Biological Conservation 111 (2003) 235–246 www.elsevier.com/locate/biocon

Reproductive biology of three sympatric endangered plants endemic to Florida scrub

Margaret E.K. Evansa,*, Eric S. Mengesb, Doria R. Gordonc aDepartment of Ecology and Evolutionary Biology, University of Arizona, Tucson, AZ 85721, USA bArchbold Biological Station, PO Box 2057, Lake Placid, FL 33870, USA cThe Nature Conservancy, Department of Botany, PO Box 118526, University of Florida, Gainesville, FL 32611, USA

Received 25 August 2001; received in revised form 24 April 2002; accepted 20 September 2002

Abstract We investigated the reproductive biology of three plants endemic to rosemary scrub habitats on the Lake Wales Ridge of Florida, USA. We used hand-pollination experiments and observations of flowers and their insect visitors to determine their mating systems and pollination. Fruit or seed set after self pollination was 94, 97, and 8% of fruit or seed set after cross pollination in Eryngium cuneifolium (Apiaceae), Hypericum cumulicola (Hypericaceae), and Liatris ohlingerae (Asteraceae) respectively, indicating that the first two are self-compatible and the last is obligately outcrossing. All three depend on insects for seed production (4–7% fruit or seed set without insects). Diverse insects visit flowers of E. cuneifolium (101 species recorded), whereas L. ohlingerae is visited pre- dominantly by butterflies and H. cumulicola by one genus of bees (Dialictus, Halictidae). Our data indicate pollinator visitation does not currently limit seed production in E. cuneifolium or H. cumulicola, but does in L. ohlingerae. Despite the features they share (habit, habitat, disturbance regime), we found unique aspects of these species’ reproductive biology yielding unique risks to popu- lation viability. We suggest that multispecies recovery plans must consider several aspects of the biology of species with superficial similarities to be successful. # 2003 Elsevier Science Ltd. All rights reserved. Keywords: Eryngium cuneifolium; Florida scrub; Hypericum cumulicola; Liatris ohlingerae; Mating system

1. Introduction change hinge upon such variation (Fisher, 1958; Huen- neke, 1991). The loss of self-incompatibility alleles in Through its effects on demography and population self-incompatible species poses a special threat to genetics, reproductive biology has important con- genetic viability (Les et al., 1991; DeMauro, 1993; sequences for the viability of rare plant populations. Schemske et al., 1994). Genetic viability may also be Seed production of plants that rely on pollinators may affected negatively by inbreeding in the presence of be limited by pollinator abundance or behavior (polli- inbreeding depression (Charlesworth and Charlesworth, nator limitation; Bierzychudek, 1981). Pollinator limi- 1987; Lande, 1995; Lynch et al., 1995) or positively by tation or, at its extreme, pollinator failure could cause inbreeding to the extent that deleterious alleles can be seed production to fall below the level necessary for purged (Husband and Schemske, 1996; Byers and Wal- population viability (Lamont et al., 1993; Bond, 1994; ler, 1999). Steiner and Whitehead, 1996; Mawdsley et al., 1998; The Lake Wales Ridge (LWR) of central Florida, Spira, 2001). Mating patterns affect the amount of USA is an ancient dune system covering 150 Â 15 km, genetic variation found in plant populations (Hamrick with more than 20 species of federally-listed, narrowly- et al., 1991; Ellstrand and Elam, 1993; Hamrick and endemic higher plants (Christman and Judd, 1990; Godt, 1996). Genetic variation is important for genetic Dobson et al., 1997; USFWS, 1999). Declines of these viability, since evolutionary responses to environmental plants have been caused primarily by habitat loss and fire suppression. More than 15 years ago, Peroni and * Corresponding author. Abrahamson (1985) estimated that just 16% of the his- E-mail address: [email protected] (M. E. K. Evans). toric extent of scrub habitat on the LWR remained. In

0006-3207/03/$ - see front matter # 2003 Elsevier Science Ltd. All rights reserved. PII: S0006-3207(02)00293-8 236 M.E.K. Evans et al. / Biological Conservation 111 (2003) 235–246 this study we examine the reproductive biology of three 1.1. The species of these species, Eryngium cuneifolium Small (Apiaceae), Hypericum cumulicola (Small) W. P. Adams (Hyper- Although all three species grow in rosemary scrub, icaceae), and Liatris ohlingerae B. L. Rob. (Asteraceae). they vary in a number of ecological traits (Table 1). All three are herbaceous perennials found in rosemary Only L. ohlingerae is commonly found in both rosemary scrub habitats and are federally listed as endangered scrub and a more mesic, oak-dominated community, species. Rosemary scrub occurs on elevated patches scrubby flatwoods. Scrubby flatwoods have few gaps (‘‘islands’’) of white sands throughout the Lake Wales between shrubs (Young and Menges, 1999), a shorter Ridge and other ridges in Florida (Menges and Hawkes, average fire return interval (5–20 years; Menges, 1999), 1998; Menges, 1999). Fire is an important ecological and are more abundant and spatially continuous on the force in rosemary scrub, which is dominated by the Lake Wales Ridge. The species also differ in postfire shrub, Florida rosemary (Ceratiola ericoides). Gaps recovery and how they respond to community-level among individuals of C. ericoides are large immediately changes with time-since-fire. Individuals of E. cuneifolium after fire and close with time-since-fire (Menges and and H. cumulicola are killed by fire, but populations Hawkes, 1998). At least 10 years are required before a recover from dormant seeds in the soil (Quintana- sufficiently continuous mass of fuel accumulates for fire Ascencio and Morales-Hernandez, 1997; Quintana- to return. Historical fire return intervals in rosemary Ascencio and Menges, 2000). In contrast, L. ohlingerae scrub probably ranged from 20 to 100 years, but fire has resprouts after fire from a belowground corm (Menges been suppressed over large parts of the LWR for the last and Kohfeldt, 1995), though resprouting is low after 60 years (Menges, 1999). some fires (Weekley, C. W., Menges, E. S., unpublished The species studied here have common and unique data). Recruitment of L. ohlingerae seedlings appears characteristics. If their similarities (habit, habitat, and episodic (Herndon, 1997). E. cuneifolium is most sensi- disturbance regime) lead to similar reproductive biol- tive to shrub encroachment into the open microsites in ogy, then detailed knowledge of the reproductive biol- which it does best (Menges and Kimmich, 1996; Menges ogy of each species would not be necessary. and Hawkes, 1998), H. cumulicola is somewhat less Alternatively, differences in their life span, reliance on sensitive to competition from shrubs (Quintana-Ascen- seeds to recover from fire, specialization to post-fire cio and Morales-Hernandez, 1997; Quintana-Ascencio gaps, and hence population dynamics could generate and Menges, 2000), and L. ohlingerae often grows different degrees of selection for reproductive assurance. among shrubs. From E. cuneifolium to H. cumulicola to Differences in reproductive biology resulting from these L. ohlingerae, these three species form a series from different selective pressures could then require species- shorter to longer-lived and greater to less demographic specific conservation approaches. To address the fluctuation (Menges et al., 2001; Menges, E. S., unpub- hypothesis that these species are reproductively uniform lished data). In addition, the first two species, which and evaluate risks to population viability arising from depend on seeds to recover from fire, appear to have reproductive biology, we investigated the floral biology, metapopulation dynamics (Quintana-Ascencio and flowering phenology, pollination, and mating systems of Menges, 1996; Quintana-Ascencio et al., 1998), which is E. cuneifolium, H. cumulicola, and L. ohlingerae. not evident in L. ohlingerae.

