Southeastern Naturalist 549 E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 22001166 SOUTHEASTERN NATURALIST 1V5o(3l.) :1554,9 N–5o7. 43 Ecology and Conservation of the Endangered Legume Crotalaria avonensis in Florida Scrub Eric S. Menges1,*, Beatriz Pace-Aldana2, Sarah J. Haller1, and Stacy A. Smith1 Abstract - We collected data from 1998 to 2014 to describe the ecology of the highly endangered Florida scrub plant Crotalaria avonensis (Avon Park Harebells), and herein address several hypotheses based on what was known of its biology and the biology of co-occurring species. This perennial herbaceous legume occurs at 3 sites and prefers microsites with more cover by bare sand than vegetation. The population at an unprotected site has declined in size, but dynamics have been more stable at the 2 protected sites. Marked plants have shown high survival, slow and inconsistent growth, and occasional plant dormancy (usually 1–2 years). Avon Park Harebells is reproductively challenged, with very low rates of fruit set and infrequent visitation by required pollinators. The hardseeded fruits germinated at a rate of 13–56%; the germination speed seemed to increase after scarification, though the overall rate was less than for unscarified seeds. Unscarified seeds remained viable in the seed bank for at least 3 years. Seedlings recruited rarely, had moderate survival, began flowering at 4 years of age or later, and reached the size of median adult plants in 6–8 years. Herbivores affected 7–53% of plants in a given year, but plants showed rapid compensatory resprouting. Caging plants reduced herbivory and increased survival, growth, and flowering. Plants resprouted after fire and mechanical disturbance and exhibited high survival and growth, but repeated disturbances by vehicles caused increased mortality. Avon Park Harebells remains extremely endangered due to its limited range, small population sizes, and poor seedling recruitment. To help this species recover, we recommend fire management, protection from herbivory, introductions and augmentations, and further study of its pollination biology. Introduction Narrowly endemic species present unique challenges to conservation and management. Many of these species have specific habitat or disturbance requirements and occur at only a few managed sites. Management activities at these sites tend to focus on ecosystem structural or functional properties or perhaps are aimed at restoring populations of economic, keystone, or foundation species. Few studies have evaluated whether broad-scale management regimes also work to benefit individual rare species (Menges and Weekley 2013). The Lake Wales Ridge (Weekley et al. 2008) in south-central Florida is a hotspot for endemism. It supports one of the highest concentrations of endemic plants in the US (Christman and Judd 1990, Dobson et al. 1997, Estill and Cruzan 2001). Most of these species occur in xeric upland habitats, including Florida scrub and sandhills. Over 85% of the natural vegetation on the Lake Wales Ridge has been lost 1Plant Ecology Program, Archbold Biological Station, 123 Main Drive, Venus, FL 33960. 2The Nature Conservancy, Disney Wilderness Preserve, 2700 Scrub Jay Trail, Kissimmee, FL 34759. *Corresponding author - emenges@archbold-station.org. Manuscript Editor: Julia Cherry Southeastern Naturalist E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 2016 Vol. 15, No. 3 550 to development (Weekley et al. 2008). Like most Florida vegetation (Slapcinsky et al. 2010), Florida scrub and sandhill biodiversity depends on periodic fire (Menges 2007, Menges and Gordon 2010). Among the many Lake Wales Ridge endemic species that are endangered or threatened, Crotalaria avonensis DeLaney & Wunderlin (Avon Park Harebells) stands out as one of the most imperiled (Turner et al. 2006). It is one of the most narrowly distributed of all Lake Wales Ridge endemics and 1 of only 2 such species—the other being Dicerandra christmanii Huck and Judd (Garrett’s Mint)—found at fewer than 5 sites (Turner et al. 2006). Avon Park Harebells was listed as one of the 8 Lake Wales Ridge species for which translocation and captive propagation may be necessary for species survival (Turner et al. 2006). Wild populations of this federally endangered and recently described plant are known from only 3 sites (DeLaney and Wunderlin 1989) covering about 518 ha (USFWS 2006). Two of the sites are protected and 1 is not. Within these sites, the distribution of Avon Park Harebells is patchy, even within appropriate habitat (USFWS 2006). A recent 5-y review of this species stated that it is “at grave danger of extinction” and that “without active and concerted conservation efforts, this species may be lost” (USFWS 2006). Recovering any species requires information on its habitat requirements, ecology, interactions with other species, and vulnerabilities to human activities. This information is lacking for Avon Park Harebells. Prior to our work, the only publication regarding this plant was the species description (DeLaney and Wunderlin 1989). Here, we draw on demographic research from 1998–2014 to paint an ecological portrait of Avon Park Harebells and suggest key issues for its conservation. At the start of this study, little was known of the ecology of Avon Park Harebells. Based on the biology of species with a similar life history and that occur in Florida scrub, we suggest the following hypotheses: • Like many diminutive Florida scrub herbaceous plants (Menges et al. 2008), Avon Park Harebells should be more likely found in open microsites. • Unprotected Florida scrub sites are not managed using fire (Menges 1999) and have other stressors; thus, we predict that population trends will be more positive in protected sites. • This species has a deep taproot (DeLaney and Wunderlin 1989), so we expect Avon Park Harebells to be long-lived, with high survival and moderate growth. • As a hard-seeded legume (DeLaney and Wunderlin 1989), we hypothesize that this species has a persistent seed bank and that scarification will increase germination speed. • As has been reported for other large-seeded plants, we expect strong seedling recruitment in Avon Park Harebells (Moles and Westoby 2004). • We hypothesize that herbivory will reduce vital rates, but that recovery from the strong taproot will allow some compensatory responses to herbivory (paralleling responses in other Florida scrub species; Kettenring et al. 2009, Tye et al. 2016). Southeastern Naturalist 551 E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 2016 Vol. 15, No. 3 • We hypothesize that the long taproot will allow Avon Park Harebells to readily resprout post-fire (Menges and Kohfeldt 1995). Field-site Description All 3 known wild populations of Avon Park Harebells occur within a small area of northern Highlands and southern Polk counties in south-central Florida (Fig. 1). Carter Creek, a tract of the Lake Wales Ridge Wildlife and Environmental Area (LWRWEA) managed by Florida Fish and Wildlife Conservation Commission (FWC), and Saddle Blanket Scrub, owned by The Nature Conservancy (TNC), are protected. Avon Park Lakes, a sprawling subdivision, supports an unprotected population just south of Saddle Blanket Scrub (Fig. 1). The Avon Park Lakes population is currently at risk due to likely further residential development. We have been studying the ecology of Avon Park Harebells since 1998, primarily at Carter Creek. We have also periodically sampled population dynamics at the 2 other sites. All 3 sites contain a variety of plant communities, including 2 types of Florida scrub (rosemary scrub and scrubby flatwoods) that support Avon Park Harebells. These types of Florida scrub are well-drained shrublands dominated by Quercus spp. (oaks), Sabal spp. (palmettos), ericads, and Ceratiola ericiodes Michx. (Florida Rosemary). These communities are affected by periodic fire at intervals generally ranging from 5 to 60 years (Menges 2007). We also noted plants in a third vegetation type: sandy roadsides. We compared much of our observational and experimental data among rosemary scrub, scrubby flatwoods, a nd roadsides. Methods Study species Avon Park Harebells is a low-growing legume found in Florida scrub (rosemary scrub, scrubby flatwoods, and roadsides) on xeric white sands (DeLaney and Wunderlin 1989). It can be clearly distinguished from other Crotalaria species (rattleboxes) by several flower characters (DeLaney and Wunderlin 1989). Avon Park Harebells differs from its relative C. rotundifolia (J.F. Gmelin) (Rabbitbells) by its curved (vs. geniculate) style, compact growth form, and pubescence. Avon Park