Fire season differentially affects resprouting vigor of pyrophytic and mesophytic hardwoods in a southeastern U.S. pine savanna Emma F. Zeitler a,*,1 , Kevin M. Robertson b , Cinnamon M. Dixon b , Marcus A. Lashley a a Department of Wildlife Ecology and Conservation, University of Florida, Gainesville, FL 32611, United States b Tall Timbers Research Station, 13093 Henry Beadel Dr, Tallahassee, FL 32312, United States A R T I C L E I N F O Keywords: Fire season Prescribed fire Hardwood encroachment Persistence equilibrium A B S T R A C T Like many savannas worldwide, pine savannas of the southeastern U.S. contain pyrophytic (fire-adapted) broadleaf woody plants (hardwoods), as well as mesophytic (fire-sensitive) hardwoods that persist through cycles of top-killing and resprouting. The persistence of mesophytes may be facilitated by anthropogenic fire regimes that top-kill hardwoods when they have higher carbohydrate reserves in roots, and the effects of fire timing may interact with herbivory. We investigated the resprouting response of pyrophytic and mesophytic hardwoods to fire in the four seasons by measuring the change in above-ground woody biomass between two fire-free intervals as a relative growth rate (RGR), with half of the plants protected from browsing by white-tailed deer ( Odocoileus virginianus ) to assess potential interactions with herbivores. Spring fires similarly reduced RGR of both pyro- phytes and mesophytes (difference between functional groups [ Δ ] = 0.02, p = 0.86), the summer ( Δ = 0.66, p < 0.0001) and fall ( Δ = 0.51, p < 0.0001) fires disproportionately disadvantaged mesophytes, and winter fires resulted in relatively high resprouting vigor for both groups ( Δ = 0.31, p = 0.06). Similar patterns were shown by the biomass to which plants were predicted to equilibrate under a given fire regime (the persistence equilibrium). Herbivore access did not influence resprouting. Our results indicate that growing season fires, which correspond to historic fire regimes, inhibit the relative growth of mesophytes more effectively than dormant season fires. We recommend that fire season be considered in the restoration and maintenance of historic woody plant compo- sition and structure in pine savanna ecosystems. 1. Introduction Broadleaved deciduous tree species, referred to as “ hardwoods ” , are an important component of fire-dependent savannas around the world (Hoffmann et al., 2003; Salis et al., 2006; Colgan et al., 2012; Mitchard and Flintrop, 2013; Veldman et al., 2013). In these ecosystems, most hardwood individuals are limited to understory vegetation by frequent fire disturbances in the "fire trap" of repeated top-killing and resprouting (Grady and Hoffmann, 2012; Robertson and Hmielowski, 2014). Hard- woods native to such community types typically have evolutionary life history traits that distinguish them as pyrophytic (i.e., resilient to and promoting frequent fire). Historically, frequent fires excluded hardwood species that are mesophytic (i.e., shade-tolerant, fire sensitive) however, in recent decades, these species have invaded and persisted in North American pine savannas (Condon and Putz, 2007; Diaz-Toribio et al., 2020). This often occurs in areas where fire is long excluded from pine or oak savannas, as mesophytic hardwoods can create a dense understory, closed-canopy overstory, and leaf-litter fuel bed that hinders fire spread (Nowacki and Abrams, 2008; Kreye et al., 2013; Varner et al., 2016; McDaniel et al., 2021). These species may also invade where fire is temporarily excluded by small-scale soil disturbances, such as those created by tilling for wildlife food plots or firebreaks (Engstrom et al., 2022; Dixon et al., 2024), or larger-scale disturbances such as disking agricultural fields, intensive logging, and site preparation for tree planting (Ostertag and Robertson, 2007; Kirkman et al., 2004; Veldman et al., 2014; Bizzari et al., 2015). Once established, some mesophytic species can persist under frequent fire regimes by resprouting (Hartnett and Krofta, 1989; Keyser, 2019; Matusick et al., 2020; Davis, 2021; Robertson et al., 2021; Dixon et al., 2024). The encroachment of mesophytes in upland pine communities, even after decades of frequent fire, raises the question: How are these species able to persist under fire regimes that historically excluded them? One * Corresponding author. E-mail address: emma.zeitler@myfwc.com (E.F. Zeitler). 