producing mapping products at spatial resolutions from tens to thousands of meters.Visible spectrum reflectance is collected by many remote sensing platforms for vegetation monitoring purposes; but alone is insufficient for the prediction of LFMC.Addition of spectral data from the shortwave infrared (SWIR) region of the electromagnetic spectrum, currently collected by only a limited number of platforms, increases the accuracy of optical LFMC estimations (Yebra et al., 2018a).One challenge with direct optical estimation of fuel moisture is that satellite and airborne platforms output reflectance data as two-dimensional coverages.In forested areas, these 2D data typically capture canopy conditions while obscuring understory characteristics (Yebra et al., 2018b), which are likely to differ widely from the overstory due to compositional and physiological differences.Similarly, these data are confounded by the presence of clouds, which are transient, yet exceedingly common, in many fire-frequent areas of the world.
Acknowledgment of these challenges has led to research on the application of microwave remote sensing for direct and indirect fuel moisture content estimation (Fan et al., 2018;Wang et al., 2019).Typically, LFMC is estimated using radar-derived vegetation optical depth, a proxy for vegetation water content (Moesinger et al., 2020).Radar systems operate in the microwave portion of the electromagnetic spectrum and can penetrate cloud cover and the upper canopy to detect the volume of liquid water in vegetation, offering an advantage relative to optical data.Nonetheless, a problem shared by direct microwave and passive optical measurements of LFMC in forests is the discrimination of signals related to the forest canopy, understory, surface litter, and soil (Gale et al., 2021).Merging microwave and optical data is a promising way forward.Rao et al. (2020) demonstrated that radar backscatter observations (from Sentinel-1) directly enhanced LFMC predictability in comparison with only using optical reflectance (from Landsat-8) for foliar, herbaceous and fine woody fuels.However, LFMC of denser woody fuels can only be estimated using longer wavelengths (i.e.L-band radar; Tanase et al., 2015).Despite these advances in mapping LFMC, little has been done to provide independent measures of dry mass and water content in the estimation of LFMC.To this end LMA (g m À2 ; also called dry matter content for remote sensing applications) and equivalent water thickness (EWT, g m À2 ) are appropriate target traits for optical sensing, capturing the effects of changing carbon allocation and moisture, respectively.Detection of EWT using optical sensors has a strong physical basis because water absorbs near infrared and shortwave infrared radiation (Yebra et al., 2013).At the leaf scale, variation in LMA has a moderate influence on reflectance in the short-wave infrared wavelengths (Feret et al., 2008) given the absorption features of structural properties found in leaves (Curran, 1989).For example, estimation of morphological and structural biochemical properties, including LMA, using reflectance spectroscopy of dried leaf material has been shown to be highly effective (e.g.Serbin et al., 2014).However, the presence of water in hydrated leaves can make the spectral estimation of dry matter content more challenging because water absorption masks the contribution of dry matter to the spectral response (Bowyer & Danson, 2004;Riano et al., 2005).Despite these challenges, the estimation of LMA at leaf and canopy scales has been shown to have reasonable accuracy using hyperspectral instruments (Ely et al., 2019;Chlus et al., 2020;Kamoske et al., 2021), likely due to covariance between EWT and LMA, as well as between leaf-level functional properties and plant and canopy growth form and structure (Ollinger, 2011).Radiative transfer models have also been inverted using multispectral optical data to simultaneously estimate LMA and EWT and, therefore, LFMC (Yebra et al., 2013).