Harebells is deeply rooted, single to multi-stemmed, and non-clonal. It generally flowers in the spring, is vegetative after June, and dies back over winter (DeLaney and Wunderlin 1989). Flowers are perfect and include anthers that release pollen at different times. Fruits are legumes containing multiple hard seeds that are released when fruits dehisce. Microhabitat preferences In 2002, we collected data on microhabitat preferences by habitat for Avon Park Harebells at both Carter Creek and Avon Park Lakes within 25-cm-radius circular quadrats. At Carter Creek, we selected all demography quadrats (see section on population dynamics and vital rates), most of which were occupied by Avon Park Harebells. We also selected unoccupied quadrats at random directions and distances (within 5 m of occupied quadrats). At Avon Park Lakes, we searched Southeastern Naturalist E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 2016 Vol. 15, No. 3 552 for occurrences of Avon Park Harebells along roadside edges and adjacent scrub. Again, we selected unoccupied quadrats at random directions and distances from occupied plots, but stayed within generally similar roadside habitat. In these small circular quadrats, we recorded ocular cover-estimates (nearest 10%, plus trace coded as 1%) on a series of microhabitat variables (percent bare sand, percent Figure 1. Map showing locations of 3 wild Crotalaria avonensis (Avon Park Harebells) populations on Florida’s Lake Wales Ridge (darker shading) in Polk and Highlands counties (lighter shading). Southeastern Naturalist 553 E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 2016 Vol. 15, No. 3 vegetation cover, percent litter, and percent ground lichens) and the presence of all vascular plants and ground lichens. We used separate t-tests for roadsides and scrub sites to compare microhabitats of occupied and unoccupied plots. We used chi-square tests to evaluate the association of the more common species with presence of Avon Park Harebells, separately for scrub and roadside sites. Occupancy and densities across landscapes We tracked Avon Park Harebells distribution and abundance across large spatial scales at its 3 known sites. During extensive searches in the spring of 2005 (Avon Park Lakes and Carter Creek), and 2005–2006 (Saddle Blanket Scrub), we recorded all known spatial locations with a Trimble® handheld GPS and marked each with a stake flag. Every 3 years thereafter (in April and May), we returned to each point at Carter Creek and Avon Park Lakes and counted the number of Avon Park Harebells plants within a 5-m (Avon Park Lakes) or 2-m (Carter Creek) radius of the GPS point. At Saddle Blanket Scrub, we made counts within randomly selected 10 m x 10 m grids cells in 2009, 2010, and 2013. For all populations, we re-counted plants in non-overlapping plots and defined a plant as all stems within 2 cm of each other. We used t-tests to compare densities among populations. Population dynamics and vital rates At Carter Creek, our demography study plots were distributed throughout the area known to support Avon Park Harebells. We set up belt transects that traversed patches of plants in 1998, 2001, 2002, 2003, and 2005. Within each belt transect, we chose five 25-cm circular quadrats with stratified random positions and permanently marked them with flags and tags. Not all established quadrats supported Avon Park Harebells. We set up a total of 105 quadrats. Within each quadrat, we initially mapped, and later marked with plastic toothpicks (of various colors and playing card suits) each Avon Park Harebells individual. This species is characterized by frequent appearances and disappearances of aboveground stems accompanied by small shifts in positions, likely reflecting its branching root system (Delaney and Wunderlin 1989), which makes it challenging to distinguish and follow individuals. A limited excavation at the beginning of the study showed that the tan-colored and thickened roots were located at least 5 cm below the soil surface and sometimes as deep as 7 cm. Relatively slender stems radiated from root crowns through the sand at various angles. We determined that the resulting aboveground shoots could emerge at distances of 0–10 cm from the root crown; we used this distance range to distinguish between presumed individuals. We conducted most of our demographic sampling during the periods of early seasonal growth, flowering, and fruiting. We sampled quadrats monthly from February through August in 1998–1999 and from February through June from 2000 to 2014. For each plant, we recorded survival and the numbers of stems, branch tips, flowers (showing yellow corolla), developing fruits (green), and mature fruits (darker). From these monthly data, we summarized annual population size and vital rates (survival, dormancy, relative growth rates, and maximum and totals for reproductive variables). Each year, we annually calculated relative growth rate on Southeastern Naturalist E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 2016 Vol. 15, No. 3 554 the basis of size (maximum number of branch tips in a year): positive values indicated growth, zero indicated the same number of branches, and negative values indicated reductions in size. We employed chi-square tests and analysis of variance (ANOVA) to compare survival and growth, respectively, among vegetation types. We estimated percent fruit initiation and fruit maturation by calculating the total number of flowers, initiated fruits, and mature fruits counted in each monthly census. These are not exact measures of fecundity because we may have missed some reproduction between censuses or we may have double-counted some reproduction. For mature fruits found on plants within demographic plots, we measured the capsule length and counted the seeds per capsule in the field. We then returned them to a an open location within the quadrat. For fruits outside the plots, we returned them to the lab and collected data, then used the seeds in germination experiments (see section on seed germination experiments). We characterized the distributions of seed number per capsule and used correlations to examine the relationship between seed number and capsule length. Pollination biology and breeding system In 2004, we made observations of pollinators at Carter Creek and attempted a breeding-system study. We recorded very few flower visitors; thus, we did not pursue analysis of these data. We were unable to conduct specific crosses in the field because of the small size of flower parts. However, we evaluated the breeding system of Avon Park Harebells by comparing fruit initiation from 10 plants (9 of which opened 75 flowers) from which we excluded pollinators with mesh bags and 10 control plants (8 of which opened 66 flowers). We marked individual flowers at 1–3-d intervals with colored thread and identified each by position on the plant. We followed subsequent fruit initiation several times per week. We defined initiation as occurring when we observed small pods. Flowers that did not initiate fruit typically dropped off after a few days. We analyzed the effects of the bagging treatment using the Mann-Whitney U-test. Seed-germination experiments The low fecundity of Avon Park Harebells made it impossible to conduct germination trials in most years. Here, we describe 2 experiments undertaken in 2004 and 2012. 2004 germination experiment. In spring 2004, we collected 317 seeds from Avon Park Harebells plants at Carter Creek (outside our study plots) for use in a germination experiment. The major factors in the experiment were scarification with sandpaper (n = 163 vs. no scarification, n = 158) and seed color (black: n = 194, purple: n = 65, green: n = 58). Scarification might be expected to enhance germination in a hard-seeded legume, and it significantly increased germination in other Crotalaria species (Wiggers 2011). We surmised that green seeds were immature (they were rather soft) and black seeds (the hardest) were the most mature. The experiment was initiated on 19 May 2004 within a week of seed collection. We sowed seeds in sand and placed the flats on an open but covered veranda at Archbold Biological Station where they were watered frequently. We monitored this experiment twice per week Southeastern Naturalist 555 E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 2016 Vol. 15, No. 3 for several months, and then less frequently for 2 years. We analyzed germination using binary logistic regression and time to germination using t-tests. 