Table 1 Ecological and life history traits of three endangered plant species endemic to the Lake Wales Ridge, Florida, USA

Species Habit Range, Habitat Specialization Post-fire Median popualtion (and max no. extant to open space regenerationa size (and size lifespan)a populationsb fluctuation)a

Eryngium Perennial 43 km Rosemary scrub Highc Seeds, limited 4280 (high) cuneifolium herb (10 year) 20 pop resprouting Hypericum Perennial 84 km Rosemary scrub Moderated Seeds 539 (high) cumulicola herb (10 year) 90 pop Liatris Perennial 109 km Rosemary scrub, none Resprouting 179 (low) ohlingerae herb (>10 year) 115 pop scrubby flatwoods

Maximum lifespan was estimated from demographic monitoring of E. cuneifolium and H. cumulicola, and from observations of L. ohlingerae. The range of each species is expressed in km north–south along the long axis of the 160 Â 20 km Lake Wales Ridge, in Florida, USA. a From Menges et al. (2001) b From Dolan et al. (1999) c From (Menges and Hawkes, 1998, Quintana-Ascencio and Menges, 2000). d From (Quintana-Ascencio & Morales-Hernandez 1997, Quintana-Ascencio & Menges 2000). M.E.K. Evans et al. / Biological Conservation 111 (2003) 235–246 237

The three species also differ in genetic variation, as Our goal was to categorize each plant’s pollination in detected by isozymes. L. ohlingerae has the most varia- two respects: as generalist (many pollinating insect spe- tion (percent of loci polymorphic, number of alleles per cies) vs. specialist (few pollinating insect species), and in polymorphic loci, and observed heterozygosity; Dolan terms of the type of insects visiting the flowers (bees, et al., 1999). Allelic diversity is also distributed most bee-flies, flies, butterflies, moths, beetles, etc.). These evenly in L. ohlingerae, indicated by a greater propor- categorizations are a first step in understanding both the tion of genetic variation found within than among plants’ pollination biology and the potential extent of populations (FST=0.120; Dolan et al., 1999). In con- gene flow through pollen movement. trast, H. cumulicola has little variation, which is strik- All observations of insect visitors were made in nat- ingly structured, both at regional (FST=0.730; Dolan et ural populations at ABS. Each observation period las- al., 1999) and local scales (Quintana-Ascencio et al., ted from 15 to 30 minutes and included a single focal 1998), whereas E. cuneifolium is intermediate in these plant. Insect behavior on flowers was recorded between respects (FST=0.445; Dolan et al., 1999). The isozyme 0720 h and 1005 h in H. cumulicola, and between 1035 h data further suggest that inbreeding is low in E. cunei- and 1415 h in E. cuneifolium, sampling the period of folium (FIS=0.078) andL. ohlingerae (FIS=0.144), and peak visitation in each species. Most observations of E. high in H. cumulicola (FIS=0.729). cuneifolium and H. cumulicola were made in 1995 at a site where the density of both species was high. Addi- tional observations of insect visitors to flowers of H. 2. Materials and methods cumulicola were made in 1996 at two populations with moderate and low densities of H. cumulicola. The 2.1. Floral biology and flowering phenology observations of H. cumulicola were made over 12days spanning May, June and July in 1995, and 3 days in The flowers of E. cuneifolium, H. cumulicola and L. August in 1996, totaling 765 min. Observations of E. ohlingerae are perfect; thus we examined the spatial and cuneifolium were made over 3 days in September, 1995 temporal arrangement of mature male and female sex- totaling 130 min. Visits to flowers of L. ohlingerae were ual parts within flowers. We define male sexual maturity so infrequent that timed watches were not practical; as the period from when anthers dehisce to when pollen instead, observations were made while pollinating flow- is no longer present or fresh, and female sexual maturity ers in the mating system experiment. The site at which as the period of stigmatic receptivity, during which these observations were made supports a relatively stigmas evolve bubbles of oxygen when submerged in large, dense population of L. ohlingerae. hydrogen peroxide (Kearns and Inouye, 1993). Peak stigmatic receptivity was determined by examining the 2.3. Mating system surface area and texture of stigmas, and the rate of oxygen evolution from samples of flowers of known age. We tested the mating systems of these plants using We observed the phenology of marked flowers of each controlled hand pollination experiments with five (E. species in early-flowering individuals in natural popula- cuneifolium and H. cumulicola) or four (L. ohlingerae) tions at Archbold Biological Station, Highlands pollination treatments during the 1995 flowering season. County, Florida (ABS). These observations, and obser- The treatments tested for spontaneous (autonomous) vations of plant- and population-level reproductive self-pollination, asexual reproduction in flowers (aga- phenology, contribute to our understanding of the mospermy), and self-compatibility, while cross and potential for pollen transfer within vs. among plants. open pollinations served for comparisons. These treat- ments were achieved using various combinations of 2.2. Pollination insect exclusion, emasculation, and hand-pollination (Table 2). It was not possible to emasculate flowers of L. We considered any insects observed on flowers ohlingerae because of their floral structure (see Section potential pollinators and recorded several aspects of 3), thus we did not test for agamospermy in this species. their behavior: their contact with male and female parts We relied on the spatial and temporal separation of of flowers; foraging and grooming behavior; the length sexes within flowers of L. ohlingerae to prevent pollen of their visit; the number of flowers or inflorescences contamination within flowers. We used plastic mesh visited; and the species of the next plant they visited. bags (Applied Extrusion Technologies, maximum mesh Rarely were observers able to record all of this infor- size 0.9 mm) to exclude insect visitors. Plants were cho- mation for a single visitor. Observers also described sen for the experiment from natural populations of each insect visitors to the best of their ability or collected species at ABS (20 plants of H. cumulicola,25ofL. voucher specimens. The insects were identified and ohlingerae and E. cuneifolium). curated by Dr. Mark Deyrup (Archbold Biological Sta- Inflorescences or flowering stalks were selected for the tion) and are housed in the insect collection at ABS. experiment according to their phenological state and 238 M.E.K. Evans et al. / Biological Conservation 111 (2003) 235–246

