Old-growth tropical and subtropical grasslands are biodiverse pyrogenic ecosystems with unique species composition, growth strategies, and disturbance regimes. These grassy ecosystems have received increasing attention due to threats from afforestation, climate change, agricultural expansion, invasive species, and alteration of the fire regime (Veldman et al. 2015; Lehmann and Parr 2016; Bond 2016, 2019). Although the global academic focus has been on hyper-diverse, extensive, and economically important tropical grasslands, there also are smaller, isolated, and evolutionary distinct old-growth subtropical and tropical grassland regions world-wide in need of further study. Among these are the subtropical treeless grasslands in south-central Florida, hereby referred to as Florida Subtropical Grasslands (FSGs). These and other grasslands once covered large parts of the southeastern coastal plain (Noss 2013) and are old, climatically buffered infertile landscapes characterized by fire-adapted heliophytic graminoids, forbs and subshrubs (Peet et al. 2018). The FSGs are part of the North American Coastal Plain biodiversity hotspot (Noss et al 2015; Pullaiah & Bridges, in prep.). Although they resemble other pyrogenic ecosystems of the North American Coastal Plain (e.g. pine savannas), we know little about their fire-adaptations and species traits.

Florida’s subtropical grasslands are among the last strongholds for these globally endangered biodiverse ecosystems, with south-central Florida having the largest extent of subtropical humid grasslands in the United States. Until recently, the distinctiveness of Florida grasslands was disputed by some (Abrahamson and Harnett 1990) as an artifact of pine clearcutting. Evidence for extensive grassy, nearly treeless regions in south-central Florida exists in presettlement land and timber surveys, early maps, and historic and contemporary accounts. Indeed, FSGs were delineated in public land surveys of the 1850’s (Bridges 2006a, b) prior to settlement and from historical accounts during early settlement (Harshberger 1914; Harper 1921, 1927). According to presettlement surveys, about 7910 km2 of the state was mapped as prairie (grassland), with an additional 158 km2 mapped as open pine savanna (Stephenson 2011). Timber cruise surveys by Consolidated Naval Stores Company from 1918-1921 prior to turpentine production and logging in south-central Florida, show large areas devoid of pine timber which coincidentally overlap with the presettlement survey mapped “prairies”. Modern accounts of Florida grasslands highlighting landscape, historical, fire regime and biota differences are numerous (Stephenson 2011; Bridges 2006a, b; Platt et al. 2006; Noss 2006, 2013; DeSelm and Murdock 1993). These grasslands were often bordered by pine savannas with a sparse canopy of Pinus palustris or P. densa over much of central Florida, which we consider collectively as the south-central Florida savanna grassland ecoregion, based on its similar species pool. The three geographically distinct grassland regions in peninsular Florida (Kissimmee River, Desoto-Glades, and Myakka) cover 5000 km2 of nearly treeless landscapes, and have been referred to as "Florida dry prairie" (Bridges 2006b). The largest contiguous grasslands in south-central Florida were on the uplands above the Kissimmee River floodplain, an area covering 2037 km2 (Fig. 1).

Fig. 1
figure 1

Location of study site at Avon Park Air Force Range in south-central Florida. The hatching delineates the presettlement extent of subtropical grasslands within the Kissimmee River region that once covered 2,037 km2 in south-central Florida

Both fire and extreme seasonal changes in hydrology interact to maintain the treeless nature of FSGs, on both a broader landscape and finer plant community level. In regions of south-central Florida with poorly-drained, flat, grassy terrain, early lightning season-ignited wildfires spread across vast areas with no major inflammable barriers. These nearly annual or biennial wildfires occurred when rainfall free intervals reached a maxima at the end of the winter dry season, coinciding with annually low groundwater levels and droughty conditions that enabled their spread (Platt et al. 2006). Upon onset of the summer wet season, rising water levels eventually saturate and often inundate the soil, producing conditions that negatively impact pine reproduction (Menges and Marks 2008) and favor herbaceous species (Platt et al. 2006). In peninsular Florida, it was found that plant community types were strongly correlated to edaphic-hydrologic variables on a landscape level (Huck 1987; Orzell and Bridges 2006a). Furthermore, it was determined that hydrology had fine scale differential effects on herbaceous and woody species richness in a Neotropical savanna (de Oliveira Xavier et al. 2019). None of these studies attempted to correlate plant traits with the hydrology gradient in FSGs.

Plant traits that enable survival from fire have been described for fire prone ecosystems world-wide (Clarke et al. 2015; Clarke 2016; Pausas et al. 2016, 2018; Freschet et al. 2021; Guerrero-Ramirez et al. 2020; Wigley et al. 2020; Paula et al. 2009). In Florida, such studies have focused on longleaf pine (Pinus palustris), a foundation species, wiregrass (Aristida beyrichiana), a keystone groundcover species, and saw-palmetto (Serenoa repens) (Abrahamson 1984b, 1995; Abrahamson and Abrahamson 2006; Fill et al. 2012; Wahlenberg 1946; Shibu et al. 2006). Post-fire recovery strategies and belowground organs were studied for Lake Wales Ridge xeric plant communities (Florida scrub and sandhills) (Abrahamson 1984a, b; Menges and Kohfeldt 1995; Maguire and Menges 2011; Menges et al. 2020; Saha et al. 2010) and for a smaller set of species from other Florida plant communities (Hierro and Menges 2002; Maliakal et al. 2000; Abrahamson et al. 2021). Some of these plants also occur in FSGs, particularly the geoxyle shrubs and subshrubs. The largest Florida dataset available for underground storage organs was obtained by Diaz-Toribio and Putz (2021), who assessed non-structural carbohydrates stored in belowground organs of 100 species from xeric pine sandhills and broadly classified belowground organs based on Pate and Dixon (1982), originally developed for southwestern Australia. Plants of FSGs may share many fire-adapted traits with those of other tropical and subtropical grasslands.,