1 Present Address: Florida Fish and Wildlife Conservation Commission, 8932 Apalachee Parkway, Tallahassee, FL, 32311. Contents lists available at ScienceDirect Forest Ecology and Management journal homepage: www.elsevier.com/locate/foreco https://doi.org/10.1016/j.foreco.2024.122478 Received 26 October 2024; Received in revised form 12 December 2024; Accepted 16 December 2024 Forest Ecology and Management 578 (2025) 122478 Available online 22 December 2024 0378-1127/© 2024 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). potential factor is that fire regimes in the southeastern U.S. have shifted in seasonality in a manner that might differentially affect pyrophytes and mesophytes. Lightning-ignited fires, which are posited to have maintained North American pine savannas before the arrival of humans, peak from late spring-early summer, whereas anthropogenic prescribed burning has most often been in late winter and early spring (Delcourt and Delcourt, 1997; Huffman et al., 2004; Outcalt, 2008; Knapp et al., 2009; Lashley et al., 2022). Several studies demonstrate that such shifts in the timing of fire can alter the relative abundance of species, popu- lation density, and above-ground productivity of woody plants (Whelan et al., 2018; Resop et al., 2023), potentially reflecting non-structural carbohydrate (NSC) allocation strategies and associated resprouting vigor following fire. For most hardwoods, NSCs are thought to be harbored in roots in the dormant season (winter to early spring) before being mobilized in the growing season (late spring to fall) to produce above-ground biomass (Bowen and Pate, 1993; Drewa et al., 2002; Schutz et al., 2009; Paroissien and Ooi, 2021). Thus, root stores are expected to be depleted during the growing season, increasing vulner- ability to top-killing disturbances, including fire (Gibson et al., 2016; Freitag et al., 2021). These concepts are supported by several studies which have reported greater decreases in hardwood growth, above-ground biomass, and plant density in response to growing-season fire as compared to dormant-season fire (Waldrop et al., 1992; Drewa et al., 2002; Hmielowski et al., 2014; Robertson and Hmielowski, 2014; Meunier et al., 2021). Furthermore, variation in post-fire responses at different points within the growing season have been observed, where early growing season (spring) fires resulted in greater declines in growth, likely because root stores had already been expended on biomass production (Glitzenstein et al., 1995; Woods et al., 1959). In contrast, late growing season (fall) fires resulted in a stronger resprouting effort, attributable to root stocks being replenished throughout the summer (Glitzenstein et al., 1995; Resop et al., 2024). Though both pyrophytes and mesophytes can be avid resprouters, they have trait differences that may cause divergences in their response to fire season. Pyrophytes have fire-resistant physiological adaptations, such as lower leaf nitrogen content, slower above-ground growth rates, and higher root-to-shoot ratios (Varner et al., 2016), which suggest that they have a more conservative strategy of storing more NSCs in the roots (Hoffmann and Solbrig, 2003; Varner et al., 2016; Nguyen et al., 2019). In contrast, mesophytes typically have a more acquisitive strategy, prioritizing above-ground growth over below-ground energy storage, at the cost of their capacity to resprout following top-killing (Waldrop, 1992; Hoffmann, 1998; Varner et al., 2016; Nguyen et al., 2019). It follows that pyrophytes may have higher growth rates than mesophytes when top-killed during the growing season when belowground reserves of NSCs are limited, while mesophytes may have greater growth rates than pyrophytes during the dormant season when NSCs are plentiful. However, such comparisons between resprouting pyrophytes versus mesophytes have not been made to date. Interactions between fire season and hardwood functional groups may have effects on resprouting vigor that accumulate over multiple fire return intervals. Past studies have reported that from one fire-free in- terval to the next, relatively small individuals tend to grow back larger than they were before being top-killed by fire, but only up to a certain size. This size has been named the "persistence equilibrium" (Fig. 3), which appears to vary according to environmental conditions, fire re- gimes, and plant species (Hoffmann and Solbrig, 2003; Grady and Hoffmann, 2012; Robertson and Hmielowski, 2014; Nguyen et al., 2019). Both the rate of resprout growth following individual fires and the persistence equilibrium over time presumably depend on the amount of non-structural carbohydrates (NSCs) stored in roots at the time of top-killing and the post-fire allocation strategy of species (Schutz et al., 2009; Clarke et al., 2013; Chen et al., 2017; Wiley et al., 2017; Smith et al., 2018; Miranda et al., 2020). Thus, these responses may be subject to effects of season of fire and differ between pyrophytes and mesophytes. Fire seasonality could also differentially affect mesophytes indirectly by interacting with herbivores (Tyler, 1995; James et al., 1997; Kim and Holt, 2012; Harper et al., 2016). Herbivores are often attracted to recently burned sites, as resprouting after top-kill makes a more nutri- tious and palatable browse available for consumption (Westlake et al., 2020; Nichols et al., 2021; Lashley et al., 2022). Some studies indicate that herbivory on plants regenerating