2012 shadehouse germination experiment. To test the effects of scarification on seed germination and to gain insight into germination requirements of unscarified Avon Park Harebells, we initiated a germination experiment in the Archbold Biological Station shadehouse. On 9 July 2012, we scarified 70 seeds and left 70 seeds unscarified as controls, using only mature (purple or black) seeds collected from plants at Avon Park Lakes. We weighed the individual seeds prior to sowing. We conducted scarification in a lab setting, using a single-blade razor to manually nick the seed coat opposite the hilum less than 1 h before sowing. In the shadehouse, we filled flats separated into individual cells with native soil and sowed a seed in each cell. We randomized sowing locations and standardized sowing depth at 1 cm. We irrigated the flats every other morning and the flats were open to rainfall. We assessed germination daily for the first 2 weeks, every other day until March 2013, and once each week through December 2013. We excluded all 21 seeds weighing less than 0.03 g from our analyses because no seeds that small germinated. We used binary logistic regression to analyze germination and a Mann-Whitney U-test to analyze time to germination. Seedling recruitment, survival, and flowering We assessed seedling recruitment as part of our monthly demographic sampling at Carter Creek. From 2004 to 2011, we sampled monthly year-round in quadrats with a history of fruit or seedling production. Following initial germination experiments in 2004, we were able to recognize seedlings in the field because they had glabrous, glandular cotyledons that possessed an evident midrib, were oval in shape, and had an opposite arrangement. Subsequent true leaves were pubescent and non-glandular. Once seedlings recruited, we followed their vital rates during regular demographic censuses. Sample sizes were too small for statistical analyses. Herbivory Avon Park Harebells is commonly eaten by animals. This herbivory often involves complete removal of stems, but sometimes just leaves and parts of stems are affected. Herbivores include mammals, ants, and the larvae of Utetheisa bella (L.) Pease (Bella Moth), an insect that sequesters toxins from Avon Park Harebells tissues and uses these chemicals to deter predators (Mattocks 1986). During each monthly demographic survey, we tracked herbivory on plants, coding it as partial herbivory (some stems eaten, some not), total herbivory (all stems clipped), or post-herbivory (plants recovering from total herbivory). We generally did not record less extreme damage to portions of stems or leaves. For data analyses, we aggregated herbivory for the entire year, with plants coded as suffering herbivory (at least once) or no herbivory. We compared herbivory among vegetation types using chi-square tests. Beginning in June 2012, we experimentally caged half (n = 23) of our study quadrats that contained living plants in 2011 or 2012. The wire cages had a mesh size of ~1 cm, were about 50 cm in diameter, 45 cm tall, and open at the top to allow free access to insects but deter small and large mammals. We hypothesized Southeastern Naturalist E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 2016 Vol. 15, No. 3 556 that these cages would reduce observed herbivory levels. We did not necessarily expect an effect of caging on vital rates because observed herbivory had no obvious effects on vital rates such as survival. We used a chi-square test to analyze effects of caging on survival and flowering, and t-tests to evaluate caging effects on relative growth rate. Effects of fire and mechanical treatments A prescribed burn took place in one of the Carter Creek study populations (the Rosemary Path population) on 19 August 2005. This burn was primarily designed as part of a larger project testing the effects of mechanical treatment and fire on rosemary–oak scrub. However, we appended a small experiment to test the effects of mechanical treatment and fire vs. fire-only on Avon Park Harebells. Before the fire, 10 quadrats (1–10) were mechanically treated using a Gyro-Trac, a large bulldozer-like machine that cut all vegetation near ground level and caused moderate soil disturbance. Five additional quadrats (11–15) were burned without prior mechanical treatment. All quadrats in both areas were thoroughly burned, although, because fire intensity varied, we noted quadrats as being lightly or intensely burned within a month of fire. We checked the response of Avon Park Harebells plants monthly, assessing survival, size, flowering, and recruitment. We analyzed treatment effects on survival with chi-square and on relative growth rate (based on number of branch tips) using one-way ANOVA. Effects of vehicles Repeated damage from vehicles at Carter Creek provided an opportunity to examine the resilience of Avon Park Harebells to different types of damage. Before 2008, the Carter Creek site was unfenced and subject to a great deal of all-terrain vehicle (ATV) traffic. Quadrats in 1 subpopulation were run over repeatedly in 2004, 2005, and 2006. Nearly all existing plants appeared to have been killed. Each year, we found the quadrats and followed the recovery of the population. During 2008, the FWC completed fencing the Carter Creek site so that damage from ATVs and truck traffic would no longer af fect the population. Other quadrats (16–25) with Avon Park Harebells plants were inadvertently damaged in 2007 and/or 2009 during pre-burn site preparation by a Gyro-Trac. The Gyro-Trac crushed plants, rather than uprooting them, as often occurs with ATV and truck traffic. The 2009 gyrotracking was followed with a prescribed burn that also included vehicle damage to Avon Park Harebells quadrats. Sample sizes affected by vehicles were too small for statistical analyses. Results Microhabitat preferences Within Florida scrub, Avon Park Harebells prefers sites with more bare sand than vegetation cover (Table 1). Similarly, in scrub locations, the species was positively associated with an open-site herb, Cnidoscolus urens var. stimulosus (Finger Rot), and negatively associated with a dominant shrub, Quercus inopina (Sandhill Oak) (Table 2). On disturbed roadsides, Avon Park Harebells occurred in microhabitats Southeastern Naturalist 557 E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 2016 Vol. 15, No. 3 that were not different than random locations, and all tested species associations along roadsides were not statistically significant (Table 2). Occupancy and densities across landscapes Occupancy percentages of Avon Park Harebells have declined in all populations over the years (Fig. 2). Initial occupancies in 2005 or 2006 were, by definition, 100% because we set up all points at Avon Park Lakes, Saddle Blanket Scrub, and Carter Creek in patches of plants. Avon Park Lakes had a strong decline in occupancy between 2005 and 2008, and a slow decline since then. The 2 managed sites, Table 1. Microhabitat differences between locations occupied by Avon Park Harebells (occ.) and random locations (not occ.) in Florida scrub and roadside habitats at 2 sites (Carter Creek and Avon Park Lakes). Percent bare sand was strongly negatively correlated with % litter; thus, we did not include the latter. Significance (Sig.) was corrected for tests of 3 variables. NS = P > 0.05. Scrub (Carter Creek only) n = 94 Roadside (2 sites) n = 92 Mean Mean P Mean Mean P Measure occ. not occ. t uncorrected Sig. occ. not occ. t uncorrected Sig. % vegetation cover 24.0 56.1 6.8 less than 0.001 less than 0.001 23.0 28.6 1.3 0.245 NS % lichen 10.0 12.4 0.9 0.383 NS 15.5 24.8 2.2 0.032 NS % bare sand 44.0 12.0 6.3 less than 0.001 less than 0.001 45.7 34.0 1.8 0.075 NS Table 2. Percent occupancy for the 12 most-common species in microhabitat plots in the area of Avon Park Harebells populations at Carter Creek and Avon Park Lakes, presented by habitat (Florida scrub or roadside) and whether microhabitat plot was occupied or unoccupied by Avon Park Harebells. P values are shown for chi-square tests for difference between unoccupied or occupied plots within each habitat. In each case, n = 94, df = 1. NT = no test performed due to ≥2 cells with an expected count less than 1. * indicates signicifance (P < 0.05). % U = % in unoccupied plots; % O = % in occupied plots. % of Florida scrub plots Roadside plots Species plots % U % O P % U % O P Cladonia leporine Fr. (Cup Lichen) 38 21 33 0.196 49 49 0.974 Cladonia subtenuis (Abbayes) Mattick 36 44 48 0.692 30 20 0.296 (Reindeer Lichen) Aristida spp. (three-awn grasses) 26 10 15 0.486 42 38 0.768 Polygonella myriophylla (Small) Horton 25 19 24 0.541 28 28 0.992 (Small’s Jointweed) Selaginella arenicola Underw. (Sand 23 12 9 0.550 36 33 0.802 