We sought to determine whether 198 grassland specialist of FSGs have similar traits (plant growth form, belowground organs, and post-fire recovery) to other Florida fire-prone ecosystems, or to other subtropical or tropical grassy ecosystems, or if they have fire-adaptation traits unique to FSGs. To do so, we address the following questions:

  1. 1)

    What plant traits and strategies (growth forms, belowground organ types, and post-fire response strategies) occur among FSG plants?

  2. 2)

    Are these plant traits and strategies phylogenetically structured?

  3. 3)

    Do plant communities differ in representation of plant growth forms, belowground organs, or post-fire recovery strategies?

  4. 4)

    Are any of these plant traits or strategies affected by hydrology in FSGs?


Climate and Fire Regime

Our study area, Avon Park Air Force Range (hereafter AP), is located in south-central Florida, (27˚35’ N, 81˚16’ W; Fig. 1). The 424 km2 installation lies north of Lake Okeechobee in the Everglades headwater region and was established during World War II for pilot training.

The climate is seasonal humid subtropical, with an annual mild, dry (winter) and hot, wet (summer) season. On average, the wet season lasts on average 134 days (May 21 to October 1) with 89 ± 27 cm yr-1 of rainfall, and the dry season lasts on average 231 days (October 2 to May 20) with 42 ± 15 cm yr-1 of rainfall (Slocum et al. 2010). The mean annual temperature at the nearest station (Avon Park 2 W; 27° 36’N, 81° 32’W) from 1942-2005 was 22.8° C, with the highest daily mean (27.8° C) in August during the wet season, and the lowest (16.3° C) in January during the dry season. Tropical depressions, storms, and hurricanes can bring strong winds and locally heavy rainfall, especially during August and September from Cape Verde tropical systems (Chen and Gerber 1990) that originate in the Atlantic Ocean. Between 1851 and 2004, fourteen land falling hurricanes have passed within 50 miles or less of Avon Park AFR (Blake et al. 2005).

A hyperseasonal climate allows for rapid fuel accumulation during the wet season and curing of fuels during the dry season. Peninsular Florida is a global hotspot for cloud-to-ground lightning strikes, which are the ignition source for wildfires during the fire season (Platt et al. 2015). Lightning-ignited wildfires during the transition (late April thru mid-June) between the dry season and the wet season historically burned large areas (Platt et al. 2015). This seasonally-cued fire timing is an important evolutionary driver, producing a fire-filtered flora and pyrogenic plant communities. AP has a drought-driven fire regime characterized by high frequency, low severity fires, with annual to biennial lightning-ignited fires prior to human settlement (Huffman 2006). From the early 1940s through the 1980s there was widespread fire suppression in central Florida (Duncan et al. 1999, 2009; Fowler and Konopik 2007). However, during this period AP had an uninterrupted history of lightning- and military-ignited wildfires, followed by frequent prescribed ignitions for natural resource management. This uninterrupted fire history, without any long period of fire suppression, makes it ideal for study of fire adaptations.

Geologic, Edaphic, and Hydrologic Setting

AP is on the Osceola Plain, a poorly to very poorly drained terrace from 15-22 m elevation, with few natural barriers to the spread of wildfires. The surficial geology is mapped as undifferentiated Quaternary sediments of fine to coarse grained sands, sandy clays and clayey sands with admixtures of organics (Green et al. 2019). The soils are acidic, nutrient poor, sandy to sandy clay Spodosols and Alfisols. Poor drainage and impermeable subsurface layers (most commonly the spodic soil horizon) result in soil saturation or shallow inundation during the summer wet season and alternate with long winter dry season droughty soils. These create stressful physiological conditions for plants due to low water retention capacity, oligotrophy, and waterlogging, with seasonally high water tables that impact the distribution of herbaceous and woody species (de Oliveira Xavier et al. 2019) in hyperseasonal grasslands (Sarmiento and Monaterio 1983).

Biodiversity Setting

AP is a cornerstone of conservation in south-central Florida due to its size and, along with nearby conservation lands, provides some protection to over 319,700 hectares of globally significant biodiversity (Stein et al. 2008; Sorrie and Weakley 2006). About 38,000 ha of natural vegetation at AP is subject to recurrent fire. The 116,510 ha of grasslands within south-central and the southern peninsular Florida comprise the largest subtropical grassland region in the United States. AP also contains one of the largest exemplary remnants of longleaf pine (Pinus palustris) savannas in the southeastern United States (Sorrie and Weakley 2006; Peet et al. 2018), supports many federally listed species (Stein et al. 2008), and has some of the largest representative landscapes of pyrogenic ecosystems in peninsular Florida.

