The reproductive response of an endemic bunchgrass indicates historical timing of a keystone process

Restoration of the Pinus palustris (longleaf pine)-wiregrass ecosystem of the southeastern United States requires information on reference conditions such as the historical fire regime. Aristida beyrichiana (wiregrass), a keystone perennial bunchgrass, was historically widespread throughout the southeast, but its dependence upon growing season fires for sexual reproduction hastened its decline in the face of decades of human fire suppression. The reproductive response of wiregrass is described by patterns of meristem allocation between competing life history strategies (i.e., vegetative growth vs. sexual reproduction). The temporal link between fire and flowering indicates this allocation was optimized to the historical fire regime through selection. In this study, we used the observed allocation of wiregrass reproductive effort to sexual reproduction as the response variable to examine reproductive response to fire season, using plant size as a covariate. Sexual reproduction was positively associated with plant size. Plants burned during early summer (May–June) produced a greater proportion of inflorescences than did those burned in early spring (March–April) or in late summer (August). Using state records of natural (lightningignited) and anthropogenic (human-ignited) fires from historical (1933–1946) and contemporary (1998– 2010) periods we found that the distribution of maximum wiregrass reproductive output most closely reflected the distribution of historical fires with natural ignition sources. Moreover, while the monthly distributions of historical and contemporary fires were different for anthropogenic ignitions, they did not differ for fires with natural ignitions. Our predictions of peak allocation based upon the biology of wiregrass provide strong support for the use of wiregrass as an indicator of the historical fire season (early summer). Efforts to restore the longleaf pine ecosystem should therefore consider the biological response of wiregrass in planning prescribed fires.


INTRODUCTION
A central goal of ecological restoration is to assist in the recovery of a natural system that has been damaged by human activities.Within this context, 'recovery' is relative to a pre-degraded reference condition, and restoration is based on re-establishing historical trajectories defined by structural and functional interactions that en-compass native species and ecological processes (Society for Ecological Restoration 2004).Therefore, successful restoration requires accurate identification of both structural (e.g., species composition) and functional reference conditions (e.g., keystone processes).
The Pinus palustris (longleaf pine) ecosystem (LLPE) epitomizes many of the difficulties confronting restoration ecology.Human activities have greatly reduced the extent of the LLPE, forcing ecologists to infer structural and functional reference conditions from disjunct landscape fragments, historical documents, and natural archives (Noel et al. 1998, Cissel et al. 1999, Swetnam et al. 1999).High-frequency fire regimes ( 3 years; Huffman 2006, Stambaugh et al. 2011) are recognized as a keystone process in the LLPE and a crucial reference condition for restoring and conserving the LLPE (Aschenbach et al. 2010).The natural fire regime required a high degree of landscape connectivity that resulted from interactions among regional topography (van Lear et al. 2005) and firefacilitating vegetation (Beckage et al. 2005) that encouraged fires to burn for extended periods of time and, under favorable climatic conditions, to spread over vast areas (Slocum et al. 2010a).An emergent property of this landscape dynamic was a fire feedback that increased the frequency an area burned by allowing ignition sources to occur at great distances (Beckage et al. 2005, 2011, van Lear et al. 2005).