Table 2 reasons, we tested the effect of pollination treatment Manipulations used to create the five pollination treatments of the and plant identity (block) on fruit or seed set with a breeding system experiment generalized linear model assuming a binomial response Manipulation Pollination treatment variable, using a logit link function (PROC GENMOD, SAS Institute), and adjusting chi-squares for over- Auto Agam Self Cross Open dispersion (Crawley, 1993). In this test, the response Caging @@ @@ variable includes both a numerator (the number of suc- Emasculation @@ cessful fruits or seeds) and a denominator (the number Hand-pollination S O of treated carpels or flowers), as in a logistic regression. Caging excludes insect visitors and emasculation removes all anthers. We analyzed the data so that the sample size per plant Pollen from the same plant (S) or another plant (O) was used during and pollination treatment combination reflected the hand-pollinations. number of experimental units, avoiding pseudoreplica- tion. For example, the seed set data from multiple size and were randomly assigned a pollination treat- flowers of H. cumulicola per flowering branch were ment. The pollination treatments were applied to multi- combined, since this is the level at which treatments ple flowers per inflorescence or flowering stalk (Table 3). were applied. Some seeds of L. ohlingerae were eaten by Flowers were emasculated (E. cuneifolium and H. grasshoppers (Melanoplus tequestae, Acrididae), so cumulicola) before the anthers dehisced and were polli- heads with more than four damaged seeds or a total of nated during peak stigmatic receptivity with pollen from fewer than 10 seeds (including empty seeds) were exclu- first-day flowers that had been protected from insect ded from data analysis. We used nÀ1 a priori contrasts visitation or (rarely) from flowers that had clearly not (where n is the number of pollination treatments) to been visited. We applied pollen to stigmas until their detect pairwise differences among pollination treat- surface was saturated. Pollen donor plants for the out- ments. These contrasts were: auto vs. self, self vs. cross, cross treatment were at least 5 m from the maternal cross vs. open, and in the data sets with a treatment plants, in order to avoid biparental inbreeding. Fruits testing for agamospermy, auto vs. agam (see treatment were harvested as they matured. Flowers of E. cuneifo- labels in Figs. 1–3). lium and L. ohlingerae have two and one carpel, respec- tively, each with a single ovule. We pressed seeds gently to determine whether they contained endosperm: empty 3. Results seeds collapsed under such pressure. In these two spe- cies, seed set data were analyzed. Each flower of H. 3.1. Floral biology cumulicola has a compound ovary with many ovules. The contents of H. cumulicola fruits were classified into Each small, head-like umbel of Eryngium cuneifolium two categories: (1) seeds that appeared to be viable; and has 10–15 small, perfect flowers with five inconspicuous (2) seeds that appeared to be inviable, unfertilized, or white petals and five stamens. Flowers of E. cuneifolium aborted. In H. cumulicola, both fruit and seed set data produce nectar, pollen and odor. The flowers are pro- were analyzed. tandrous: the anthers emerge from the petals and The mating system experiment has a randomized dehisce throughout the first morning of flowering, and complete blocks design, with each pollination treatment the stamens abscise within two days. The flowers within represented at least once in each plant (i.e. block). The an umbel mature in rapid sequence, such that each fruit and seed set data could not be transformed to umbel passes through a male phase, then a brief sexu- approximate normality because of many zero values. ally inactive phase, and then a female phase. Each These zeros were concentrated in certain pollination inflorescence takes about 7–9 days to complete this treatments, causing heterogeneity of variance. For these sequence. Because there are numerous umbels per plant

Table 3 Statistics describing the number of flowers treated per plant in the mating system experiment

Pollination treatment Eryngium cuneifolium Hypericum cumulicola Liatris ohlingerae

Auto 22 (14–27, N=21) 45 (3–84, N=19) 23 (19–32, N=19) Agam 23 (12–28, N=20) 18 (5–34, N=20) Self 20 (8–33, N=20) 19 (8–34, N=20) 21 (10–49, N=18) Cross 19 (5–28, N=21) 16 (4–32, N=20) 22 (13–40, N=14) Open 21 (9–27, N=19) 130 (24–241, N=19) 25 (14–46, N=19)

The average number of flowers per plant and pollination treatment is followed in parentheses by the range of flower numbers and the number of plants. Pollination treatments are described in the text and correspond to labels in Table 2 and Figs. 1–3. M.E.K. Evans et al. / Biological Conservation 111 (2003) 235–246 239

Fig. 1. Seed set of Eryngium cuneifolium resulting from pollination treatments testing for autonomous pollination (AUTO), agamospermy (AGAM), and self-compatibility (SELF). Cross-pollinated (CROSS) and open-pollinated (OPEN) flowers serve for comparison. See Table 2 for manipulations. Both plant identity (F20, 164=32.25, P=0.0407) and pollination treatment (F4, 164=399.21, P<0.0001) significantly affected seed set.

(mean=121, range 1–1560, S.D.=192, n=225), flowers of each gender are generally present within a plant. Flowering occurs for about a month in September. Flowers of Hypericum cumulicola are also perfect, with small, bright yellow petals and numerous stamens (mean=27, S.D.=2.3, n=13). The flowers are borne in panicles of spikes. Pollen is the only reward for insect visitors. Each flower opens with the morning light and withers by late morning (around 1100 h). During this Fig. 2. Fruit (A) and seed (B) set of Hypericum cumulicola as a result interval, both male and female sexual parts are mature. of pollination treatments. Treatment labels are as in Fig. 1. Both plant

Thus, flowers of H. cumulicola lack the temporal identity (F19, 73=170.00, P<0.0001) and pollination treatment (F4, 73 separation of genders seen in flowers of E. cuneifolium =172.28, P<0.0001) affected fruit set in a highly significant manner, (Table 4). However, the genders are separated spatially whereas plant (F19, 51=29.47, P=0.0589) and pollination treatment (herkogamy). The styles (3–4) in flowers of H. cumuli- (F3, 51=7.02, P=0.0712) had only marginally significant effects on seed set. cola are typically bent at a 90 angle near the ovary, positioning the stigmas away from the upright-standing cluster of stamens. The degree of spatial separation is part of the stylar lobe emerges from the corolla mid-day variable among flowers within and among individuals. on the second day of flowering, and is receptive both that Flower production per plant can be quite large (up to day and the third day of flowering. Thus, flowers of L. 4218 flowers; Quintana-Ascencio, P. F., unpublished ohlingerae are both protandrous and herkogamous. data), but occurs over an extremely long flowering Flowers in male and female phase are present within season lasting from April to November. On all but the heads, and among heads within a plant. Like E. cuneifo- largest plants, from zero to a few flowers open per day. lium, flowering lasts for about a month, but in August. Peak flowering occurs in August. L. ohlingerae has from 20 to 25 tubular, bright purple, 3.2. Pollination perfect disk flowers per head, and several heads per plant. The flowers produce nectar and pollen and are The diversity of insects visiting flowers of E. cuneifo- typical of the Asteraceae. Five fused anthers surround lium was the striking feature of this plant’s pollination. the two-lobed style. The anthers dehisce on the first day This was obvious after observing flowers for a limited the flower opens, and the pollen is carried on the stylar time; thereafter, we focused our efforts on documenting lobes as they grow up and out of the corolla, emerging this diversity. We collected insects from five orders, 39 by late morning on the first day of flowering. The stig- families, 80 genera, and 101 species from E. cuneifolium matic surface is located on both margins of each stylar flowers (Appendix). Flowers of E. cuneifolium were vis- lobe below the region where the pollen is deposited. This ited frequently in the 130 min of timed observations: an 240 M.E.K. Evans et al. / Biological Conservation 111 (2003) 235–246

boscis, but did not appear to carry pollen on its body. It visited flowers throughout the day, including times that flowers had neither pollen nor receptive stigmas. Flowers of L. ohlingerae produce nectar; this, in addition to their color (purple) and morphology (tubu- lar) give the expectation that they are butterfly polli- nated (Faegri and van der Pijl, 1979). Butterflies were observed visiting the flowers, especially skippers, sulfurs, and swallowtails (Hesperiidae, Pieridae, and Papilioni- dae, respectively). Their visits were infrequent, mostly occurring between 1100 h and 1500 h. Pollen was visible on the butterflies’ hairy bodies, and even on their pro- boscides. The butterflies were often observed to move among plants of L. ohlingerae, rather than to flowers of other species, at this relatively high-density site.