Vegetation Sampling

We classified 198 vascular plant species of FSGs as to: 1) plant growth forms; 2) belowground organs; and 3) post-fire recovery strategies. We used datasets created from quantitative vegetation plot sampling at burned sites over a 28-year period (1995-2023) and 6 years (2017-2022) of post-fire recovery sampling at AP. The vegetation plot sampling was used in assessment of community level relationships of plant growth forms and belowground organ types. Post-fire recovery sampling data allowed us to classify post-fire recovery strategies.

Quantitative vegetation sampling was timed to correspond with peak post-fire flowering of graminoids and forbs at the end of the wet season. 731 plots were sampled within intact groundcover at AP, in linear transects of 50 to 100 meters, with 1.0 m2 plots sampled every 5 meters. Transects were placed in areas burned during the year of sampling at sites lacking obvious signs of artificial disturbances or hydrologic impediments. Percent cover was estimated for each species from vertical projection of the aboveground plant material with the square meter, and for unvegetated plot substrate (i.e., bare ground or surface rhizomes). All vascular plant species within each plot were recorded and percentage cover, bare ground and surface rhizomes were visually estimated at each plot. Field data was recorded on a field sampling form that was later computerized into Microsoft Access for processing and analysis. Plant nomenclature follows Weakley et al. (2023).

Hydrology classes were assigned to plots based upon soil sampling, field evaluation of the highest wet season water table levels, published soil series drainage classes, and hydraulic conductivity descriptions. We recognized hydrologic variables at two scales, a coarse plant community with four types, and a finer-scale with nine vegetation types (Table S1). The species composition and environmental characteristics of the first six of these are described in detail by Orzell and Bridges (2006a).

Post-fire Recovery Sampling

We described and classified post-fire recovery traits for FSG vascular plant species. We sampled 14 sites averaging 3.0 ha each on dry-mesic to wet-mesic soil types at AP which were burned at various times between December and July, with sampling starting within one week post-fire at each site.

Following fire, measurements of leaf length, stem height, and/or number of culms were compiled, and phenology data were collected for up to 10 individuals of each species at each site. In the first year post-fire, these were sampled at 7-10 day intervals until about 100 days post-burn, at which point sampling frequency decreased gradually to about once a month by the end of the first growing season, and continued once every two months through the second year post-burn. Sampling began in 2017 and has continued through 2023, with all sites experiencing additional fires, some in different seasons.

At each site, a central photo point was marked using an angled aluminum stake. Four additional stakes were placed 5 meters from the central stake in the four cardinal directions. At the central stake, a camera was placed atop a 75 cm tripod and photos were taken of the surrounding vegetation at each sampling event, with a 2 meter PVC pipe marked at 20cm intervals placed at the directional stakes for each photo. This provided both consistency and scale for photo-documentation of the general aspect and changes in phenology and height of the vegetation.

To classify post-fire recovery strategies we used the systems of Pausas et al. (2016) and Clarke et al. (2015), independently assessing species with each classification in our initial evaluation, based on resprouting and seedling observations from 6 to 18 months post-fire. Pausas et al. (2016) describe four post-fire recovery strategies: obligate resprouters, facultative seeders, obligate seeders, and post-fire colonizers. Clarke et al. (2015) describe five post-fire recovery strategies: resprouters, non-resprouters, variable [resprouters], post-fire seeders, and post-fire non-seeders. By adapting criteria provided in both papers, we ultimately classified the FSG flora into three broad categories: obligate resprouters, facultative resprouters, and obligate seeders (Supplemental Information S1). Obligate resprouters rely on resprouting following fire and often lack fire resistant seedbanks. Facultative resprouters are species which post fire regeneration strategies include both re-seeding recruitment and resprouting recovery. Obligate seeders do not exhibit any resprouting following fire, but instead rely on seeding for post-fire regeneration.

Belowground Storage Organs & Plant Growth Forms Sampling

We collected belowground storage organs of 198 vascular plant species following the protocols in Klimešová et al. (2019), and Pausas et al. (2018), and used the rhizome terminology in Klimešová and Klimeš (2007). Most excavations were made during the winter dry season of 2016-2017, with supplemental excavations in 2021. For each plant species, representative individuals were hand-excavated to the deepest practical depth. We have not unearthed an entire genet of any of the large geoxyle species, since their areal extent, tough woody tissues, and mixed growth with other species make this impractical. Each specimen was photographed on a white or beige background with a ruler or a scale bar with 1 cm increments, before being dried and stored. Using the published literature, photos, and observations at earlier dates, each species was then classified by growth form (Palmquist et al. 2014; Pausas et al. 2018) and type of belowground organ (Klimešová et al. 2019) (Supplemental Information S2 & S3). Definitions of specialized terms are in S2. For belowground organs we also distinguished two rhizome textures (herbaceous and woody), where soft tissue is herbaceous in contrast to hardened lignified/woody tissue. Therefore, we call woody rhizomes woody but also we include lignified rhizome as in saw-palmetto (Serenoa repens) as a woody rhizomatous shrub, despite it lacking classic secondary thickening of the vascular cambium. We also classified rhizome extent based three lengths of lateral spread in rhizomes: short (up to 5 cm). medium (from 5 to 25 cm), and long (more than 25 cm). Rhizome type refers to epigeogenous or hypogeogenous rhizomes (Klimešová and Klimeš 2007).

