ABSTRACT

WALL, WADE ALAN. Population Genetics and Demography of Two Rare Species Endemic to the Longleaf Ecosystem. (Under the direction of William A. Hoffmann and Thomas R. Wentworth.)

Astragalus michauxii and brevifolia are two rare plant species endemic to the Fall‐line

Sandhills region of the Gulf and Atlantic Coastal Plain in the southeastern United States that are currently considered vulnerable to extirpation. The Fall‐line Sandhills, a large area of relictual dunes, are part of the longleaf pine ecosystem, a temperate savanna dominated by Pinus palustris and maintained by frequent fire. The longleaf pine ecosystem covered 37 million hectares across the southeastern United States at the time of European settlement, but land use changes led to the loss or degradation of 97% of the original area. In this dissertation, we used a variety of methods to better understand the biogeography, population genetic structure, and the effects of fire on the population dynamics of A. michauxii and P. brevifolia. Pyxidanthera brevifolia () is an evergreen subshrub that occurs on xeric ridgetops in the Fall‐line Sandhills of North Carolina and

South Carolina (USA), with the majority of identified populations occurring on Fort Bragg Military

Installation, NC. Fort Bragg is also the only location for the other taxon within the genus, P. barbulata. We developed a germination protocol for P. brevifolia, testing for effects of light and temperature on germination success and rates. Both rates were highest under conditions of low temperature and low light. We used AFLP and cpDNA genetic markers developed for P. brevifolia to assess the taxon’s recent phylogeogeography and taxonomic relationship to P. barbulata. Results indicated there is significant morphological overlap between the two taxa very little genetic differentiation. In addition, evidence suggested that the two taxa did not exhibit any evidence of a range shift following the Pleistocene and that the northern populations of Pyxidanthera were most likely present during the Pleistocene. To investigate the population genetic variation of Astragalus michauxii, we developed eight microsatellites and genotyped 355 individuals across 22 populations.

Genetic evidence indicates that within population genetic variation accounts for 92% of the total observed genetic variation and that the species encountered a genetic bottleneck within the past.

To explore the influence of fire on the population dynamics of A. michauxii and P. brevifolia, we inventoried established demographic monitoring plots from 2007‐2010. Demographic modeling demonstrated that fire negatively affected both species in the short term by increasing mortality of smaller individuals and reducing fruit set. Results indicated that under simulated annual burning, both species would have reduced population growth rates, and that the “ideal” fire return interval may be 2‐4 years. In summary, while anecdotal evidence suggests that fire is indirectly necessary to maintain an open habitat, A. michauxii and P. brevifolia did not respond as “positively” to fire as might be expected for two species that are endemic to a fire‐dependent ecosystem. However, evidence based on climate data and pollen records indicates that the Fall‐line Sandhills were much colder and drier during the Pleistocene and that there has been a substantial shift in the dominant vegetation. If the two taxa were present in the Fall‐line Sandhills since at least the Pleistocene, they would have experienced much colder and drier conditions and, based on the inferred species composition, a much less frequent fire return interval. In conclusion, while fire may be necessary to maintain an open habitat that may have formerly been maintained climatically, it is not apparent that the two species are “fire‐adapted” in the narrow sense of this term and that a more nuanced use of the concept of fire adaptation may be appropriate. Population Genetics and Demography of Two Rare Plant Species Endemic to the Longleaf Pine Ecosystem

by Wade Alan Wall

A dissertation submitted to the Graduate Faculty of North Carolina State University in partial fulfillment of the requirements for the degree of Doctor of Philosophy

Plant Biology

Raleigh, North Carolina

2013

APPROVED BY:

______William A. Hoffmann Thomas R. Wentworth Committee Co‐Chair Committee Co‐Chair

______Kevin Gross Ignazio Carbone BIOGRAPHY

Wade Wall was born in Asheville, NC and spent his formative years in Salt Lake City, UT, and

Marion, NC. After attending the University of North Carolina, Chapel Hill, he spent a number years doing a wide range of jobs, including building treehouses, teaching ethics to prisoners, and teaching

English to non‐native speakers. In 2005, Wade returned to school at North Carolina State University, obtaining a Master of Science degree in Botany in 2008. He currently lives in Champaign, IL with his wife, Katie, and his son, Asa.

ii ACKNOWLEDGMENTS

First and foremost, I thank Kathleen Bernadette Coyle, my wife, for her support and understanding during these past few years. Without her by my side I would not have been able to complete this process. My mother and father, Louis and Edna Wall, have always been there for me and I appreciate what they have taught me about how to live a life.

Drs. Thomas R. Wentworth and William A. Hoffmann have been sources of inspiration and great role models my years at North Carolina State University. My graduate advisory committee has also been supportive and understanding.

Many past and present students of the Plant Biology Department at NC State University have been influential and supportive. In particular, Andy Walker, Renee Marchin, and Kristen Kostelnik provided comic relief and emotional support through the whole process. I will always remember my days in the field with Andy Walker, who will remain a lifelong friend whether he likes it or not.

Norm Douglas and Jenny Xiang have been overly patient, teaching me population genetic laboratory and data analysis techniques.

The support of Sue Vitello and the Plant Biology departmental staff has been tremendous; without their assistance no graduate student would be able to complete a degree. We all owe them a debt of gratitude that can never be repaid.

I am grateful to Matthew Cleary, Brendan Dawal, Jacob Hilton, and Sherrie Emerine for excellent assistance in the field and in the laboratory. They definitely made field work more entertaining.

iii Finally, thanks to Matthew G. Hohmann, United States Army Corps of Engineers (ERDC‐CERL), and

Janet Gray, Endangered Species Branch, Fort Bragg Military Installation. Without their unwavering support and knowledge, this project would never have been completed.

iv TABLE OF CONTENTS

LIST OF TABLES .……………………………………………………………………………………………………………………………...vii LIST OF FIGURES ...... x CHAPTER 1 OVERVIEW ……………………………………………………………………………………………………………………..1 1.1 Introduction ...... 1 1.2 References ...... 6 CHAPTER 2 EFFECTS OF LIGHT AND TEMPERATURE ON GERMINATION OF PYXIDANTHERA BREVIFOLIA WELLS (DIAPENSIACEAE) ……………..………………………………………………………………………………..9 2.1 Abstract ...... 9 2.2 Introduction ...... 10 2.3 Methods ...... 13 2.4 Results ...... 14 2.5 Discussion ...... 15 2.6 References ...... 19 CHAPTER 3 EVIDENCE FOR RANGE STASIS DURING THE LATTER PLEISTOCENE FOR THE ATLANTIC COASTAL PLAIN ENDEMIC GENUS, PYXIDANTHERA MICHAUX…………………………………………………………24 3.1 Abstract ...... 24 3.2 Introduction ...... 25 3.3 Methods ...... 29 3.3.1 Sampling and Morphological Measurements ...... 29 3.3.2 Molecular methods ...... 30 3.3.3 cpDNA data analysis ...... 31 3.3.4 AFLP data analysis ...... 34 3.4 Results ...... 35 3.4.1 Morphology ...... 35 3.4.2 cpDNA ...... 36 3.4.3 AFLP ...... 38 3.5 Discussion ...... 39

v 3.5.1 ...... 39 3.5.2 Phylogeography of the genus Pyxidanthera ...... 41 3.6 References ...... 45 CHAPTER 4 EVIDENCE OF POPULATION BOTTLENECK IN THE ENDEMIC PLANT SPECIES ASTRAGALUS MICHAUXII (KUNTZE) F.J. HERM. …………….………………………………………………………………………………………71 4.1 Abstract ...... 71 4.2 Introduction ...... 72 4.3 Methods ...... 74 4.3.1 Population genetic methods ...... 74 4.3.2 Genetic structure and diversity ...... 75 4.3.3 Evidence of genetic bottlenecks across multiple temporal scales ...... 77 4.3.4 Estimating gene flow between populations ...... 78 4.4 Results ...... 79 4.5 Discussion ...... 81 4.6 References ...... 86 CHAPTER 5 DEMOGRAPHIC EFFECTS OF FIRE ON TWO ENDEMIC PLANT SPECIES IN THE LONGLEAF PINE‐WIREGRASS ECOSYSTEM...... 102 5.1 Abstract ...... 102 5.2 Introduction ...... 103 5.3 Methods ...... 105 5.3.1 Study area and species ...... 105 5.3.2 Field Methods ...... 106 5.3.3 Vital rates data analysis ...... 107 5.3.4 Matrix construction and analysis ...... 108 5.4 Results ...... 111 5.4.1 Effects of plant size and fire on vital rates ...... 111 5.4.2 Population growth rates, elasticities, and LTRE in relation to fire ...... 113 5.5 Discussion ...... 114 5.6 References ...... 118

vi LIST OF TABLES

Table 3.1: Chloroplast haplotype accession numbers as archived in Genbank for the atpI‐atpH intergenic spacer region (partial sequence) and the psbD‐trnT intergenic spacer region (partial sequence). Chloroplast haplotypes H1 through H12 refer to unique composite sequences from the two intergenic spacer regions...... 57 Table 3.2: Polymorphisms of the 12 cpDNA haplotypes based on the cpDNA regions atpI‐atpH and psbD‐trnT in the genus Pyxidanthera. Numbers below cpDNA regions denote position in sequence. N refers to the number of individuals identified as the corresponding haplotype, while dashes represent correspondence to consensus haplotype...... 58 Table 3.3: Analyses of Molecular Variance (AMOVA) results for using cpDNA sequences and AFLP markers. *** indicates p‐value < 0.001, ** p‐value < 0.01, * p‐value < 0.05, and NS indicates non‐significance of variation...... 59 Table 3.4: Genetic diversity indices for P. barbulata and P. brevifolia based on cpDNA sequences and AFLP markers. AFLP genetic diversity indices were only calculated for populations with more than 7 genotyped individuals (437 total specimens). %P represents the

number of polymorphic loci, DW is a measure of rare alleles per population, and He is a measure of expected heterozygosity based on the AFLP markers. π is a measure of cpDNA nucleotide diversity. N represents the number of specimens for each population for AFLP markers and cpDNA sequences (in parentheses)...... 60 Table 3.5: Summary of model statistics for the 24 IMa2 models. Included for each model (left to right) are the negative log of the probability, the number of parameters, the degrees of freedom when compared to the full model, AIC and ΔAIC, the likelihood of the model, the model probability, and the evidence ratio for each model, calculated according to Burnham and Anderson (2002)...... 62 Table 4.1: Eight polymorphic loci identified and developed for Astragalus michauxii. Column are primer pair name, sequence, repeat, microsatellite range (with fragment size in parentheses), and number of alleles observed...... 91 Table 4.2: Proportion of linkage disequilibrium (LD) tests that detected significant LD for eight polymorphic microsatellite loci...... 92 Table 4.3: Genetic variation in Astragalus michauxii populations from North Carolina and Georgia based on eight polymorphic microsatellite loci. Column headings are: N, number of

individuals; A, average number of alleles across loci; AR, average allelic richness; P,

number of private alleles; PR, private allelic richness; HO, observed heterozygosity; HE,

vii expected heterozygosity; H‐W disequilibrium, loci identified as not in Hardy‐Weinberg equilibrium...... 93 Table 4.4: Analysis of molecular variance (AMOVA) results for Astragalus michauxii populations from North Carolina and Georgia (USA). Populations were defined by Natural Heritage Program protocols. Grouping of populations into regions follows Table 2...... 94 Table 4.5: Tests for genetic bottlenecks in Astragalus michauxii using Bottleneck version 1.2 in populations with > 20 gene copies and M ratio for all populations as calculated in Arlequin 3.1. Results indicate no recent genetic bottleneck events, but evidence of a severe genetic bottleneck in the more distant past...... 95 Table 4.6: A. michauxii individuals identified by GeneClass 2 as being the result of possible interpopulation gene flow (p‐value >= 0.001). Fourteen individuals across eleven populations were identified...... 96 Table 5.1: Number of A. michauxii individuals used to estimate survivorship, transition probabilities, and reproductions for each size class. TSB refers to time since burn, with “0” indicating year that population was burned...... 122 Table 5.2: Transition matrices for A. michauxii individuals that were recently burned, 1 year post‐ fire, and 2 years post‐fire. Stages delineated based on the tallest stem: small (0.01 – 20 cm), small‐medium (>20‐40 cm), medium (>40‐80 cm), and large (> 80 cm)...... 123 Table 5.3: Number of P. brevifolia individuals used to estimate survivorship, transition probabilities, and reproductions for each size class. Populations were sampled before the burning season, so TSB = “1” indicates first measurements after population was burned...... 124 Table 5.4: Transition matrices for P. brevifolia during the 2008‐2009 and 2009‐2010 transition intervals. Seedling class is an age class, while the other 10 size classes are defined by plant area (cm2)...... 125 Table 5.5: Elasticity values for A. michauxii transition matrices. Size classes are based on tallest stem and are small (0.1 – 20 cm), small‐medium (>20‐40 cm), medium (>40‐80 cm), and large (> 80 cm)...... 127 Table 5.6: Elasticity estimates for P. brevifolia during the 2008‐2009 and 2009‐2010 transition intervals. Seedling class is an age class, while the other 10 size classes are defined by plant area (cm2)...... 128 Table 5.7: Population growth rates, stable stage distributions, and reproductive values for the three transition matrices (burned, one year post‐fire, and two or more years post‐fire) estimated for A. michauxii. Stable stage distribution values represent proportions and

viii sum to 1; reproductive values are scaled to the smallest size class, which is equal to 1...... 130 Table 5.8: Population growth rates, stable stage distributions, and reproductive values for the 6 transition matrices (2 time steps and burned, 1 year post‐fire, and 2 years post‐fire) estimated for P. brevifolia. Stable stage distribution values represent proportions and sum to 1; reproductive values are scaled to the smallest size class, which is equal to 1. Stable stage distribution does not include the seed stage and the reproductive values are scaled to seedling stage...... 131

ix LIST OF FIGURES

Fig. 2.1 Winter‐flowering Pyxidanthera brevifolia in full bloom on Fort Bragg (left) and ex situ propagated P. brevifolia seedling (right). Left photo taken 25 March 2009...... 22 Fig. 2.2: Effects of light and temperature on percentage of Pyxidanthera brevifolia seeds germinating. Error bars represent 1(+/‐) standard error, with 5 replicates (10 seeds per replicate) for each treatment. There were significant effects of both light and temperature on the percentage of germinating seeds, but also an interaction between light and temperature...... 23 Fig. 3.1: Morphological variation in leaf length, leaf width, and pubescence of P. barbulata (circles) and P. brevifolia (triangles). Solid triangles represent P. brevifolia specimens that had pubescence for half or less than half of the leaf; open triangles represent P. brevifolia specimens with pubescence greater than half of the leaf. Although there are statistically significant differences between the two varieties for leaf length, leaf width, and pubescence, there is considerable overlap between the two species...... 63 Fig. 3.2: Geographic distribution (shaded in grey) and statistical parsimony network for 12 haplotypes from 2 cpDNA regions of Pyxidanthera. State names are in bold abbreviations and numbers represent haplotypes from Table 3.2. Black dots in the haplotype network represent mutational steps; associated letters (S for South, N for North) are the most likely (>95% probability) geographic origins of mutations, inferred using Genetree 9.0. Light grey shading of haplotype network represents proportion of the associated haplotype comprised of southern individuals, and darker grey shading represents proportion comprised of northern individuals. Inset map: Sampling of Pyxidanthera populations on Fort Bragg Military Reservation. Pyxidanthera barbulata populations are represented by closed circles and P. brevifolia populations are represented by open circles...... 64 Fig. 3.3: Isolation by distance for cpDNA (left side) and AFLP (right side) markers across the entire range of Pyxidanthera barbulata. cpDNA demonstrates no isolation by geographic distance (R = 0.01, p‐value = 0.39), while AFLP markers demonstrate weak but significant (R = 0.27, p‐value = 0.02) isolation by distance at shorter distances with effects of genetic drift more evident at greater distances...... 65 Fig. 3.4: Parameter estimates for θ (southern, northern, and ancestral populations), time since divergence, and migration (gene flow) between northern and southern populations of the genus Pyxidanthera based on results from IMa2...... 66 Fig. 3.5: Non‐metric multidimensional scaling ordination of P. barbulata and P. brevifolia population genetic distances (Nei’s D) based on amplified fragment length polymorphism markers. In

x the legend, letters in parentheses represent US states. Little separation is evident among populations defined according to either taxonomic status or geographical location...... 67 Fig. 3.6: Population genetic structure for the genus Pyxidanthera as inferred from the program STRUCTURE for K 2 through 9. The individual columns represent individual genetic samples and the colors represent proportion of ancestry assigned to the different ancestral populations. Little genetic structuring is evident based on either geographic location (South vs. North) or taxonomic identity (var. barbulata vs. var. brevifolia)...... 68 Fig. 3.7: Log likelihood (Ln P(D)) and standard deviation results from program STRUCTURE for K 1 through 9. Runs included a burn‐in length of 10 000 and a post burn‐in length of 25 000 with admixture and correlated allele frequencies...... 70 Fig. 4.1: Historic range and collection sites of Astragalus michauxii. Historical range determined based on voucher specimens (UNC Herbarium Flora of the Southeast; http://www.herbarium.unc.edu/seflora). Current range is greatly restricted, with most sites in North Carolina. Survey of Georgia populations located 13 individuals and there are only two known sites in South Carolina and Alabama...... 97 Fig. 4.2: Non‐metric multidimensional scaling ordination of Astragalus michauxii population genetic distances based on eight polymorphic microsatellite loci. The Georgia populations appear separate from the North Carolina populations, while little separation appears between the North Carolina populations from Fort Bragg and Camp Mackall...... 98 Fig. 4.3: Population genetic structure for Astragalus michauxii as determined by the program STRUCTURE (K = 2‐9). Individual columns represent genetic samples and colors represent proportion of ancestry assigned to different ancestral populations. Results reflect low genetic population differentiation in A. michauxii...... 99 Fig. 4.4: Pairwise population isolation by distance for sampled Astragalus michauxii populations based on eight polymorphic microsatellite loci. Genetic distances demonstrate significant (R = 0.43, P<0.001) isolation by distance. When GA populations are removed isolation by distance is not evident (R = 0.05, P=0.36)...... 100 Fig. 4.5: M ratio values (black circles) estimated for 22 Astragalus michauxii populations across North Carolina and Georgia (USA). Horizontal lines represent 95% CIs, the vertical line is

the threshold indicative of a past genetic bottleneck, the open triangles are the critical Mc