Spikemoss) Quercus inopina Ashe (Sandhill Oak) 22 50 28 0.031* 4 5 NT Licania michauxii Prance (Gopher Apple) 14 19 20 0.920 9 8 NT Serenoa repens (W. Bartram) Small 14 23 22 0.891 6 5 NT6 (Saw Palmetto) Cladonia evansii Abbayes (Powder-puff 13 27 21 0.547 2 0 NT Lichen) Stipulicida setacea Michx. (Pineland 11 6 17 0.093 11 8 0.563 Scaleypink) Cnidoscolus urens (L.) Arthur var. 9 2 20 0.006* 9 3 NT stimulosus (Michx.) Govarts (Finger Rot) Quercus geminata Small (Sand Live-oak) 8 6 13 NT 6 5 NT Southeastern Naturalist E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 2016 Vol. 15, No. 3 558 Saddle Blanket Scrub and Carter Creek, have had a consistently slow rate of decline in percent occupancy. This overall decline has occurred despite the fact that about 15% of all occupied plots that became unoccupied became occupied once again in subsequent samples. In contrast, the number of plants per occupied plot has not generally shown consistent patterns (Table 3). Mean and median numbers of plants have been consistent at Avon Park Lakes despite the loss in occupied plots, but have increased at Saddle Blanket between 2009 and 2010 and changed little from 2010 to 2013. At Carter Creek, there was a substantial increase in the mean but not the median between 2009 and 2012 (Table 3). Plant densities were significantly higher at Carter Creek (median ± SE = 1.42 plants/m2, ± 0.32, P < 0.05) than either Avon Park Lakes (0.12, ± 0.02) or Saddle Blanket Scrub (0.10, ± 0.02); the latter 2 sites had statistically similar densities. Population dynamics Avon Park Harebells populations at Carter Creek tended to have slow declines from 1998–2009, but have remained fairly stable or increased from 2009 to 2014 Figure 2. Percent occupancy of Avon Park Harebells at 1 unprotected site (Avon Park Lakes) and 2 protected sites (Saddle Blanket Scrub and Carter Creek) from 2005 to 2014. By definition, all studies started with 100% occupancy because points were located at all known patches of plants. The Saddle Blanket Scrub initial survey occurred over 2 years (2005–2006). Southeastern Naturalist 559 E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 2016 Vol. 15, No. 3 (Fig. 3). Within years, plant numbers generally increased from February through April or May, although patterns varied annually. Vital rates Annual survival. Median annual survival for Avon Park Harebells has been consistently high. Including plant dormancy, annual survival has been about 86%, with Table 3. Mean, standard error, and median number of Avon Park Harebells plants per occupied plot at Avon Park Lakes, Saddle Blanket Scrub, and Carter Creek. Plots are in a 5-m-radius circle at Avon Park Lakes, 10 m x 10 m square at Saddle Blanket Scrub, and 2-m-radius circle at Carter Creek. Mean densities per m2 are also shown. These data were not collected for the initial years of the study (2005, 2006) when plots were set up and occupancy was recorded. Population Year Mean # SE Median # Mean density Avon Park Lakes 2008 17.0 3.0 11 0.22 2011 19.1 2.9 11 0.24 2014 19.2 4.7 10 0.24 Saddle Blanket 2009 9.8 1.3 5 0.10 2010 14.9 1.8 9 0.15 2013 15.2 1.8 9 0.15 Carter Creek 2009 15.8 4.1 6 1.25 2012 24.2 7.1 6 1.92 Figure 3. Population trends in Avon Park Harebells study plots at Carter Creek during monthly monitoring February through June from 1998 to 2015. Each line shows the number of plants for the set of plots initiated in the year shown. The x-axis labels indicate the start of each sample year. Southeastern Naturalist E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 2016 Vol. 15, No. 3 560 little variation (range = 82–92%; Fig. 4). Annual survival varied among vegetation types (rosemary scrub, scrubby flatwoods, and roadsides) in 6 of 16 years (Fig. 5). In 5 of 6 cases in which annual survival differed significantly, plants in rosemary scrub had higher survival. In early years, roadside plants often had the lowest survival, likely due to vehicle damage. Dormancy. Both within-year and full-year plant dormancy are characteristic of Avon Park Harebells. Based on data from 1999–2013 (dormancy cannot be calculated directly from the first and last years of our longitudinal data), about 12% of plants (95 of 802) in our dataset had at least a full year of dormancy, and the median annual dormancy was 7% of living plants (Fig. 4). Just over half of plants recorded as dormant at some point during our study were dormant for only 1 years (54%), but we recorded 1 case each of plants that were dormant for 7 and 8 years. The mean length of dormancy for plants with at least 1 years of dormancy was 1.99 years. Only 8 plants were dormant twice, and 1 plant was dormant 3 times. Growth. Avon Park Harebells plants do not show consistent positive growth (increase in size from year to year). On average, growth was slightly negative and more plants regressed (42%) than grew (38%) (Table 4). These patterns varied year to year. In 2 years, half or more plants became smaller, and in 3 years, half or more plants grew larger. A consistent proportion (14–24% by year) maintained the same number of branches. Vegetation type had an effect on growth in only 5 years (of Figure 4. Total percent of Avon Park Harebells plants surviving (including dormant), surviving aboveground, and surviving plants that were dormant annually from 1998–1999 through 2013–2014 at Carter Creek. Year on X axis refers to second year of annual survival (e.g., 2004 indicates survival from 2003–2004). Southeastern Naturalist 561 E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 2016 Vol. 15, No. 3 16 y), with plants in scrubby flatwoods showing lower growth than plants along roadsides or in rosemary scrub (Table 4). Flowering and fruiting. Avon Park Harebells at Carter Creek are reproductively challenged. Flowering peaked in April in 10 of 14 years since 2000, and fruits were most abundant in May or June. Nonetheless, few fruits were produced. Fruits were initiated from less than ¼ of flowers in most years, and maturation rates were often less than 5% (Fig. 6). Low fecundity in Avon Park Harebells is clearly one of the factors contributing to poor recruitment. Slightly more than half of Avon Park Harebells plants had consistent flowering status from year to year. For example, 61% of plants that were vegetative in 2012 remained so in 2013 and 2014, while the remaining plants flowered once or twice. Fifty-three percent of plants that flowered in 2012 flowered in all 3 years, but the remainder reverted to being vegetative for 1–2 years. However, between 2012 and 2014, 41% of all plants switched between vegetative or flowering . Fruit lengths were slightly right-skewed from normal, with a range of 1.7–3.1 cm), a mode of 2.7 cm, and a mean and median of 2.5 cm. Seed number also Figure 5. Annual survival aboveground for Avon Park Harebells plants by vegetation type at Carter Creek. Year refers to the second year of the annual sequence (e.g., 2005 refers to 2004–2005 annual survival). Plants moving into or out of dormancy are not included. Bars are shown in cases with n > 9 plants. Southeastern Naturalist E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 2016 Vol. 15, No. 3 562 exhibited a range of 4–18 seeds per fruit (mean = 9.3, median = 8, mode = 7) and were slightly left-skewed. The number of seeds was not significantly correlated with capsule length (r = 0.151, n = 56, P = 0.262). Pollination biology and breeding system In 2003, we observed a diversity of floral visitors (mainly bees) on Avon Park Harebells, although insect visits were quite infrequent. Flowers seemed to open about 10:00 AM and were produced on the same plant for many weeks. Flowers were protandrous. Casual observations in subsequent years confirmed that insect visitation was quite rare. None of the 75 flowers from which we excluded pollinators initiated fruits (0 fruits on 9 plants). We observed low rates of fruit initiation on control plants—8 fruits developed from 66 flowers (12.1%) on 3 of the 8 plants that flowered. Although this is a small experiment and only 3 of 17 plants initiated fruits, the treatment difference was nearly significant (Mann-Whitney U = 22.5, P = 0.051). Fruit production among plants was highly variable, a pattern that also occurred in plants followed in our demography quadrats. Table 4. Growth statistics for Avon Park Harebells. For each year, the number of plants and the mean and standard deviation of annual relative growth rate (RGR), based on number of branches, is shown. We also show percent of plants regressing, growing, or remaining with the same number of branches, as well as the effects of vegetation type on growth in 1-way