Data Analysis

To examine species based on their belowground organs, plant growth form, and post-burn recovery strategy (Table S1), we calculated their relative occupancy rates (total cover of a category across all sites divided by the total cover of all categories together) when vegetation plots were sampled within 12 months post burn. Overall effects of site characteristics (hydrology and vegetation type, (Table S2) on species abundances were tested using multivariate generalised linear models (Wang et al. 2012; Warton et al. 2012). We fitted the models assuming negative binomial distribution with a logarithmic link function and unknown over dispersion parameter, which has been shown to have desirable properties when analysing multivariate abundance data (Wang et al. 2012; Warton et al. 2012). Statistical inference was based on Akaike’s information criterion (AIC), not on P-values (due to very long computation times to estimate these). We considered two or more AIC units as a decisive difference between the model including site characteristics as a predictor and the null model (with no predictors). Because the considered site characteristics provided overlapping information in some cases (Table S1), the explanatory power of all predictors of species abundances was examined using commonality analysis and hierarchical partitioning (Lai et al. 2022). For each predictor, these statistical tools estimate it’s unique (separated from joint effects with other predictors) and shared (the effect common to all predictors) contributions. These contributions can be used to calculate the independent effect of a predictor (sensu hierarchical partitioning; Chevan and Sutherland 1991; Lai et al. 2022) to assess the relative importance of predictors with some degree of collinearity.

To test whether species with different belowground organs, plant growth forms or post-burn recovery strategies tended to prefer or avoid sites with different hydrology or vegetation type, we performed fourth-corner analysis (Legendre et al. 1997; Dray and Legendre 2008). To study trait-environment links, fourth corner analysis measures the relationship between three matrices, one describing species x site abundances (L matrix), one describing site (environmental) variables (R matrix) and one describing species traits (Q matrix). Significance testing was based on chi-squared (X2) statistic and 999 permutations, and Holm correction (Holm 1979) for multiple testing was applied. We used the permutation model that permutes both values of sites and species because it does not have inflated type I error (Dray and Legendre 2008). All site and species characteristics were categorical variables (factors). In order to obtain as precise fourth corner results as possible, species belonging to very rare categories were omitted. The arbitrary threshold to keep a category for the analysis was if four and more species belonged to this category. For example, when linking belowground organ to hydrology and vegetation type, only annuals, species with epi- and hypogeogenous rhizomes and species with main root were considered (Table S2).

All analyses and graphics in R 4.2.1 (R Core Team 2022) were used in the packages: ade4 1.7-19 (Dray and Dufour 2007) to perform fourth corner analysis, ggplot2 3.4.0 (Wickham 2016) and ggtree 3.4.2 (Yu et al. 2017) to draw graphics, mvabund 4.2.1 (Wang et al. 2012) to perform multivariate generalised linear models, rdacca.hp 1.0-8 (Lai et al. 2022) to perform commonality analysis and hierarchical partitioning, and vegan 2.6-2 (Oksanen et al. 2022) to compute scores for ordination diagrams. We used Pagel’s lamba (Pagel 1999) as a measure of phylogenetic signal, which was estimated using the fitDiscrete function (geiger 2.0.10, Pennell et al. 2014). Species phylogeny was obtained using the V.PhyloMaker2 package (Jin and Qian 2022).


Species Level Analysis

The most common belowground organs were rhizomes (81 species with epigeogenous and 60 species with hypogeogenous rhizomes) followed by nonclonal perennial plants with main roots (23 species) and annual herbs (23 species). Clonality in herbs and shrubs is high in 151(76%) of the sampled species. Other types of belowground organs (root sprouters, stolons, stem tubers and bulbs) were represented by one to 4 species each (Table S2a). Among rhizomatous plants, long and short rhizomes were similarly abundant while herbaceous rhizomes were more common than woody ones.

The most common growth forms were herbaceous forbs (98 species), with the most common categories being caulescent herbs (43) and rosette herbs (31). Other non-graminoid herbs include species with elongate (medium to long) rhizomes (often clonal) (7), bulbs (1), corms (2), stem tubers (2), xylopodia (9), and hemiparasites (1) (Table S2b). Graminoids are represented by 70 species, these mostly cespitose (27), having single or few culms (23) or rhizomatous (12). The remaining species are shrubs (19), trees (4), ferns (2), vines (4), and lianas (1) (Supplemental Information S2).

The studied species were mainly characterized by an obligate resprouting strategy (133 species) while facultative resprouters (30) and obligate seeders (33) were less common (Table S2).

Evolutionary and Taxonomic Analysis

Florida subtropical grasslands have a species pool of warm-temperate and Neotropical floristic affinities, from which we assessed 198 species in 49 plant families (Table 1), and found the most diverse families being Poaceae, Asteraceae and Cyperaceae (Supplemental Information S3).

Table 1 Summary of plant families

Belowground organs were clearly associated with phylogeny in our species pool (Fig. 2A), reflecting deep phylogenetic differences between monocots and dicots. Epigeogenous rhizomes were highly prevalent in monocots (especially in Poaceae) but quite rare in dicots (except for Asteraceae, (Fig. 2A). Post-burn recovery strategies were relatively evenly distributed across the phylogeny (Fig. 2B). Taxonomic classifications underscored the observed differences in growth forms; graminoids were predominately found among monocots, while herbs were chiefly represented in dicots (Fig. 2C).