(90% SSM), and the gray triangles are the critical Mc (80% SMM) (Garza and Williamson 2001). Population numbers are in parentheses...... 101 Fig. 5.1: time line for collection of demographic data for Astragalus michauxii (top) and Pyxidanthera brevifolia (bottom). Grey boxes represent prescribed burn season, long dashes represent data collection, and letters along x‐axis represent seasons. Data collection for A. michauxii

xi occurred during the burn season, with burned populations measured at the end of the growing season (short dashed line). Data collection for P. brevifolia occurred before the burning season; measurements were roughly 9 months post‐fire in burned populations.132 Fig. 5.2: Mortality of Astragalus michauxii (top) and Pyxidanthera brevifolia (bottom) as a function of time since last burn and size class. Error bars represent standard error of the mean .133 Fig. 5.3: Mean number of fruits produced as a function of time since last burn and size class for Astragalus michauxii (top) and Pyxidanthera brevifolia (bottom). Error bars represent standard error of the mean...... 134 Fig. 5.4: Modified box plot of post‐fire recovery rates for Astragalus michauxii (top) and Pyxidanthera brevifolia (bottom) individuals as a function of pre‐burn size. Dotted line represents recovery to pre‐burn size. Smaller individuals recover at a faster rate relative to larger individuals for both species...... 135 Fig. 5.5: Projected stochastic population growth rates under different fire‐return intervals (1‐4 years) for Astragalus michauxii (left) and Pyxidanthera brevifolia (right) using a matrix selection approach. Error bars represent bootstrapped 95% confidence intervals...... 136 Fig. 5.6: Contributions of growth, survivorship, and fecundity by size class to the difference in the population growth rate between unburned and burned Astragalus michauxii (top) and Pyxidanthera brevifolia (bottom) populations. Size classes for A. michauxii are small = 1‐20 cm, small‐medium = >20‐40 cm, medium = >40‐80 cm, and large > 80 cm. Size classes for P. brevifolia are small = 1‐50 cm2, medium = >50‐400 cm2, and large > 400 cm2...... 137

xii CHAPTER 1 OVERVIEW 1.1 Introduction

Understanding the abiotic and biotic factors that contribute to the persistence of plant populations within contemporary plant communities is important from both basic and applied science standpoints. A number of different and overlapping theories have been presented for the coexistence of multiple species within the same trophic level (Silvertown 2004), with the topic still drawing scientific attention today (Chase and Leibold 2003; Hubbell 2001). The topic is not merely an academic pursuit. Identifying the abiotic and biotic factors that influence population persistence and coexistence with other species is critical for developing active management plans for threatened and endangered species. This is becoming increasingly important as the global loss of species increases at an alarming rate due to anthropogenic causes (Pimm et al. 1995).

Numerous abiotic and biotic factors can influence population persistence at multiple spatial scales. At the landscape level, climate is one of the main drivers of population persistence within a landscape (Prentice et al. 1991). At finer scales, abiotic factors such as fire (Glitzenstein et al. 2003), soil characteristics (e.g. pH (Dodd et al. 1994; Gough et al. 2000), nutrient availability (Grime 1973), soil texture (Williams et al. 1996), and hydrology (Dimick et al. 2010) can greatly influence the population dynamics and persistence of plant species. Biotic factors related to resource competition

(e.g. for light or nutrient availability) can result in plant species failing to occupy otherwise suitable habitats (Tilman 1982). All of these factors operate across time and space to influence patterns of species persistence and, ultimately, species diversity patterns.

The species considered in this project are members of the longleaf pine ecosystem. The longleaf pine ecosystem stretched from East Texas to Florida and north to southeastern Virginia at

1 the time of European settlement, covering an estimated 37 million hectares (Frost 1993). Most of the longleaf pine ecosystem was found within the Gulf and Atlantic Coastal Plain physiographic region (GACP), and to a certain extent the two constitute a single biogeographic region. The GACP, and by extension the longleaf pine ecosystem, has been noted for its high species diversity (Estill and Cruzan 2001; Walker and Peet 1983) Changes in land use over the last few centuries have led to the loss of 97% of ecosystem (Frost et al. 1986). The dominant canopy tree over most of the area was Pinus palustris, with an often species‐rich herbaceous understory and little to no woody midstory component.

A temperate savanna physiognomy has been maintained in the longleaf pine ecosystem by a high fire frequency, with fire return intervals estimated at 1‐10 years across the region (Myers and

White 1987; Frost 1993; Huffman 2006). The ecological role of fire in the longleaf pine ecosystem has been a focal point of interest for almost a century (Andrews 1917; Chapman 1932) and has been recognized to influence the population dynamics, species diversity, and vegetation structure. Fire suppression leads to the development of a woody midstory and, ultimately, replacement of the longleaf pine canopy. Most herbaceous species in the system are relatively shade intolerant and currently rely upon frequent fire to maintain an open habitat. Shading by midstory species and accumulation of dense litter suppress many of these fire‐dependent herbaceous species.

The longleaf pine ecosystem contains the second‐highest levels of endemism in the continental United States (Takhtajan 1986), with 22 identified regions of endemism (Sorrie and

Weakley 2001). One of the endemic patterns identified occurs in the Fall‐line Sandhills region, an area of relictual dunes that occurs on the northeastern edge of the GACP, extending from North

Carolina into Georgia. The Fall‐line Sandhills consist of two main geologicformations: the Pinehurst

2 and the Middendorf. In North Carolina, the Pliocene‐aged Pinehurst formation occurs on many of the ridge tops and is generally understood to be Aeolian in origin (Bartlett 1967), while the late‐

Cretaceous Middendorf appears to be of deltaic origin (Sohl and Owens 1991). The Pinehurst formation consists mainly of sand, while the underlying Middendorf formation includes a higher percentage of clay and silt. Weathering into the Middendorf has led to differential drainage rates and results in numerous seepages that can occur on middle slopes. This creates a matrix of wetland and dry habitats across the landscape. In contrast to most the GACP, the Fall‐line Sandhills have more topography and include numerous incised streamheads.

Nine taxa have been identified as endemic to the Fall‐line Sandhills: Astragalus michauxii,

Liatris cokeri, pyrophilum, Lobelia batsonii, Lycopus cokeri, Physalis lanceolata, Pyxidanthera brevifolia, Vaccinium crassifolium var. sempervirens, and Hexastylis sorriei. Interestingly, three of these species have been described within the last decade. These taxa occur across numerous habitat types that span the hydrologic gradient. Astragalus michauxii, Liatris cokeri, P. lanceolata and P. brevifolia tend to occur in the upland, more xeric habitats. Lilium pyrophilum, V. crassifolium var. sempervirens, and H. sorriei occur in wet areas such as Sandhills seeps and wet pine savannas, while

Lycopus cokeri is generally found in Sandhills streamhead swamps.

Within a given region, the species composition of communities is generally thought of in atemporal terms, with the biogeographic history of individual plant species and past vegetation assemblages generally left to the paleo‐ecologists. However, since the Last Glacial Maximum climate has changed dramatically with ensuing range shifts for many species (Soltis et al. 2006), the disappearance of former vegetation assemblages (Jackson and Williams 2004; Overpeck et al. 1992), and the creation of new ones (Watts 1979). This is especially true in the mid and higher latitudes,

3 because climate change appears to have been more dramatic in these areas. Seen through the lens of paleo‐ecological research, communities are composed of species with differing biogeographic histories. Some species are remnants from past vegetation assemblages, while others have migrated because of climatic changes. Understanding the past biogeography of plant taxa is not simply an academic pursuit, but has wider implications for species persistence under projected climate change.

Presumably, the nine endemic taxa that were identified above have been present in the

Sandhills since the Pleistocene, although there is always the possibility that they migrated into the

Sandhills following the Last Glacial Maximum. During the Pleistocene the climate of the Fall‐line

Sandhills was colder and drier compared to the current climate, with pollen records indicating the presence of Pinus spp. (most likely Pinus banksiana) and Picea spp. with an herbaceous understory that resembles prairie assemblages (Watts 1980). Geomorphology of rivers in the Sandhills suggests that many were braided rather than meandering during the Pleistocene, becoming channelized during the wetter climatic conditions of the Holocene (Leigh 2006). Braided rivers generally occur when a sediment load threshold is reached (Leopold et al. 1964) and in some areas have been associated with exposed river banks (Tal et al. 2004), suggesting that there may have been a discontinuous vegetation layer during the Pleistocene. If we accept that these taxa maintained populations in the Sandhills during the Pleistocene, each has experienced changes in climatic conditions that in contemporary spatial terms spans thousands of km. To reiterate, the climate that these endemic taxa experienced during the Pleistocene was vastly different than that of the current period.

4 Thus, there is a certain conflict between understanding the longleaf pine ecosystem (and the species within it) as “fire‐adapted” and the fact that, at least in the northern parts of the longleaf pine ecosystem, the current assemblage of species (and the dominance of P. palustris) may only date to the middle Holocene (Goman and Leigh 2004). It is within this context that we explore the population genetics and population demography of two rare plant species that are endemic to the southeastern United States – Astragalus michauxii Sargent and Pyxidanthera brevifolia Wells.

Chapter 2 explores seed germination and seed ecology of Pyxidanthera brevifolia, a species for which a germination protocol had not been developed and little was known regarding the conditions necessary for successful germination. Chapter 3 explores the phylogeography of the genus Pyxidanthera. Chapter 4 is an analysis of the population genetic diversity within Astragalus michauxii using microsatellite markers. Finally, chapter 5 reports on the effect of fire on the short‐ term population dynamics of P. brevifolia and A. michauxii.

5 1.2 References

Andrews EF (1917) Agency of fire in propagation of longleaf . Botanical Gazette 64:497–508.

Bartlett CS (1967) Geology of the Southern Pines quandrangle. Master of Science thesis, University of North Carolina at Chapel Hill

Chapman HH (1932) Is the longleaf type a climax? Ecology 13:328–334.

Chase JM, Leibold MA (2003) Ecological niches: linking classical and contemporary approaches. University of Chicago Press

Dimick BP, Stucky JM, Wall W, et al. (2010) Plant‐Soil‐Hydrology Relationships in Three Carolina Bays in Bladen County, North Carolina. Castanea 75:407–420.

Dodd ME, Silvertown J, McConway K, et al. (1994) Stability in the plant communities of the Park Grass Experiment: the relationships between species richness, soil pH and biomass variability. Philosophical Transactions of the Royal Society of London Series B: Biological Sciences 346:185–193.

Estill JC, Cruzan MB (2001) Phytogeography of rare plant species endemic to the southeastern United States. Castanea 66:3–23.

Frost CC (1993) Four centuries of changing landscape patterns in the longleaf pine ecosystem. In: Hermann SM (ed) Proceedings of the Tall Timbers Fire Ecology Conference. Tall Timbers Research Station, Tallahassee, FL, pp 17–37

Frost CC, Walker J, Peet RK, et al. (1986) Fire‐dependent savannas and prairies of the Southeast: Original extent, preservation and status. Wilderness and Natural Aeas in the eastern United States: A Management Challenge. Center for Applied Studies, School of Foresty, Stephen F. Austin St. University, Nacogdoches, Texas, pp 348–357

Glitzenstein JS, Streng DR, Wade DD (2003) Fire frequency effects on longleaf pine (Pinus palustris P. Miller) vegetation in South Carolina and northeast Florida, USA. Natural Areas Journal 23:

Goman M, Leigh DS (2004) Wet early to middle Holocene conditions on the upper Coastal Plain of North Carolina, USA. Quaternary Research 61:256–264.

Gough L, Shaver GR, Carroll J, et al. (2000) species richness in Alaskan arctic tundra: the importance of soil pH. Journal of Ecology 88:54–66.

Grime JP (1973) Competitive exclusion in herbaceous vegetation. Nature 242:344–347.

6 Hubbell SP (2001) The unified neutral theory of biodiversity and biogeography. Princeton University Press, Princeton N.J.

Huffman JM (2006) Historical fire regimes in the southeastern pine savannas. Loiusisana State University

Jackson ST, Williams JW (2004) Modern analogs in Quaternary paleoecology: Here today, gone yesterday, gone tomorrow? Annual Review of Earth and Planetary Sciences 32:495–537.

Leigh DS (2006) Terminal Pleistocene braided to meandering transition in rivers of the Southeastern USA. Catena 66:155–160.

Leopold LB, Wolman MG, Miller JP (1964) Fluvial processes in geomorphology. WH Freeman, San Francisco

Myers RL, White DL (1987) Landscape history and changes in sandhill vegetation in north‐central and south‐central Florida. Bulletin of the Torrey Botanical Club 114:21–32.

Overpeck JT, Webb RS, Webb III T (1992) Mapping eastern North American vegetation change of the past 18 ka: No‐analogs and the future. Geology 20:1071.

Pimm SL, Russell GJ, Gittleman JL, Brooks TM (1995) The future of biodiversity. Science 269:347– 350.

Prentice IC, Bartlein PJ, Webb III T (1991) Vegetation and climate change in eastern North America since the last glacial maximum. Ecology 2038–2056.

Silvertown J (2004) Plant coexistence and the niche. Trends in Ecology & Evolution 19:605–611.

Sohl N, Owens J (1991) Cretaceous stratigraphy of the Carolina coastal plain. In: Horton JW, Zullo VA (eds) Geology of the Carolinas: Carolina Geological Society fiftieth anniversary volume. University of Tennessee Press, Knoxville, pp 191–220

Soltis DE, Morris AB, McLachlan JS, et al. (2006) Comparative phylogeography of unglaciated eastern North America. Molecular Ecology 15:4261–4293.

Sorrie BA, Weakley AS (2001) Coastal plain vascular plant endemics: Phytogeographic patterns. Castanea 66:50–82.

Takhtajan A (1986) Floristic Regions of the World. University of California Press, Berkeley, CA

Tal M, Gran K, Murray AB, et al. (2004) Riparian vegetation as a primary control on channel characteristics in multi‐thread rivers. In: Bennett SJ, Simon A (eds) Riparian vegetation and fluvial geomorphology: Hydraulic, hydrologic, and geotechnical interactions. pp 43–58

7 Tilman D (1982) Resource competition and community structure. Princeton University Press, Princeton, N.J.

Walker J, Peet RK (1983) Composition and species diversity in pine‐wiregrass savannas of the Green Swamp, North Carolina. Vegetatio 55:166–179.

Watts WA (1979) Late Quaternary vegetation of central Appalachia and the coastal plain. Ecological Monographs 427–469.

Watts WA (1980) The late Quaternary vegetation history of the southeastern United States. Annual Review of Ecology and Systematics 11:387–409.

Williams RJ, Duff GA, Bowman D, Cook GD (1996) Variation in the composition and structure of tropical savannas as a function of rainfall and soil texture along a large‐scale climatic gradient in the Northern Territory, Australia. Journal of Biogeography 23:747–756.