Fire season is another important component of the LLPE reference condition (Mulligan et al. 2002, van Lear et al. 2005, Aschenbach et al. 2010).Both climatic patterns of natural ignition sources (Beckage et al. 2005, Slocum et al. 2007, Slocum et al. 2010b) and congruence among historical descriptions of the LLPE and the longterm effects of high-frequency growing season fires (Waldrop et al. 1992, Streng et al. 1993) indicate that natural fire seasons varied regionally but were generally limited to the growing season.Researchers have used floristic responses to the timing of prescribed fires to infer the historical fire regime (Platt et al. 1988).This approach is based on the premise that target species are adapted to a historically predictable, high-frequency fire regime (Platt 1999).Thus, they respond positively to a fire prescription that most closely matches the natural fire regime (Platt et al. 1988, Streng et al. 1993, Platt 1999, Beckage et al. 2005).However, Beckage et al. (2005) cautioned against using an over-simplified biotic approach to infer the natural fire regime due to the four following issues: (1) the temporal response of a species to the timing of a fire may reflect different life history strategies, with some species exhibiting lag times that can obscure signals; (2) pyrogenic species may have redis-tributed in response to anthropogenic landscape change, and thus their present distribution may not reflect a location's historical fire regime; (3) natural fire regimes likely varied regionally due to the historical range of variability (HRV), characterizing large-scale ecological phenomena (see Keane et al. 2009 for review) and thus inferences are spatially limited; (4) humanmodified fire regimes may exert strong selection, rapidly alter population genetics, and obscure or erase any signal of the natural fire regime.
In this study, we addressed these four issues by using the reproductive response of Aristida beyrichiana (wiregrass, Trinius and Ruprecht) to the timing of prescribed fires, life history theory, and historical and contemporary observations of the source and timing of wildfire ignitions to infer the natural fire season for a portion of the South Carolina Coastal Plain.The ecological role, life history traits, life form, and distribution of wiregrass are strongly tied to the natural fire regime, allowing wiregrass to serve as an indicator of this regime (Noss 1989, van Eerden 1997).By limiting our study to a single species, we avoided problems associated with interspecific variation in life history strategies and response lag-times (see issue 1 above).Further, life history traits of wiregrass (i.e., limited dispersal, sensitivity to soil disturbance, and low seed viability; Clewell 1989, Noss 1989) reduce the probability that the species has undergone range expansion in response to anthropogenic landscape change (Donohue et al. 2000; see issue 2 in Introduction, third paragraph).
Wiregrass is an ecosystem engineer in the LLPE (Clewell 1989, Hardin and White 1989, Noss 1989, Jones et al. 1994) that facilitates the spread of fire at a landscape scale (Noss 1989).Historically, wiregrass was a dominant feature of many southeastern Coastal Plain landscapes.Vegetation/fire feedbacks tightly link longleaf pine, wiregrass, and fire, and appear to be an emergent property of the system.When coupled with its congener, Aristida stricta (wiregrass), these two species cover approximately half of the historical range of the LLPE (Platt 1999) and provide a standardized signal to capture the HRV of natural fire regimes at a regional scale (see issue 3 in Introduction, third paragraph).
Inflorescence development and sexual repro-duction in wiregrass are temporally linked to fires; more inflorescences are produced when burning occurs outside the dormant season and inflorescences are rarely produced greater than one year after a plant has been burned (Mulligan et al. 2002).As fire exclusion came to dominate landscape dynamics during the 20th century, the strong link between fire and inflorescence development nearly eliminated sexual reproduction in wiregrass over the extent of its range.Sexual reproduction became so suppressed that many researchers considered wiregrass a sterile species by the mid 20th century (Clewell 1989).Because traits must be expressed for selection to act on them, the absence of sexual reproduction excluded phenotypic expressions of sexual reproductive strategies, functionally removing these traits from direct selection to anthropogenic fire regimes of the 20th century (Lahti et al. 2009).
In essence, the reproductive disjunction that resulted from fire exclusion froze selection on these traits and preserved signals of the historical fire regime (see issue 4 in Introduction, third paragraph).
In the absence of fire, growth and maintenance of a phalanx growth form enable wiregrass to locally persist for long durations without sexual reproduction (Clewell 1989).During this vegetative growth, the apical meristem remains close to ground level where it is protected from grazing and fire, and developing tillers persist by producing vegetative structures (i.e., leaves).However, at the onset of flowering, the apical meristem is elevated above ground level and is terminated through the normal development and senescence of the inflorescence (flowering culm).The finite number of meristems within a wiregrass clump (Doust 1989) constrains wiregrass reproduction, creating a tradeoff between the allocation of meristems to either vegetative growth or sexual reproduction (Bonser and Aarssen 1996).Within the context of life history theory, the reproductive response of wiregrass is described by patterns of meristem allocation between competing life history strategies (i.e., vegetative growth vs. sexual reproduction), and the cost of sexual reproduction must be offset by a fitness advantage (Huber and During 2001).The temporal link between fire and flowering indicates that this allocation was optimized to the historical fire regime through selection (Ackerly et al. 2000).In this study, we used the proportion of meristems allotted to flowering per wiregrass clump as a measure of competing life history tradeoffs.