Fig. 2
figure 2

Species pool phylogenies with the distributions of A. belowground organ forms, B. post-burn recovery strategy and C. plant growth form across the tree. Rare trait categories (represented by less than four species) are merged (the “other” category) to reduce over plotting. Five the most species-rich families of the species pool are indicated to allow orientation. See Supplemental Information S1 and S2 for definitions of terminology

Community Level Aanalysis

Species with epigeogenous and hypogeogenous rhizomes dominated the sampled plots (Fig. 3A). In terms of growth form, graminoid matrix, shrub geoxylic (e.g. Quercus minima) and shrub rhizomatous (i.e. Serenoa repens) were the most dominant groups (Fig. 3B). In terms of post-burn recovery strategy, obligate resprouters dominated (Fig. 3C). Considering rhizomatous species only, more species had short (Fig. 4A) and herbaceous (Fig. 4B) rhizomes but the differences were slight. Including any of the site characteristics (hydrology and vegetation type) improved the model fit of species abundances (Table 2) and 72% or more species showed changes in their abundance according to these site characteristics (AIC improved by at least two units compared to a null model with no predictors). In total, hydrology and vegetation type explained around 16% of variability in species abundances. Unique contributions and independent effects were higher for vegetation type (both coarse and fine resolution, Table 2) while the effect of hydrology was mostly shared with them – this is because finer vegetation type levels were basically a combination of hydrology and coarse vegetation type (Table S1). Although vegetation type was a better predictor of species abundances than hydrology, it was hydrology only that was significantly associated with any species traits measured, namely rhizome texture (Table 3b). Species (Table 4) with woody rhizomes avoided wet sites (Fig. 5). In all cases, the information about evolutionary contingency- lambda model - improved AIC (by 15 units and higher) compared to a non-phylogenetic model. Lambda values (belowground organ = 0.88, post-burn recovery strategy = 0.74 and growth form = 0.97) mostly suggest an association of these traits with phylogeny.

Fig. 3
figure 3

Total observed proportion (relative occupancy rates) of species belonging to different categories in subtropical Florida grassland plant communities. A - belowground organ; B - plant growth form; C - post-burn recovery strategy. See Supplemental Information S1 and S2 for definitions of terminology

Fig. 4
figure 4

Total observed proportion (relative occupancy rates) of species belonging to different categories in subtropical Florida grassland plant communities. A - rhizome length; B - rhizome texture (herbaceous and woody)

Table 2 Explanatory power of three community aspects (site hydrology and vegetation type) of plant species abundances. Total variation explained by these community aspects was 0.164. AIC sum = overall model’ Akaike’s information criterion (the sum of AIC values of all species), AIC improvement = total number of species (and relative number in percentages) in which the tested community aspect improved AIC by at least two units (in comparison with a null model with no predictors), unique contribution = individual effect of a predictor without joint effects with other predictors, shared contribution = total average shared effect of a predictor with other predictors, independent effect = percentage independent effect of a predictor (hierarchical partitioning).
Table 3 Significance testing of the fourth corner analysis examining whether plant traits (belowground organ, or post-burn recovery strategy) affected species abundances according to site hydrology and vegetation type
Table 4 Example rhizomatous species with the strongest affiliation to the hydrology
Fig. 5
figure 5

Ordination diagram of the responses of the abundances of herbaceous and woody rhizomatous species (arrows) by community type and site hydrology


Within studied Florida subtropical grasslands, plant communities had similar growth forms, belowground organs and post-fire response strategies but differed in rhizome texture, with rhizomes being the most common belowground bud bearing and storage organ. In communities typical for wet sites, woody rhizomes were absent while both woody and herbaceous rhizomes occurred in communities typical of mesic and dry sites. Herbaceous and woody species are uniquely adapted to cope with hyperseasonal water availability and fire (Silva and Batalha 2011). Herbaceous species were more often seeders (annuals) compared to woody species but resprouters prevailed in all community types and among all perennial species (both woody and herbaceous). The studied traits were mainly phylogenetically conservative, i.e. they probably developed recently in the evolutionary history of represented taxonomical groups (monocots vs dicots).