after fire can alter the relative abundance of those plants in the community through unequal strength of top-down control across plant species (Lashley et al., 2015; Cherry et al., 2017; Forbes et al., 2019; Westlake et al., 2020). In these sce- narios, herbivory may exacerbate the negative impacts fire has on tree growth, and thus, the NSC storage strategies that influence post-fire recovery may also impact hardwood species ’ resilience to herbivory. If fire seasonality and herbivory interact in a way that differs between pyrophytes and mesophytes, this may affect any competitive advantages that fire season confers to one group over the other and thus, assessing the relative effects of fire season with and without herbivory is an important consideration when evaluating the net effects of fire timing on plants. Based on the work reviewed above, we hypothesized that the sea- sonal timing of fire would differentially affect post-fire aboveground growth of pyrophytic versus mesophytic hardwoods because of differ- ences in their life history traits, namely the phenological strategies of below-ground resource allocation. This entails that 1) both pyrophytes and mesophytes commit most of their belowground resources to aboveground growth at the beginning of the growing season, limiting the energy available for resprout growth following top-kill at that time; 2) pyrophytes have a conservative strategy by which more NSCs are stored in root reserves than mesophytes throughout the growing season, providing more energy for resprout growth following top-kill at that time; 3) root reserves are replenished for both pyrophytes and meso- phytes near the end of the growing season and in the dormant season, providing larger energy reserves for resprouting following top-kill at those times. To test this hypothesis, we followed the growth of resprouts of two pyrophytic and two mesophytic tree species following top-kill by fire in each of four seasons (winter, spring, summer, and fall) over two fire-free intervals to determine the change in aboveground biomass from one interval to the next. To control for potential impacts of herbivory on the growth rates of pyrophytic and mesophytic resprouts, we used fencing to exclude the largest herbivore in the region, white-tailed deer (Odocoi- leus virginianus), from half the focal individuals, allowing us to evaluate the degree to which observed differences were attributable to fire- herbivore interactions. We predicted that 1) fires in the early growing season (spring) would result in less growth of both pyrophytic and mesophytic resprouts from one fire-free interval to the next; 2) fires in the mid-growing season (summer) will result in relatively greater growth of pyrophytic resprouts than mesophytic resprouts; and 3) fires at the end of the growing season (fall) and dormant season (winter) will result in relatively greater growth of both pyrophytic and mesophytic resprouts. This study seeks to elucidate the possible role of woody plant life history traits shaped by historic fire regimes in predicting their resprouting responses to altered fire regimes. The results have poten- tially important implications for prescribed fire management seeking to minimize off-site species and favor historic plant species composition and associated ecosystem function of this and similar ecosystems. 2. Methods 2.1. Study site The study was part of a multi-year season of fire experiment at Tall Timbers Research Station (hereafter “ Tall Timbers ” ) in Tallahassee, Florida (30.6563, 84.2089). Tall Timbers is located in the Red Hills Region of northern Florida and southern Georgia and covers approxi- mately 2500 ha. The climate is subtropical with average annual E.F. Zeitler et al. Forest Ecology and Management 578 (2025) 122478 2 precipitation of 1575 mm and average temperatures for January and July (coldest and warmest months) of 10.6 ◦ C and 27.2 ◦ C, respectively (1960-present; National Weather Service, Southeastern Regional Climate Center). Elevation of research plots at the study site ranged from 52 to 75 m above mean sea level. The overstory of the site is dominated by shortleaf pine ( Pinus echinata ) and scattered hardwood trees, mostly post oak ( Quercus stellata ), southern red oak ( Q. falcata ), and mockernut hickory ( Carya tomentosa ). The study site represents a native shortleaf pine-oak-hickory community (Clewell, 2013) managed with 2-year in- terval prescribed fire in a section of the Tall Timbers property called the Anders North Course (Fig. 1). Soils are in the Ultisol order and consist of the Faceville (Typic Kandiudult) and Dothan (Plinthic Kandiudult) series (USDA Natural Resource Conservation Service Web Soil Survey (https:// websoilsurvey.nrcs.usda.gov/)). When the property was purchased by Tall Timbers in 1990 it had been fire-excluded for some years, after which most canopy hardwoods were removed and biennial prescribed burning was re-initiated, leaving shortleaf