The objective of this study was to infer the historical fire season using two sources of data.First, we applied life history theory to the current reproductive response of wiregrass to variation in fire season.We examined wiregrass reproductive response to fire timing by sampling plants that had been burned during spring (March-April), early summer (May-June), and late summer (August).We hypothesized, based on life history theory, that wiregrass would allocate a greater relative proportion of meristems to reproduction when fire prescriptions were most congruent with the historical fire season.Secondly, we used the historical and contemporary timing of lightning fire ignitions in South Carolina to determine the historical fire season.This approach allowed us to characterize the historical fire season in terms of a natural fire season driven by lightning ignition fires, and to examine which of the three burn treatments (spring, early summer, and late summer) most closely coincided with the historical fire season.We assume that the near absence of sexual reproduction in wiregrass over much of the 20th century largely removed these reproductive traits from direct selection and thus preserved a temporal signal of the historical fire regime.

Study area
We sampled wiregrass at the James W. Webb Wildlife Center in Hampton County, South Carolina (32.58998N, 81.310668 W; approximately 70 km from the coast).The 2374-ha property is located in the Coastal Plain (11.9 m elevation) and has been managed by the South Carolina Department of Natural Resources as a wildlife management area since 1941.The area has an average temperature of 9.88C in winter and 26.48C in summer.Total annual precipitation is 122 cm, about 60% of which falls in April through September.Primary habitats include longleaf pine savannas (flatwoods; Fig. 1), Pinus taeda (loblolly pine) forests, oak-hickory mixedpine hardwoods, hardwood bottoms, and cypress-tupelo swamp forests associated with the Savannah River floodplain.Fire has been used as a primary management tool for over a century, maintaining high vascular plant diversity, with more than 600 species occurring on site (Porcher and Rayner 2001).Further, the site harbors an abundance of longleaf-pine endemic wildlife, and supports colonies of the federally endangered Picoides borealis (Red-cockaded Woodpecker).Game management, which includes maintenance of agricultural food plots in upland habitats, focuses on Colinus virginianus (Northern Bobwhite) and Odocoileus virginianus (whitetailed deer).

Wiregrass data collection
We sampled wiregrass reproductive response to timing of prescribed fires at eight sites.All sites shared a common land use history that included an extensive timber harvest in the early 20th century and the subsequent application of high frequency prescribed fires (2-5 years) during winter and early spring months.Conse- v www.esajournals.orgquently, there was little fire suppression at the site.Stands had similar structure (21.00 6 0.99% canopy closure) dominated by longleaf pine and abundant wiregrass in the ground layer (0.76 6 0.06 m distance between wiregrass clumps).
We stratified wiregrass sampling by fire season, i.e., we sampled pine flatwood sites that had been burned during one of the following seasons (1) spring (March-April), (2) early summer (May-June), and (3) late summer (August).Sampling was conducted on 2-3 November 2010.Within each site, we followed a pointquarter sampling scheme.Beginning at our estimate of the center of the stand, we located fifteen points sequentially by choosing a random azimuth (0-340, at 20 degree intervals) and distance (50 m maximum) from one point to the next.At each point, we measured the distance from the center point to the nearest wiregrass clump in each of four quadrants.We then randomly selected one of the four bunchgrasses at each point and cut across its base using a knife and stored it in a plastic bag.Prior to processing, we stored wiregrass samples in either a refrigerator or a À708C freezer.For each sample (i.e., wiregrass clump), we separated the inflorescences (reproductive meristems) from the non-reproductive tillers (vegetative meristems), counting as reproductive only those meristems that had inflorescences (i.e., not meristems that simply appeared to be elongating).