Herbs and Shrubs

The most diverse growth forms in FSGs are caulescent and rosette herbs and cespitose graminoids. In terms of belowground storage organs, epigeogenous and hypogeogenous rhizomes prevail, and obligate resprouters are the most common post-fire recovery strategy. In terms of community dominance, the proportion of growth forms differ from those based on species diversity. Non-dominant small or dwarf interstitial plants (i.e. single culm, rosette, geophytes, matrix graminoids & hemiparasites) are emblematic of grasslands with recurrent fires, declining when fire is less frequent (Palmquist et al. 2014). In south-central peninsular Florida, herbs and shrubs with temperate and tropical affinities mix (Long 1974; Long et al. 1969; Orzell and Bridges 2006b) along transition zones producing many geographically unique species assemblages (Allen et al. 2019) in FSGs across subtle edaphic hydrologic gradients (Huck 1987; Carr et al. 2009, 2010). Shrub geoxyles (Ilex glabra, Quercus minima, Geobalanus oblongifolius), a pyrogenic acaulescent palm (Serenoa repens), matrix graminoids in the Cyperaceae (Rhynchospora, Scleria, etc.) and Poaceae (Aristida, Andropogon, etc.) comprise most of the dominant species. Two or three geoxyles are often co-dominant across large parts of the FSG landscape, the so-called geoxyle grasslands, clearly differentiating them from pine savannas. FSGs form unique species associations with forbs, dominant C4 grasses and a diversity of Xyridaceae, Cyperaceae, and Eriocaulaceae species which further serves to differentiate them from floristically similar Florida pine savannas while better aligning them with tropical rather than temperate North American grasslands. Rhynchospora is one of the richest genera in the southeastern coastal plain flora in endemics (Sorrie 2008), with 29 species found in FSGs (Orzell and Bridges 2006a), and 13 in our sample. The number of Xyris species and their abundance in FSGs differentiates these grasslands from other southeastern US grasslands. FSGs also have a modest diversity of six species of carnivorous plants, representing three families (Droseraceae, Lentibulariaceae, Sarraceniaceae) and four genera (Drosera, Pinguicula, Utricularia, Sarracenia). Moist sandy soils are known for their diversity of Ericaceae (Peet et al. 2018) and such is the case in FSGs where Ericaceae is represented by eight species in four genera (Bejaria, Lyonia, Gaylussacia, Vaccinium), all shrubby and rhizomatous.

Shrubs are not as species-rich in FSGs as herbs, an important difference in comparison with tropical grasslands that are characterized by numerous woody taxa (e.g. Brazilian tropical grassland Cerrados, Pausas et al. 2018). This perhaps is an effect of the relatively small size and climatic isolation of Florida subtropical grasslands (or suboptimal conditions) thereby the niche of resprouting shrubs is occupied by fewer species. In this respect, FSGs are more similar to the Rio de la Plata grasslands of southern Brazil, Uruguay and central-eastern Argentina, which cover a sizeable area (750,000 km2). This grassland region and FSGs both harbor a diverse herbaceous flora, but lack diversity of tropical woody species, especially tropical shrubs, a hallmark of Cerrado savanna grasslands (Overbeck and Pfadenhauer 2007; Simon et al. 2009).

Saw-palmetto (Serenoa repens), a monotypic genus in the Arecaceae (subfamily Coryphoideae) endemic to the southeastern United States Coastal Plain, has multiple adaptations to recurrent fire (Fig. 6). First, although its leaves are exposed aboveground to fire, most of its biomass is a semi-subterranean, epigeogenous rhizome that is protected from the direct effects of fire. The long petioles, which position the flammable leaf canopy away from the apical meristem, serve to dissipate heat during leaf combustion. When fire is long-excluded, S. repens can assume an ascendant small tree growth form (Fisher and Tomlinson 1973). Second, the rhizome is often covered with persistent, closely packed charred leaf bases, and thereby is insulated from fire by thickened hypodermal tissue (Fisher and Jayachandran 1999). Third, the rhizomes contain abundant storage of carbohydrates for resprouting, a common trait in resprouters (Diaz-Toribio and Putz 2021). Fourth, resprouting is both from terminal apical meristems and epicormic buds, the latter derived from axillary buds (i.e., suckering sensu Fisher and Tomlinson 1973). Resprouting from epicormic buds is a post-fire evolved recovery strategy in high fire frequented ecosystems (Pausas and Keeley 2014, 2017). The apical bud of Serenoa repens is not killed by most fires due to insulation by leaf bases (Abrahamson 1995; Abrahamson and Abrahamson 2006; Fisher and Tomlinson 1973). Insulation of apical buds (McPherson and Williams 1998) is a trait similar to that found in other fire-tolerant monocots, such as the “grass trees” of Australia (Xanthorrhoea spp.) (Lamont et al. 2004).

Fig. 6
figure 6

Saw-palmetto (Serenoa repens) is often dominant in peninsular Florida fire prone ecosystems. Note the semi-subterranean, epigeogenous rhizomes which easily reach several meters in length. The blackened rhizome and re-sprouting leaves are 21 days since burning. Photo by Steve Orzell

Serenoa repens exhibits very rapid post-fire leaf regeneration (Abrahamson 1984b; Menges and Kohfeldt 1995; Hierro and Menges 2002; Abrahamson 2007) due to its carbohydrate reserves (Hough 1965, 1968). Its leaves are larger in area, but are reduced in number when compared to sympatric groundcover species in FSGs, however, its leaf size is smaller than most other coryphoid palms (Abrahamson 2007). Our post-burn data show that S. repens resprouts with emerging new fronds in as little as 7-14 days post burn. Its rapid post-fire leaf recovery contributes to long-term persistence under recurrent fire, with some clonal genets of S. repens over 10,000 years old, and possibly much older (Takahashi et al. 2011, 2012). Similarly Liesenfeld and Vieira (2018) found post-fire resprouting greater in clonal Neotropical palms. Despite these fire adaptations, we consider Serenoa repens to be an aeroxyle (e.g., Pausas et al. 2018) since aeroxyles can resprout both apically and epicormically. Other dwarf acaulescent, fire-adapted palms include Allagoptera campestris, Sabal etonia, Sabal miamiensis, and Syagrus petraea. The Mediterranean dwarf palm (Chamaerops humilis), with its underground rhizomes and upright stems radiating from a clumped base, is adapted to recurrent fire in shrublands and also has apical buds which are protected by persistent leaf bases (Garcia et al. 2018; Pausas 2016). Dowe (2010) noted similar adaptations to fire retained leaf bases, insulation of the terminal bud in some palms from Australia, including Livistona humilis, L loriphylla, and L. muelleri.