pine in the overstory and the native plant community largely intact. Ground layer vegetation is dominated by little bluestem ( Schizachyrium scoparium ), other perennial native C 4 grasses, a high diversity of forbs, and resprouting hardwoods. To date, problematic invasive species have been limited to the legume showy crotalaria ( Crotalaria spectabilis ) and the annual grass sweet tanglehead ( Heteropogon melanocarpus ), but these are mostly limited to areas of recent soil disturbance. In the early 1990s, eight replicate square blocks each 0.4 hectare (ha) in size were planted with longleaf pine ( Pinus palustris ), which currently have a basal area ranging from 6 to 12 m 2 ha-1, with younger shortleaf pine also present. We divided these blocks into four square 0.1 ha plots (30 m x 30 m; Fig. 2A). Within each block, we randomly assigned plots to have treatments with fire at one of four seasons: spring (April), summer (June), fall (October), and winter (January). The spring burns were after all individuals had leafed out, which mainly occurred during the month of March, and the fall burns were before the beginning of leaf senescence and the first frost of the season. Winter burns were during the dormant season. Fine fuel loads were dominated by herbaceous biomass, pine needle litter, and broadleaf litter senesced or deposited during the Fig. 1. A photo of a fire season treatment plot, located at the Anders North Course of Tall Timbers Research Station in Tallahassee, Florida, USA. The site represents a native shortleaf pine-oak-hickory community, managed with a 2-year interval prescribed fire. The overstory is dominated by shortleaf pine ( Pinus echinata ), post oak ( Quercus stellata ), southern red oak ( Q. falcata ), and mockernut hickory ( Carya tomentosa ), and the surface vegetation dominated by little bluestem ( Schizachyrium scoparium ), other perennial C 4 grasses, a high diversity of forbs, and resprouting hardwoods. E.F. Zeitler et al. Forest Ecology and Management 578 (2025) 122478 3 current or most recent dormant season. We burned the plots in their respective seasons in 2021 – 2022 and one year later in 2022 – 2023 (see Table 1 for specific dates). Burns were complete or near complete in each season. Burns were low intensity surface fires, characteristic of prescribed fires regularly conducted on the property (Reid et al., 2012; Robertson et al., 2014). We initiated the experiment just prior to the second set of treatment fires, when plants in each treatment had been burned and top-killed one year prior. We collected data from March 2022-January 2024. 2.2. Hardwood responses to fire in different seasons To assess the difference in resprouting responses between pyrophytes and mesophytes, we selected two species to represent each functional group (USDA FEIS, 2023; Fig. 2C). We chose mockernut hickory (MH) and post oak (PO) to represent pyrophytic hardwoods. These species are Fig. 2. A conceptual depiction of the experimental design. Boxes b-d describe the experimental treatments. A) A map of the eight experimental blocks, located at Tall Timbers Research Station in Tallahassee, Florida. B) To determine the effects of fire seasonality, each block was divided into four plots where prescribed fire was administered in each of the four seasons. Colors indicate the season of fire. C) Four species of hardwood trees were monitored in every plot: pyrophytes mockernut hickory ( Carya tomentosa ) and post oak ( Quercus stellata ) and mesophytes sweetgum ( Liquidambar styraciflua ) and water oak ( Quercus nigra ). D) To isolate the effects of herbivory, trees (n = 419) received one of two treatments: protection from herbivory using a metal exclosure, and exposure to herbivory. Table 1 Application dates of prescribed fire treatments within each experimental block. Burn series Season of fire Block 1 Block 2 Block 3 Block 4 Block 5 Block 6 Block 7 Block 8 1 Spring 4 APR 2021 6 APR 2021 13 APR 2021 7 APR 2021 7 APR 2021 7 APR 2021 7 APR 2021 4 APR 2021 Summer 3 JUN 2021 3 JUN 2021 3 JUN 2021 4 JUN 2021 4 JUN 2021 4 JUN 2021 4 JUN 2021 4 JUN 2021 Fall 17 OCT 2021 17 OCT 2021 19 OCT 2021 20 OCT 2021 19 OCT 2021 19 OCT 2021 19 OCT 2021 20 OCT 2021 Winter 30 JAN 2022 30 JAN 2022 30 JAN 2022 30 JAN 2022 30 JAN 2022 30 JAN 2022 30 JAN 2022 30 JAN 2022 2 Spring 11 APR 2022 11 APR 2022 10 APR 2022 10 APR 2022 10 APR 2022 11 APR 2022 10 APR 2022 11 APR 2022 Summer 3 JUN 2022 3 JUN 2022 20 JUN 2022 20 JUN 2022 20 JUN 2022 21 JUN 2022 8 JUN 2022 20 JUN 2022 Fall 18 OCT 2022 18 OCT 2022 18 OCT 2022 18 OCT 2022 19 OCT 2022 19 OCT 2022 19 OCT 2022 19 OCT 2022 Winter 16 JAN 2023 16 JAN 2022 16 JAN 2022 16 JAN 2022 16 JAN 2022 16 JAN 2022 16 JAN 2022 16 JAN 2022 E.F. Zeitler et al. Forest Ecology and Management 578 (2025) 122478 4 considered pyrophytic because they facilitate fire by producing leaf litter in which leaves are large, thick, strongly curled, and quick to dry (Kane et al., 2008; Varner et al., 