8 EFFECTS OF LIGHT AND TEMPERATURE ON GERMINATION OF PYXIDANTHERA BREVIFOLIA WELLS

(DIAPENSIACEAE)

*Previously published in the Journal of the Torrey Botanical Society, 137(4): 348‐354. 2010. Used by permission

2.1 Abstract

Pyxidanthera brevifolia is an evergreen semi‐woody cushion plant endemic to the Sandhills of North and South Carolina, with the majority of populations occurring on Fort Bragg Military Reservation in

North Carolina. Currently the species is listed as Endangered in North Carolina and is designated as a

Species at Risk (SAR) by the US Department of Defense. Previous studies have suggested that seeds may not be viable because they failed to germinate under controlled conditions. Our objectives in this study were to attempt germination of Pyxidanthera brevifolia seeds, determine the best temperature conditions for germination, and understand more about germination requirements to aid in future restoration efforts. Using seeds that had been stored at room temperature for six months, we performed a germination experiment at the NCSU Phytotron with six treatments, all combinations of three temperature regimes (low (18oC day/14°C night), medium (22/18°C), and high

(26/22°C)) and two light conditions (light and dark). We monitored the experiment for 13 weeks, recording the number of seeds germinating per dish and the number of days to germination for seeds in each treatment. We found that Pxyidanthera brevifolia produces germinable seeds and that there are significant effects of light and temperature on germination. Highest germination occurred under low temperature and high light conditions (78%); the combination of high temperature and

9 no light produced the lowest germination (6%). Seeds exposed to light germinated significantly earlier at the coolest temperature, compared to medium and high temperatures. These results indicate that it is possible to germinate seeds of this rare plant and suggest that germination of

Pyxidanthera brevifolia likely occurs in late fall and is dependent on adequate light availability.

2.2 Introduction

Developing effective conservation strategies for rare plant species demands a comprehensive understanding of the numerous biotic and abiotic factors that can limit population growth. Efforts to conserve rare can benefit from ex situ, or off‐site, propagation (Schemske et al. 1994; Guerrant et al. 2004), which can be used to provide a source of plants for population augmentation (Brumback et al. 2004; Mooney and McGraw 2007), reintroduction (Bowles and

McBride 1996), or for studying questions relevant to species conservation. Ex situ propagation of rare plants has aided studies of physiological tolerance (Wang et al. 2006; Kimball and Campbell

2008; Marchin et al. 2009), phenotypic plasticity (Picotte et al. 2007), pollination (Hackney and

McGraw 2001), mating system evolution (Moeller and Geber 2005), and conservation genetics

(Levin et al. 1979). Population augmentation and reintroduction can ameliorate negative consequences of commercial over‐harvesting of wild populations (Kharkwal et al. 2008) and introduction of exotic pathogens and pests (Lee et al. 1995; Hebard 2001). Information gleaned from studies of ex situ propagation can also provide insights about the response of a species to management (Schwartz and Hermann 1999; Rhoades et al. 2009) and global climate change

(Maschinski et al. 2006). Often little is known about the propagation of rare plants because of a lack of study or underreporting of failed attempts. In the case of Pyxidanthera brevifolia (Sandhills

10 pixiemoss) there is a long history of failed propagation attempts and observations of the species’ apparent lack of recruitment from sexual reproduction (Wells 1929; Reynolds 1966; Primack and

Wyatt 1975).

Pyxidanthera brevifolia is a winter‐flowering evergreen cushion plant (Fig. 1) that occurs in a narrow geographic range within the Sandhills region of North and South Carolina (Wells 1929;

Weakley 2011). Currently, P. brevifolia is found in six North Carolina counties and two South

Carolina counties, with approximately 90% of the known populations occurring on Fort Bragg

Military Reservation, NC (Buchanan and Finnegan 2008). The species is listed by the state of North

Carolina as Endangered, designated by the US Department of Defense as a Species At Risk, and ranked by NatureServe as S2 (imperiled) and S3 (vulnerable) in South Carolina and North Carolina, respectively. Pyxidanthera brevifolia occupies Xeric Sandhill Scrub (Schafale and Weakley 1990), a habitat characterized by an exposed topographic position and excessively well‐drained soils that create harsh growing conditions, especially during the summer season (Wells and Shunk 1931). The more wide‐ranging congeneric P. barbulata is found in the outer Atlantic Coastal Plain from northeastern South Carolina to southeastern Virginia, with disjunct populations in the Pine Barrens of New Jersey and older dunes on Long Island, New York. The two species co‐occur on Fort Bragg, but are separated ecologically, with P. brevifolia occurring in the xeric uplands and P. barbulata occurring in ecotones between streamhead pocosins and adjacent pine‐dominated woodlands

(Schafale and Weakley 1990; Sorrie et al. 2006).

Successful seed germination in the field has not been observed for P. brevifolia, and there has been some question about seed production and viability. Wells and Shunk (1931) observed low seed‐set, and the seeds produced appeared to be inviable. Moreover, they observed no seedlings in

11 the field and speculated that P. brevifolia populations were clonal relicts maintained by vegetative propagation, though there is scant evidence for this. A subsequent study also observed low seed‐set for the species and production of seeds of questionable viability (Reynolds 1966). More recently,

Primack and Wyatt (1975) did not find low seed‐set; percentages of seed‐producing capsules collected from six populations ranged from 22‐83%. Nevertheless, all studies conducted thus far have reported a failure to germinate P. brevifolia seeds in situ (Wells and Shunk 1931; Reynolds

1966; Primack and Wyatt 1975). Attempts at transplanting P. brevifolia have also largely failed

(Primack and Wyatt 1975) or have not yet documented survival beyond four years (Hohmann et al. unpublished data).

Successful propagation of P. brevifolia is important because of its limited range, relatively low population numbers, and the pressures that it faces both currently and in the near future.

Because of base realignments, Fort Bragg Military Reservation has recently experienced unprecedented expansion, with more than 32,000 new troops being assigned to United States Army

Forces Command (FORSCOM). In addition, the nearby city of Fayetteville has also been expanding, and several road projects have led to the destruction of P. brevifolia populations. Our objectives in this study were to germinate P. brevifolia seeds to aid in possible future population augmentation, restoration, or establishment efforts, to identify the optimal germination conditions for P. brevifolia seeds, and to identify possible environmental controls on seed germination for this endemic

Sandhills species.

12 2.3 Methods

Seeds were collected during the last two weeks of April 2008 from at least 5 individuals in each of 24 randomly selected P. brevifolia subpopulations representing 19 populations (as delineated by the North Carolina Natural Heritage Program) on Fort Bragg, NC. Seeds were stored in paper envelopes under ambient laboratory conditions at North Carolina State University until

November 2008. Seeds from the 24 subpopulations were thoroughly mixed and placed on moistened filter paper in 8.5 cm diameter Petri dishes. Each dish was randomly assigned to one of six treatments of a factorial experiment consisting of three temperature settings (low, 18°C day

/14°C night; medium, 22/18°C; and high, 26/22°C) and two light settings (complete darkness and 12 hr daily exposure to fluorescent light). The temperature settings selected represented available growth chambers. We included five replicate Petri dishes per treatment with 10 seeds per replicate.

To simulate complete darkness, Petri dishes in the dark treatment were placed in breathable fabric bags custom designed for the North Carolina State University Phytotron to be both breathable and eliminate all light; seeds in the dark treatment were only light‐exposed when monitoring for germination and watering. We monitored the experiment for 13 weeks, recording germination success and time to germination for each seed in the experiment on average every 4.3 days. The experiment was conducted at the North Carolina State University Phytotron.

We tested for main effects of light and temperature on seed germination percentages using a generalized linear model with a binomial error structure and a logit link function as implemented in the statistical program R (R Development Core Team 2012). We tested for significant effects of light and temperature on the seed germination percentage with a Bonferroni correction using the

13 glht() function from the R package multcomp (Hothorn et al. 2008). We tested for differences in average time to germination using a linear mixed effects model in the R package NLME.

2.4 Results

Percentage of germinating seeds across all treatments was 47%, but it varied considerably among treatments. Overall, germination percentage in the light was twice that in darkness (60% vs

30%, = 20.56, P<0.001, Fig. 2). Germination percentage was also significantly affected by

temperature ( = 45.8, P<0.001), but this effect depended on light conditions (temperature X

light interaction, =9.03, P=0.011), such that temperature had a much greater effect on the percentage of germinating seeds in darkness than in light. Under the dark condition, germination percentage was higher at low temperature (52%) than at medium (30%, P<0.001) or high (6%,

P<0.001) temperatures; there were also significant differences in germination percentage under dark conditions between the medium and high temperatures (P<0.001). Similar effects were found under light conditions. Germination percentage was significantly higher at low temperature (78%) than at medium (60%, P<0.001) and high (61%, P<0.001) temperatures.

Average time to germination was 55.8 days (+/‐ 18.3 s.d.), but, as with germination percentage, this varied among treatments (F2 df = 6.55, P<0.001). We only compared time to germination for seeds exposed to light because of low germination percentage under dark conditions at high temperatures. Under low temperature, seeds germinated after an average of 47.3 days (+/‐ 2.73 s.e.), which was significantly different from average time to germination under

14 medium (58.5 days (+/‐ 3.45 s.e.), P<0.001) and high (63.8 days (+/‐ 2.80 s.e.), P<0.001) temperatures. Average time to germination did not differ between medium and high temperatures.

2.5 Discussion

Our study dispels speculation that Pyxidanthera brevifolia has a low capacity to produce viable seeds. Our observations (unpublished) also support the finding of Primack and Wyatt (1975) that P. brevifolia does not suffer from low seed‐set, with Fort Bragg populations producing a mean of 10 seeds per capsule. In contrast to previous studies, we found that P. brevifolia does produce viable seeds, with germination rates as high as 78% when 6 month old seeds were exposed to light and low (18/14° C) temperature. It is not clear why previous attempts to germinate this species were unsuccessful, but the relatively long time to germination (7‐9 weeks) may have been a factor if trials were not continued for sufficient lengths of time.

Previous attempts to propagate P. brevifolia from seed can be separated into studies conducted in situ and those conducted ex situ. The reason why in situ efforts have largely failed is difficult to evaluate, but we can explore why ex situ experiments have failed. Although our study was not specifically designed to investigate dormancy and represents a snapshot of a continually varying process, the available information points to the importance of after‐ripening of P. brevifolia seeds for breaking dormancy. Unfortunately, there is great variability in the methodological details for previous attempts to germinate P. brevifolia seeds. Primack and Wyatt (1975) reported a failure to germinate seeds, but provided no detail on whether seeds were fresh or after‐ripened. Two subsequent studies used freshly harvested seeds from P. brevifolia populations on Fort Bragg; both studies failed to document any germination (Hohmann et al. 2002; Shapiro et al. 2000). Although

15 gibberellic acid is known to promote germination in seeds exhibiting physiological dormancy and ordinarily requiring light (Kahn et al. 1957; Finch‐Savage and Leubner‐Metzger 2006), application of gibberellic acid at varying concentrations failed to induce germination in P. brevifolia seeds

(Hohmann, unpublished data). It is also possible that this last study was not of sufficient duration to observe germination, given that our results showed delayed germination at higher temperatures.

However, Shapiro et al. (2000) reported germination of P. brevifolia seeds from the seed bank after soil was collected near (< 1 m) adult plants in July, moistened and placed under growth lights for 12 weeks.

The slow germination of this species should substantially limit establishment success during the warmer months in its typical habitat. From late spring to early fall, xeric soils of the Sandhills region are unlikely to remain continuously moist for the time necessary for germination and successful establishment (Wells and Shunk 1931; Doublin and Grundstein 2008). This is particularly true for surface soils where germination must occur if the seedling is to survive. Owing to the small seeds (0.6‐0.8 mm) of this species (Radford et al. 1968), if germination were to occur in deeper soils that are buffered from rapid drying, the cotyledons would be unable to reach the surface, and seedlings would suffer high mortality rates. Previous studies have also demonstrated that small‐ seeded species in general require light for germination (Grime et al. 1981). Although there was a significant effect of light on the percentage of germinating seeds (Fig. 2), P. brevifolia is capable of some germination in the dark. This result may be due to the exposure of dark‐treatment seeds to light during monitoring episodes. It was not possible to exclude all light, and although we attempted to minimize light exposure, all dark‐treatment seeds were light‐exposed to some extent.

16 The conditions that favor germination of P. brevifolia in the NC Sandhills are most prevalent in the late fall and early winter, when temperatures are cooler and rainfall is only slightly less than in the summer months. Although the temperature range of the present study did not include a temperature low enough to capture decreased germination percentages due to colder temperatures, it is clear that germination percentages decrease above 18° C (Fig. 2). Pyxidanthera brevifolia appears to have adapted to the environmental constraints of the Sandhills climate by having germination cued to cool temperatures and high light conditions. This relatively low temperature optimum, and the fact that P. brevifolia does not appear to need cold stratification, points to a fall or early winter germination. Cues for fall germination would be advantageous and allow for a longer growing period before seedlings experience extreme summer conditions, which are characterized by frequent soil moisture deficits (Doublin and Grundstein 2008). Pyxidanthera brevifolia is a slow‐growing species with small seeds, and spring germination would most likely not allow enough time for root establishment before onset of the higher temperatures and drought conditions that occur during the summer months in the Sandhills region. Although our data support fall germination, laboratory conditions do not replicate natural settings; field trials will be necessary to test this hypothesis rigorously.

Our phytotron data suggest that germination of P. brevifolia occurs in the cooler months of fall and winter. If this proves the case in natural populations, we can find interesting parallels in other habitats. Seedling germination at relatively low temperatures has been observed in

Mediterranean environments as well (Thanos et al. 1989) and has been termed the

“Mediterranean” germination syndrome (Fenner and Thompson 2005), although these effects are not limited to the Mediterranean region (Grime et al. 1981). In this syndrome, increasing

17 temperature positively affects germination percentage to a certain threshold, at which point the percentage of germinating seeds decreases (Schütz 1997). In Mediterranean areas, summer drought is more likely to cause seedling mortality than cold, and so many species are cued to germinate in the fall or winter.

In conclusion, our results indicate that P. brevifolia produces viable seeds, suggesting that it is possible to propagate this species for restoration purposes. This is an important finding, because attempts at relocation have largely failed. Our discovery of positive germination response to low temperatures suggests that germination in nature likely occurs in late fall and that germination appears to depend on adequate light to break dormancy. Germination under cooler temperatures may be an adaptation necessary for seedling survival, ensuring that early growth occurs under less stressful conditions. Although further investigation will be necessary to confirm this hypothesis, the germination protocol we produced will be beneficial to future research and conservation efforts focused on Pyxidanthera brevifolia.

18 2.6 References

Bowles ML, McBride J (1996) Pitcher’s thistle (Cirsium pitcheri) reintroduction. In: Falk DA, Millar CI, Olwell M (eds) Restoring diversity: strategies for reintroduction of endangered plants. Island Press, Washington, D.C., pp 423–432

Brumback WE, Weihrauch DM, Kimball KD (2004) Propagation and transplanting of an endangered alpine species: Robbins’ cinquefoil, Potentilla robbinsiana (Rosaceae). Native Plants Journal 5:91–97.

Buchanan M, Finnegan JT (2008) Natural Heritage Program List of Rare Plant Species of North Carolina. 140.

Doublin JK, Grundstein AJ (2008) Warm‐season soil‐moisture deficits in the southern United States. Physical Geography 29:3–18.

Fenner M, Thompson K (2005) The ecology of seeds. Cambridge Univ Press, New York

Finch‐Savage WE, Leubner‐Metzger G (2006) Seed dormancy and the control of germination. New Phytologist 171:501–523.

Grime JP, Mason G, Curtis AV, et al. (1981) A comparative study of germination characteristics in a local flora. The Journal of Ecology 69:1017–1059.

Guerrant E., Havens K, Maunder M (2004) Ex situ plant conservation : supporting species survival in the wild. Island Press, Washington, DC, p 504

Hackney EE, McGraw JB (2001) Experimental demonstration of an Allee effect in american ginseng. Conservation Biology 15:129–136.

Hebard FV (2001) Backcross breeding program produces blight‐resistant American chestnuts (Virginia). Ecological Restoration 19:252–254.

Hohmann MG, Bates M, Gray J (2002) Germination and seedling survivial of Sandhills Pyxie (Pyxidanthera barbulata var. brevifolia). 5.

Hothorn T, Bretz F, Westfall P (2008) Simultaneous inference in general parametric models. Biometrical Journal 50:346–363.

Kahn A, Goss JA, Smith DE (1957) Effect of gibberellin on germination of lettuce seed. Science 125:645–646.

19 Kharkwal AC, Kushwaha R, Prakash O, et al. (2008) An efficient method of propagation of Podophyllum hexandrum: an endangered medicinal plant of the Western Himalayas under ex situ conditions. Journal of Natural Medicines 62:211–216.

Kimball S, Campbell D (2008) Physiological differences among two Penstemon species and their hybrids in field and common garden environments. New Phytologist 181:478–488.