Another typical component of tropical and subtropical savannas grasslands are geoxyles. In addition to their dominance in savanna grasslands they also occur in xeric sandy uplands. However, they are rarely mentioned in the Florida and the southeastern coastal plain literature (Abrahamson et al. 2021; Noss 2018; Peet et al. 2018). Many global studies (Bond and Parr 2010; Pennington and Hughes 2014; Gomes et al. 2019, 2021; Lamont et al. 2017; Meller et al. 2021, 2022; Corkalo 2023; Rickenback 2022) provide phylogenetic evidence for fire adaptation in geoxyles (Simon et al. 2009; Simon and Pennington 2012; Maurin et al. 2014). Their aerial stems are highly susceptible to annual or biennial fires, while their well-protected extensive belowground suffrutices with carbohydrate reserves afford rapid resprouting from bud banks (Fig. 7) (Pausas et al. 2018). The geoxylic suffrutex has been considered an evolutionary response to recurrent fire (Peet et al. 2018). Geoxyles contribute to ecosystem function, persistence, and resource storage; their clonal underground storage organs promote resilience through rapid resprouting in fire-prone open ecosystems (Tsakatos et al. 2022; Meller et al. 2021).

Fig. 7
figure 7

Runner oak (Quercus minima) a geoxyle: a - 60 days post burn, note the flush of young leaves; b – partially excavated woody geoxylic suffrutex of Q. minima, pole with 20 cm incremental marks for scale. Photo by Steve Orzell

Phylogenetically, the geoxyle flora of FSGs has evolved in multiple lineages and often have trees or tall shrubs as congeners (White 1976; Maurin et al. 2014; Gomes et al. 2019; Pennington and Hughes 2014; Girolda et al. 2017). While geoxyles in tropical grasslands originated in tropical or subtropical plant lineages (Pennington and Hughes 2014), in Florida there are geoxyles derived from mostly temperate families (Ericaceae, Fagaceae, and Aquifoliaceae), mostly tropical families (Chrysobalanaceae and Annonaceae), and families with no clear temperate or tropical affinity (Myricaceae). While geoxyles are considered typical of savanna regions characterized by high precipitation (Zupo et al. 2021) our study shows that they avoid wet and long-term flooded habitats within those regions (see results concerning woody rhizomes). Within the southeastern US coastal plain, peninsular Florida is the center of geoxyle diversity and evolution, where humid subtropical grasslands and savannas are maintained by some of the highest fire return intervals in the world. Frequent fire and high precipitation, both advantageous for geoxyles (Zupo et al. 2021), prevail in peninsular Florida more than elsewhere in the southeastern US.

Rhizomes, Moisture and Ffire

Belowground stems (rhizomes) with adventitious roots are the most common belowground organs not only in the subtropical savanna of Florida but also in grasslands with temperate flora (Klimešová and Klimeš 2008). Herbs and shrubs place their stems belowground to protect their bud bank and to store carbohydrates to utilize during periods of seasonal adversity (cold or dry) and disturbance (grazing or fire). These organs enable movement through the community, clonal multiplication, and fast regeneration (Klimešová et al. 2018). In the open pyrogenic ecosystems of peninsular Florida, frequent fires, a seasonal climate, hyperseasonal hydrology, and historical mammalian herbivory probably shaped and selected for belowground diversification (Figs. 8 & 9) and clonality. Belowground bud banks provide for post-fire regeneration for many grassland species and disproportionately account for belowground biomass allocation, an adaptive strategy in grasslands with recurrent fire (Fidelis et al. 2014; Buisson et al. 2019; Kang et al. 2013).

Fig. 8
figure 8

Woody plant belowground organs: a –Temperate plant family clonal geoxyle of Ilex glabra (Aquifoliaceae) with woody hypogeogenous rhizome, pole marked at 20cm increments, photo by James Cheak, b –Tropical plant family clonal geoxyle of Geobalanus oblongifolius (Chrysobalanaceae), a monotypic genus endemic to the southeastern US with woody hypogeogenous rhizome, photo by Brett Budach

Fig. 9
figure 9

Herbaceous plant belowground organs: aAristida beyrichiana (Poaceae) a caespitose or bunch C4 grass with its short rhizomes, epigeogenous rhizome (EB), b - Hypoxis juncea (Hypoxidaceae) main root (EB), c - Litrisa carnosa (Asteraceae) epigeogenous rhizome, cushion or rosette with centrally located apical meristemic growth protected from scorching in a recent fire (EB), d - Pityopsis tracyi (Asteraceae) hypogeogenous rhizome (EB), e - Rhexia mariana (Melastomataceae) root sprouter (EB), fCalopogon pallidus (Orchidaceae) geophyte with a stem tuber (SO), gLilium catesbaei (Liliaceae) geophyte with bulb (BB), hAsclepias connivens (Apocynaceae) main root, xylopodium (BB). Photos credits: EB (Edwin Bridges), BB (Brett Budach), SO (Steve Orzell)