At smaller scales, techniques exist that allow for high-temporal resolution vegetation carbon and water inventories.For instance, novel spectroscopic approaches leveraging high-resolution spectratrait modeling (Serbin & Townsend, 2020;Burnett et al., 2021) provide the opportunity to more directly link spectral signatures with underlying physiology and plant condition (e.g.water and carbon content) in 2D (e.g.Dahlin et al., 2013;Singh et al., 2015) and 3D (Chlus et al., 2020;Kamoske et al., 2021).Likewise, dualwavelength lidar shows promise in moisture content estimation due to its ability to distinguish forest layers through ranging (Gale et al., 2021).These approaches are currently not available on space-based platforms, limiting spatial coverage.However, these systems can be used together with ground-based systems (e.g.imaging spectroscopy) to calibrate airborne and spaceborne hyperspectral (e.g.EnMAP, Guanter et al., 2016;SBG, Cawse-Nicholson et al., 2021; OzFuel, ANU Institute for Space, 2021), thermal (ECOSTRESS, Anderson et al., 2021), and lidar (GEDI, Rishmawi et al., 2021) missions.
While current remote sensing LFMC products lack immediate predictive capability beneficial for fire management, they can be used to improve next-generation process models directly (parameterization) or indirectly (benchmarking predictions).This will lead to improved simulations of fire behavior and spread under changing climate conditions.A challenge posed by the availability of remote sensing data is ensuring integration of data from disparate sources with existing databases for fuel characteristics and LFMC.New 'big' data systems are needed for standardized fusion of data and scalable dynamic updates to fuel data in a changing environment.

VI. Conclusions and future directions
Fire behavior and effects are intrinsically associated with plant physiology through water and carbon cycles.Integration of plant water and carbon process models with fine-scale fire models will allow for the representation of process-level feedbacks between fire behavior and effects, enabling the use of virtual experiments to explore vegetation responses to global change scenarios (increased temperature and vapor pressure deficit, change in precipitation) and resulting fire outcomes.Recent advances in remote sensing, in combination with upcoming sensor deployments, will enable highresolution mapping of plant water and carbon status across regions and seasons, providing parameterization and benchmarking for next-generation models.Several areas are ripe for development across the disciplines of plant physiology, fire modeling, and remote sensing to fully enable these advances in understanding and simulation of vegetation-fire interactions.Furthermore, while the research presented here focuses on woody plants (trees and shrubs), additional efforts are needed to understand the interactions between climate, fire, and physiology of other vegetation types that contribute to fire regimes (Knapp, 1985;D'Antonio & Vitousek, 1992;Taylor et al., 2014;Simpson et al., 2016).Parallel advances on all frontiers will be critical to meet the urgency of the wildfire crisis.
With respect to fire behavior, we need to better understand the impacts of plant carbon and water on heat transfer and their interaction with fuel structure beyond the leaf scale.For fire effects, more studies are necessary to disentangle the physiological impacts of fire on tissue-level water and carbon status, along with interactions across organs to enable scaling to whole plant function and mortality (Hood et al., 2018;Michaletz, 2018;Kleynhans et al., 2021).We need benchmark datasets of physiological mortality mechanisms across organs, species, size class, life history, geographic region, season, and under climate change scenarios (Hood et al., 2018).New terrestrial lidar approaches for evaluating fuels and fire effects that are based on structure and visual-spectrum imagery may be a useful approach for determining these New Phytologist (2023) 238: 952-970 www.newphytologist.comÓ 2023 The Authors New Phytologist Ó 2023 New Phytologist Foundation New Phytologist benchmarks and understanding vertical heterogeneity in fire impacts on plant organs at these scales (Gallagher et al., 2021;Pokswinski et al., 2021).
Fire model implementation needs to address the nonlinear biophysical-processes of whole-plant heat transfer and resulting fire behavior and effects based on dynamic physiology, including refined fire-effects predictions stemming from an improved understanding of mechanisms (Kleynhans et al., 2021).Advances in machine learning can provide estimations of unknown process and enhance computational efficiency but require us to understand underlying processes and diagnose cases of overfitting.Directly computing many of these dynamics may be increasingly tractable as computation progresses.We additionally need to understand how the interaction among these dynamics affects fire and tree mortality at scales relevant to human decision-making.Using the inherent advantage of each scale of fire models, nested model design can work to incorporate new understanding at multiple scales (for example, plant physiology, stand management of plant density, landscape fire spread and suppression, and global carbon balance).Developing and validating nested model designs will be crucial to understand how each scale of organization influences the larger scale of organization.