Lee JC, Yang X, Schwartz M, et al. (1995) The relationship between an endangered North American tree and an endophytic fungus. Chemistry & Biology 2:721–727.

Levin DA, Ritter K, Ellstrand NC (1979) Protein polymorphism in the narrow endemic Oenothera organensis. Evolution 33:534–542.

Marchin RM, Bhandari RK, Wall WA, et al. (2009) Are rare species less shade tolerant than common species in fire‐prone environments? A test with seven Amorpha (Fabaceae) species. Plant Ecology 205:249–260.

Maschinski J, Baggs JE, Quintana‐Ascencio PF, Menges ES (2006) Using population viability analysis to predict the effects of climate change on the extinction risk of an endangered limestone endemic shrub, Arizona cliffrose. Conservation Biology 20:218–228.

Moeller DA, Geber MA (2005) Ecological context of the evolution of self‐pollination in Clarkia xantiana: population size, plant communities, and reproductive assurance. Evolution 59:786–799.

Mooney EH, McGraw JB (2007) Effects of self‐pollination and outcrossing with cultivated plants in small natural populations of American ginseng, Panax quinquefolius (Araliaceae). American Journal of Botany 94:1677–1687.

Picotte JJ, Rosenthal DM, Rhode JM, Cruzan MB (2007) Plastic responses to temporal variation in moisture availability: consequences for water use efficiency and plant performance. Oecologia 153:821–832.

Primack RB, Wyatt R (1975) Variation and taxonomy of Pyxidanthera (Diapensiaceae). Brittonia 27:115–118.

R Development Core Team (2012) R: a Language and environment for statistical computing. Vienna, Austria

Radford AE, Ahles HE, Bell CR (1968) Manual of the vascular flora of the Carolinas. University of North Carolina Press, Chapel Hill

Reynolds JD (1966) Morphological Studies in Diapensiaceae. The Embryology of Pyxidanthera Mich. Doctoral dissertation, University of South Carolina

20 Rhoades C, Loftis D, Lewis J, Clark S (2009) The influence of silvicultural treatments and site conditions on American chestnut (Castanea dentata) seedling establishment in eastern Kentucky, USA. Forest Ecology and Management 258:1211–1218.

Schafale MP, Weakley AS (1990) Classification of the natural communities of North Carolina: Third approximation. pp. 321.

Schemske DW, Husband BC, Ruckelshaus MH, et al. (1994) Evaluating approaches to the conservation of rare and endangered plants. Ecology 75:584–606.

Schütz W (1997) Are germination strategies important for the ability of cespitose wetland sedges (Carex) to grow in forests? Canadian Journal of Botany 75:1692–1699.

Schwartz MW, Hermann SM (1999) Is slow growth of the endangered Torreya taxifolia (Arn.) normal? Journal of the Torrey Botanical Society 126:307–312.

Shapiro A‐M, McAllister B, Hohmann MG, et al. (2000) Ecological Studies on the Sandhills Pyxie Moss. 36.

Sorrie BA, Gray JB, Crutchfield PJ (2006) The vascular flora of the longleaf pine ecosystem of Fort Bragg and Weymouth Woods, North Carolina. Castanea 71:129–161.

Thanos CA, Georghiou K, Skarou F (1989) Glaucium flavum seed germination‐an ecophysiological approach. Annals of Botany 63:121–130.

Wang GG, Bauerle WL, Mudder BT (2006) Effects of light acclimation on the photosynthesis, growth, and biomass allocation in American chestnut (Castanea dentata) seedlings. Forest Ecology and Management 226:173–180.

Weakley AS (2011) Flora of the Southern and Mid‐Atlantic States. Working Draft, Chapel Hill, NC

Wells BW (1929) A new pixie from North Carolina. Journal of the Elisha Mitchell Scientific Society 44:238–239.

Wells BW, Shunk IV (1931) The vegetation and habitat factors of the coarser sands of the North Carolina coastal plain: An ecological study. Ecological Monographs 1:465–520.

21

Fig. 1.1. Winter‐flowering Pyxidanthera brevifolia in full bloom on Fort Bragg (left) and ex situ propagated P. brevifolia seedling (right). Left photo taken 25 March 2009.

22

Fig. 1.2. Effects of light and temperature on percentage of Pyxidanthera brevifolia seeds germinating. Error bars represent 1(+/‐) standard error, with 5 replicates (10 seeds per replicate) for each treatment. There were significant effects of both light and temperature on the percentage of germinating seeds, but also an interaction between light and temperature.

23 CHAPTER 3 EVIDENCE FOR RANGE STASIS DURING THE LATTER PLEISTOCENE FOR THE ATLANTIC COASTAL PLAIN ENDEMIC GENUS, PYXIDANTHERA MICHAUX

*Previously published in Molecular Ecology, 19, 4302‐4314, 2010. Used by permission.

3.1 Abstract

The general phylogeographic paradigm for eastern North America (ENA) is that many plant and animal species retreated into southern refugia during the last glacial period, then expanded northward after the last glacial maximum. However, some taxa of the Gulf and Atlantic Coastal Plain

(GACP) demonstrate complex yet recurrent distributional patterns that cannot be explained by this model. For example, eight co‐occurring endemic plant taxa with ranges from New York to South

Carolina exhibit a large disjunction separating northern and southern populations by >300 kilometers. Pyxidanthera (Diapensiaceae), a plant genus that exhibits this pattern, consists of two taxa recognized as either species or varieties. We investigated the taxonomy and phylogeography of

Pyxidanthera using morphological data, cpDNA sequences, and AFLP markers. Morphological characters thought to be important in distinguishing P. barbulata and P. brevifolia demonstrate substantial overlap with no clear discontinuities. Genetic differentiation is minimal and diversity estimates for northern and southern populations of Pxyidanthera are similar with no decrease in rare alleles in northern populations. In addition, the northern populations harbor several unique cpDNA haplotypes. Pyxidanthera appears to consist of one morphologically variable species that persisted in or near its present range at least through the later Pleistocene, while the vicariance of the northern and southern populations may be of comparatively recent origin. This work demonstrates that the refugial paradigm is not always appropriate and GACP endemic plants, in

24 particular, may exhibit phylogeographical patterns qualitatively different from those of other ENA plant species.

3.2 Introduction

The alternating glacial and interglacial periods that characterized the Pleistocene had major impacts on the biogeography and genetic diversity of plant species in the Northern Hemisphere

(Comes & Kadereit 1998; Hewitt 2000). The last glacial maximum (LGM), approximately 18,000 years

BP, saw the Laurentide ice sheet reach its southern extent in eastern North America (ENA) (Ehlers &

Gibbard 2004). The primary scenario describing plant species’ ranges during and following the LGM in ENA includes (1) range contraction to southern refugia (Delcourt & Delcourt 1981) and (2) subsequent recolonization of northern habitats after the retreat of the glaciers (Dorken & Barrett

2004). Previous studies have identified the resulting phylogeographic patterns of plant species in

ENA and made inferences about possible refugia during the glacial maxima (Soltis et al. 2006).

Several such patterns have been identified in the ranges of ENA plant and animal species. These include east‐west divisions between Gulf Coast and Atlantic Coast populations (Mylecraine et al.

2004), phylogeographical separation by river drainage systems (Church et al. 2003), and the identification of refugia closer to the glacial front, either in the southern Appalachians (Gonzales et al. 2008) or even farther north (McLachlan, et al. 2005; Magni et al. 2005).

Eastern North America has generally been divided into four physiogeographic regions:

Interior Lowlands, Appalachian Highlands, Piedmont, and Gulf and Atlantic Coastal Plain (Fenneman

1938). Compared to the other regions, the Gulf and Atlantic Coastal Plain (GACP) is well defined geologically and floristically (Takhtajan 1986), but little is known about the phylogeography of the

25 many widespread species that are endemic to the GACP. Previous phylogeographic studies of GACP plant species have generally focused on narrow endemics with small latitudinal ranges (Evans et al.

2000; Oliveira et al. 2007) or species with ranges that cross multiple physiographic regions (Morris et al. 2008; however see Mylecraine et al. 2004). This is unfortunate; over 1300 species and 47 genera are endemic to the region (Sorrie & Weakley 2001), the second‐highest concentration in the United

States and only exceeded by the California Floristic Province (Flora of North America Editorial

Committee 1993). Without taking into account the endemic species of the GACP and their distributional patterns, any attempt to understand the post‐glacial phylogeography of ENA is limited.

Sorrie and Weakley (2001) documented twenty‐seven different recurrent distributional patterns for endemic plant species of the GACP. One of the most interesting of these patterns is the disjunct distribution of eight taxa (Calamovilfa brevipilis, Dichanthelium hirstii, Eupatorium resinosum, Gentiana autumnalis, Lobelia canbyi, Narthecium americanum, Pyxidanthera barbulata, and Rhynchospora pallida) that occur in New York and New Jersey and eastern North and South

Carolina, but not in the intervening areas of Maryland, Delaware, and most of Virginia. In addition to these eight taxa, numerous species exhibit the same disjunction between New Jersey and the southern GACP, but are more widespread in the southern part of the GACP. Common distributional patterns may be the result common biogeographic processes, but there is always the possibility of pseudocongruence (Hafner & Nadler 1990; Cunningham & Collins 1994).

We focus here on the genus Pyxidanthera as a case study to investigate the refugial paradigm in the GACP. Pyxidanthera is interesting in its own right as a genus that has presented taxonomic challenges and that contains a rare species. Pyxidanthera Michaux. is in Diapensiaceae, a

26 small family with a circumboreal distribution, with some taxa extending southward into eastern Asia and eastern North America. The genus includes two recognized species; both are woody, winter‐ flowering, evergreen cushion plants. Populations of the more widespread P. barbulata occur on

Long Island in New York, the Pine Barrens of New Jersey, several locations in southeastern Virginia, and the coastal plain of North Carolina and South Carolina. P. brevifolia has a more limited range; it has only been documented in six counties in the Sandhills region of North Carolina and South

Carolina. P. brevifolia, currently under intensive study as a species at risk (SAR) by the US

Department of Defense, is considered vulnerable to extinction in North Carolina, with over 80% of the North Carolina populations confined to Fort Bragg Military Reservation, NC (Buchanan &

Finnegan 2008). P. brevifolia is nearly restricted to xeric sandhill scrub communities within the long leaf pine ecosystem (Schafale & Weakley 1990; Sorrie et al. 2006), one of the most imperiled ecosystems in North America, with approximately 2% of the historical area currently extant (Frost

2006). In addition to containing most of the remaining P. brevifolia populations, Fort Bragg Military

Reservation is also one of the few places where the two species of Pyxidanthera co‐occur. When sympatric, P. barbulata and P. brevifolia occupy non‐overlapping ecological habitats, with P. barbulata occupying wetter sites such as pocosin ecotones and P. brevifolia occurring on extremely xeric sand ridges.

Pyxidanthera was monotypic until 1929, at which time P.brevifolia was separated from sympatric populations of P. barbulata in the Sandhills region of North Carolina and upper South

Carolina based on habitat differences, shorter leaves, and dense pubescence relative to the more‐ widespread P. barbulata (Wells 1929). Differing ecological niches and morphological characters of P. barbulata and P. brevifolia led to a debate regarding the proper taxonomic status of the two taxa. In

27 1964, P. brevifolia was reduced to a variety of P. barbulata without comment (Ahles 1964).

Afterward, several studies investigated the appropriate taxonomic status of P. brevifolia. An embryological study (Reynolds 1966) concluded that both notable developmental similarities and differences existed between the two species, and ultimately relied on the ecological and morphological differences to support the continued recognition of two species. Primack and Wyatt

(1975) found correlation between leaf length and soil moisture of P. brevifolia and P. barbulata at a single site in South Carolina and concluded that the difference in leaf length between the two species is clinal, suggesting that P. brevifolia is simply a morphological variant of P. barbulata. More recently, an allozyme study ‐ restricted to the populations from the southern range of the genus ‐ found that the two species share similar levels of genetic diversity, with very little inter‐taxa genetic differentiation (Godt & Hamrick 1995). However, recent floras for the region (Sorrie et al. 2009;

Weakley 2008) have continued to recognize two species, emphasizing the morphological, ecological, and embryological differences between them.

In this study we use cpDNA sequences, amplified fragment length polymorphism (AFLP) markers, and morphological measurements to investigate the taxonomy and phylogeography of both P. barbulata and P. brevifolia across the entire range of the genus. We attempt to determine if clear morphological and genetic differences exist between the two species and if the morphological, ecological, and embryological variation previously observed in the southern populations of

Pyxidanthera (in the text, Pyxidanthera will refer to both P. barbulata and P. brevifolia) correlate with greater genetic diversity in the south. Using genetic data, we attempt to distinguish between two plausible phylogeographic scenarios. The first scenario represents typical refugial patterns described for numerous species in ENA; the genus Pxyidanthera was isolated in one or more

28 southern refugia during the Pleistocene and subsequently recolonized northern areas after the LGM.

Genetic patterns typically observed supporting this scenario include reduced genetic diversity in northern populations (Hewitt 2000), recolonized areas containing only a subset of refugial population alleles (Broyles 1998), and putative refugia having a greater number of rare alleles, which may reflect historical processes better than genetic diversity estimates (Comps et al. 2001; Paun et al. 2008) Alternatively, the two species in Pyxidanthera persisted in their present ranges through the later Pleistocene rather than retreating into one or more glacial refugia. Genetic patterns that would suggest this second scenario include no reduction in genetic diversity or rare alleles in northern populations and the presence of alleles restricted to northern populations.

3.3 Methods

3.3.1 Sampling and Morphological Measurements

We collected leaf tissue samples of 423 individuals from 29 P. brevifolia populations

(defined as all P. brevifolia individuals that occurred within 0.75 km of each other) and 178 individuals from 14 P. barbulata populations, across the ranges of both species. A priori taxonomic identity was determined based on habitat differences, State Natural Heritage Program records, and geographic region (P. brevifolia is restricted to the Sandhills region of North and South Carolina). For each sample we measured the longest leaf length and width and categorized the leaf pubescence into one of two categories: pubescence covering more than half of the leaf, and pubescence covering half or less of the leaf. We evaluated differences in leaf length and leaf width means between P. barbulata and P. brevifolia using t‐tests and evaluated differences in leaf pubescence categories using a chi‐square test.

29 3.3.2 Molecular methods

DNA was extracted from 319 P. brevifolia individuals across 17 populations and 157 P. barbulata individuals across 14 populations using the CTAB method with minor modifications (Doyle

& Doyle 1987). After an initial screening of 16 cpDNA regions known to be highly polymorphic (Shaw et al. 2007), we amplified two polymorphic regions – atpI‐atpH and psbD‐trnT(GUU) – of 63 and 42 samples from 14 and 10 populations of P. barbulata and P. brevifolia, respectively, using universal primer pairs (Shaw et al. 2007). PCR conditions followed Shaw et al. (2005) in 12.5 µL solutions using the following protocol: 1 hold (5 min/ 80° C), 30 cycles ((1 min/ 95° C), (1 min/ 50° C), (4 min/ 65° C)),

1 hold (5 min/ 65° C). PCR products were cleaned prior to sequencing using Antarctic Phosphatase

(0.5 Units), Exonuclease I (0.2 Units), and 1 µL 10x Antarctic Phosphatase buffer (New England

BioLabs, Ipswich, MA, USA) at 37° C (15 min) and 80° C (15 min). We sequenced in the forward direction for the atpH and psbD regions using the Big Dye 3.1 kit (Applied Biosystems, USA) and analyzed the products using an ABI 3730 DNA sequencer (Applied Biosystems, USA). We edited and aligned sequences using Sequencher 4.2.2 (Gene Codes Corporation, Ann Arbor, MI, USA) and MEGA version 4 (Tamura et al. 2007). Sequences were submitted to GenBank under accession numbers nos. HM564379‐HM56491 (Table 3.1: Chloroplast haplotype accession numbers as archived in

Genbank for the atpI‐atpH intergenic spacer region (partial sequence) and the psbD‐trnT intergenic spacer region (partial sequence). Chloroplast haplotypes H1 through H12 refer to unique composite sequences from the two intergenic spacer regions.

AFLP markers are appropriate for examining low levels of genetic divergence within and between closely related taxa (McKinnon et al. 2008; Coart et al. 2002) and have been successfully used in phylogeographic studies (Meudt & Bayly 2008; Perez‐Collazos et al. 2009). AFLP genotyping

30 followed the multiplexing protocol described by Trybush et al. (2006), with the minor modification that the restriction and ligation steps were combined in a single reaction at a total volume of 10µL.