Fig. 3. Seed set of Liatris ohlingerae as a result of pollination treat- 3.3. Mating system ments. Treatment labels are as in Fig. 1. Both plant identity (F20, 54 =39.78, P=0.0053) and pollination treatment (F3, 54=81.60, P<0.0001) significantly affected seed set. Pollination treatment had a highly significant effect on fecundity in all three species (Figs. 1–3; except average of 0.31 visits per min per plant. Most foraging Fig. 2b). Seed set in E. cuneifolium (4%, Fig. 1) and fruit visitors focused on nectar, but halictid and megachilid set in H. cumulicola (2%, Fig. 2a) were very low when bees were among the few observed to collect pollen. we tested for agamospermy by emasculating and isolat- These bees, as well as the syrphid flies, were the most ing flowers. Seed set in E. cuneifolium (4%, Fig. 1) and common visitors (88% of visits). In the dense popu- fruit set in H. cumulicola (7%, Fig. 2a) were also very lation where visitation was observed, insects often low when we tested for spontaneous self-pollination. moved among flowers on a plant, or traveled short dis- Seed (E. cuneifolium) and fruit set (H. cumulicola) in this tances (<5 m) when they moved among plants. second treatment was a small fraction of that when Flowers of H. cumulicola were visited by native soli- flowers were hand-pollinated with self pollen (6 and tary bees (Dialictus and Augochloropsis, Halictidae) 16%, respectively). Flowers given self pollen produced bumblebees (Bombus, Apidae), and a small bee-fly 94% of the seeds (E. cuneifolium, Fig. 1) and 98% of the (Geron sp., Bombyliidae). The average rate of visitation fruits (H. cumulicola, Fig. 2a) produced by flowers given was 0.11 visits per min per plant in 765 minutes of cross pollen. Seed and fruit set, respectively, was not observations. Visits by Dialictus constituted 81% of all significantly different between the self and cross treat- visits. Foraging by Dialictus, and the second most fre- ments in either of these species (F1, 164=0.76, P=0.38, quent visitor, Augochloropsis, left most flowers cleaned E. cuneifolium and F1, 73=0.19, P=0.66, H. cumulicola). of pollen within the first hour after opening. Individual While fruit set was not affected by pollen type in H. flowers lost most of their pollen after a single visit. cumulicola, seed set from outcross pollinations was greater Dialictus are small enough that they often did not con- than seed set from self pollinations (F1, 51=6.05, P=0.01) tact the stigmas while standing on the anthers gathering or open pollinations (F1, 51=3.43, P=0.06; Fig. 2b). pollen. When Dialictus were observed to contact stigmas Uncontrolled insect visitation resulted in high seed set it was often as they first arrived on the flower. Augo- (80%, E. cuneifolium) and moderate fruit set (44%, H. chloropsis and Bombus are large enough that they con- cumulicola)inthesetwospecies,eitherhigherthanthe tacted anthers and stigmas simultaneously. The cross treatment (E. cuneifolium, F1, 164=7.22, P=0.01) or bombyliid may have foraged for pollen using its pro- no different (H. cumulicola, F1, 73=0.27, P=0.61).

Table 4 Traits affecting the likelihood of self-pollination and predicted mating systems of three Florida scrub endemics

Trait Eryngium cuneifolium Hypericum cumulicola Liatris ohlingerae

Dichogamous (protandrous) Yes No Yes Herkogamous Yes Yes Yes Flowering flowers/plant Many Few Intermediate Autonomous pollination No Some (7% fruit set) No Self-compatible Yes Yes No Predicted mating system Mixed Mixed Outcrossing M.E.K. Evans et al. / Biological Conservation 111 (2003) 235–246 241