2016; McClure et al., 2022; Kreye et al., 2023), are strongly associated with many native pine savanna commu- nities in the region (Ostertag and Robertson, 2007; Carr et al., 2010), and historically occupied upland ecosystems frequented by fire (Schwartz, 1994). We chose sweetgum ( Liquidambar styraciflua , SG) and water oak ( Q. nigra , WO) to represent mesophytic hardwoods as they are known to inhibit fire because of their flat leaves that retain moisture in the litter layer (Varner et al., 2016; Kreye et al., 2018), have thin bark with relatively low insulation against heating (Hare, 1965), and were not recorded as occurring in upland pine savannas in surveys conducted in the early 1800s (Schwartz, 1994). Each of these species can become a forest canopy tree (ca. 25 m height) at full growth in the region. Just prior to the second set of treatment fires beginning in 2022, we sought to locate six individuals of each species in each of the 24 plots, but numbers varied in some cases (2 7) with a total average of six. We chose hardwoods that appeared to have been top-killed by the previous fire one year prior and then resprouted, based on their size and the presence of the top-killed stem. Additionally, we chose individuals that were within the height range of 0.5 – 2.5 m. Considering that the area containing plots had received biennial prescribed fire in the spring for over two decades prior to the experiment ’ s initiation, the focal trees had likely undergone multiple cycles of top-kill and resprouting. Most in- dividuals had multiple stems, so we assumed that the stems belonged to the same genetic individual if they were rooted within 20 cm of each other, based on partial excavations of plants not in this study. We marked plant locations with fire-proof aluminum flags and tagged them with aluminum tags with a unique identification number. Total numbers of plants with species combined were 194 for spring, summer, and fall and 190 for winter. To test the effects of season of fire on resprouting hardwoods, we examined the change in above ground biomass from the end one fire-free interval to the end of the next fire-free interval, where both intervals were one year. At the end of the first fire-free interval, we measured the diameter at the base (mm) of all resprouting stems judged to have been top-killed by the previous fire one year prior, based on the presence of top-killed stems and sizes of living stems. We made these measurements one month prior to the subsequent fire, applied in seasons corresponding to their treatments (spring 2022 - winter 2023). Post-fire measurements were taken one year after these fires within each treatment (spring 2023 - winter 2024). To measure the effect of herbivore access on changes in tree growth from one fire-free interval to the next, we excluded deer from half of the hardwood individuals to compare plant responses to fire season with and without herbivore access. To protect plants, we used 1.5 m tall, graduated wire fencing to construct 1 m diameter herbivore exclosures. Fencing gaps ranged from 5 cm 2 at the bottom to 15 cm 2 at the top. Within a week following the fire treatment, we placed these exclosures around half of the individuals, which we randomly chose within each species in each plot, then maintained for the duration of the experiment. Considering that individual trees had recently lost all biomass to fire, these exclosures were sufficient in protecting the trees from deer- mediated herbivory for most of the experiment. There was potential for some browsing by deer in the latter portion of the fire-free interval (months 9 – 12 after experiment initiation) when a small proportion of leaves had extended between the mesh gaps. By comparing the change in growth of individuals with and without white-tailed deer access, this design allowed us to detect if fire-herbivore interactions substantially contributed to the net effects of fire season. We used previously published allometric equations (Robertson and Ostertag, 2009) to calculate the woody biomass of each individual hardwood tree. Because most individuals had multiple stems, we calculated the woody biomass from the diameter measurements of each stem and summed them to derive the total biomass for the individual resprout. Although allometric equations for leaf biomass were also available, we only used woody biomass as a relative measure of growth since the species are deciduous and leaf biomass varies greatly with herbivore pressure and leaf senescence over the course of the year. Any tagged individuals that were not top-killed one month after fire were removed from the study. If an individual did not produce live stems or could not be located for two consecutive post-fire measurements, we recorded these individuals as dead and removed it from the study. We removed a total of 25 plants in the spring, 35 in the summer, 54 in the fall, and 22 in the winter treatments (Table 2). 