For the pre‐amplification reaction we used EcoRI+A and MseI+C primers. Three selective primer pairs were chosen after a trial based on the number of reproducible polymorphic markers produced:

Eco‐ACC/MseI‐CAT (Hex), Eco‐ATG/MseI‐CAT (FAM), and Eco‐AGG/MseI‐CAT (NED). Selective amplification products were separated and analyzed using an ABI 3730 DNA sequencer (Applied

Biosystems, USA) and automatically scored with Genemarker version 1.8 (Softgenetics LCC, USA) using the default settings, with the exception that we normalized the FAM‐dyed markers and set the allele evaluation peak score to “pass” if it was >= 1 (Holland et al. 2008). The reproducibility of the

AFLP profiles was evaluated by running eight duplicate samples for each 96‐well plate. Error rates between duplicates were calculated using a Euclidean distance measure (Bonin et al. 2004). To reduce the error rate we removed bands with 10 or more errors when comparing duplicate samples

(Zhang et al. 2010).

3.3.3 cpDNA data analysis

Because the chloroplast represents a single non‐recombining locus, sequences of the two sampled regions were concatenated. We recoded insertions or deletions (indels) that did not violate the assumptions of the infinite sites model (Kimura 1969) as identified by SNAP Workbench (Price &

Carbone 2005). We performed three separate tests of neutrality to test for evidence of population expansion or selection in the cpDNA – Fu and Li’s D* and F*(Fu & Li 1993) and Fu’s FS (Fu 1997) – using SNAP Workbench. Fu and Li’s D* and F* neutrality tests are more powerful for detecting background selection, while Fu’s Fs is more powerful for detecting population growth (Ramos‐Onsins

31 & Rozas 2002). We estimated an unrooted haplotype network using the haploNet function as implemented in the pegas package (Paradis 2009) in R (R Development Core Team 2009). This package implements the statistical parsimony method for network reconstruction (Templeton et al.

1992). We performed two analyses of molecular variance (AMOVA) using Arlequin version 3.01

(Excoffier et al. 2005) with the data set hierarchically partitioned by region and individual populations within regions (with two regions defined as New Jersey and New York, hereafter referred to as northern populations, and Virginia, North Carolina, and South Carolina, hereafter referred to as southern populations) and by species (P. barbulata and P. brevifolia) and populations within species. We estimated the nucleotide genetic diversity (π) (Nei 1987) for each population using DnaSP version 5 (Librado & Rozas 2009).

To test for phylogeographic structure in the data set, we compared two measures of genetic differentiation between populations – GST ,based on haplotype frequency, and NST, by similarities between haplotype sequences – using PERMUT 2.0 (Pons & Petit 1996) with 1000 permutations. If

NST is significantly greater than GST, it is taken as evidence of a phylogeographic signal in the data set.

To test for isolation by distance, we performed Mantel tests using the R package vegan (Oksanen et al. 2009) between the log‐transformed geographical distance matrix and the pairwise population NST matrix as calculated in DnaSP version 5 (Librado & Rozas 2009) for all populations, and for the southern and northern populations separately.

We reconstructed the gene genealogy for the sampled chloroplast regions using Genetree version 9.0 (Bahlo & Griffiths 2000) as implemented in SNAP Workbench (Price & Carbone 2005).

We estimated the population mutation rate (θ), using Watterson’s method (1975) as calculated in

Genetree for both geographical regions and used the average between the two regions as the

32 starting θ. Because of the larger geographical area covered by the two species of Pyxidanthera in the southern populations, we assumed a model of unequal population sizes, with the southern population twice as large as the northern population, and non‐exponential population growth. We performed ten independent simulations with different starting values of 106 iterations, selecting the rooted genealogy and mutation age estimates with the highest probability.

To simultaneously analyze the effects of incomplete lineage sorting and gene flow on the genetic structure of the northern and southern Pyxidanthera, we employed an isolation with migration model of population divergence (Nielsen & Wakeley 2001) implemented in the program

IMa2 (Hey & Nielsen 2007). IMa2 estimates the following parameters based on the genetic data: θ for all populations (extant and ancestral), migration parameters (m) for gene flow between populations, and t, time in coalescent units since divergence of the extant populations. We performed three independent runs with ten chains each under an infinite sites model with a burnin period of 150000 steps. 500000 genealogies were sampled, saving one genealogy every 100 steps.

We evaluated proper mixing based on the absence of trends in plotted parameter estimates and congruence of parameter estimates between runs. 100 000 of the 500 000 saved genealogies were combined to evaluate twenty‐four models that were either nested within the full model or that restrained select parameters by setting them equal to each other (e.g., equal migration between populations). We compared the different model posterior probabilities using an information‐ theoretic approach recently extended to phylogeographic data (Carstens et al. 2009). Information theory statistics were calculated according to Burnham and Anderson (Burnham & Anderson 2002).

33 3.3.4 AFLP data analysis

We calculated the percentage of polymorphic loci (P%) and Nei’s expected heterozygosity

(Nei 1987) using AFLPsurv version 1.0 (Vekemans et al. 2002) and the “frequency down weighted marker score” (DW) (Schönswetter & Tribsch 2005) using the R script AFLPdat (Ehrich 2006); several population genetic diversity measures were included to ensure consistency between methods. DW is calculated by summing each occurrence of a particular marker in a population and dividing that value by the sum of the marker across all populations. For each population these values are then averaged across all markers. Populations that have been isolated are expected to accumulate rare markers and thus their DW scores should be higher. We first removed populations that contained fewer than seven samples to minimize effects of low sample size (Bonin et al. 2007), leaving a total of 437 samples from 25 populations. We tested for effects of sample size on all the calculated genetic diversity estimates by regressing estimated diversity on sample size and we also tested for correlation between all possible pairings of the included diversity measures. Populations were grouped according to taxonomic identity and region, and we tested for significant differences between diversity estimates using t‐tests in R.

Population differentiation and structure were explored by first running an ordination using non‐metric multidimensional scaling (NMDS) to graphically display population pairwise genetic distances (D) (Nei 1972) in a reduced dimensional space using the R package labdsv(Roberts 2010).

We included all populations regardless of sample size for the analysis. In addition, we explored population genetic structure using STRUCTURE 2.3.2.1 (Pritchard et al. 2000; Falush et al. 2007). For

K 1 through 9 we performed three runs with a burn‐in length of 10 000 and post burn‐in length of 25

000, assuming admixture and correlated allele frequencies. We determined the most likely number

34 of populations by graphically analyzing the model log likelihoods for each K. Because Ln P(D) did not increase monotonically to the optimal K (Herrera & Bazaga 2008) we did not use the methods of

Evanno et al. (2005). Three analyses of molecular variance (AMOVA) were performed using Arlequin version 3.01 (Excoffier et al. 2005), with partitioning of the data following the chloroplast AMOVAs.

To test for isolation by distance, we performed a Mantel test between the population genetic distance matrix and the log‐transformed geographical distance matrix using the R package vegan

(Oksanen et al. 2009).

3.4 Results

3.4.1 Morphology

P. barbulata has significantly longer leaf lengths (6.3 mm vs. 4.5 mm respectively, P < 0.001) and widths (1.9 mm vs. 1.3 mm, P < 0.001) compared to P. brevifolia, but there is considerable overlap between the two species in both traits (Fig. 3.1). The variation in leaf length is continuous between P. barbulata and P. brevifolia with no obvious break, certainly not at the 3.5‐4 mm size suggested in taxonomic keys (Sorrie et al. 2009). There is a significant difference in leaf pubescence between the two species (P < 0.001). All P. barbulata had pubescence covering less than half of their leaves, but 49% of P. brevifolia also had pubescence covering more than half of their leaves. As with leaf length and width, there is considerable variation within taxa.

35 3.4.2 cpDNA

The two sampled cpDNA regions for 105 individuals yielded 975 characters, of which 14 were polymorphic (Table 3.2). The data set included 12 substitutions and 2 indels that did not violate the infinite sites model. None of the three neutrality tests (Fu and Li’s D* and F* and Fu’s Fs) were significant (P >0.05), indicating that there is no evidence of either population growth or background selection. A statistical parsimony haplotype tree identified 12 haplotypes (Fig. 3.2). The interior haplotypes of the network (H7, H11) are geographically widespread compared to the derived haplotypes, which tend to be both less frequent and geographically restricted (Fig. 3.2;

Table 3.2). Results from Genetree indicate that H11 is the haplotype with the highest probability of being ancestral (average relative likelihood 69.5%); however both of the interior haplotypes were almost equally common in the northern and southern populations (Fig. 3.2). Two of the 4 haplotypes derived from H7 (H4 and H9) only occur in New Jersey and New York and had the highest probability of a northern origin, while the other 2 derived haplotypes (H6 and H8) occur in both northern and southern population. Five of the 6 haplotypes derived from the second interior haplotype, H11, only occur in southern populations and most likely are of southern origin. Only H12 has a higher probability of a northern origin; it is a private haplotype restricted to one northern population.

Region explains a small but statistically significant percentage of the genetic variation when used as the highest grouping variable in a hierarchical AMOVA (17.27%, P < 0.05), revealing that genetic variation is not evenly spread across the northern and southern populations (Table 3.3).

Populations within regions explain most of the variation (56.58%, P < 0.001); within‐population genetic differences and region account for a smaller but still significant percentage of the variation

36 (26.15%, P < 0.001). When species is used as the highest grouping variable, AMOVA results demonstrate significant genetic differences among populations (72.22%, P < 0.001), but not significant differences between the two species (0%, P>0.05). The nucleotide genetic diversity (π) averages 0.0004 across all specimens with no significant differences between means for either regions or species (P > 0.05; Table 3.4).

Geographically distant populations are not more differentiated from each other than populations in closer geographical proximity (Fig. 3.3). NST is significantly greater than GST (0.788 vs.

0.695, P<0.01), indicating that there is a phylogeographic signal in the chloroplast data; in other words, haplotypes within populations are more similar to each other than expected. However,

Mantel tests for isolation by distance (IBD) find no significant signal across the range of Pyxidanthera populations (Fig. 3.3, R = 0.01, P = 0.39). This pattern generally arises when genetic drift exerts more influence than gene flow at the regional scale (Hutchison & Templeton 1999). When northern and southern populations are analyzed separately there is not significant IBD in the northern populations (R = ‐0.05, P = 0.47) but there is marginally significant IBD in the southern populations, although the effect is weak (R = 0.13, P =0.049).

The highest posterior density for θSouth was higher than both θ Ancestral and θNorth, although there is significant overlap between the 95% confidence intervals (Fig. 3.4). The highest posterior density parameter estimate for migration from south to north is 2.09, while the estimate for migration from northern populations into southern populations is 0.01, indicating there has been gene flow between the two regions, with possibly greater migration from the southern populations into the northern. Past gene flow between the two populations is also supported by the model selection exercise; the worst‐performing models constrained both migration parameters to 0 (Table

37 3.5). Time since divergence, t, was poorly estimated and failed to converge; this typically reflects a lack of a signal available in analyses that incorporate only a single locus with limited informative characters (J. Hey, pers. comm.).

3.4.3 AFLP

310 polymorphic bands were scored based on the 3 primer pairs. Each individual produced a unique AFLP profile and the Euclidean error rate (based on 47 replicate pairs) was 4.2%, within the margin of acceptable error rates (Bonin et al. 2004). All genetic diversity indices were highly correlated and sample size was not significantly correlated with any of the genetic diversity values

(all P > 0.05). The population genetic diversity estimates for P. barbulata and P. brevifolia populations do not differ significantly for %P, DW, or He (all P > 0.05). In addition, there are no significant differences between regional genetic diversity estimates for percentage %P, He, or DW

(all P > 0.05; Table 3.4). The percentage of polymorphic loci (%P) ranges from 23.2% to 51.3%, with a mean of 37.8 %, while Nei’s population genetic diversity (He) ranges from 0.08 to 0.16 with a mean of 0.12 (Table 3.4).

Non‐metric multidimensional scaling (NMDS) ordination based on the population genetic distances (D) reveals no discrete grouping of populations based on either region or species (Fig. 3.5).

Results from STRUCTURE also demonstrate little population genetic structure based on either geographical location or taxonomic identification (results not shown) and there was no graphical evidence for an optimal number of K distinct genetic groups (Fig. 3.7). The hierarchical AMOVAs grouped according to species (P. barbulata vs. P. brevifolia) and geographic region (North vs. South) found small but significant variation was explained by species (1.60%, P < 0.01) and region (3.20%, P

38 < 0.01; Table 3.3), while within‐population variation remained high (90.53% and 89.01%, respectively, P < 0.001). There is evidence for a weak but significant effect of isolation by distance

(IBD) in the AFLP data (Fig. 3.3; R = 0.27, P = 0.02). Genetic differentiation between populations increases with geographic distance, indicating low to moderate levels of short distance gene flow but little evidence of long‐distance gene flow; both gene flow and genetic drift influence the pattern depending on the geographical scale (Hutchison & Templeton 1999). At shorter distances, gene flow is dominant increasing the correlation between genetic and geographic distances, while at greater distances genetic drift predominates.

3.5 Discussion

3.5.1 Taxonomy

There does not appear to be clear separation between P. barbulata and P. brevifolia based on either the morphological or genetic data (Fig. 3.1 and Fig. 3.5). Although P. brevifolia in general has shorter and narrower leaves than the more widespread P. barbulata, there is significant overlap between the two species for both leaf length and width. Previously published work on the leaf morphology indicated that the differentiation between the two species was because of hydrological differences between the habitats that P. barbulata and P. brevifolia occupy (Primack & Wyatt 1975), with leaf length increasing continuously with increasing soil moisture. Although P. brevifolia individuals tend to be more pubescent than P. barbulata individuals (Fig. 3.1), there is significant variation in the pubescence of P. brevifolia both at the taxonomic level and within populations (data not shown), with both glabrous and pubescent individuals represented in most populations.

39 Interestingly, there are herbarium specimens of P. barbulata from xeric habitats of the Outer

Coastal Plain of North Carolina that exhibit the shorter leaves of P. brevifolia specimens, but that are not pubescent; the extreme pubescence appears to be restricted to P. brevifolia.

Although several authors have suggested that P. brevifolia may represent a pre‐adapted P. barbulata ecotype that moved into the Sandhills region from the Outer Coastal Plain (Wells & Shunk

1931; Primack & Wyatt 1975), the current study using cpDNA sequences and AFLP markers and a previous study using allozymes (Godt & Hamrick 1995) do not support this hypothesis. P. barbulata and P. brevifolia populations in the Sandhills are not genetically distinct from each other, with P. barbulata populations on Fort Bragg sharing cpDNA haplotypes with nearby P. brevifolia populations

(Fig. 3.2, inset). In addition, there is no separation between P. barbulata and P. brevifolia populations in their AFLP profiles (Fig. 3.5). We cannot rule out the possibility that P. brevifolia is a recently derived ecotype of P. barbulata, restricted to the Sandhills, and that a few mutations have led to local adaptation, but this would need to have been recent enough that genetic differentiation is not apparent in AFLP profiles. Even though P. brevifolia appears to be an extreme morphological variant of P. barbulata associated with sandy, xeric sites, in our estimation it warrants continued active management – specifically the regular prescribed fire schedule that Fort Bragg Military

Reservation maintains – and further study because of its potentially critical role as an early season pollen and nectar provider and as a system for studying physiological adaptation to drought stress and phenotypic plasticity.

40 3.5.2 Phylogeography of the genus Pyxidanthera

Contrary to the well‐documented trends of range contraction observed in many temperate plant species during the last glacial period in eastern North America, we found little evidence for either a southern refugium or range expansion following the last glacial maximum (LGM) in the genus Pyxidanthera. Genetic diversity estimates for both the AFLP and cpDNA markers were not significantly different for northern and southern Pyxidanthera barbulata populations (Table 3.4) and northern populations contained several cpDNA haplotypes that did not occur in the southern populations (Fig. 3.2). More pointedly, estimates of the number of rare AFLP markers (DW), which may be more helpful in identifying refugial phylogeographic patterns (Paun et al. 2008), did not demonstrate significant differences between northern and southern populations. Finally, the two interior haplotypes – H7 and H11 – were widespread in both northern and southern populations with comparable frequencies. These genetic patterns are contrary to what would be expected if there was a southern refugium for Pyxidanthera (Ikeda et al. 2008; Comps et al. 2001; Paun et al.

2008). Thus, it appears that the most likely scenario includes range stasis through the later

Pleistocene. Furthermore, evidence of gene flow between geographically close populations suggests a possible explanation for low levels of genetic differentiation between northern and southern populations; these populations may not have been as geographically isolated in the recent past, with populations in the intervening area facilitating gene flow.

Several studies of tree species have also demonstrated the absence of typical refugial patterns (Palme et al. 2003; Maliouchenko et al. 2007), indicating that some species may have persisted closer to the ice sheet than previously thought. There is increasing evidence for “cryptic refugia” in more northern latitudes for a number of mammal and plant species (Stewart & Lister

41 2001). Although mid‐latitude refugia are possible, several alternatives have also been put forth. In the case of Salix caprea, which demonstrates little phylogeographical patterning, Palme et al. (2003) posit high rates of dispersal, hybridization with other Salix species, and high mutation rates as possible reasons. These explanations are not very probable in the case of Pyxidanthera.