Flowers of L. ohlingerae isolated from insects were spatial structuring of allelic diversity is consistent with also unsuccessful (4% seed set, Fig. 3). In contrast to E. our observation of butterfly visitation to L. ohlingerae cuneifolium and H. cumulicola, seed production from flowers, since butterflies can carry pollen substantial hand-pollinations using self pollen was low in L. ohlin- distances. The two species with mixed mating systems, gerae (4%), 8% of what flowers given cross pollen pro- E. cuneifolium and H. cumulicola, have less allelic duced. The difference between self and cross treatments diversity and heterozygosity and more patchily dis- was highly significant (F1, 54=43.39, P<0.0001). In tributed allelic diversity. Allelic diversity is distributed addition, seed set was significantly lower in open polli- most patchily in H. cumulicola (FST=0.730), which we nated flowers than flowers given cross pollen (F1, 54 observed to be visited predominantly by small-bodied =5.46, P=0.02). solitary bees that are unlikely to transfer pollen over large distances. Boyle and Menges (2001) also reported that visitation of H. cumulicola was dominated by small 4. Discussion solitary bees. It is also possible that high differentiation among populations of H. cumulicola is generated by 4.1. Patterns of mating metapopulation dynamics (Quintana-Ascencio and Menges, 1996; Quintana-Ascencio et al., 1998). Patterns of reproduction, i.e. asexual vs. sexual and Superficially, the isozyme data conflict with our sug- selfing vs. crossing, have important consequences for the gestion that E. cuneifolium should show the highest amount and distribution of genetic variation (Hamrick degree of selfing. Instead, the isozyme data suggest that et al., 1991; Karron, 1991). These three herbaceous spe- selfing is most prevalent in H. cumulicola (FIS=0.729), cies showed little evidence of asexual reproduction or followed by L. ohlingerae (FIS=0.144), and last by E. spontaneous pollination within flowers (‘‘agam’’ and cuneifolium (FIS=0.078). However, this species-level ‘‘auto’’ treatments, Figs. 1–3). Reproduction without statistic obscures heterogeneity found among loci and insect visitors occurred at its highest rate in H. cumuli- populations of E. cuneifolium that is not found in the cola, in which male and female parts mature synchro- other two species. In all cases where genotype frequencies nously within flowers, but was still minimal (7% fruit of H. cumulicola and L. ohlingerae deviated significantly set, Fig. 2a). from random expectation, the deviation was positive, Limits to pollen transfer within plants increase from indicating inbreeding. But in E. cuneifolium, 75% of the E. cuneifolium to H. cumulicola to L. ohlingerae. The first cases of significant deviation from random expectation two are self-compatible: seed set of E. cuneifolium and were positive and 25% were negative (Dolan et al., fruit set of H. cumulicola in flowers given self pollen was 1999). Thus the low species-level value of FIS in E. cunei- indistinguishable from that in flowers given cross pollen. folium does not indicate a consistent pattern of out- Inflorescences of E. cuneifolium are sequentially male breeding. The heterogeneity of inbreeding vs. outbreeding and then female. This sort of pattern of synchronized in E. cuneifolium and the evidence of strong selfing in H. protandry is known in other members of the Apiaceae, cumulicola are surprising and worth further investigation. such as Aralia hispida (Thompson & Barrett, 1981; Bar- rett, 1984). But because plants of E. cuneifolium typically 4.2. Limits on seed production: pollen quantity and have many inflorescences, blooming within a short quality flowering season, there is ample opportunity for selfing. In contrast, on all but the largest plants of H. cumuli- Seed production may be limited by the number of cola, few flowers are open per plant, limiting opportu- pollen grains that reach stigmas (pollinator or pollen nities for pollen transfer within plants. Our data indicate limitation; Bierzychudek, 1981). Much evidence indi- that L. ohlingerae is self-incompatible: seed set from cates that pollinator limitation is a widespread phe- hand-pollination with self pollen was a small fraction of nomenon, though often variable (see the reviews of seed set from hand-pollination with cross pollen. Burd, 1994, and Larson and Barrett, 2000). All of the Genetically based mate incompatibility systems are plants we studied rely on insects for pollen transfer and known from many other species in the Asteraceae are subject to pollinator limitation. We did not detect (Richards, 1997). We conclude that E. cuneifolium and pollinator limitation in E. cuneifolium or H. cumulicola. H. cumulicola are likely to have mixed mating systems, Uncontrolled insect visitation led to high (80% seed set) whereas L. ohlingerae is obligately outcrossing (Table 4). or moderate (44% fruit set) fecundity in these two spe- Our results regarding patterns of mating and gene cies, respectively, which was not improved upon by flow via pollination are generally consistent with the hand pollination (Figs. 1 and 2). Both were observed to isozyme data for these species (Dolan et al., 1999). The have high insect visitation rates. In contrast, seed set in obligate outcrosser L. ohlingerae has higher allelic L. ohlingerae flowers given open access to insect visitors diversity, more heterozygosity, and less spatial structur- was significantly improved upon by hand-pollination ing of allelic diversity than the other two species. Low with cross pollen, and insect visitation rates were low. 242 M.E.K. Evans et al. / Biological Conservation 111 (2003) 235–246

Pollinator limitation or failure is thought to be more sion (Brown and Kephart, 1999; Fischer and Matthies, likely in plants that are pollinated by one or few insect 1997; Bosch and Waser, 1999). Self-incompatibility in species (Bond, 1994; Kearns and Inouye, 1997; Johnson L. ohlingerae could limit its seed productivity. We found and Steiner 2000; Spira, 2001). The diversity of floral no evidence of inbreeding depression in E. cuneifolium, visitors we observed decreased from E. cuneifolium to L. but seed set of H. cumulicola is higher given pure out- ohlingerae to H. cumulicola. E. cuneifolium seems to cross pollen compared to either self or open pollinated have a generalist pollination system, whereas L. ohlin- fruits (Fig. 2b). Inbreeding depression reduced seed gerae may rely upon one guild of insects: butterflies. We production in H. cumulicola by 13% [1Àws/wx; Eq. (1) caution that our data for these two species are pre- in Johnston and Schoen, 1994]. Potential limits on seed liminary. Visitation of Hypericum cumulicola flowers is production increase from E. cuneifolium to H. cumuli- dominated by Dialictus spp., among a handful of other cola to L. ohlingerae (Table 5). native solitary bees (this study and Boyle and Menges, The differences in reproductive biology that we dis- 2001). A study of the effectiveness of Dialictus bees vs. covered among these three rare species are consistent other, larger-bodied bees in causing fruit and seed set with selection for reproductive assurance, and illustrate a would resolve whether the pollination of H. cumulicola pattern of ‘‘compensation’’ described by Bond (1994). L. is as specialized as it appears. But the halictid bees that ohlingerae has a belowground organ (corm) that resists visit H. cumulicola are not specialized with respect to the fire-induced mortality and confers longer lifespan. This, plants they visit: they are known to visit flowers of other combined with its lack of specialization to post-fire open species (Deyrup, M., unpublished data). Generalization microhabitats, leads to less turnover of individuals per from the insect’s point of view should reduce the risk of unit time and hence the need for fewer seeds to maintain reciprocal declines between pollinators and plants, as a stable population size. This species has the most traits with generalization from the plant’s point of view that can limit seed production (self-incompatibility, pol- (Deyrup and Menges, 1997). lination by a limited guild of insects). The species with Pollinator limitation may also arise via low pollinator minimal belowground storage organs, E. cuneifolium and visitation rates in small, isolated, or low-density popula- H. cumulicola, are shorter-lived, are specialized to post- tions. Evidence has accumulated for these patterns in fire open microhabitats, rely on seeds to recover from plant populations (Schaal, 1978; Jennersten, 1988; fire, and fluctuate considerably with fire. These species Kunin, 1993, 1997; Aizen and Feinsinger, 1994; Agren, have high demographic dependence on seeds (E. cunei- 1996; Groom, 1998; Oostermeijer et al., 1998; Kearns et folium and H. cumulicola), and ‘‘compensate’’ for this al., 1998; Spira, 2001). In fact, there is evidence that visi- risk with low risk of pollinator failure (generalist polli- tation to flowers of H. cumulicola is positively density- nation in E. cuneifolium) or moderate dependence on dependent, and that seed set increases with visitation rate pollination for seed production (self-compatibility, (Boyle and Menges, 2001). Plant densities of E. cuneifo- though not autonomously pollinating). lium and H. cumulicola decline with time-since-fire (Menges, E. M., Quintana-Ascencio, P. F., unpublished 4.3. Implications for conservation data), suggesting that pollinator visitation could cycle with fire and plant population dynamics. Plant density is Our study of reproductive biology and the survey of chronically low in L. ohlingerae (Dolan et al., 1999; isozymes by Dolan et al. (1999) identify three genetic Menges et al., 2001), which is particularly problematic effects that could affect viability in the study species. since seed set with self pollen or without insects is so poor. Evolutionary responses to environmental change could Seed set can be limited by pollen quality, via a be constrained in H. cumulicola by its lack of genetic genetically based self-incompatibility system (Byers and diversity, and inbreeding depression in H. cumulicola Meagher, 1992; Byers, 1995; Ramsey and Vaughton, reduces its seed productivity. Successful matings in L. 2000; Wolf and Harrison, 2001) or inbreeding depres- ohlingerae require allelic diversity at the self-incompat-