From the non-reproductive tillers of each sample, we selected twenty tillers with the stipulation that tillers had a thickened base and at least two leaves.This method eliminated the chance that (1) we would count leaves (absence of a thickened base) as tillers, and overestimate the number of non-reproductive tillers in our subsequent proportion calculations and (2) that by including tillers that had potentially lost their leaves, we would underestimate the number of non-reproductive tillers in our calculations.We cut the inflorescences, the twenty tillers, and the rest of the non-reproductive biomass of each sample into smaller pieces and wrapped each group separately in foil.All samples were placed in a drying oven at 608C for 24-48 hrs, and afterwards weighed to the nearest 0.01 g.
We calculated the total number of meristems per clump by first dividing the total nonreproductive biomass by the mass of the sub-sample of tillers, and multiplying this number by the number of tillers weighed in the subsample.This estimated the total number of tillers assumed to be non-reproductive in the sample.This number represented the number of vegetative meristems, which was added to the number of inflorescences to yield the total number of meristems for each sample.We calculated size by averaging the longest and shortest diameters of the plant, and using the resulting radius in the equation for the area of a circle.

Documentary fire data collection
We obtained contemporary data on wildfire ignitions occurring in South Carolina from 1998 to 2010 from the South Carolina State Forestry Commission.These records included the cause and month of the fire, as well as the county in which the fire occurred.Additionally, we used historical information on wildfire occurrence collected from fire district and county reports between 1933 and 1946 at the South Carolina Department of Archives and History, which also documented fire occurrence by cause and month in each county.

Statistical analyses
All statistical analyses were performed using SAS software, version 9.2.1.We used a generalv www.esajournals.orgized mixed negative binomial model in the GLIMMIX procedure (SAS Institute 2008) to examine the effects of fire season on meristem allocation to reproduction, which was expressed as a percentage (i.e., proportion of reproductive meristems per sample, multiplied by 100).The model included fire season (spring, early summer, late summer) as a fixed effect and plant size as a covariate.We specified site as a random effect to account for lack of independence among observations from the same site.We assessed model fit by examining residuals and model deviance.We used a Tukey-Kramer test for multiple comparisons of reproductive response among seasons.
We used contingency tables to compare historical and contemporary temporal fire frequencies (number of fires per month by ignition source, summed across counties and years for each period).For anthropogenic fires, we used a Pearson chi-square test to compare monthly frequencies of contemporary and historical fires.We analyzed natural fires similarly, but used a Fisher's exact test for comparisons due to small sample size.To assess whether our observations of wiregrass meristem allocation to reproduction reflected contemporary or historical fire regimes, we superimposed meristem allocation estimates of least squares means derived from the negative binomial model over monthly fire distributions.Because least squares means estimate the marginal means over a balanced population (SAS Institute 2008), we visually assessed mean estimates relative to fire frequency distributions to make inferences about the link between historical and contemporary fire regimes, their ignition sources, and wiregrass reproductive phenology.
Monthly distributions of historical and contemporary fires were different for anthropogenic ignitions (v 2 ¼ 750.36, df ¼ 11, p , 0.0001), but we failed to detect differences for fires with natural ignitions ( p .0.05).The temporal distribution of natural fires appeared as the inverse of the anthropogenic fires for both historical and contemporary fire regimes (Fig. 3).Similarly, contemporary and historical distributions of anthropogenic fires, which were more common during the dormant season, were dissimilar to wiregrass reproductive response (Fig. 3).Rather, the distribution of wiregrass reproductive output was similar to that of the timing of natural fires, and most closely reflected the distribution of historical fires with natural ignition sources (Fig. 3).