2.3. Data analysis For hardwood individuals that were top-killed and resprouted, we used the change in biomass relative to the pre-burn biomass to deter- mine the effects of fire season and herbivory. For our first set of analyses, we calculated relative growth of each resprout from end of the first fire- free interval to end of the second fire-free interval using the relative growth rate (RGR) equation: RGR = ln ( x 2 ) ln ( x 1 ) ( t 2 t 1 ) where x 1 and x 2 are the biomass at the end of the first and second in- tervals, respectively, and t 1 and t 2 are the dates of pre-burn measure- ments. RGR is useful for comparing growth differences in response to experimental treatments while controlling for the initial size (South, 1995). All RGR analyses were performed using R Statistical Software (v4.2.2; R Core Team, 2022). After calculating the RGR of individuals within each seasonal treatment, we identified and removed outliers via the Rosner test (spring = 2, summer = 1, fall = 1, winter = 1) to correct for human error and improve model convergence (Rosner, 1975). We did this using the rosnerTest function in the EnvStats package (Millard, 2013). To test the effects of fire season and herbivore treatments on RGR, we used a generalized linear mixed model (GLMM) in which RGR was the response variable and functional group (pyrophyte or meso- phyte), season of fire, herbivory treatment, initial biomass measure- ment, and their interactions were fixed effects, and the block was a random effect. The initial (end of the first fire-free interval) biomass measurement (natural log transformed) was included as a covariable because, according to the persistence equilibrium concept, it has an important effect on changes in growth from one fire-free interval to the next (Hoffmann and Solbrig, 2003; Grady and Hoffmann, 2012; Rob- ertson and Hmielowski, 2014; Nguyen et al., 2019). Models were fit using the glmmTMB function in the glmmtmb package (Brooks et al. 2017). The response variable was evaluated with Gaussian distributions. All models were analyzed with an analysis of variance (ANOVA) using the Anova function in the car package (Fox and Weisberg, 2019). We conducted linear contrasts using post hoc Tukey ’ s test via the emmeans function in the emmeans package (Lenth, 2023). In a second set of analyses, we calculated the persistence equilibrium for each combination of functional group, season of fire, and herbivore treatment, as well as for individual plant species within each season of fire treatment. We did this by running regression analyses fitting the linear model relating the natural log of biomass at the end of the first fire-free interval to the natural log of biomass at the end of the second fire-free interval, then solving for its intersection with the line repre- senting a 1:1 relationship (Fig. 3B). To determine the standard error, we subtracted the persistence equilibrium value from the independent variable (X axis value) and re-ran the regression, such that the Y inter- cept and its standard error were the same as the persistence equilibrium. We derived confidence interval from the standard error. We considered the persistence equilibrium to predict a longer-term effect of the fire regime on average biomass of resprouting individuals (Grady and Hoffmann, 2012). Although our use of the natural log of pre-fire biomass as a covariable in the first analysis partially accounts for this relation- ship, it is related to the dataset as a whole and may not reflect more E.F. Zeitler et al. Forest Ecology and Management 578 (2025) 122478 5 subtle responses specific to functional groups, among species, and fire season treatments. Also, since the persistence equilibrium is based on fitting data to a line relating pre- and post-burn biomass, it should be less sensitive to variations in initial biomass among treatments. Given that it is represented by a single value for each combination of factors, there are no statistical tests that can be applied. However, values compliment the RGR analysis for predicting the long-term responses of resprouting hardwoods. 3. Results Overall, pyrophytes had a larger RGR than mesophytes (Fig. 4). However, the season of fire influenced the magnitude of the difference between groups, such that pairwise comparisons showed significant differences only for the summer (mesophyte = 0.70 ± 0.11, pyrophyte = 0.042 ± 0.12, p < 0.0001) and fall (mesophyte = 0.08 ± 0.11, pyrophyte = 0.53 ± 0.11, p < 0.0001) fire treatments. Both groups had similarly positive growth rates in the winter fire treatment (mesophyte = 0.39 ± 0.11, pyrophyte = 0.70 ± 0.13, p = 0.06), and similarly negative responses in the spring fire treatment (mesophyte = 0.26 ± 0.09; pyrophyte = 0.24; p = 0.86). Individual species generally followed the pattern of their respective functional groups, although the responses of post oak (pyrophyte) and water oak (mesophyte) were not statistically different from each other in any fire season treatment (Table