Pyxidanthera seeds lack obvious morphological adaptations for dispersal, although ants have been observed transporting seeds (W. Wall, pers. obs.). Hybridization with other species is implausible, since Pyxidanthera is well differentiated from all other taxa within Diapensiaceae (Ronblom &

Anderberg 2002). Although we have not estimated mutation rates, this alone would not generate the observed patterns.

That P. barbulata would persist, rather than retreat, during the climatic oscillations of the

Pleistocene is consistent not only with the genetic data but also with our knowledge of Pleistocene habitats and the species’ natural history. The Gulf and Atlantic Coastal Plain physiographic region, relative to more interior physiographic regions, may have been climatically buffered during the

Pleistocene because of the moderating influence of the Atlantic Ocean (Rahmstorf 2002); moderation of climatic extremes could have allowed persistence closer to the ice sheet during glacial periods for some GACP species. Still, P. barbulata populations in New Jersey and New York would have experienced much colder conditions through much of the last glacial period (Jacobson et al. 1987; French et al. 2003; French et al. 2007). The vegetation community of the late Pleistocene in some of the areas of ENA does not have a modern analog; most likely it would have been a relatively open spruce (Picea spp.) forest with an herbaceous understory dominated by Carex spp. (Overpeck et al. 1992). The most important factors in determining the ecological niche of P. barbulata may be high light levels and an absence of competition, rather than temperature or moisture. The

42 frequently burned habitats of the Sandhills of North and South Carolina and the Pine Barrens of New

Jersey provide this habitat; it is conceivable that environments near the glacial boundary that lacked a dominant canopy cover during the last glacial period did as well. sea levels during glacial periods may have periodically increased available habitats for Atlantic Coastal Plain species such as P. barbulata on the exposed continental shelf (Hobbs III 2004).

The present‐day disjunction in the range of P. barbulata may be related to regional geomorphology. The Atlantic Coastal Plain is characterized by a series of alternating arches and embayments (Ward 1992); Pyxidanthera populations occur on the Cape Fear, Norfolk, and South

New Jersey Arches, but are absent in the intervening Salisbury Embayment. The current disjunction in the range of the genus Pyxidanthera may be the result of oscillating sea levels that inundate embayment areas while arches remain above sea level (Bloom 1983; Sorrie & Weakley 2001). It is unlikely that long‐distance gene flow between the northern and southern populations without intermediate populations would be high enough to prevent genetic differentiation. This suggests that the current vicariance between northern and southern populations may be recent and that during periods of relatively low sea levels, suitable habitat was exposed on the continental shelf, connecting northern and southern populations and allowing gene flow to minimize genetic differentiation.

The Gulf and Atlantic Coastal Plain floristic province contains the second‐highest level of endemism in North America north of Mexico, yet the endemic plant species have been relatively understudied. Despite subtle topographic variation across the region, complex vegetation patterns exist and the biogeographical processes involved elude simple characterization. Although more phylogeographical studies of GACP endemic plant species are needed to determine whether the

43 recent phylogeographic history of the genus Pyxidanthera is representative of multiple taxa or is simply an isolated case, It is apparent that the simple refugial model cannot account for the phylogeographic pattern in the genus Pyxidanthera. If similar phylogeographic patterns are found in similarly distributed GACP endemics, it would suggest a common mechanism was responsible and the remaining challenge would be to explain why only these taxa were thusly affected. Refugia are generally thought of as existing in the past; it could be the case that contemporary distributional patterns represent modern‐day refugia for many Atlantic Coastal Plain endemic plant species.

44 3.6 References

Ahles HE (1964) New combinations for some vascular plants of southeastern United States. Journal

of the Elisha Mitchell Scientific Society 80:172–173.

Akaike H (1973) Information theory and an extension of the maximum likelihood principle. In: Petrov

BN, Csaki F (eds) Second international symposium on information theory. Akademia Kaido,

Budapest, pp 267–281

Bahlo M, Griffiths RC (2000) Inference from gene trees in a subdivided population. Theoretical

Population Biology 57:79–95.

Bloom AL (1983) Sea level and coastal changes. Late‐Quaternary Environments of the United States

2:42–51.

Bonin A, Bellemain E, Bronken Eidsen P, et al. (2004) How to track and assess genotyping errors in

population genetics studies. Molecular Ecology 13:3261–3273.

Bonin A, Ehrich D, Manel S (2007) Statistical analysis of amplified fragment length polymorphism

data: a toolbox for molecular ecologists and evolutionists. Molecular Ecology 16:3737–3758.

Broyles SB (1998) Postglacial migration and the loss of allozyme variation in northern populations of

Asclepias exaltata (Asclepiadaceae). American Journal of Botany 85:1091–1097.

Buchanan M, Finnegan JT (2008) Natural Heritage Program List of Rare Plant Species of North

Carolina. 140.

45 Burnham KP, Anderson DR (2002) Model Selection and Multimodel Inference: a Practical

Information‐Theoretic Approach. Springer Verlag

Carstens BC, Stoute HN, Reid NM (2009) An information‐theoretical approach to phylogeography.

Molecular Ecology 18:4270–4282.

Church SA, Kraus JM, Mitchell JC, et al. (2003) Evidence for multiple Pleistocene refugia in the

postglacial expansion of the eastern tiger salamander, Ambystoma tigrinum tigrinum.

Evolution 57:372–383.

Coart E, Lamote V, De Loose M, et al. (2002) AFLP markers demonstrate local genetic differentiation

between two indigenous oak species [Quercus robur L. and Quercus petraea (Matt.) Liebl.] in

Flemish populations. Theoretical and Applied Genetics 105:431–439.

Comes HP, Kadereit JW (1998) The effect of Quaternary climatic changes on plant distribution and

evolution. Trends in Plant Science 3:432–438.

Comps B, Gomory D, Letouzey J, et al. (2001) Diverging trends between heterozygosity and allelic

richness during postglacial colonization in the European beech. Genetics 157:389–397.

Cunningham CW, Collins TM (1994) Developing model systems for molecular biogeography:

vicariance and interchange in marine invertebrates. In: Schierwater B, Streit B, Wagner P,

DeSalle R (eds) Molecular Ecology and Evolution: Approaches and Applications. Birkhauser

Verlag, Basel, Switzerland, pp 405–433.

46 Delcourt PA, Delcourt HR (1981) Vegetation maps for eastern North America: 40,000 yr BP to the

present. In: Romans RC (ed) Geobotany II. Plenum, New York, pp 123–165.

Dorken ME, Barrett SCH (2004) Chloroplast haplotype variation among monoecious and dioecious

populations of Sagittaria latifolia (Alismataceae) in eastern North America. Molecular

Ecology 13:2699–2707.

Doyle JJ, Doyle JL (1987) A rapid DNA isolation procedure for small quantities of fresh leaf tissue.

Phytochemical bulletin 19:11–15.

Ehlers J, Gibbard PL (2004) Quaternary Glaciations: Extent and Chronology. Elsevier, Amsterdam

Ehrich D (2006) AFLPdat: a collection of R functions for convenient handling of AFLP data. Molecular

Ecology Notes 6:603–604.

Evanno G, Regnaut S, Goudet J (2005) Detecting the number of clusters of individuals using the

software structure: a simulation study. Molecular ecology 14:2611–2620.

Evans MEK, Dolan RW, Menges ES, Gordan DR (2000) Genetic diversity and reproductive biology in

Warea carteri (Brassicaceae), a narrowly endemic Florida scrub annual. American Journal of

Botany 87:372–381.

Excoffier L, Laval G, Schneider S (2005) Arlequin (version 3.0): An integrated software package for

population genetics data analysis. Evolutionary Bioinformatics 1:47–50.

47 Falush D, Stephens M, Pritchard JK (2007) Inference of population structure using multilocus

genotype data: dominant markers and null alleles. Molecular Ecology Notes 7:574–578.

Fenneman NM (1938) Physiography of Eastern United States. McGraw‐Hill, New York.

Flora of North America Editorial Committee (1993) Flora of North America: North of Mexico. Oxford

University Press, New York.

French HM, Demitroff M, Forman SL (2003) Evidence for Late‐Pleistocene permafrost in the New

Jersey Pine Barrens (Latitude 39 degrees N), Eastern USA. Permafrost and Periglacial

Processes 14:259–274.

French HM, Demitroff M, Forman SL, Newell WL (2007) A chronology of Late‐Pleistocene permafrost

events in southern New Jersey, eastern USA. Permafrost and Periglacial Processes 18:49–59.

Frost C (2006) History and future of the longleaf pine ecosystem. In: Jose S, Jokela EJ, Miller DL (eds)

The Longleaf Pine Ecosystem: Ecology, Silvaculture, and Restoration. Springer Science, New

York, pp 9–42.

Fu YX (1997) Statistical tests of neutrality of mutations against population growth, hitchhiking and

background selection. Genetics 147:915–925.

Fu YX, Li WH (1993) Statistical tests of neutrality of mutations. Genetics 133:693–709.

Godt MJW, Hamrick JL (1995) Low levels of allozyme differentiation between Pyxidanthera (pyxie‐

moss) taxa (Diapensiaceae). Plant Systematics and Evolution 195:159–168.

48 Gonzales E, Hamrick JL, Chang S‐M (2008) Identification of glacial refugia in south‐eastern North

America by phylogeographical analyses of a forest understorey plant, Trillium cuneatum.

Journal of Biogeography 35:844–852.

Hafner MS, Nadler SA (1990) Cospeciation in host‐parasite assemblages: comparative analysis of

rates of evolution and timing of cospeciation events. Systematic Biology 39:192–204.

Herrera CM, Bazaga P (2008) Population‐genomic approach reveals adaptive floral divergence in

discrete populations of a hawk moth‐pollinated violet. Molecular Ecology 17:5378–5390.

Hewitt G (2000) The genetic legacy of the Quaternary ice ages. Nature 405:907–913.

Hey J, Nielsen R (2007) Integration within the Felsenstein equation for improved Markov chain

Monte Carlo methods in population genetics. Proceedings of the National Academy of

Sciences 104:2785–2790.

Hobbs III CH (2004) Geological history of Chesapeake Bay, USA. Quaternary Science Reviews 23:641–

661.

Holland BR, Clarke AC, Meudt HM (2008) Optimizing Automated AFLP Scoring Parameters to

Improve Phylogenetic Resolution. Systematic Biology 57:347.

Hutchinson GE (1957) Concluding remarks. Cold Springs Harbor Symposia on Quantitative Biology

22:415–427.

49 Hutchison DW, Templeton AR (1999) Correlation of pairwise genetic and geographic distance

measures: inferring the relative influences of gene flow and drift on the distribution of

genetic variability. Evolution 53:1898–1914.

Ikeda H, Senni K, Fujii N, Setoguchi H (2008) Survival and genetic divergence of an arctic‐alpine plant,

Diapensia lapponica subsp obovata (Fr. Schm.) Hulten (Diapensiaceae), in the high

mountains of central Japan during climatic oscillations. Plant Systematics and Evolution

272:197–210.

Jacobson GL, Webb III T, Grimm EC (1987) Patterns and rates of vegetation change during the

deglaciation of eastern North America. In: Ruddiman WF, Wright, Jr. HE (eds) North America

and Adjacent Oceans during the Last Deglaciation. The Geology of North America.

Geological Society of America, Boulder, Colorado, pp 277–288.

Kimura M (1969) The number of heterozygous nucleotide sites maintained in a finite population due

to steady flux of mutations. Genetics 61:893–903.

Librado P, Rozas J (2009) DnaSP v5: a software for comprehensive analysis of DNA polymorphism

data. Bioinformatics 25:1451–1452.

Magni CR, Ducousso A, Caron H, et al. (2005) Chloroplast DNA variation of Quercus rubra L in North

America and comparison with other Fagaceae. North America and comparison with other

Fagaceae Molecular ecology 14:513–524.

50 Maliouchenko O, Palme AE, Buonamici A, et al. (2007) Comparative phylogeography and population

structure of European Betula species, with particular focus on B. pendula and B. pubescens.

Journal of Biogeography 34:1601–1610.

McKinnon GE, Vaillancourt RE, Steane DA, Potts BM (2008) An AFLP marker approach to lower‐level

systematics in Eucalyptus (Myrtaceae). American Journal of Botany 95:368–380.

McLachlan JS, Clark JS, Manos PS (2005) Molecular indicators of tree migration capacity under rapid

climate change. Ecology 86:2088–2098.

Meudt HM, Bayly MJ (2008) Phylogeographic patterns in the Australasian genus Chionohebe

(Veronica s.l., Plantaginaceae) based on AFLP and chloroplast DNA sequences. Molecular

Phylogenetics and Evolution 47:319–338.

Morris AB, Ickert‐Bond SM, Brunson DB, et al. (2008) Phylogeographical structure and temporal

complexity in American sweetgum (Liquidambar styraciflua; Altingiaceae). Molecular

Ecology 17:3889–3900.

Mylecraine KA, Kuser JE, Smouse PE, Zimmermann GL (2004) Geographic allozyme variation in

Atlantic white‐cedar, Chamaecyparis thyoides (Cupressaceae). Canadian Journal of Forest

Research 34:2443–2454.

Nei M (1972) Genetic distance between populations. American Naturalist 106:283–292.

Nei M (1987) Molecular evolutionary genetics. Columbia University Press, New York.

51 Nielsen R, Wakeley J (2001) Distinguishing migration from isolation: A Markov chain Monte Carlo

approach. Genetics 158:885–896.

Oksanen J, Kindt R, Legendre P, et al. (2009) vegan: Community Ecology Package.

Oliveira LO, Huck RB, Gitzendanner MA, et al. (2007) Molecular phylogeny, biogeography, and

systematics of Dicerandra (Lamiaceae), a genus endemic to the southeastern United States.

American Journal of Botany 94:1017–1027.

Overpeck JT, Webb RS, Webb III T (1992) Mapping eastern North American vegetation change of the

past 18 ka: No‐analogs and the future. Geology 20:1071.

Palme AE, Semerikov V, Lascoux M (2003) Absence of geographical structure of chloroplast DNA

variation in sallow, Salix caprea L. Heredity 91:465–474.

Paradis E (2009) Pegas: Population and Evolutionary Genetics Analysis System.

Paun O, Schönswetter P, Winkler M, et al. (2008) Historical divergence vs. contemporary gene flow:

evolutionary history of the calcicole Ranunculus alpestris group (Ranunculaceae) in the

European Alps and the Carpathians. Molecular Ecology 17:4263–4275.

Perez‐Collazos E, Sanchez‐Gomez P, Jimenez JF, Catalan P (2009) The phylogeographical history of

the Iberian steppe plant Ferula loscosii (Apiaceae): a test of the abundant‐centre hypothesis.

Molecular Ecology 18:848–861.

52 Pons O, Petit RJ (1996) Measuring and testing genetic differentiation with ordered versus unordered

alleles. Genetics 144:1237–1245.

Price EW, Carbone I (2005) SNAP: workbench management tool for evolutionary population genetic

analysis. Bioinformatics 21:402–404.

Primack RB, Wyatt R (1975) Variation and taxonomy of Pyxidanthera (Diapensiaceae). Brittonia

27:115–118.

Pritchard JK, Stephens M, Donnelly P (2000) Inference of population structure using multilocus

genotype data. Genetics 155:945–959.

R Development Core Team (2012) R: a Language and environment for statistical computing. Vienna,

Austria

Rahmstorf S (2002) Ocean circulation and climate during the past 120, 000 years. Nature 419:207–

214.

Ramos‐Onsins SE, Rozas J (2002) Statistical properties of new neutrality tests against population

growth. Molecular Biology and Evolution 19:2092–2100.

Reynolds JD (1966) Morphological Studies in Diapensiaceae. The Embryology of Pyxidanthera Mich.

Doctoral dissertation, University of South Carolina

Roberts DW (2010) labdsv: Ordination and Multivariate Analysis for Ecology.

53 Ronblom K, Anderberg AA (2002) Phylogeny of Diapensiaceae based on molecular data and

morphology. Systematic Botany 27:383–395.

Schafale MP, Weakley AS (1990) Classification of the natural communities of North Carolina: Third

approximation. pp. 321.

Schönswetter P, Tribsch A (2005) Vicariance and dispersal in the alpine perennial Bupleurum

stellatum L.(Apiaceae). Taxon 725–732.

Shaw J, Lickey EB, Beck JT, et al. (2005) The tortoise and the hare II: Relative utility of 21 noncoding

chloroplast DNA sequences for phylogenetic analysis. American Journal of Botany 92:142–

166.

Shaw J, Lickey EB, Schill EE, Small RL (2007) Comparison of whole chloroplast genome sequences to

choose noncoding regions for phylogenetic studies in angiosperms: The tortoise and the

hare III. American Journal of Botany 94:275–288.