ANOVAs. In years when some plants were caged, we conducted separate analyses for caged and uncaged plants and overall. * indicates ≥1/2 of the plants were either regressing or growing in size. Vegetation types: RS = rosemary scrub, SF = scrubby flatwoods, Road = roadsides. Annual RGR Percent of plants Veg. type Greatest Years n Mean SD Regressing Same Growing effect (P) growth 1998–99 101 -0.49 0.74 69 14 17 0.221 - 1999–00 103 -0.05 0.77 41 19 40 0.409 - 2000–01 91 0.21 0.69 29 21 50 0.002 RS 2001–02 162 0.16 0.70 30 23 47 0.035 Road 2002–03 211 -0.18 0.72 49 22 29 0.726 - 2003–04 229 0.09 0.65 33 17 50 0.380 - 2004-05 217 -0.09 0.65 48 21 31 0.003 Road 2005–06 227 -0.18 0.71 48 16 35 0.026 RS 2006–07 221 0.15 0.80 37 12 51 0.000 Road 2007–08 212 -0.08 0.60 50 18 32 0.353 - 2008–09 202 0.04 0.69 43 13 44 0.934 - 2009–10 197 0.02 0.63 36 23 41 0.068 - 2010–11 185 -0.07 0.69 45 17 38 0.051 - 2011–12 182 -0.14 0.69 46 24 30 0.433 - 2012–13 191 0.20 0.69 29 20 51 0.855 - 2012–13 Caged 97 0.32 0.77 27 14 59 0.884 2012–13 Uncaged 94 0.09 0.58 31 26 43 0.945 2013–14 201 -0.17 0.69 44 26 30 0.750 - 2013–14 Caged 107 -0.06 0.64 36 31 33 0.835 2013–14 Uncaged 94 -0.29 0.63 53 20 27 0.520 Mean 183 -0.04 0.69 42 19 38 - - Southeastern Naturalist 563 E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 2016 Vol. 15, No. 3 Seed germination 2004 germination experiment. Germination began in the first few weeks of the study, with 37 seeds (11.3%) germinating in the first month. Three more seeds (0.9%) germinated in the first summer, and 3 seeds germinated in the second year of the experiment. These results demonstrated that Avon Park Harebells has a persistent seed bank, although germination rates after 1 year were below 1%. The overall germination rate through May 2007 was 13.2%. Germination percentages were much lower for green seeds (1.7%) than for purple (12.3%) or black (17.0%) seeds. Scarification reduced germination percentages (10.1% vs. 16.5%) and eliminated germination of green seeds. Seed color, but not scarification, was a significant predictor of germination in binary logistic regression (log likelihood = 124.2, df = 2, P < 0.001). When we removed green seeds from the analysis, neither the color of the seeds (purple vs. black) nor the scarification treatment affected germination (binary logistic regression, P > 0.15 for each variable). For black seeds alone, scarification did not affect germination percentage. Scarified seeds germinated marginally more quickly than non-scarified seeds (t = 2.04, df = 26, P = 0.052) in a mean time of 0.25 months vs. 2.5 months for non-scarified seeds. The effect of scarification on germination time was somewhat more pronounced for black seeds (t = 2.19, df = 16, P = 0.044); purple seeds only germinated in the first month. All scarified seeds germinated Figure 6. Estimated percentage of Avon Park Harebells flowers initiating or maturing fruits during the period 2000–2014 at Carter Creek, inferred from monthly counts of flowers, initiated fruits, and mature fruits. Southeastern Naturalist E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 2016 Vol. 15, No. 3 564 within 2 months, while non-scarified seeds germinated in both the first 2 months (23 seeds) and in the second year (3 seeds). 2012 shadehouse germination-experiment. For seeds heavy enough (>0.03 g) to germinate (no seeds less than 0.03 g ever germinated), germination percentages through December 2013 (17 mo) were high (56.3%). Over 70% of the germinants had emerged within 3 months during the summer of 2012 (Fig. 7). However, germination beyond 12 months (about 6% of germinants) again confirmed that Avon Park Harebells can persist in the seed bank. Both scarification treatment (B = 4.00, Wald = 6.09, df = 1, P = 0.014) and seed dry-weight (B = 370.1, Wald = 4.77, df = 1, P = 0.029) but not their interaction (B = -346.6, Wald = 2.02, df = 1, P = 0.155) predicted whether seeds germinated. Germination percentages were much higher for non-scarified seeds (77%) than scarified seeds (36%) during this 17-mo experiment. Heavier seeds were more likely to germinate than lighter seeds. Scarification affected time to germination (Fig. 7); scarified seeds germinated significantly faster post-sowing than non-scarified seeds (median days to germination = 9 vs. 58; Mann-Whitney U = 16.0, P ≤ 0.001). Non-scarified seeds continued to germinate throughout the experiment and during most months of the year. Figure 7. Number of Avon Park Harebells germinants by week after sowing for scarified and nonscarified seeds in the 2012 Archbold Biological Station shadehouse experiment. No seeds germinated during the final 8 weeks of the experiment (wee ks 63–71). Southeastern Naturalist 565 E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 2016 Vol. 15, No. 3 Seedling recruitment, survival, and flowering We observed seedlings recruiting sporadically and in most years from 2004– 2010 (Table 5). Less-frequent sampling after 2010 probably explains part of the drop in seedling numbers in the most recent years. Seedlings recruited throughout the year, but mostly appeared during late spring and summer. Seedlings survived well but grew slowly. Seven of the 32 seedlings found since 2004 were still alive through 2014 (Table 5). Seedling survival was moderate after the first year, but then rose to >80% from the second year onward (Fig. 8). However, seedlings had variable and often fairly slow growth. Mean number of branches by age increased from 1.5 in the first year to 2.6 by the third year, and to 6.0 by the eighth year. Although sample sizes are small, 6–8-year-old seedlings had similar mean and median numbers of branches as the overall population (5.5 and 3, respectively). Avon Park Harebells does not become reproductively mature for several years. Over the years, we have noted flowering in 4 plants first observed as seedlings. The age of first flowering has varied from 4 to 7 years (Table 5), with a median age at first flowering of 5 years. One seedling flowered twice and 1 seedling flowered 3 times. However, 3 plants surviving to 2014 never flowered, despite surviving for 6–9 years. The percent of plants flowering increased slowly from age 4 through ages 6–8, although less than 1/3 of the oldest seedlings flowered in any given year (Fig. 8). Herbivory Herbivory is a common but variable event for Avon Park Harebells. About 16% of our study plants had evidence of herbivory in any given year (Table 6); this rate varied widely from 7% to 53% from 1998 to 2014. The 2 years with the most herbivory were 2008 and 2009. We believe that the high level of herbivory in 2009 was mainly due to Bella Moth larvae because we observed many caterpillars. Herbivory Table 5. Summary of Avon Park Harebells seedling recruitment in the Carter Creek population since 2004. Seedlings were marked in their recruitment year and subsequently followed monthly through June 2011, and from February through June each year since then. Ages of flowering = seedling recruitment year defined as age zero, NA = not applicable (no plants survived to flowering), and V = still vegetative. Year of seeding # of seedlings Maximum # of Ages when emergence # of seedlings alive in 2014 branches in 2014 flowering (y) 2004 2 0 - - 2005 0 0 - NA 2006 8 3 8, 9, 1 4, 7, V 2007 2 1 2 V 2008 2 1 5 V 2009 14 2 4, 6 4, V 2010 2 0 - NA 2011 0 - - - 2012 2 0 - NA 2013 0 - - - 2014 0 - - - All cohorts 32 7 1–9 4–7 Southeastern Naturalist E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 2016 Vol. 15, No. 3 566 did not consistently vary among the 3 vegetation types (Table 6). Prior to caging, herbivory was highest in rosemary-scrub habitats from 1998 to 2000, scrubby flatwoods from 2008 to 2010, and roadside plants from 2008 onward. Caging, begun in 2012, significantly reduced herbivory in both subsequent years (Table 6). Caged plants had 8% herbivory in 2014, which was significantly lower than the 44% herbivory in uncaged plants (χ2 = 42.7, df = 1, P < 0.001). The difference was also significant in 2013 (the first year of caging), when caged plants had only 4% herbivory (vs. 21% in uncaged plants; χ2 = 13.3, df = 1, P < 0.001). These observed effects of caging might be underestimates because the percentages only include plants that emerged and were observed during monthly surveys. If plants emerged and then were immediately eaten (or died back from drought or other causes), we might not have recorded them in our monthly surveys. Caging significantly increased annual survival from 2013 to 2014 (χ 2 = 9.5, df = 1, P = 0.002). Caged plants had 93% survival vs. 79% survival for uncaged plants. Caged plants had slightly