S3). The overall pattern of lowest RGR in spring, mixed levels of RGR in summer, and higher levels of RGR in fall and winter (Fig. 4) resulted in significant differences for the season of fire treatment (p < 0.0001) and a significant interaction between functional type and season of fire (p = 0.002) (Table 3). The covariable representing the natural log of pre-fire biomass also had a significant negative effect on RGR (p < 0.0001), i.e., plants with initially smaller biomass had higher RGR and vice versa, as expected based on the persistence equilibrium concept. The persistence equilibrium, calculated for each functional group and individual species within each season of fire treatment (Fig. 5A), showed similar patterns to those shown by the RGR analysis (Fig. 4A). Relative to the RGR analysis, the individual species showed greater distinction between the pyrophyte and mesophyte groups in the summer (mesophyte = 1.44, pyrophyte = 2.89) and fall (mesophyte = 2.72, pyrophyte = 4.27) season of fire treatments (Fig. 5B). In particular, whereas RGR was similar between post oak (pyrophyte) and water oak (mesophyte) among all season of fire treatments (Fig. 4B), post oak had a lower persistence equilibrium than water oak in spring (PO = 1.81, WO = 2.52) and winter (PO = 4.16, WO = 4.78) whereas the opposite was seen in summer (PO = 2.55, WO = 1.78) and fall (PO = 3.81, WO = 2.99, Fig. 5B). Adjusted r 2 values for the regression equations relating the natural logarithms of pre-burn biomass to post-burn biomass ranged from 0.110 to 0.668 with an average of 0.303, with p < 0.0001 for each regression analysis. Scatterplots and linear models are provided in the Supplementary Materials (Figures S1-S2). The herbivore access treatments did not significantly affect RGR within functional groups, regardless of group or fire treatment (Table S2). However, there was a tendency for hardwoods with herbi- vore access to have greater growth rates than those without herbivore access (Figure S3). The persistence equilibrium calculated for each functional group, season of fire, and herbivore access treatment showed the same general pattern among seasons of fire and pyrophytes versus mesophytes (Figure S2). Some cases presented fairly large differences between the herbivore access treatments, particularly in the fall treat- ment for pyrophytes (Table S3, Figure S4). Table 2 The sample sizes of hardwood tree species, functional groups, and fire season treatments that were included in final analyses. Species Functional Group Total Mockernut hickory Post oak Sweetgum Water oak Mesophyte Pyrophyte Spring fire 39 40 41 47 88 79 167 Summer fire 34 41 43 42 85 75 160 Fall fire 27 33 39 42 81 60 141 Winter fire 39 46 38 46 84 85 169 Total 139 160 161 177 338 299 637 Fig. 3. The persistence equilibrium concept, shown using hypothetical data, where Biomass 1 represents the biomass of a resprouting hardwood at the end of one fire- free interval and Biomass 2 represents the biomass of the resprouting hardwood at the end of the following fire-free interval after being top-killed by fire, shown for both (B) untransformed numbers and (A) log-transformed numbers. The persistence equilibrium has the same value for Bomass 1 (X axis) and Biomass 2 (Y axis) and represents the level below which resprouts tend to come back larger which resprouts tend to come back smaller following topkilling by fire. For a given plant, this equilibrium is expected to eventually be obtained after some number of fires (Grady and Hoffmann, 2012) and may vary according to plant species, environmental conditions, and fire regime (Hoffmann and Solbrig, 2003; Robertson and Hmielowski, 2014; Nguyen et al., 2019). E.F. Zeitler et al. Forest Ecology and Management 578 (2025) 122478 6 4. Discussion 4.1. Effects of fire season on biomass regeneration of pyrophytes and mesophytes Our data broadly support our hypothesis that the seasonal timing of fire differentially affects pyrophytic and mesophytic species of resprouting hardwoods in the pine savanna studied. The negative effect of spring fires on the RGR and persistence equilibrium of both pyro- phytes and mesophytes, attributable to low NSCs following the spring flush, was consistent with our first prediction. This result concurs with studies of deciduous resprouting woody plants in the southeastern U.S. that showed growing season fires to be associated with higher whole plant mortality (Glitzenstein et al., 1995; Resop et al., 2024) and lower NSC root concentration (Woods et al., 1959) of resprouting hardwoods compared to winter measurements. The greater resprout growth of pyrophytes than mesophytes following summer fire, which was consistent with our second prediction, supports our hypothesis that pyrophytes have a more conservative strategy for belowground NSC storage at that time of year. This result is consistent with previous studies that report summer fires, in comparison to winter or spring fires, reduce the