Rhizome type correlated with soil moisture in previous studies from central Europe. While belowground formed rhizomes (hypogegoenous) prefer wet soil with fine structure, the rhizomes formed at the soil surface (epigeogenous) can also withstand dry and stony soil (Klimešová et al. 2012; Sosnová et al. 2010). In the present study hydrology was also found to be significantly associated with species traits, but only with rhizome texture partitioning of herbaceous versus woody rhizomatous species, not with rhizome length or rhizome type. Woody rhizomes may be prone to rot in wet conditions while herbaceous rhizomes may be short-lived and escape rotting, moreover, herbaceous rhizomes may be suitable for water storage, which allows herbs to persist in drying soils and to resprout during dry periods (Silva and Rossatto 2019).

Both belowground organs and growth form were associated with phylogeny. In general, two trait syndromes related to frequent fire and phylogeny can be seen in our dataset. The five most species-rich families were Poaceae (41), Asteraceae (27), Cyperaceae (26), Ericaceae (8), and Xyridaceae (8). In terms of belowground organs, a high prevalence of epigeogenous rhizomes in Poaceae, a family usually with highly flammable species (Cui et al. 2020). In contrast, Asteraceae were mostly rhizomatous herbs and species from this family tend to be less flammable (Cui et al. 2020). Both Poaceae and Asteraceae are two of the most species-rich families of angiosperms in Florida (Allen et al. 2019; Weakley et al. 2023), suggesting that both low and high shoot flammability are successful strategies in fire-prone ecosystems. Within Asteraceae, the subtribe Liatrinae is diverse in the southeastern US, with 19 species known from Florida’s pyrogenic ecosystems, of which at least seven are known from FSGs. Zanzarini et al. (2022) attribute maintenance of the openness in savanna grasslands to highly flammable grasses. Rapid resprouting from belowground organs post-fire enables forbs and shrubs to exploit recently burned habitat created by flammable grasses, and there is strong evidence of selection for C4 grasses driven by fire regime (Scheiter et al. 2012). In the tribe Andropogonae flammability traits are associated with high fire frequency (Visser et al. 2012), and are phylogenetically clustered, so diversification of Andropogon and Schizachyrium in peninsular Florida, where 20 taxa are tentatively recognized, is not surprising. With the phylogenetic signal in these traits, fire regime could be a crucial driver of the spatial patterns of community phylogenetic diversity observed in Florida (Allen et al. 2019).

Post fire Regeneration Strategies

When comparing post-fire recovery strategies and growth forms of subtropical savanna grasslands in Florida, resprouting is the main but not exclusive post-fire regeneration strategy. It is accompanied by re-seeding, resisting fire, or a combination of recovery modes. Resprouting is most the prevalent strategy among dominant species in Florida’s pyrogenic (scrub and savanna grasslands) ecosystems (Maliakal et al. 2000; Hierro and Menges 2002; Abrahamson 1984a). Extensive belowground biomass with carbohydrate and water reserves in frequently burned Florida ecosystems promotes resprouting and clonal growth (Abrahamson et al. 2021, Olano et al. 2006; Diaz-Toribio and Putz 2021; Zhou 2023). The post fire regeneration study which is most comparable to FSGs is from dry wiregrass pineland on the Lake Wales Ridge (Maliakal et al. 2000), sharing 34 species with our study. Maliakal et al. (2000) found 89% resprouters and 11% obligate seeders, close to our FSGs values of 82% and 18% respectively. Neotropical grassy ecosystems (savannas & grasslands) maintained by frequent fire, such as the Brazilian Cerrado and subtropical south Brazilian grasslands, share many growth forms, post-fire strategies, and belowground organ types with FSGs. As in FSGs the forbs, cespitose grasses, shrubs, and rhizomatous geophytes are mainly resprouters (ca 90%, Overbeck and Pfadenhauer 2007; Fidelis et al. 2021; Pilon et al. 2021, Zupo et al. 2021) while only about 10% of species were reseeders (Overbeck and Pfadenhauer 2007; Pilon et al. 2021). Reseeders are typically ephemeral herbs that germinate during the wet season in fire-created gaps, however some may germinate from the existing seed bank with or without fire.


The urgency to recognize, protect, and manage the remaining worldwide pyrogenic old-growth grasslands has taken center stage in recent years. They are strongholds for herbaceous plant species richness, where belowground biomass promotes rapid post-fire resprouting and fine fuel replacement, and long-term community persistence. While FSGs grasslands share many plant traits with tropical grasslands, FSGs have a significantly lower diversity of tropical and subtropical woody specialists that are prevalent in other Neotropical grasslands in comparison to FSGs. We hope that our application of terminology on plant growth form, belowground organs, and post-fire recovery strategies will foster continued research in subtropical grasslands of the southeastern US coastal plain. Our sub-sample of these grassy ecosystems should be viewed as a foundation to build knowledge about fire-adapted traits of plants from elsewhere in Florida and the southeastern US, where there are grasslands with unique species assemblages from those in south-central Florida.