Soltis DE, Morris AB, McLachlan JS, et al. (2006) Comparative phylogeography of unglaciated eastern

North America. Molecular Ecology 15:4261–4293.

Sorrie BA, Gray JB, Crutchfield PJ (2006) The vascular flora of the longleaf pine ecosystem of Fort

Bragg and Weymouth Woods, North Carolina. Castanea 71:129–161.

Sorrie BA, Weakley AS (2001) Coastal plain vascular plant endemics: Phytogeographic patterns.

Castanea 66:50–82.

54 Sorrie BA, Weakley AS, Neesom GL Pyxidanthera. Flora of North America North of Mexico. pp 336–

337

Stewart JR, Lister AM (2001) Crypitc refugia and the origins of the modern biota. Trends in Ecology

and Evolution 16:608–613.

Takhtajan A (1986) Floristic Regions of the World. University of California Press, Berkeley, CA

Tamura K, Dudley J, Nei M, Kumar S (2007) MEGA4: Molecular evolutionary genetics analysis

(MEGA) software version 4.0. Molecular Biology and Evolution 24:1596–1599.

Templeton AR, Crandall KA, Sing CF (1992) A cladistic analysis of phenotypic associations with

haplotypes inferred from restriction endonuclease mapping and DNA sequence data. III.

Cladogram estimation. Genetics 132:619–633.

Trybush S, Hanley S, Cho K‐H, et al. (2006) Getting the most out of fluorescent amplified fragment

length polymorphism. Canadian Journal of Botany 84:1347–1354.

Vekemans X, Beauwens T, Lemaire M, Roldán‐Ruiz I (2002) Data from amplified fragment length

polymorphism (AFLP) markers show indication of size homoplasy and of a relationship

between degree of homoplasy and fragment size. Molecular Ecology 11:139–151.

Ward LW (1992) Tertiary molluscan assemblages from the Salisbury embayment of Virginia. Virginia

Journal of Science 43:85–100.

55 Watterson GA (1975) On the number of segregating sites in genetical models without

recombination. Theoretical Population Biology 7:256.

Weakley AS (2008) Flora of the Carolinas, Virginia, Georgia, northern Florida, and surrounding areas

(Working Draft of 7 April 2008). Univerisity of North Carolina Herbarium, Chapel Hill, NC

Wells BW (1929) A new pixie from North Carolina. Journal of the Elisha Mitchell Scientific Society

44:238–239.

Wells BW, Shunk IV (1931) The vegetation and habitat factors of the coarser sands of the North

Carolina coastal plain: An ecological study. Ecological Monographs 1:465–520.

Zhang YY, Zhang DA, Barrett SCH (2010) Genetic uniformity characterizes the invasive spread of

water hyacinth (Eichhornia crassipes), a clonal aquatic plant. Molecular Ecology 19:1774–

1786.

56 Table 3.1: Chloroplast haplotype accession numbers as archived in Genbank for the atpI‐atpH intergenic spacer region (partial sequence) and the psbD‐trnT intergenic spacer region (partial sequence). Chloroplast haplotypes H1 through H12 refer to unique composite sequences from the two intergenic spacer regions.

chloroplast atpI‐atpH psbD‐trnT haplotype haplotype accession number haplotype accession number H1 atp1 HM564385 psb1 HM564379 H2 atp2 HM564386 psb2 HM564380 H3 atp2 HM564386 psb3 HM564381 H4 atp3 HM564387 psb4 HM564382 H5 atp4 HM564388 psb2 HM564380 H6 atp5 HM564389 psb2 HM564380 H7 atp6 HM564390 psb2 HM564380 H8 atp6 HM564390 psb6 HM564384 H9 atp6 HM564390 psb4 HM564382 H10 atp7 HM564391 psb2 HM564380 H11 atp1 HM564385 psb2 HM564380 H12 atp1 HM564385 psb5 HM564383

57 Table 3.2: Polymorphisms of the 12 cpDNA haplotypes based on the cpDNA regions atpI‐atpH and psbD‐trnT in the genus Pyxidanthera. Numbers below cpDNA regions denote position in sequence. N refers to the number of individuals identified as the corresponding haplotype, while dashes represent correspondence to consensus haplotype.

chloroplast atpI‐atpH psbD‐trnT haplotype N 370 390 399 455 509 599 763 829 851 502 504 672 782 867 consensus ‐ T T 1 G A 1 C T A G G T G G H1 6 ‐ ‐ ‐ ‐‐‐‐‐‐ ‐‐ G ‐ ‐ H2 12 ‐ ‐ ‐ ‐‐‐‐C T ‐ ‐ ‐ ‐‐ H3 3 ‐ ‐ ‐ ‐‐‐‐C ‐ A ‐ ‐ ‐‐ H4 2 ‐ ‐ ‐ ‐C ‐ G ‐‐ ‐‐ ‐ A ‐ H5 8 ‐ ‐ ‐ ‐ ‐ 2 ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ H6 4 ‐ ‐ ‐ C C ‐‐‐‐ ‐‐ ‐ ‐‐ H7 9 ‐ ‐ ‐ ‐C ‐‐‐‐ ‐‐ ‐ ‐‐ H8 8 ‐ ‐ ‐ ‐C ‐‐‐‐ ‐‐ ‐ ‐A H9 11 ‐ ‐ ‐ ‐C ‐‐‐‐ ‐‐ ‐ A ‐ H10 3 C C 2 ‐‐‐‐‐‐ ‐‐ ‐ ‐‐ H11 37 ‐ ‐ ‐ ‐‐‐‐‐‐ ‐‐ ‐ ‐‐ H12 2 ‐ ‐ ‐ ‐‐‐‐‐‐ ‐A ‐ ‐ ‐

58 Table 3.3: Analyses of Molecular Variance (AMOVA) results for Pyxidanthera barbulata using cpDNA sequences and AFLP markers. *** indicates p‐value < 0.001, ** p‐value < 0.01, * p‐ value < 0.05, and NS indicates non‐significance of variation.

AFLP cpDNA

Source of variation d.f. variance % of d.f. variance % of

variation variation

Grouped by species

Between species 1 0.37 1.60** 1 ‐0.01 ‐1.71 NS

Among populations within species 23 1.85 7.87*** 22 0.63 72.22***

Within populations 412 21.30 90.53*** 81 0.26 29.49***

Grouped by region (North vs. South)

Between regions 1 0.76 3.20** 1 0.17 17.27*

Among populations within regions 23 1.86 7.79*** 22 0.56 56.58***

Within populations 412 21.30 89.01*** 81 0.26 26.15***

59 Table 3.4: Genetic diversity indices for P. barbulata and P. brevifolia based on cpDNA sequences and AFLP markers. AFLP genetic diversity indices were only calculated for populations with more than 7 genotyped individuals (437 total specimens). %P represents the number of polymorphic loci, DW is a measure of rare alleles per population, and He is a measure of expected heterozygosity based on the AFLP markers. π is a measure of cpDNA nucleotide diversity. N represents the number of specimens for each population for AFLP markers and cpDNA sequences (in parentheses).

Population Species State N %P DW He π Haplotypes

NC_1 barbulata NC 14(5) 41.0 12.99 0.15 0.0019 H1,H6,H11

NC_2 barbulata NC 0(5) ‐‐ ‐0.0000 H11

NC_3 barbulata NC 0(4) ‐‐ ‐0.0000 H2

NC_4 barbulata NC 9(4) 50.0 7.89 0.14 0.0000 H11

NC_6 barbulata NC 19(1) 26.5 7.40 0.08 0.0000 H7

NC_8 barbulata NC 12(4) 23.2 4.09 0.09 0.0005 H7,H11

NC_9 barbulata NC 19(6) 44.2 18.91 0.15 0.0013 H5,H10

NJ_CB barbulata NJ 0(5) ‐‐ ‐0.0010 H7,H11,H12

NJ_CW barbulata NJ 17(5) 34.2 14.41 0.11 0.0012 H4,H8,H9

NJ_WB barbulata NJ 15(6) 27.1 9.48 0.10 0.0000 H11

NJ_WG barbulata NJ 18(5) 38.7 18.69 0.12 0.0004 H4,H9

NY_1 barbulata NY 0(5) ‐‐ ‐0.0004 H7,H9

SC_1 barbulata SC 8(3) 50.3 12.67 0.16 0.0000 H3

VA_1 barbulata VA 13(5) 32.6 10.08 0.12 0.0019 H1,H6

SC_HP brevifolia SC 8(5) 25.2 1.83 0.08 0.0004 H7,H8

SC_SL brevifolia SC 26(5) 26.1 8.33 0.08 0.0000 H11

002A brevifolia NC 19(5) 41.0 16.25 0.12 0.0000 H11

10 brevifolia NC 17(0) 47.4 23.21 0.15 ‐ ‐

20 brevifolia NC 21(5) 51.3 23.17 0.14 0.0008 H2,H11

24 brevifolia NC 19(0) 46.1 17.26 0.14 ‐ ‐

026D brevifolia NC 19(1) 39.0 9.66 0.12 0.0000 H1

03_25 brevifolia NC 30(5) 36.1 15.18 0.11 0.0008 H2,H11

028E brevifolia NC 0(4) ‐‐ ‐0.0000 H7

60 Table 3.4, continued

Population Species State N %P DW He π Haplotypes

33 brevifolia NC 33(0) 31.3 22.40 0.11 ‐ ‐

038D brevifolia NC 15(0) 41.9 13.65 0.14 ‐ ‐

057Y brevifolia NC 24(3) 41.9 25.43 0.13 0.0000 H8

058B brevifolia NC 22(0) 40.6 18.61 0.14 ‐ ‐

065N brevifolia NC 15(5) 31.6 7.24 0.11 0.0000 H11

066A brevifolia NC 0(4) ‐‐ ‐0.0000 H11

092B brevifolia NC 8(0) 34.2 3.26 0.11 ‐ ‐

93_115 brevifolia NC 17(0) 43.2 20.41 0.14 ‐ ‐

Overall mean 37.8 13.70 0.12 0.0004

barbulata 36.8 11.66 0.12 0.0006

brevifolia 38.5 15.06 0.12 0.0002

Northern 33.3 14.20 0.11 0.0006

Southern 38.5 13.63 0.12 0.0004

61 Table 3.5: Summary of model statistics for the 24 IMa2 models. Included for each model (left to right) are the negative log of the probability, the number of parameters, the degrees of freedom when compared to the full model, AIC and ΔAIC, the likelihood of the model, the model probability, and the evidence ratio for each model, calculated according to Burnham and Anderson (2002).

Model log(P) Pars df AIC ΔAIC L wi wi/wj

ABCD0 ‐4.537 4 1 17.074 0.000 1.000 0.241 1.000

AACDD ‐6.009 3 2 18.018 0.944 0.624 0.151 1.603

ABCDE ‐4.537 5 0 19.074 2.000 0.368 0.089 2.718

ABCDD ‐5.818 4 1 19.636 2.562 0.278 0.067 3.600

ABC0E ‐5.87 4 1 19.740 2.666 0.264 0.064 3.792

AAA0E ‐7.953 2 3 19.906 2.832 0.243 0.059 4.121

AACDE ‐5.968 4 1 19.936 2.862 0.239 0.058 4.183

ABB0E ‐7.205 3 2 20.410 3.336 0.189 0.046 5.302

ABA0E ‐7.428 3 2 20.856 3.782 0.151 0.036 6.626

AAC0E ‐7.645 3 2 21.290 4.216 0.121 0.029 8.232

AACD0 ‐7.84 3 2 21.680 4.606 0.100 0.024 10.004

AAADE ‐7.953 3 2 21.906 4.832 0.089 0.022 11.201

AAADD ‐9.135 2 3 22.270 5.196 0.074 0.018 13.437

AAAE0 ‐9.155 2 3 22.310 5.236 0.073 0.018 13.708

ABBDE ‐7.205 4 1 22.410 5.336 0.069 0.017 14.411

ABADE ‐7.428 4 1 22.856 5.782 0.056 0.013 18.011

ABBDD ‐8.772 3 2 23.544 6.470 0.039 0.009 25.406

ABADD ‐8.918 3 2 23.836 6.762 0.034 0.008 29.400

ABAD0 ‐9.149 3 2 24.298 7.224 0.027 0.007 37.040

ABBD0 ‐9.149 3 2 24.298 7.224 0.027 0.007 37.040

AAA00 ‐11.33 1 4 24.660 7.586 0.023 0.005 44.389

AAC00 ‐10.36 2 3 24.720 7.646 0.022 0.005 45.741

ABB00 ‐10.59 2 3 25.180 8.106 0.017 0.004 57.570

ABA00 ‐11.23 2 3 26.460 9.386 0.009 0.002 109.180

ABC00 ‐10.36 3 2 26.720 9.646 0.008 0.002 124.338

62

Fig. 3.1. Morphological variation in leaf length, leaf width, and pubescence of P. barbulata (circles) and P. brevifolia (triangles). Solid triangles represent P. brevifolia specimens that had pubescence for half or less than half of the leaf; open triangles represent P. brevifolia specimens with pubescence greater than half of the leaf. Although there are statistically significant differences between the two varieties for leaf length, leaf width, and pubescence, there is considerable overlap between the two species.

63

Fig. 3.2. Geographic distribution (shaded in grey) and statistical parsimony network for 12 haplotypes from 2 cpDNA regions of Pyxidanthera. State names are in bold abbreviations and numbers represent haplotypes from Table 3.2. Black dots in the haplotype network represent mutational steps; associated letters (S for South, N for North) are the most likely (>95% probability) geographic origins of mutations, inferred using Genetree 9.0. Light grey shading of haplotype network represents proportion of the associated haplotype comprised of southern individuals, and darker grey shading represents proportion comprised of northern individuals. Inset map: Sampling of Pyxidanthera populations on Fort Bragg Military Reservation. Pyxidanthera barbulata populations are represented by closed circles and P. brevifolia populations are represented by open circles.

64

Fig. 3.3. Isolation by distance for cpDNA (left side) and AFLP (right side) markers across the entire range of Pyxidanthera barbulata. cpDNA demonstrates no isolation by geographic distance (R = 0.01, p‐value = 0.39), while AFLP markers demonstrate weak but significant (R = 0.27, p‐value = 0.02) isolation by distance at shorter distances with effects of genetic drift more evident at greater distances.

65

Fig. 3.4. Parameter estimates for θ (southern, northern, and ancestral populations), time since divergence, and migration (gene flow) between northern and southern populations of the genus Pyxidanthera based on results from IMa2.

66

Fig. 3.5. Non‐metric multidimensional scaling ordination of P. barbulata and P. brevifolia population genetic distances (Nei’s D) based on amplified fragment length polymorphism markers. In the legend, letters in parentheses represent US states. Little separation is evident among populations defined according to either taxonomic status or geographical location.

67 Fig. 3.6. Population genetic structure for the genus Pyxidanthera as inferred from the program STRUCTURE for K 2 through 9. The individual columns represent individual genetic samples and the colors represent proportion of ancestry assigned to the different ancestral populations. Little genetic structuring is evident based on either geographic location (South vs. North) or taxonomic identity (var. barbulata vs. var. brevifolia).

68

69

Fig. 3.7. Log likelihood (Ln P(D)) and standard deviation results from program STRUCTURE for K 1 through 9. Runs included a burn‐in length of 10 000 and a post burn‐in length of 25 000 with admixture and correlated allele frequencies.

70 CHAPTER 4 EVIDENCE OF POPULATION BOTTLENECK IN THE ENDEMIC PLANT SPECIES ASTRAGALUS MICHAUXII (KUNTZE) F.J. HERM.

4.1 Abstract

The Gulf and Atlantic Physiographic Region (GACP) in the southeastern United States has the second highest levels of endemism in the US, but many species are threatened by fire suppression and land use change. Much of the GACP was covered by the longleaf pine ecosystem; The 37 million hectares of presettlement longleaf pine ecosystem has been reduced to 2% of the original range, resulting in catastrophic population declines for many species dependent on this ecosystem. Astragalus michauxii is a legume endemic to the longleaf pine ecosystem and is considered vulnerable to extinction, with populations generally consisting of small numbers of individuals that are geographically isolated. In order to explore the genetic diversity, structure, and contemporary gene flow between populations, we developed eight polymorphic microsatellites and genotyped 355 individuals from 22 populations. Our results suggest that genetic diversity is comparable across regions and populations and that within‐population genetic variation accounted for 92% of the total genetic variation. For the North Carolina populations, results suggest that there is contemporary gene flow. Finally, the mean ratios of alleles to the allelic range (M ratio) are low for A. michauxii populations, suggesting that the species experienced a severe genetic bottleneck in the past.

Although genetic factors can affect the survival of plant species and A.michauxii has few rare alleles, maintaining habitats through prescribed burning may be more critical to population persistence since gene flow is continuing to occur and population differentiation is low.