Table 5 Influences on seed production in three Florida scrub endemics

Trait Eryngium cuneifolium Hypericum cumulicola Liatris ohlingerae

Fecundity in open pollinated flowers 80% (seed) 44% (fruit) 65% (seed) 33% (seed) Insect visitation rates High High Low Pollination Generalist Specialist (Dialictus spp.) Moderate specialist (butterflies) Plant density (plants/m2) 1.84 0.46 0.19 Ratio of self and cross pollinated treatments 0.94 (seed) 0.97 (fruit) 0.89* (seed) 0.08* (seed)

Both E. cuneifolium and L. ohlingerae have single-seeded fruits. Statistically significant differences between fecundity in the self vs. cross treatments are indicated with an asterisk. M.E.K. Evans et al. / Biological Conservation 111 (2003) 235–246 243 ibility locus, as in other rare, self-incompatible plants Packham, Helen Violi, Joyce Voneman, and Rebecca (Les et al., 1991; DeMauro, 1993; Byers, 1995; Weekley Yahr for field assistance. Maria Clauss, Eric Dyreson, Jill and Race, 2001; Wolf, 2001). The latter two problems Miller, Robert Steidl, and D. Lawrence Venable advised are most likely to threaten viability in small popu- our statistical analyses. We thank Pedro F. Quintana- lations. In such populations, plant numbers and hence Ascencio for sharing unpublished data, and Susan overall seed production is low, at the same time that Kephart, Michael S. Singer, Mark Schwartz, and one inbreeding increases. Small populations are expected to anonymous reviewer for helpful comments on the lose self-incompatibility (SI) alleles. Populations of H. manuscript. This work was funded by The Nature cumulicola and L. ohlingerae are currently most threa- Conservancy’s Ecosystem Research Program, the Florida tened by habitat loss and fire suppression, which can be Chapter of The Nature Conservancy, the Florida Division reversed by habitat preservation and fire management. of Forestry, and Archbold Biological Station. Our study of reproductive biology also reveals certain risks for demographic viability. Seeds are critical for persistence of E. cuneifolium and H. cumulicola popula- Appendix tions, since fire kills all or most plants. Seed set in these two species may be increasingly limited by pollinator Insect visitors to flowers of three plants endemic to abundance or behavior with time-since-fire, as discussed Florida scrub: Eryngium cuneifolium, Hypericum cumu- above. H. cumulicola may be particularly vulnerable to licola, and Liatris ohlingerae. pollinator limitation, since its flowers are visited by a In parentheses are the initials of the collector, either few species of native solitary bees. The threats to L. MD (Dr. M. Deyrup, Archbold Biological Station) or ohlingerae are different. Individuals of L. ohlingerae ME (M. E. K. Evans, University of Arizona). often survive fire, and population declines between fires Eryngium cuneifolium are less rapid, but pollinator limitation may be chronic. COLEOPTERA Though less dramatic, these declines are also serious, CANTHARIDAE since plant density and seed production are chronically Chauliognathus marginatus (MD) low and seedling recruitment is episodic. Factors influ- CERAMBYCIDAE encing recruitment in L. ohlingerae should be a focus of Strangalia sexnotata (MD) further investigation. Prescribed fire programs tailored COCCINELLIDAE to the population dynamics of these plants and reserve Exochima marginipennis (ME) systems designed to accommodate extinction-recoloni- CUCURLIONIDAE zation dynamics should address both the positive and Notolomus basalis (ME) negative effects of fire on these plants. Odontocorynus pulverulentus (ME) Although the species we studied are all herbaceous MELOIDAE perennials subject to natural disturbance via fire in Pyrota sp. (MD) rosemary scrubs, our study reveals species-specific risks P. lineata (ME) to population viability related to reproductive biology. SCARABAEIDAE Based on this work, we suggest that reproductive biol- Trigonopeltastes delta (MD, ME) ogy and concomitant risks to population viability can- DIPTERA not be predicted from coarse-grained similarities among BIBIONIDAE species, in this case, from habit, shared habitat, and Plecia nearctica (MD, ME) shared disturbance regime. Single conservation strate- BOMBILIIDAE gies for multiple sympatric species that are grouped Chrysanthrax dispar (MD) based on a few such coarse-grained characteristics are C. mira (MD) unlikely to be successful. Instead, knowledge of several Exprosopa fasciata (MD, ME) aspects of population biology is necessary to predict E. fascipennis (MD) risks to population viability of rare species. The sheer Geron sp. (MD, ME) number and distribution of rare species requires that G. vitripennis (MD) common conservation strategies be sought, hence further Paravilla separata (MD) investigation should focus on whether strategies based on Poecilognathus sp. (MD) multiple aspects of population biology can be successful. Systoechus solitus (MD) Villa sp. (MD) CALLIPHORIDAE Acknowledgements Cochliomyia macellaria (MD, ME) Chrysomya sp. (ME) The authors thank Scott Bergen, Dawn Berry, Sonia CHLOROPIDAE Honeydew, George Landman, Margaret Mayfield, Helen Liohippelates pusio (MD) 244 M.E.K. Evans et al. / Biological Conservation 111 (2003) 235–246

Olcella sp. (MD) Oxybelus emarginatus (ME) Thaumatomyia sp. (ME) EUCHARITIDAE CONOPIDAE Orasema sp. (ME) Pysoconops brachyrhynchus (MD) EUMENIDAE P. fronto (MD) Euodynerus castigatus (ME) Zodion americanum (ME) E. boscii (MD) EUPHYDRIDAE Pachydynerus erynnis (ME) Hydrochasma incisum (MD) HALICTIDAE MILICHIIDAE Augochlora pura (ME) Milichiella sp. (MD) Augochlorella aurata (ME) Paramyia nitens (MD) Augochloropsis sumptuosa (ME) MUSCIDAE Dialictus miniatulus (MD) Atherigona orientalis (MD) D. nymphalis (ME) OTITIDAE D. placidensis (MD) Euxesta sp. (MD) Sphecodes heraclei (MD) SARCOPHAGIDAE LEUCOSPIDIDAE Gymnprosopa sp. (MD) Anacrabro ocellatus (MD) Oxysarcodexia sp. (MD) Leucospis affinis (MD) Sarcodexia sternodontis (MD) L. robertsoni (MD) Senotania rubriventris (ME) L. slossonae (ME) S. trilineata (MD) MEGACHILIDAE SEPSIDAE Anthidiellum perplexum (ME) Palaeosepsis pusio (MD) Coelioxys mexicana (MD) STRATIOMYIDAE C. octodentata (ME) Hedriomorpha sp. (MD) C. sayi (MD) SYRPHIDAE Megachile albitarsis (MD) Copestylum pusillum (MD, ME) M. brevis (ME) Palpada agrorum (MD, ME) M. mendica (MD) P. vinetorum (MD, ME) M. texana (ME) P. pusilla (ME) NYSSONIDAE Pseudodoros clavatus (ME) Bembix sayi (MD) Toxomerus floralis (MD) Bicyrtes capnoptera (MD) TACHINIDAE SCOLIIDAE Archtas sp. (MD) Campsomeris plumipes (MD) Crocinosoma cernuale (MD) SPHECIDAE Gnadochaeta sp. (MD) Isodontia exornata (MD) Gonia sp. (MD) TIPHIIDAE Masiphya sp. (MD) Myzinum carolinianum (MD) Paradidyma melania (MD) M. maculatum (MD) Phytomyptera sp. (MD) M. sp. (MD) Ptilodexia sp. (MD) LEPIDOPTERA Trichopoda pennipes (MD) ARCTIIDAE Vanderwulpia sequens (MD) Cisseps fulvicollis (MD) HETEROPTERA Cisthene subjecta (ME) CORIMELAENIDAE Utetheisa bella (MD) Corimelaena lateralis (ME) HESPERIDAE PHYMATIDAE Euphyes arpa (ME) Phymata sp. (ME) LYCAENIDAE HYMENOPTERA Hemiargus ceraunus (MD) APIDAE Strymon melinus (MD) Bombus pennsylvanicus (ME) NYMPHALIDAE COLLETIDAE Junonia coenia (ME) Colletes sp. (MD) PAPILIONIDAE C. mandibularis (ME) Eurytides marcellus (MD) CRABRONIDAE PIERIDAE Ectemnius decemmaculatus (ME) Eurema daira (MD) M.E.K. Evans et al. / Biological Conservation 111 (2003) 235–246 245