Advancements in model implementation require finer spatial and temporal resolution observations of important processes to fire behavior and effects.Concerning remotely sensed drivers, we need to disentangle the numerous contributions to foliar reflectance signatures, along with advanced techniques for detection of subcanopy vegetation (Gale et al., 2021).Improved spatial and spectral resolution will reduce uncertainty in characterization of water and carbon status across regions and seasons, and better statistical methods will make outcomes more robust across larger areas and disparate biomes (Yebra et al., 2013).Advancements in data reduction and/or model ingestion will be needed to allow effective use of this higher resolution data.Ground-based experiments and observations can be used to better describe vegetation-fire interactions, and every prescribed burn provides an opportunity for data collection and model validation.Nelson (2001) suggested that 'the complexity of the task may have precluded all attempts to develop a reasonably complete physics-and physiology-based model' of live fuel moisture.With an increasing wealth of drought physiology research, the advent of next-generation models, and new spaceborne sensor deployments on the horizon, we are finally poised to tackle the problem of providing a dynamic and mechanistic description of fire behavior and effects through vegetation processes.Doing so will better equip us and the ecosystems we depend on to survive and thrive in a future made uncertain by global change.development and writing.AB and ZJR were consulted for subject matter expertise on the manuscript.All authors contributed to manuscript writing and editing.
Introduction 953 II.Existing models of fire behavior and effects 955 III.Physiological controls of fire behavior and effects 956 IV.Integrating fine-scale physiology with fire models 961 V. Remote sensing of plant water and carbon to inform process modeling 963 VI. Conclusions and future directions 964 Fuel models Stylized set of fuel bed characteristics used in fire models Landscape fire succession Spatial simulation models of fire and vegetation dynamics at stand to landscape scales Operational fire Computationally inexpensive models that rely on empirical representations of fire behavior and fixed, stylized fuel models to determine flame length and resulting mortality Process-based fire effects Models which predict injury to different plant compartments based on heat transfer Vegetation demography Models which capture dynamic, size-structured vegetation Remote sensing Hyperspectral Acquired in narrow, contiguous wavelength bands; high spectral resolution allows for material characterization, inference of chemical/biological processes, and novel signature identification Lidar Light detection and ranging; active optical sensing method using return time of a pulsed laser to measure distance Microwave Detects backscattering of actively transmitted radiation in the microwave (1 cm -1 m) portion of the electromagnetic spectrum; longer wavelengths allow penetration of clouds, rain, and surfaces Multispectral Acquired in broad, often discontinuous, wavelength bands; low spectral resolution reduces cost and complexity for monitoring known signatures Optical Detects reflected radiation in the visible (VIS, 380-780 nm), near-(NIR, 780-1000 nm) and short-wave infrared (SWIR, 1000-2500 nm) portions of the electromagnetic spectrum; primarily passive Thermal Detects emitted radiation in the thermal infrared (TIR, 780 nm -1 mm) portions of the electromagnetic spectrum; primarily passive

Fig. 1
Fig.1Model classification according to scale and process representation.In this conceptual figure, we show the classification applied to different models of fire behavior and effects discussed in this review.We also highlight our proposed linkage between fine-scale computational fluid dynamics (CFD) and biophysical fire effects models.