higher annual 2012–2013 survival than uncaged plants (89 vs. 85%), but the difference was not significant (χ2 = 0.59, df = 1, P = 0.44). However, the cages were not installed until halfway through the 2012 calendar year. The significant cage effect on survival and the lack of prior observed herbivory effects on survival (Menges et al. 2013) suggest that monthly censuses do not capture all herbivory events. Figure 8. Percent of known Avon Park Harebells seedlings surviving or flowering annually at Carter Creek, as a function of age, determined by following naturally recruiting seedlings over time. Plants 6–8 y old are binned under “6”. Sample sizes are n > 9 for each point. Southeastern Naturalist 567 E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 2016 Vol. 15, No. 3 Caging also had positive effects on relative growth rate (based on branch numbers) from 2012 to 2013 and 2013 to 2014 (Table 4). For 2012–2013, caging increased growth (0.32 vs. 0.08, t = 2.31, P = 0.022). From 2013 to 2014, caged plants had less-negative relative growth rates (-0.06) than uncaged plants (-0.29); this difference was significant (t = 2.41, P = 0.017). Caging also affected flowering in both 2013 and 2014. Caging had a significant effect on flowering in 2013 (χ2 = 8.7, df = 1, P = 0.003); caged plants were more likely to flower than uncaged plants (41% vs. 24%). In 2014, caged plants flowered 36% of the time, in contrast to uncaged plants, which flowered 20% of the time; the difference was again significant (χ2 = 7.8, df = 1, P = 0.005). Effects of fire and mechanical treatments Fire and mechanical treatments had no discernable negative effects on Avon Park Harebells. Plants rapidly resprouted (typically between 2–4 months) after the August 2005 disturbances (fire, Gyro-Trac followed by fire). Fire had neither a negative nor a positive effect on initial survival of Avon Park Harebells. Between 2005 and 2006, survival did not differ significantly among the various treatments (χ2 = 3.3, df = 2, P = 0.195). Survival was similar in the burn-only quadrats (95.6%), the Gyro-Trac + burn (87.5%), and unburned rosemary-scrub quadrats elsewhere at Carter Creek (matching the vegetation type of the burned quadrats; 80.6%). However, in the second year (2006–2007), survival was highest in burned quadrats (95.6%), lowest in control quadrats (71.4%), and intermediate in combination Table 6. Percent of Avon Park Harebells plants at Carter Creek with herbivory at some point in the year, and effects of vegetation, by year. When vegetation effects were significant, the vegetation type with the highest herbivory is indicated by +. x = P < 0.1, *P < 0.05; **P < 0.01; ***P < 0.001, and NS = not significant. Year Overall Rosemary scrub Scrubby flatwoods Roadside P 1998 9.0 17.1+ 2.7 3.4 *** 1999 11.5 24.0+ 3.4 0.9 *** 2000 9.1 18.7+ 1.3 2.6 *** 2001 10.3 9.8 12.1 8.6 NS 2002 12.5 7.8 18.8+ 12.1 ** 2003 11.6 11.0 13.4 10.3 NS 2004 15.9 19.2 14.1 12.9 NS 2005 17.6 22.4+ 13.4 13.8 * 2006 6.9 3.2 14.7+ 1.8 ** 2007 10.8 11.1 14.4 5.0 NS 2008 30.1 19.8 25.9 51.6+ *** 2009 52.8 75.0+ 30.3 43.8 *** 2010 24.6 16.2 44.8+ 20.9 *** 2011 20.2 8.8 18.9 41.9+ *** 2012 16.1 11.3 13.8 26.2+ * 2013 all plants 12.9 9.5 4.0 23.5+ ** 2013 uncaged 21.4 16.4 7.7 32.6 x 2014 all plants 27.0 31.3+ 3.8 25.8 * 2014 uncaged 44.4 56.3+ 7.1 36.6 ** Median (uncaged) 16.1 16.2 13.8 12.9 - Southeastern Naturalist E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 2016 Vol. 15, No. 3 568 quadrats (80.9%); the difference was significant (χ2 = 6.1, df = 2, P = 0.047). Subsequently, survival was greater in each of the treated areas (100% in 2007–2008) than in the control (70.2%; χ2= 21.2, df= 2, P < 0.001). Beyond 2008, the advantages of the disturbances were inconsistent. Growth was affected by management treatments (fire, Gyro-Trac followed by fire) for 2 of the years following management. The combination of Gyro-Trac and fire produced higher initial growth relative to the burn only and the control (Table 7). Two years after treatments, there was usually little difference among treatments, although plants in the area that was Gyro-Trac–chopped and then intensely burned sometimes had lower growth. Effect of vehicles Repeated, but not single, vehicle disturbances caused marked mortality of Avon Park Harebells. At the Carter Creek plots that were run over repeatedly by ATVs between 2004 and 2006, only 4 plants (15%) recovered, despite the removal of traffic due to fencing in 2008. Burial of plot markers indicated that some of the disturbances may have involved vehicles that dug well into the sandy soil. In contrast, plants crushed by a single Gyro-Trac event during preparation for prescribed burns survived more frequently. Nine of 19 plants (47%) crushed in 2007 and 7 of 11 plants crushed in 2009 recovered (some recovery was delayed until 2010). Plants affected in both years rarely survived (2 of 19 marked plants). Our analysis may underestimate the resilience of this species because their positions may have shifted due to the mechanical disturbance and new plants later recruited into these areas. Discussion Support for hypotheses Most of our hypotheses addressing the ecology of Avon Park Harebells were supported by our long-term study. Supported hypotheses included the plant’s preference for open microsites, more-positive population trends in protected sites, Table 7. Effect of management treatments (burn only, Gyro-Trac [G] followed by light or intense fire) made in August 2005 on subsequent relative growth rate (based on number of branch tips). For the comparison, we only considered unburned rosemary-scrub sites with the same vegetation as occurred in the area of the burn and Gyro-Trac treatments. * indicates relatively high growth in years with significant (P < 0.05) treatment effects. Year Unburned Burn only G + light burn G + intense burn F P 2005–2006 -0.38 0.08 0.23* 0.14* 7.6 0.000 2006–2007 -0.09 0.44 0.57* 0.73* 4.7 0.005 2007–2008 0.11 -0.09 -0.04 -0.03 1.1 0.335 2008–2009 0.21 -0.16 0.00 -0.06 1.5 0.214 2009–2010 -0.27 0.11* 0.14* -0.33 3.4 0.021 2010–2011 0.20 0.10 0.14 -0.33 1.6 0.199 2011–2012 0.19* -0.32 -0.23 -0.56 4.1 0.008 2012–2013 0.15 0.17 0.25 0.11 0.2 0.874 2013–2014 -0.07 0.06 -0.28 -0.31 1.8 0.158 Southeastern Naturalist 569 E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 2016 Vol. 15, No. 3 high annual survival, a persistent seed bank, increase in germination speed with scarification, negative effects of herbivory, and post-fire resprouting. However, in contrast to our hypotheses, Avon Park Harebells had minimal growth from year to year and had limited seedling recruitment. Life history and demography of Avon Park Harebells Avon Park Harebells is a long-lived perennial herb with fairly high (about 85%) and consistent annual survival, but inconsistent and often non-directional growth trajectories. Seedling survival is >80% after the first year, seedlings grow slowly, and flower between 4 and 7 years of age. Seed production and seedling recruitment are both fairly limited. This species also is a strong resprouter after fire, herbivory, and mechanical disturbances that remove its aboveground parts. All these traits are typical of organisms on the slow end of the fast–slow continuum (Franco and Silvertown 1996, Jones et al. 2008). As a low-growing herbaceous plant growing in a shrubland (Florida scrub), Avon Park Harebells is potentially vulnerable to aboveground and belowground competition from the shrubs (oaks, palmettos, ericads) that dominate its habitat (Abrahamson 1985). Little is known of rooting patterns of Florida scrub plants, although some shrubs are deeply rooted with high aboveground:belowground ratios (Saha et al. 2010). Avon Park Harebells is also deeply rooted (DeLaney and Wunderlin 1989), but, aboveground, it can be easily overtopped by shrubs. Although plants in this study showed a preference for open microsites, the species is capable of growing beneath a shrub canopy. This characteristic contrasts with many Florida scrub herbs that strongly specialize in gaps (Dee and Menges 2014; Menges et al. 1999, 2008). Other herbs that are able to grow well in shrub matrices include Liatris ohlingerae (S.F. Blake) B.L. Rob. (Scrub Blazing-star) (Petru and Menges 2003) and Asclepias curtissii (A.Gray) (Curtiss’ Milkweed) (Mondo et al. 2010). Many long-lived herbaceous plants exhibit plant dormancy (vegetative dormancy, prolonged dormancy; Shefferson 2009), during which plants spend an entire growing season belowground. This habit may reflect an adaptive bet-hedging trait to counter the effects of environmental stress and stochasticity (Shefferson et al. 