DISCUSSION
Our application of life history theory to wiregrass reproductive allocation patterns demonstrated that the fitness advantage of sexual reproduction was greatest after growing season (late May and early June) fires.Although the frequency of historical and contemporary lightning-ignited fires was highest between May and August, our observed peak in wiregrass reproductive effort in May-June was dramatic in contrast to the effort observed in August-burned plants.Thus, despite the paucity of lightning-fire records predating the period of human fire exclusion, our results support the recognition of summer fires as the historical peak fire season in the longleaf pine-wiregrass ecosystem in South Carolina.
Our results provide further support for the consistent pattern of increased reproductive response to fires occurring in the season that matches the historical, natural fire regime.Platt et al. (1988) concluded that growing season fires (April-August) increased flowering within and among longleaf pine savanna species relative to fires between growing seasons.In their foundational review and experimental study, Streng et al. (1993) demonstrated a significant effect of fire season on the fruiting and flowering of almost all dominant grasses, including wiregrass; those burned during the growing season (April-August) exhibited significantly greater reproduction (Streng et al. 1993).Reproductive success in Pityopsis graminifolia (Narrowleaf silkgrass) was greatest after May fires (Brewer and Platt 1994), and many other fall-flowering species such as those in the genera Stylosanthes and Liatris respond positively to growing season fires (Platt et al. 1991).Moreover, seed production of the dominant grasses, including wiregrass, is greatest after fires occurring May-July (Rodriguez andBohn 2011, Shepherd et al. 2011).Not only plants, but also many endemic animals have adapted to fires that primarily occurred during the summer.Ground nesting birds, such as the  v www.esajournals.orgendemic Peucaea aestivalis (Bachman's Sparrow; Tucker et al. 2004), exhibit nesting and migration strategies that are strongly tied to conditions created by growing season fires (Cox and Widener 2008).Moreover, amphibian reproductive phenology and migration patterns (e.g., Lithobates capito (gopher frog); Roznik and Johnson 2009) could reflect the response of vegetation to fire season (Means et al. 2004).
Our results support the greater influence of growing-season lightning fires relative to humanignited fires in shaping evolutionary responses in this ecosystem (Bond andKeeley 2005, Pausas andKeeley 2009).This perspective challenges the current paradigm, in which lightning and anthropogenic fires are of equal importance in shaping the life histories of pine savanna species.Southeastern Native Americans tended to burn during the autumn, late winter, or early spring primarily for subsistence purposes: to open the woods for hunting, prepare land for crops, and remove litter for wild food collection (Fowler and Konopik 2007).After the arrival of Europeans, post-contact fire regimes mimicked these practices (Fowler and Konopik 2007;Fig. 3).Using dendrochronology, Huffman (2006) found that in coastal mainland pine savanna between 1592 and 1883, only three fires occurred in the dormant season in the 1800s, after European settlement of the local area, and that all pre-settlement fires occurred during the growing season (Huffman 2006).Within the twentieth century, shifts in federal policy from fire suppression (late 1800s-1940s) to prescribed fire management (Stephens andRuth 2005, Fowler andKonopik 2007) slightly altered, but did not change, the predominance of dormant season fires.Taken in combination, the life history response of wiregrass to lightning ignitions and its role as a keystone species encourages a rethinking of how issues related to the timing of fires are approached in this ecosystem.
Wiregrass historically occurred from southeastern North Carolina to southern Florida, west to Georgia, Alabama, and coastal Mississippi (Wells and Shunk 1931, Clewell 1989, Peet 1993), coincident with approximately half of the historical range of the longleaf pine ecosystem.However, this extent was greatly reduced by logging and other forestry practices that involved severe soil disturbance, and once eradicated, wiregrass is not known to recolonize (Clewell 1989).Although wiregrass does successfully recruit through seed (Mulligan et al. 2002), its requirement of growing season burns for sexual reproduction was unmet through decades of fire suppression, further contributing to its decline.The importance of the fire regime to the life history of wiregrass underscores the role of fire as a critical ecosystem process, rather than as a disturbance (Pausas and Keeley 2009).Indeed, fire exclusion appears to function as a disturbance in the longleaf pine ecosystem (Landers et al. 1989).