Hypericum cumulicola Byers, D.L., Meagher, T.R., 1992. Mate availability in small popula- DIPTERA tions of plant species with homomorphic sporophytic self-incom- BOMBILIIDAE patibility. Heredity 68, 353–359. Byers, D.L., Waller, D.M., 1999. Do plant populations purge their Geron sp. (ME) genetic load? effects of population size and mating history on G. vitripennis (MD) inbreeding depression. Annual Review of Ecology and Systematics HYMENOPTERA 30, 479–513. APIDAE Charlesworth, D., Charlesworth, B., 1987. Inbreeding depression and Bombus pennsylvanicus (ME) its evolutionary consequences. Annual Review of Ecology and Sys- tematics 18, 237–268. HALICTIDAE Christman, S.P., Judd, W.S., 1990. Notes on plants endemic to Florida Augochloropsis sumptuosa (ME) scrub. Florida Scientist 53, 52–73. Dialictus miniatulus (MD) Crawley, M.J., 1993. GLIM for Ecologists. Blackwell Scientific, D. nymphalis (MD) Oxford, UK. D. placidensis (ME) DeMauro, M.M., 1993. Relationship of breeding system to rarity in the lakeside daisy (Hymenoxys acaulis var. glabra). Conservation Biology 7, 542–550. Liatris ohlingerae Deyrup, M., Menges, E.S., 1997. Pollination ecology of the rare scrub COLEOPTERA mint Dicerandra frutenscens (Lamiaceae). Florida Scientist 60, 143– MELYRIDAE 157. Collops nigriceps (MD) Dobson, A.P., Rodriguez, J.P., Roberts, W.M., Wilcox, D.S., 1997. Geographic distribution of endangered species in the United States. DIPTERA Science 275, 550–553. BOMBYLIIDAE Dolan, R.W., Yahr, R., Menges, E.S., Halfhill, M.D., 1999. Conserva- Poecilognathus sulphuea (MD) tion implications of genetic variation in three rare species endemic to HYMENOPTERA Florida scrub. American Journal of Botany 86, 1556–1562. HALICTIDAE Ellstrand, N.C., Elam, D.R., 1993. Population genetic consequences of small population size: implications for plant conservation. Annual Dialictus nymphalis (MD) Review of Ecology and Systematics 24, 217–242. LEPIDOPTERA Faegri, K., van der Pijl, L., 1971. The Principles of Pollination Ecol- HESPERIIDAE ogy, second ed. Pergamon Press, Oxford, UK. Atrytone delaware (MD) Fischer, M., Matthies, D., 1997. Mating structure and inbreeding and PIERIDAE outbreeding depression in the rare plant Gentianella germanica (Gentianaceae). American Journal of Botany 84, 1685–1692. Phoebis sennae (MD) Fisher, R.A., 1958. The Genetical Theory of Natural Selection, second revised ed. Dover Publications, New York, NY. Groom, M.J., 1998. Allee effects limit population viability of an annual plant. The American Naturalist 151, 487–496. References Hamrick, J.L., Godt, M.J.W., Murawski, D.A., Loveless, M.D., 1991. Correlations between species traits and allozyme diversity: implica- Agren, J., 1996. Population size, pollinator limitation, and seed set tions for conservation biology. In: Falk, D.A., Holsinger, K.E. in the self-incompatible herb Lythrum salicaria. Ecology 77, 1779– (Eds.), Genetics and Conservation of Rare Plants. Oxford Uni- 1790. versity Press, New York, pp. 75–86. Aizen, M.A., Feinsinger, P., 1994. Forest fragmentation, pollination, Hamrick, J.L., Godt, M.J.W., 1996. Effects of life history traits on and plant reproduction in a Chaco dry forest, Argentina. Ecology genetic diversity in plant species. Philosophical Transactions of the 75, 330–351. Royal Society of London B 351, 1291–1298. Barrett, S.C.H., 1984. Variation in floral sexuality of diclinous Aralia Herndon, A., 1997. Life history studies in the scrub endemic Liatris (Araliaceae). Annals of the Missouri Botanical Garden 71, 278–288. ohlingerae. Unpublished Report to the Florida Game and Fresh- Bierzychudek, P., 1981. Pollinator limitation of plant reproductive water Fish Commission. Tallahassee, FL, USA. effort. The American Naturalist 117, 838–840. Huenneke, L.F., 1991. Ecological implications of genetic variation in Bond, W.J., 1994. Do mutualisms matter? assessing the impact of pol- plant populations. In: Falk, D.A., Holsinger, K.E. (Eds.), Genetics linator and disperser disruption on plant extinction. Philosophical and Conservation of Rare Plants. Oxford University Press, New Transactions of the Royal Society of London Series B 344, 83–90. York, pp. 31–44. Bosch, M., Waser, N.M., 1999. Effects of local density on pollination Husband, B.C., Schemske, D.W., 1996. Evolution of the magnitude and reproduction in Delphinium nuttallianum and Aconitum colum- and timing of inbreeding depression in plants. Evolution 50, 54–70. bianum (Ranunculaceae). American Journal of Botany 86, 871–879. Jennersten, O., 1988. Pollination in Dianthus deltoides (Car- Boyle, O.D., Menges, E.S., 2001. Pollinator visitation to Hypericum yophyllaceae): effects of habitat fragmentation on visitation and cumulicola (Hypericaceae), a rare Florida scrub endemic. Florida seed set. Conservation Biology 2, 359–366. Scientist 64, 107–117. Johnson, S.D., Steiner, K.E., 2000. Generalization versus specializa- Brown, E., Kephart, S., 1999. Variability in pollen load: implications tion in plant pollination systems. Trends in Ecology and Evolution for reproduction and seedling vigor in a rare plant, Silene douglasii 15, 140–143. var. oraria. International Journal of Plant Sciences 160, 1145–1152. Johnston, M.D., Schoen, D.J., 1994. On the measurement of inbreed- Burd, M., 1994. Bateman’s Principle and plant reproduction—the role of ing depression. Evolution 48, 1735–1741. pollen limitation in fruit and seed set. Botanical Review 60, 83–139. Karron, J.D., 1991. Genetic variation and breeding systems in rare Byers, D.L., 1995. Pollen quantity and quality as explanations for low plants. In: Falk, D.A., Holsinger, K.E. (Eds.), Genetics and Con- seed set in small populations exemplified by Eupatorium (Asteraceae). servation of Rare Plants. Oxford University Press, New York, pp. American Journal of Botany 82, 1000–1006. 87–98. 246 M.E.K. Evans et al. / Biological Conservation 111 (2003) 235–246