dominance or above-ground growth of deciduous woody species (Drewa et al., 2002; Waldrop et al., 1992; Hmielowski et al., 2014; Little, 2023; Resop et al., 2024), and have more negative impacts on mesophytes than pyrophytes (Ferguson, 1957; Lewis and Harshbarger, 1976, Glitzenstein et al., 1995). In this light, our Fig. 4. The mean relative growth rate (RGR) of woody biomass with 95 % confidence intervals, compared across the fire seasons and hardwood tree A) functional groups and B) species. Colors indicate the functional groups (orange = pyrophytes, blue = mesophytes) and species (dark orange = mockernut hickory, light orange = post oak, light blue = sweetgum, dark blue = water oak). Common names of trees have been abbreviated (PO = Post oak, MH = Mockernut hickory, WO = Water oak, SG = Sweetgum). The dotted line marks where RGR = 0, the threshold point at which post-fire growth reached or surpassed the initial biomass. Table 3 ANOVA outputs derived from generalized linear mixed model analyzing the effects of experimental factors on the woody biomass relative growth rate of mesophytic and pyrophytic hardwood trees. df x 2 p variance Hardwood functional group 1 40.3 < 0.001 * ** Fire season 3 150.63 < 0.001 * ** Herbivory 1 2.46 0.12 ln(initial biomass) 1 271.37 < 0.001 * ** Functional group x Fire season 3 14.41 0.002 * * Functional group x Herbivory 1 0.003 0.96 Fire season x Herbivory 3 5.53 0.14 Functional group x Fire season x Herbivory 3 0.51 0.92 ln(initial biomass) x Functional group 1 0.08 0.93 ln(initial biomass) x Fire season 3 5.80 0.12 ln(initial biomass) x Herbivory 1 0.17 0.89 ln(initial biomass) x Functional group x Fire season 3 4.62 0.20 ln(initial biomass) x Functional group x Herbivory 1 3.23 0.07 ln(initial biomass) x Fire season x Herbivory 3 1.12 0.77 Full interaction 3 1.36 0.72 Block 0.02 * Asterisks indicate significance levels. * = ≤ 0.5 | * * = ≤ 0.01 | * ** = ≤ 0.001. E.F. Zeitler et al. Forest Ecology and Management 578 (2025) 122478 7 results suggest that the historic fire regime, characterized by lightning-initiated fires, mostly in the summer (Paul and Waters, 1978; Taylor, 1981; Cohen et al., 2007; Rother et al., 2018), was more effective at minimizing dominance by woody plants in general and mesophytic woody species in particular. In contrast to our third prediction, that fires at the end of the growing season (fall) would result in relatively high resprouting vigor for both pyrophytic and mesophytic hardwoods, mesophytes as a group still had lower RGR following fall fires (Fig. 3A). This result suggests that me- sophytes lagged behind pyrophytes in replenishing NSC root stocks. It is also consistent with mesophytes having a more acquisitive strategy, in which resources are allocated to above-ground growth throughout the growing season, leaving less available for post-fire recovery. As it is not uncommon for lightning fires to be initiated in the late summer and early fall, this difference between groups may have also disadvantaged me- sophytes under a lightning-fire driven regime (Cohen et al., 2007; Knapp et al., 2009; Rother et al., 2018; Taylor, 1981). It should be noted that significant differences in RGR between the groups were driven mostly by two species, mockernut hickory in the pyrophyte group and sweetgum in the mesophyte group, while the two oaks species had intermediate responses. The similar responses shown by the two oak species, despite being in different functional groups, may reflect common traits more broadly associated with the genus Quercus Conversely, these two species differed in their persistence equilibria in a manner that reflected the broader patterns for their functional groups (Fig. 4A). Although not statistically tested, the differences in persistence equilibria may be more important, as they predict long-term differences in above-ground biomass without being as strongly influenced by initial size distribution among treatments at the time of our experiment. Further studies may examine additional species within each group and clarify what is likely a gradient of resprouting responses among decid- uous woody plant species in pine savannas. 4.2. Role of fire-herbivore interactions in hardwood resprouting Herbivore access did not significantly affect hardwood resprouting responses in this experiment regardless of the season of fire. However, we did detect lower growth rates in individuals inaccessible to white- tailed deer. This negative effect may be explained by the cages inter- fering with lateral growth of hardwood branches, or by the shading ef- fect of vines growing over some exclosures. Altogether, we did not detect the top-down control of herbivory on resprouting plants that other studies demonstrate, wherein the combined effects of fire and herbivory limited plant recovery and shifted community composition more than fire alone (Okello et al., 2008; Fuhlendorf et al., 2009; Sensenig et al., 2010; Giljohann et al., 2017; Pelleg