2012). Dormancy may also be effective because plants can obtain carbon without having aboveground parts (Shefferson 2009) or remobilize structural carbon for subsequent re-growth (Gremer et al. 2010). Avon Park Harebells can be dormant for an entire growing season or longer, but fewer than 10% of plants exhibit this plant dormancy and most are only dormant for 1 year. In contrast, some orchid populations have as many as 30% of plants dormant with many plants dormant for multiple years (e.g., Hutchings 2010). Species exhibiting dormancy tend to grow on dry sites, with high-light levels and infertile soil (Reintal et al. 2011); these traits also characterize Avon Park Harebells. Reproductive biology Avon Park Harebells is reproductively challenged. This non-clonal species produces limited numbers of fruits and seeds, at least at our demographic study Southeastern Naturalist E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 2016 Vol. 15, No. 3 570 site (Carter Creek). At our site, flowers were consistently produced over a period of months, but few fruits were initiated and matured. Our breeding-system experiments demonstrated that no fruit production occured in flowers denied access to pollinators. This failure of bagged flowers to set fruit could be due to genetic selfincompatibility or to the need for insect vectors to transfer self or outcross pollen. Fruit production in unbagged, open-pollinated flowers could have resulted from pollen movement among plants (outcrosses), between flowers of the same plant (geitonogamy), or from insect-vectored transfer of pollen within a flower (facilitated autogamy). Although our observations suggest that fruit set requires insect pollinators, they do not indicate whether or not Avon Park Harebells is genetically self-compatible. Pollinator visits to Avon Park Harebells plants appear rare, although its pollination biology remains largely unstudied. Seeds mature inside capsules and turn color from green to purple/brown and black. Ripe pods dehisce and drop black seeds to the ground. Green seeds are generally not viable, and germination of purple/brown seeds is limited to a few months after sowing. In contrast, the mature, hard, black seeds can germinate immediately or stay dormant for at least 2 y. Results from our 2 shadehouse experiments showed the strong effects of scarification in breaking seed dormancy and facilitating rapid germination. However, germination percentages were lower with scarification, in part because only non-scarified seeds germinated in the second year after seed production. Both germination experiments showed peaks in germination during the first few months after seed maturation (in the summer). Our results also showed a small, persistent seed bank with germination in the summer of the second year. These trends are consistent with a field germination experiment started as part of an experimental introduction in 2012 (Smith and Menges 2013). Mechanisms for scarification in the field probably include sand movement and heat from fires. Fire breaks physical dormancy in species of many families (Baskin and Baskin 1998), and fire promotes germination of buried seeds in other legumes (Bell et al. 1993, Bradstock and Auld 1995). Promotion of germination by fire-induced scarification may allow seedlings a greater chance of survival and growth. Likewise, breakdown of the hard seed coat over time by rain may help synchronize most germination to occur during the wetter summer months. Herbivory Avon Park Harebells is affected by both mammalian and insect herbivory. One insect herbivore is the Bella Moth larva, which specializes on Crotalaria species and utilizes pyrrolizidine alkaloids obtained from the plants for its defense (Mattocks 1986). Damage due to insects appears variable in space and time. In some years, damage from the Bella Moth was greater in spring vs. winter months (O’Chaney 2007). Population densities of the Bella Moth have been inflated by the spread of exotic (African) species of Crotalaria (Mark Deyrup, Archbold Biological Station, Venus, FL, pers. comm.). The caterpillar feeds preferentially on Crotalaria fruits, although it selects leaves if no fruits are in the vicinity. Based on the feeding rate of a captive caterpillar, each one is capable of completely eating an Avon Park Southeastern Naturalist 571 E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 2016 Vol. 15, No. 3 Harebells fruit in a day. However, herbivory by Bella Moth appears episodic. Avon Park Harebells are also damaged by beetles and caterpillars (O’Chaney 2007). Herbivory of entire stems is very common. We attribute this herbivory to mammals such as Sylvilagus floridanus (J.A. Allen) (Eastern Cottontail) or Odocoileus virginianus (Zimmermann) (White-tailed Deer). Caging plots reduced observed herbivory by 5-fold, and plants protected by cages had higher survival, growth, and flowering in subsequent years. This result was unexpected, because we frequently observed strong resprouting of Avon Park Harebells after complete herbivory, and because observed vital rates were not related to whether we had observed herbivory in the previous year. Disturbance ecology Avon Park Harebells is found only in Florida scrub, which is a pyrogenic ecosystem. The natural fire regime for Florida scrub is characterized by infrequent, patchy, intense fires ignited largely in the late spring and early summer (Menges 2007). Avon Park Harebells individuals and populations are resilient to fire. Plants resprout quickly after fires, and populations do not show markedly altered survival, growth, or fecundity. In fact, some measures of survival and growth were higher for burned than unburned plots perhaps due to reduced competition or increased nutrient availability post-fire. This species is also resilient to some human-caused disturbance events, although either frequent or severe disturbances can impact populations. Multiple disturbances by vehicles (often within the same year), especially causing soil disturbances, reduced populations several times during this study. Land-management practices that merely mow or crush plants are less damaging than vehicle disturbances that churn up the sandy soil, but even these kinds of impacts caused population declines when they occurred only 2 years apart. Population trends In protected areas such as our study site (Carter Creek), Avon Park Harebells populations appear fairly stable. Densities in our study plots have showed stability or a slow decline over the years. Broader surveys at both protected sites have also shown stable populations. In contrast, the single unprotected site (Avon Park Lakes) showed a drastic decline in geographic extent (as measured by occupancy) during a time when housing activity was high. During the recent economic recession, new housing construction ceased and the population stabilized. However, at this site, nearly all plants grow on roadsides and not in the very dense, overgrown, long-unburned scrub. We have initiated an introduction of Avon Park Harebells to a protected site using seeds and plant material from Avon Park Lakes. The introduction is less than 3 years old, but to date has been successful (E. Menges, pers. observ.; Smith and Menges 2013). Conservation and management recommendations Wild populations of Avon Park Harebells have been documented at only 2 protected sites; thus, active conservation and land management is essential to its Southeastern Naturalist E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 2016 Vol. 15, No. 3 572 persistence (Turner et al. 2006). Based on our research findings, we recommend the following: • Continue managing the Florida scrub sites with prescribed fire. Avon Park Harebells responds well to fires, and fires will limit the development of vegetation structure (tall shrubs and trees) that may harm this species in the long run. Fire management is consistent with many conservation objectives in Florida ecosystems (Slapcinsky et al. 2010). • Consider additional introductions and augmentations of Avon Park Harebells to the protected sites within its limited range. This action will provide a bit of conservation bet-hedging against catastrophes at its current sites and provide useful information on its management. • Where feasible in small populations, take measures to reduce herbivory. Caging of plants has been very successful in reducing herbivory and increasing vital rates, and we routinely cage any introduced plants to limit damage. Actions to control likely herbivores will be difficult to accomplish, given the degree of suburban development within the range of Avon Park Harebells. • Use management