The extent to which wiregrass reproduction appears to be tied to the historical fire regime could reflect several potential benefits of this reproductive strategy.Potential benefits of tying reproduction to fire include increased light availability (van Eerden 1997), litter removal and microsite availability (van Eerden 1997), reduced competition (Outcalt et al. 1999, Mulligan andKirkman 2002), and increased nutrient availability (Bond and van Wilgen 1996).Moreover, we propose that wiregrass reproduction is closely tied to the phenology of the natural fire regime because the occurrence of lightningignited fires was more predictable than that of anthropogenic (i.e., dormant-season) fires.If fire were a stochastic process in this ecosystem, it is unlikely that such strong selection would not be so readily apparent or persistent in wiregrass over centuries.We also propose that synchronizing reproduction after predictable fires would increase survival and continued dominance.Wiregrass requires at least two years for successful seedling establishment (Mulligan and Kirkman 2002).Under a frequent fire regime (Huffman 2006) with a two to five year return interval (Stambaugh et al. 2011), wiregrass that reproduces after a growing season fire could depend on sufficient time for seedling regeneration.
The persistence of a reproductive signal over centuries has also been observed in Pinus banksiana (Jack pine), whose geographical variation in cone serotiny corresponds to geographical variation in the fire regimes that occurred well before European alteration of the fire regime in the 1860s (Radeloff et al. 2004).In the LLPE, the fire phenology is a key influence on the timing of flowering in wiregrass and other pine savanna v www.esajournals.orgdominants (Platt et al. 1991, Shepherd et al. 2011).Therefore it is possible that these species have evolved mechanisms by which to reinforce the continued predictability of fire (Gagnon et al. 2010, Keeley et al. 2011).For instance, wiregrass experiences year-round root production, with much greater root growth occurring during the winter relative to Schizachyrium scoparium (little bluestem), which experiences root growth primarily during the growing season (West 2002).Perhaps this root proliferation in wiregrass represents not only nutrient uptake as suggested by West (2002), but also a ''shunting'' of nutrients from the leaves to belowground structures prior to the times that summer fires were most likely to occur.This would not only contribute to the survival and growth of wiregrass post-fire, but it would also make the leaves more flammable during the fire season (Gagnon et al. 2010).
Our study demonstrated the dominant role of lightning ignitions in the phenology of pine savanna fires in the South Carolina Coastal Plain.Moreover, the reproductive response of wiregrass, a dominant keystone bunchgrass, underscores the importance of fire phenology to ecosystem functioning.Successful restoration of longleaf pine savannas and woodlands depends on both the reinstatement of its drivers (fire), and the persistence of ecosystem engineers and keystone species (Landers et al. 1989).Therefore, it is important for those involved in the restoration of this ecosystem to incorporate this information into their activities.

Fig. 3 .
Fig. 3.The distributions of historical (1933-1946) and contemporary (1998-2011) fires ignited by (A) anthropogenic sources and (B) natural sources (i.e., lightning).The secondary y-axis corresponds to Aristida beyrichiana (wiregrass) reproductive response to fire season, measured using a negative binomial model of the effects of fire season and plant size on the proportion of meristems allocated toward sexual reproduction.Means (6 standard error) are presented for three fire seasons, spring (March-April), early summer (May-June), and late summer (August).

Table 1 .
Parameter estimates (b) for fixed effects from generalized mixed negative binomial model examining the effects of fire season, i.e., early spring (March-April fires), early summer (May-June fires), and late summer (August fires), and plant size (cm 2 ) on Aristida beyrichiana (wiregrass) reproductive response (proportion of meristems allocated to sexual reproduction for each plant).Estimates for late summer are not presented due to linear dependence with other parameters.SE ¼ standard error; LCL ¼ lower 95% confidence limit; UCL ¼ upper 95% confidence limit.