Fig. 2
Fig. 2 Live fuel moisture content (LFMC, black), an important driver of landscape-scale fire behavior (a), is a function of plant carbon (C, orange) and water (H 2 O, blue) (b, c), which can have opposing influences on fire behavior in leaf level burn experiments (d, e).(a) Rate of spread (ROS), area burned, and fire occurrence (N fires ) as a function of LFMC, as shown by Dennison & Moritz (2009), Martin-StPaul et al. (2018), Pimont et al. (2019a,b), and others.(b) Monthly variation in foliar dry mass (orange) and water (blue) density, and foliar moisture content (dry mass/water mass, LFMC; black) extracted from Jolly et al. (2016) with WEBPLOTDIGITIZER (Rohatgi, 2021).Dashed vertical line indicates divergence between LFMC and water density driven by increasing dry mass.(c) Conceptual model of LFMC as a function of plant water and carbon cycles redrawn from Jolly & Johnson (2018).(d) Time to ignition as a function of leaf carbon (leaf mass per area, LMA) and water (moisture) as shown in Grootemaat et al. (2015) and Bianchi et al. (2019).(e) Flame and smolder duration as a function of leaf carbon (dry mass) and water (moisture), as shown in Grootemaat et al. (2015) and Bianchi et al. (2019).

Fig. 5
Fig.5Plant carbon (C) and water (H 2 O) dynamics provide the mechanistic linkage between fire behavior and effects.In our conceptual framework, plant carbon and water process models (a; adapted fromZhou et al., 2020), which can be parameterized by remotely sensed estimates of canopy carbon and water contents (b; e.g.Dahlin et al., 2013;Singh et al., 2015), provide input to 3D fuel models (c; e.g.Pimont et al., 2016;Parsons et al., 2018).Detailed fuel properties are then exported to physics-based fire models(d; adapted from O'Brien et al., 2018;B€ ar et al., 2019) that provide heat transfer estimates for input to the plant water and carbon model (a), which feeds into simulation of fire effects (e; adapted fromB€ ar et al., 2019).These effects, along with remotely sensed estimates (b), drive subsequent plant water and carbon (a), feeding 3D fuels (c), and the cycle repeats.

Table 1
Glossary of pyro-ecophysiology terms.Canopy bulk density (CBD, mass per unit volume) Measure of how closely canopy fuels are packed, reflects likelihood that fire can move through the forest canopy Carbon starvation Plant mortality resulting from inability of NSC to meet metabolic demands Cavitation Process by which excessive water tension causes expansion of dissolved air to form bubbles; in plants, this causes a break in the water column and a decrease in hydraulic conductivity Conduction Heat transfer through a material from a region of higher temperature to a region of lower temperature Consumption (mass per unit area) Amount of biomass consumed during fire Convection Aeat transfer by the movement of a gas or liquid Embolism Blockage of a vessel by a mass; in plants, caused by air bubbles formed in xylem via cavitation Equivalent water thickness (EWT, g m À2 ) Measure of leaf water content Fire radiative energy density (FRED, MJ m À2 ) Measure of the intensity of radiative energy released from fuel during a fire Hydraulic conductivity (water mass or volume per unit time per unit area) (Jolly & Johnson, 2018proach to live fuel research that considers how plant water and carbon cycles independently and collectively interact at the leaf and whole plant level to regulate flammability and subsequent fire behavior(Jolly & Johnson, 2018) Radiation Heat transfer through a gas or vacuum other than by heating of the intervening space Relative water content (RWC, %) Plant water content relative to its fully hydrated state Senescence Process of biological aging; can be stress-induced or developmental Traits Morphological, physiological or phenological features measurable at the individual level (Violle et al., 2007) Transpiration Loss of water vapor from a living body; in plants, evaporation via stomata Vapor pressure deficit (VPD, kPa) The difference between the amount of moisture in the air and how much moisture the air can hold when saturated; measure of atmospheric aridity Water potential (MPa) Pressure potential required to remove a water molecule from its matrix (e.g.xylem); a measure of plant water status New Phytologist (2023) 238: 952-970 www.newphytologist.comÓ 2023 The Authors New Phytologist Ó 2023 New Phytologist Foundation

Table 2
Glossary of models and remote sensing methods.
note that in neotropical savannas, where fire is common and accelerated fire frequencies and prolonged droughts New Phytologist (2023) 238: 952-970 www.newphytologist.comÓ 2023 The Authors New Phytologist Ó 2023 New Phytologist Foundation