that encourages pollinators (e.g., prescribed fire, avoidance of herbicides) and consider further studies on the identities and roles of pollinators on fecundity of Avon Park Harebells. Acknowledgments We thank Viani Menges, Carl Weekley, Amanda Brothers, Marcia Rickey, Christine Bertz, Evan Batzer, Sadie Watts, Austin Ritenour, Stephanie Koontz, Kelly Peterson, John Benning, Ryan Cressey, Britta Countryman, Devon Picklum, and Jamie Peeler for assistance in the field. Amanda Brothers conducted the pollination and breeding systems experiment. We are grateful to the co-operating agencies and landowners: Nicole Ranalli and Wade Ulrey (Florida Fish and Wildlife Conservation Commission), and Steve Morrison (The Nature Conservancy). We appreciate the funding support from the Endangered and Threatened Native Flora Conservation Grants Program administered by the Division of Plant Industry, the National Science Foundation (DEB98-15370, DEB02-33899, DEB08- 12717, DEB-1347843), and the US Fish and Wildlife Service. We thank the Endangered Plant Advisory Council and Dave Bender (USFWS) for advice and support. This manuscript was improved by the comments of Christine Bertz and Stephanie Koontz. Literature Cited Abrahamson, W.G. 1984. Post-fire recovery of Florida Lake Wales Ridge vegetation. American Journal of Botany 71:9–21. Baskin, C.C., and J.M. Baskin. Seeds: Ecology, Biogeography, and Evolution of Dormancy and Germination. Academic Press, San Diego, CA. 666 pp. Bell, D.T., J.A. Plummer, and S.K. Taylor. 1993. Seed-germination ecology in southwestern western Australia. Botanical Review 59:1–55. Bradstock, R.A., and T.D. Auld. 1995. Soil temperatures during experimental bushfires in relation to fire intensity: Consequences for legume germination and fire management in southeastern Australia. Journal of Applied Ecology 32:76–84. Southeastern Naturalist 573 E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 2016 Vol. 15, No. 3 Christman, S.P., and W.S. Judd. 1990. Notes on plants endemic to Florida scrub. Florida Scientist 53:52–73. Dee, J.R., and E.S. Menges. 2014. Gap ecology in Florida scrubby flatwoods: Effects of time-since-fire, gap size, gap aggregation, and microhabitat on species occurrence and composition. Journal of Vegetation Science 25:1235–1246. DeLaney, K.R., and R.P. Wunderlin. 1989. A new species of Crotalaria (Fabaceae) from the Central Florida Ridge. Sida 13:315–324. Dobson, A.P., J.P. Rodriguez, W.M. Roberts, and D.S. Wilcove. 1997. Geographic distribution of endangered species in the United States. Science 275:550–553. Estill, J.C., and M.B. Cruzan. 2001. Phytogeography of rare plant species endemic to the southeastern United States. Castanea 66:3–23. Franco, M., and J. Silvertown. 1996. Life-history variation in plants: An exploration of the fast–slow-continuum hypothesis. Philosophical Transactions: Biological Sciences 351:1341–1348. Gremer, J.R., A. Sala, and E.E. Crone. 2010. Disappearing plants: Why they hide and how they return. Ecology 91:3407–3413. Hutchings, M.J. 2010. The population biology of the Early Spider Orchid, Ophrys sphegodes Mill. III. Demography over three decades. Journal of Ecology 98:867–878. Jones, O.R., J.-M. Gaillard, S. Tuljapurkar, J.S. Alho, K.B. Armitage, P.H. Becker, P. Bize, J. Brommer, A. Charmantier, M. Charpentier, T. Clutton-Brock, F.S. Dobson, M. Festa- Bianchet, L. Gustafsson, H. Jensen, C.G. Jones, B. Lillandt, R. McCleery, J. Merilä, P. Neuhaus, M.A.C. Nicoll, K. Norris, M.K. Oli, J. Pemberton, H. Pietiäinen, T.H. Ringsby, A. Roulin, B.-E. Saether, J.M. Setchell, B.C. Sheldon, P.M. Thompson, H. Weimerskirch, E.J. Wickings, and T. Coulson. 2008. Senescence rates are determined by ranking on the fast–slow life-history continuum. Ecology Letters 11:664–673. Kettenring, K.M., E.S. Menges, and C.W. Weekley. 2009. Herbivory delays flowering and reduces fecundity of Liatris ohlingerae, an endangered, endemic plant of the Florida scrub. Journal of the Torrey Botanical Society 136:350–362. Mattocks, A.R. 1986. Chemistry and Toxicology of Pyrrolizidine Alkaloids. Academic Press, London, UK. 393 pp Menges, E.S. 1999. Ecology and conservation of Florida scrub. Pp. 7–22, In R.C. Anderson, J.S. Fralish, and J. Baskin (Eds.). The Savanna, Barren, and Rock Outcrop Communities of North America. Cambridge University Press, Cambridge, UK. 470 pp. Menges, E.S. 2007. Integrating demography and fire management: An example from Florida scrub. Australian Journal of Botany 55:261–272. Menges, E.S., and D.R. Gordon. 2010. Should mechanical treatments and herbicides be used as fire surrogates to manage Florida’s uplands? A review. Florida Scientist 73:145–172. Menges, E.S., and N.M. Kohfeldt. 1995. Life-history strategies of Florida scrub plants in relation to fire. Bulletin of the Torrey Botanical Club 122:282–297. Menges, E.S., and C.W. Weekley. 2013. Demographic responses of two endemic plants to sandhill restoration on the Lake Wales Ridge. Journal of the Torrey Botanical Society 140:480–492. Menges, E.S., P.J. McIntyre, M.S. Finer, E. Goss, and R. Yahr. 1999. Microhabitat of the narrow Florida scrub endemic Dicerandra christmanii, with comparisons to its congener D. frutescens. Journal of the Torrey Botanical Society 126:24–31. Menges, E.S., A. Wally, J. Salo, R. Zinthefer, and C.W. Weekley. 2008. Gap ecology in Florida scrub: Species occurrence, diversity, and gap properties. Journal of Vegetation Science 19:503–514. Southeastern Naturalist E.S. Menges, B. Pace-Aldana, S.J. Haller, and S.A. Smith 2016 Vol. 15, No. 3 574 Menges, E.S., S.A. Smith, and S.M. Koontz. 2013. Conservation research on state-listed plants endemic to the Lake Wales Ridge. Final Report to Florida Division of Plant Industry, Gainesville, FL. 113 pp. Moles, A.T., and M. Westoby. 2004. What do seedlings die from and what are the implications for evolution of seed size? Oikos 106:193–199. Mondo, P., K.D.M. Mattson, and C.C. Bennington. 2010. The effect of shrubs on the establishment of an endangered perennial (Asclepias curtissii) endemic to Florida scrub. Southeastern Naturalist 9:259–274. O’Chaney, M. 2007. The effects of herbivory on rattlebox species (Crotalaria spp., Fabaceae) by Bella Moth (Utetheisa ornatrix L.) caterpillars. M.Sc. Thesis. Florida International University, Miami, FL. 57 pp. Petru, M., and E.S. Menges. 2003. Seedling establishment in natural and experimental Florida scrub gaps. Journal of the Torrey Botanical Society 130:8–100. Reintal, M., K. Tali, M. Haldna, and T. Kull. 2011. Habitat preferences as related to prolonged dormancy of perennial herbs and ferns. Plant Ecology 210:111–123. Saha, S., A. Catenazzi, and E.S. Menges. 2010. Does time since fire explain plant-biomass distribution in the Florida, USA, scrub ecosystem? Fire Ecology 6:13–25. Shefferson, R.P. 2009. The evolutionary ecology of vegetative dormancy in mature herbaceous perennial plants. Journal of Ecology 97:1000–1009. Shefferson, R.P., T. Kull, K. Tali, and K.M. Kellett. 2012. Linking vegetative dormancy to fitness in two long-lived perennials. Ecosphere. Available online at http://dx.doi. org/10.1890/ES11-00328.1. Accessed 8 September 2016. Slapcinsky, J.L., D.R. Gordon, and E.S. Menges. 2010. Responses of rare plant species to fire across Florida’s fire-adapted communities. Natural Areas Journal 30:4–19. Smith, S.A., and E.S. Menges. 2013. Introduction of the critically endangered Crotalaria avonensis onto protected lands. Final Report to US Fish and Wildlife Service, Vero Beach, FL. 43 pp. Turner, W.R., D.S. Wilcove, and H.M. Swain. 2006. State of the scrub: Conservation progress, management responsibilities, and land-acquisition priorities for imperiled species of Florida’s Lake Wales Ridge. Archbold Biological Station, Lake Placid, FL. 160 pp. Tye, M.R., E.S. Menges, C. Weekley, P.F. Quintana-Ascencio, and R. Salguero-Gomez. 2016. A demographic ménage à trois: Interactions between disturbances both amplify and dampen population dynamics of an endemic plant species. Journal of Ecology. DOI: 10.1111/1365-2745.12642. US Fish and Wildlife Service (USFWS). 2006. Avon Park Harebell (Crotalaria avonensis) 5-year review: Summary and evaluation. Available online at https://www.fws.gov/sou theast/5yearreviews/5yearreviews/avonpark_harebellsfinal.pdf. Accessed 8 September 2016. Weekley, C.W., E.S. Menges, and R.L. Pickert. 2008. An ecological map of Florida’s Lake Wales Ridge: A new boundary delineation and an assessment of post-Columbian habitat loss. Florida Scientist 71:45–64. Wiggers, M.S. 2011. Some like it hot: Fire and legume germination in the Longleaf Pine Ecosystem. M.Sc. Thesis. Auburn University, Auburn AL. 92 pp.