Kearns, C.A., Inouye, D.W., 1993. Techniques for Pollination Biolo- Peroni, P.A., Abrahamson, W.G., 1985. A rapid method for deter- gists. University Press of Colorado, Niwot, Colorado, USA. mining losses of native vegetation. Natural Areas Journal 5, 20–24. Kearns, C.A., Inouye, D.W., 1997. Pollinators, flowering plants, and Quintana-Ascencio, P.F., Dolan, R.W., Menges, E.S., 1998. Hypericum conservation biology. Bioscience 47, 297–307. cumulicola demography in unoccupied and occupied Florida scrub Kearns, C.A., Inouye, D.W., Waser, N.M., 1998. Endangered mutu- patches with different time-since-fire. Journal of Ecology 86, 640–651. alisms: the conservation of plant-pollinator interactions. Annual Quintana-Ascencio, P.F., Menges, E.S., 1996. Inferring metapopulation Review of Ecology and Systematics 29, 83–112. dynamics from patch-level incidence of Florida scrub plants. Con- Kunin, W.E., 1993. Sex and the single mustard: population density servation Biology 10, 1210–1219. and pollinator behavior effects on seed set. Ecology 74, 2145–2160. Quintana-Ascencio, P.F., Menges, E.S., 2000. Competitive abilities of Kunin, W.E., 1997. Population size and density effects in pollination: three narrowly endemic plant species in experimental neighborhoods pollinator foraging and plant reproductive success in experimental along a fire gradient. American Journal of Botany 87, 690–699. arrays of Brassica kaber. Journal of Ecology 85, 225–234. Quintana-Ascencio, P.F., Morales-Hernandez, M., 1997. Fire-medi- Lamont, B.B., Klinkhamer, P.G.L., Witkowski, E.T.F., 1993. Popu- ated effects of shrubs, lichens and herbs on the demography of lation fragmentation may reduce fertility to zero in Banksia goodii— Hypericum cumulicola in patchy rosemary scrub. Oecologia 112, a demonstration of the Allee effect. Oecologia 94, 446–450. 263–271. Lande, R., 1995. Mutation and conservation. Conservation Biology 9, Ramsey, M., Vaughton, G., 2000. Pollen quality limits seed set in 782–791. Burchardia umbellata (Colchicaceae). American Journal of Botany Larson, B.M.H., Barrett, S.C.H., 2000. A comparative analysis of 87, 845–852. pollen limitation in flowering plants. Biological Journal of the Lin- Richards, A.J., 1997. Plant Breeding Systems. Second ed. Chapman & nean Society 69, 503–520. Hall, London, UK. Les, D.H., Reinartz, J.A., Esselman, E.J., 1991. Genetic consequences Schaal, B.A., 1978. Density dependent foraging on Liatris pycnos- of rarity in Aster furcatus (Asteraceae), a threatened, self-incompa- tachya. Evolution 32, 452–454. tible plant. Evolution 45, 1641–1650. Schemske, D.W., Husband, B.C., Ruckelshaus, M.H., Goodwillie, C., Lynch, M., Conery, J., Burger, R., 1995. Mutational meltdowns in Parker, I.M., Bishop, J.G., 1994. Evaluating approaches to the sexual populations. Evolution 49, 1067–1080. conservation of rare and endangered plants. Ecology 75, 584–606. Mawdsley, N.A., Compton, S.G., Whittaker, R.J., 1998. Population Spira, T.P., 2001. Plant-pollinator interactions: a threatened mutual- persistence, pollination mutualisms, and figs in fragmented tropical ism with implications for the ecology and management of rare landscapes. Conservation Biology 12, 1416–1420. plants. Natural Areas Journal 21, 78–88. Menges, E.S., 1999. Ecology and conservation of Florida scrub. In: Steiner, K.E., Whitehead, V.B., 1996. The consequences of specializa- Anderson, R.C., Fralish, J.S., Baskin, J. (Eds.), The Savanna, Bar- tion for pollination in a rare South African shrub, Ixianthes ren, and Rock Outcrop Communities of North America. Cambridge retziodes (Scrophulariaceae). Plant Systematics and Evolution 201, University Press, Cambridge, pp. 7–22. 131–138. Menges, E.S., Dolan, R.W., Yahr, R., Gordon, D.R., 2001. Com- Thompson, J.D., Barrett, S.C.H., 1981. Temporal variation of gender parative genetics of seven plants endemic to Florida’s Lake Wales in Aralia hispida Vent. (Araliaceae). Evolution 35, 1094–1107. Ridge. Castanea 66, 98–114. US Fish and Wildlife Service, 1999. South Florida Multi-species Menges, E.S., Hawkes, C.V., 1998. Interactive effects of fire and Recovery Plan. USFWS, Atlanta. microhabitat on plants of Florida scrub. Ecological Applications 8, Weekley, C.W., Race, T., 2001. The breeding system of Ziziphus celata 935–946. Judd and D. W. Hall (Rhamnaceae), a rare endemic of the Lake Menges, E.S., Kimmich, J., 1996. Microhabitat and time since fire: Wales Ridge, Florida, USA: implications for recovery. Biological effects on demography of Eryngium cuneifolium (Apiaceae), a Flor- Conservation 100, 207–213. ida scrub endemic plant. American Journal of Botany 83, 185–191. Wolf, A., 2001. Conservation of endemic giants in serpentine land- Menges, E.S., Kohfeldt, N.M., 1995. Life history strategies of Florida scapes. Biological Conservation 100, 35–44. scrub plants in relation to fire. Bulletin of the Torrey Botanical Club Wolf, A.T., Harrison, S.P., 2001. Effects of habitat size and patch 122, 282–297. isolation on reproductive success of the serpentine morning glory. Oostermeijer, J.G., Luijten, S.H., Krenova, Z.V., den Nijs, H.C.N., Conservation Biology 15, 111–121. 1998. Relationships between population and habitat characteristics Young, C.C., Menges, E.S., 1999. Gap-phase dynamics and micro- and reproduction of the rare Gentiana pneumonanthe L. Conserva- habitat specialization: post-fire regeneration of scrubby flatwoods tion Biology 12, 1042–1053. communities on the Lake Wales Ridge. Florida Scientist 62, 1–10.