71 4.2 Introduction

Before European settlement of the southeastern United States, the longleaf pine (Pinus palustris Mill.) ecosystem covered roughly 37 million hectares and stretched from east Texas to

Florida and north to southeastern Virginia (Frost 1993), with most of the longleaf pine occurring in the Gulf and Atlantic Coastal Plain (GACP) physiographic region. Across most of its range, longleaf pine was the dominant canopy tree, but below this species‐poor canopy existed a diverse array of habitats and species (Peet 2006), with the GACP having the second‐highest levels of endemism in the United States (Sorrie and Weakley 2001). Among temperate regions, the longleaf pine ecosystem has been recognized as one of the most species‐rich at multiple spatial scales (Walker and Peet 1983). Since European settlement, approximately 98% of the original habitat has been destroyed due to development, fire suppression, and conversion to agriculture and plantations

(Frost 2006). Conversion and degradation of the longleaf pine ecosystem has led to population decline and population isolation for many species that are endemic to the region. Astragalus michauxii (Kuntze) F.J. Herm. (Sandhills milkvetch), a rare legume and the subject of this paper, is among these.

A. michauxii is an herbaceous, long‐lived perennial that is found in what has been characterized as the pine/scrub oak sandhill vegetation community (Schafale and Weakley 1990).

The largest populations are found in the loamy soil variant of this community type, known locally as

“pea swales” or “bean dips”, due to their high diversity of Fabaceae spp. (James 2000). The species is endemic to the Fall‐line Sandhills region of North Carolina, South Carolina, Georgia, and Alabama

(USA) (Sorrie and Weakley 2001). The Fall‐line Sandhills, located at the boundary between the

72 Coastal Plain and the Piedmont physiographic regions, are characterized by a rolling topography with excessively‐well drained, sandy soils in the interfluvial areas.

At the current time, A. michauxii is vulnerable to extinction and has been given a global status of G3 by NatureServe (NatureServe 2012). In Georgia it has an S2 ranking, indicating that it is present in 20 or fewer populations and is considered imperiled. Astragalus michauxii has an S3 ranking in North Carolina and South Carolina. In North Carolina most of the populations are found on Fort Bragg Military Reservation and the Sandhills Gamelands. In South Carolina, A. michauxii is only known from two populations. Population sizes are generally small, with fewer than 20 individuals (W. Wall, unpublished data). As such, many of the included populations may not be viable due to habitat fragmentation and low population size.

Temporal changes in population size leave a genetic signature. Reduced genetic diversity is more common in species with restricted ranges and reduced population sizes, relative to more widespread species (Hamrick and Godt 1996). Reduced population size can lead to inbreeding or genetic drift (Ellstrand and Elam 1993), which can further reduce genetic diversity. Lower genetic diversity has been observed in threatened species relative to more common species (Spielman et al.

2004), suggesting that genetic factors can increase the likelihood of extinction of species by reducing fitness (Leimu et al. 2006)and restricting evolutionary potential (Franklin 1980). An understanding of the partitioning of genetic diversity is one of the cornerstones of conservation biology (Frankham et al. 2010), and for many species provides key insights into the biology and management of threatened and endangered species (Avise 1989). To investigate the effect of these various factors on the genetic structure and variation of A. michauxii, we developed eight polymorphic microsatellite markers. We estimated the within‐ and between‐population partitioning of genetic

73 variation, assessed gene flow between populations and analyzed the data for evidence of past genetic bottleneck events.

4.3 Methods

4.3.1 Population genetic methods

Leaf tissues (whole leaves) were collected from populations on Fort Bragg and Camp

MacKall Military Reservations (North Carolina) and from two populations in Georgia, and these were stored in a ‐80° C freezer. DNA was extracted from 358 individuals across 24 populations (38 subpopulations) on Fort Bragg and Camp MacKall, and 9 individuals from Georgia (Fig. 4.1), using the

CTAB method with minor modifications (Doyle and Doyle 1987). Possible polymorphic microsatellite regions were identified using a recently published protocol (Jennings et al. 2011). Briefly, the extracted DNA was first sheared using a sonicator (Diagenode Inc., Denville, NJ, USA) and DNA libraries were created (Cronn et al. 2008). For each microsatellite hybridization selection, three dinucleotide microsatellite probes were added. The mixture was heated and incubated for hybridization. After hybrid capture and elution, the microsatellite‐enriched library was concentrated, cleaned, and quantified using the Nanodrop 1000 spectrophotometer (Thermo

Scientific,Wilmington, DE, USA). Libraries were then pooled and sequenced using an Illumina

Genome Analyzer II at 5 pM concentration. The resulting microreads were then sorted by the 4 bp barcodes, searched for dinucleotide and trinucleotide motifs, and filtered for redundant reads and placement of microsatellites (center of reads to optimize primer development). The resulting filtered sequences were then analyzed using BatchPrimer3 to identify PCR primer sites (You et al.

2008).

74 We optimized the microsatellite‐containing sequences and identified eight polymorphic microsatellite‐containing sequences. We used Batch Primer3 to design marker amplification primers. 6µL multiplexed PCR reactions were performed with fluorescently labeled primers (6‐FAM,

HEX , NED) as follows: 3 µL Qiagen multiplex PCR master mix (Qiagen, Hilden, Germany), 0.6 µL Q‐

Solution, 0.8 µL H20, 0.6 µL multiplexed primer pair mix, and 1.0 µL diluted (1:8 DNA:H20) DNA. PCR cycling conditions were 95˚ C for 15 minutes; 45 cycles at 94˚ C for 30 sec, 58˚ C for 1 minute 30 sec., and 72˚ C for 1 minute; 60˚ C for 30 minutes. PCR products were genotyped on an ABI 3730 sequencer (Applied Biosystems, USA). PCR was performed again for unamplified loci. After a second failed amplification, a locus was marked as missing. Genotyping was performed using GeneMarker

1.8 (Softgenetics LLC, State College, PA, USA). Exported peak heights were binned using TANDEM

(Matschiner and Salzburger 2009). We genotyped 5% of the individuals twice in order to assess data quality. Consistency across all loci in duplicate samples was 95.5%.

4.3.2 Genetic structure and diversity

We tested for significant departures from Hardy‐Weinberg equilibrium using Arlequin 3.1 to perform Fisher’s exact test with a chain of 1,000,000 iterations and a burn‐in of 100,000 iterations

(Guo and Thompson 1992). We tested for linkage disequilibrium within populations and across loci using Arlequin 3.1 using Fisher’s exact test with 10,100 permutations. Since adjustments for multiple comparisons can make it difficult to find significance, even when it may exist (Moran 2003), we assessed evidence of linkage disequilibrium using p‐values both unadjusted and adjusted by the

Bonferroni correction for multiple comparisons (Bonferroni 1935). We calculated the average number of alleles and absolute number of private alleles for each of the populations (Table 2).

75 Because populations consisted of varying numbers of individuals, we used a rarefaction method

(Kalinowski 2004) for calculating allelic richness and private allelic richness available in HP‐Rare 1.1

(Kalinowski 2005). We calculated expected and observed heterozygosity (Nei 1987)using Arlequin

3.1 (Excoffier et al. 2005).

Population differentiation and structure were first explored by performing a non‐metric multi‐dimensional scaling (NMDS) ordination using the R (R Development Core Team 2012) package labdsv (Roberts 2010). Pairwise population genetic distances were calculated using Nei’s unbiased D

(Nei 1972) as implemented in GenAlex (Peakall and Smouse 2006). We performed an analysis of molecular variance (AMOVA) using the locus‐by‐locus method as implemented in Arlequin 3.1, with the samples hierarchically partitioned into ultrapopulations and populations within ultrapopulations

(see groupings in Table 2). We explored population structure in A. michauxii populations using the software program STRUCTURE (Falush et al. 2007; Pritchard et al. 2000). For population numbers 1 through 9 (K=1‐9), we performed three runs with a burn‐in length of 750,000 and a post burn‐in length of 250,000, with the assumptions of admixture and correlated allele frequencies. We identified the number of populations with the maximum likelihood graphically using the log likelihoods for each K. Ln P(D) did not increase monotonically to the optimal K (Herrera and Bazaga

2008), so we did not use the methods of Evanno et al. (2005). To test for isolation by distance (IBD), we performed a Mantel test using the R package vegan (Oksanen et al. 2009), with the geographic distances log‐transformed and the pairwise population genetic distances (unbiased Nei’s D).

76 4.3.3 Evidence of genetic bottlenecks across multiple temporal scales

Identification of genetic bottlenecks is important for conservation because bottlenecks can increase the risk of extinction. To detect the genetic imprint of recent bottlenecks in A. michauxii, we used the software package BOTTLENECK 1.2.02 (Piry et al. 1999), an implementation of the method described by Cornuet and Luikart for detecting recent bottleneck events (1996). Assuming a population is at mutation drift equilibrium, a population that has experienced a recent bottleneck event (<4Ne generations) should have both a reduced number of alleles and reduced heterozygosity

(He). The allele deficiency is influenced by four factors: time since the bottleneck event, effective population size ratio before and after the bottleneck, the mutation rate of the locus, and the sample size of genes (Maruyama and Fuerst 1985). Although allele number and He both decrease during a bottleneck event, allele number decreases more rapidly than He (Cornuet and Luikart 1996). The heterozygosity at mutation‐drift equilibrium (Heq) is calculated from the allele number, so as allele number decreases more rapidly than He, Heq should be less than He for a limited number of generations following the bottleneck. BOTTLENECK 1.2.02 detects evidence of recent bottlenecks by testing whether Heq < He, which would result from an excess of heterozygosity. We fit two models: the stepwise mutation model (SMM), in which microsatellites gain or lose only one repeat per time step, and the two‐phase model (TPM) mutation model with 95% of the mutations single‐step and a variance of 12 (Piry et al. 1999). For both models we ran 10,000 simulations and used a one‐tailed

Wilcoxon signed rank test to detect significant heterozygote excess in populations because this test is most appropriate when the sample size is less than 30 and the number of loci is less than 10. We only tested for evidence of recent bottlenecks in population that had > 20 gene copies.

77 To test for bottleneck events that may have occurred over longer time periods (>100 generations), we used the M ratio test (Garza and Williamson 2001) as implemented in Arlequin

3.11. This implementation of the M ratio test excludes monomorphic loci because these can erroneously increase the M ratio. The M ratio is the mean number of alleles in a population divided by the allelic size range. When alleles are lost from a population, the number of alleles decreases at a faster rate than the allelic size range, so small M ratio values are indicative of populations that have gone through a genetic bottleneck at some time in the past. We estimated 95% confidence intervals through resampling the M ratio estimates for each locus within a population for 10,000 replicates (Swatdipong et al. 2009). We also calculated the Mc estimate for each population. Mc is a value estimated through 10,000 simulations such that M > Mc in 95% of the simulations based on the mean size of non‐stepwise mutations and θ. Settings were mean size of non‐stepwise mutations

= 3.5, θ = 10, and 10,000 iterations, as recommended (Garza and Williamson 2001).

4.3.4 Estimating gene flow between populations

Understanding gene flow between populations in recent time scales is of fundamental concern to conservation biologists, as these processes are more likely to affect future gene flow patterns. To detect possible first generation migrants between populations we used GeneClass2

(Piry et al. 2004), an implementation of the Bayesian method of Rannala and Mountain (1997). This method assumes no disequilibrium and HWE within populations. We computed probabilities for each individual being a recent migrant using the simulation algorithm of Paetkau et al (2004) with a type I error threshold of 0.01, 1000 simulated individuals, and all loci included. Assignment procedures like GeneClass2 are particularly useful when genetic differentiation is low; they allow

78 one to identify possible recent migration events (Manel et al. 2005). Gene flow between populations may follow a source‐sink metapopulation model, with gene flow from large populations to small populations. To test whether the probability of being a migrant was correlated with population size

(i.e. whether larger populations contribute disproportionately to the number of migrants), we performed a logistic regression with migration as the response variable and population size as the explanatory variable. We tested for significant differences in population size for resident and assigned populations of the identified migrants by simulating 10,000 replicate data sets with replacement and calculating the 95% confidence interval for mean population size for each group of populations (resident or assigned).

4.4 Results

The microsatellite amplification generated 100 candidate loci, from which we developed eight polymorphic microsatellite loci that amplified consistently (Table 4.1). Allele frequencies detected significant departures from Hardy‐Weinberg expectations in 19 out of 176 tests at the population level. The northern populations on Fort Bragg had more loci out of Hardy‐Weinberg equilibrium compared to populations from other regions. Microsatellite marker AM29 was out of

Hardy‐Weinberg equilibrium in eight of 22 populations and three of four regions, indicating that this locus may be influenced by null alleles or allelic dropout. After correcting for multiple tests, none of loci were significantly out of Hardy‐Weinberg equilibrium (p‐value> 0.05). Linkage disequilibrium was detected in 91 of 616 tests, although only one comparison was significant after using Bonferroni adjustment. Significant evidence of linkage disequilibrium was not isolated to pairs of loci, but rather was encountered in all loci compared and was never greater than 30% (Table 4.2). The linkage

79 disequilibrium observed across loci could be due to a number of non‐exclusive factors, including sampling bias, admixture of populations, inbreeding, or genetic drift due to a bottleneck event (Ohta

1982).

Genetic diversity in A. michauxii averaged 10.88 alleles per locus across all populations, with larger populations having a greater number of alleles compared to smaller populations (R2 = 0.78, p‐ value < 0.001; Table 4.3). After adjusting for sample size, allelic richness was similar across populations (R2 = 0.11, p‐value=0.07). Allelic richness ranged from 2.43 to 3.8 across populations. As with allele number, the number of private alleles increased with population size (R2 = 0.0.39, p‐ value=0.001), but after rarefaction there was not a significant correlation (R2 = 0, p‐value>0.05).

Ordination results using non‐metric multidimensional scaling graphically demonstrated the low population differentiation in A. michauxii populations (Fig. 4.2). While the two Georgia populations separated in ordination space from the North Carolina populations, little separation occurred among the North Carolina populations. AMOVA results indicated that that within‐ population genetic variation accounted for 91.57% of the total genetic variation (Table 4.4), with region and population accounting for small but significant percentages of the overall genetic variation (2.6% and 5.8%, respectively; P<0.001). STRUCTURE results indicated that the optimal K was 1, with low genetic differentiation between populations and no clear boundaries between geographic areas (Fig. 4.3). In addition to low genetic differentiation between populations, isolation by distance (IBD) results indicated that the population genetic distance between populations increased with the log of the geographic distance (r = 0.4553, p‐value = 0.003; Fig. 4.4) when all populations were included. Removal of the GA populations indicated that population genetic distance did not increase with the log of the geographic distance (R = 0.05, P=0.36).

80 BOTTLENECK results did not indicate evidence of a recent genetic bottleneck in any population using either the stepwise mutation model (SMM) or the two‐phase mutation model

(TPM) based on Wilcoxon’s test (Table 4.5). The mode‐shift test identified a distortion of the L‐ shaped allele frequency distribution in only one population, ASMI049. Contrary to BOTTLENECK results, Critical M results indicated a severe bottleneck in A. michauxii populations. M ratio values across all populations averaged 0.48 and were lower than the 0.68 threshold identified by Garza and

Williamson as indicative of a past genetic bottleneck (2001), with the upper 95% CI for all estimate

M ratio values less than the 0.68 threshold in all populations (Fig. 4.5). The Mc values (90% SMM) were greater than the observed M ratio values in 21 out of 22 populations and the Mc values

(80%SMM) were greater than the observed M ratio values in 14 out of 22 populations. Populations with critical Mc values less than the observed M ratio had smaller population sizes (and sample sizes) relative to populations with M ratio values less than the critical Mc values.

For the Fort Bragg and Camp Mackall populations, GeneClass2 results identified 14 putative first‐generation migrants out of a total of 346 individuals (Table 4.6). The average distance between all sampled NC populations was 15.6 km, and the average distance between resident and putative original population was 15.8 km. There was not a significant correlation between migration and population size, with the population sizes of putative source populations similar to overall population sizes ( 18.5 ± 2.6 s.d. vs. 17.3 ± 3.4 respectively; p‐value> 0.05).

4.5 Discussion

Habitat fragmentation and degradation can result in reduced population sizes and reduced genetic diversity within populations for many plant species. For Astragalus michauxii, a rare legume

81 endemic to the longleaf pine ecosystem in the southeastern United States, populations do not demonstrate evidence of reduced genetic diversity relative to other plant species. Estimates of expected (He) and observed (Ho) heterozygosity for A. michauxii (0.68 and 0.57 respectively) were comparable to the average He (0.61±0.21 ) and Ho (0.58± 0.22) found in a review of plant microsatellite data sets (Nybom 2004). Within‐population genetic variation accounted for 92% of the total genetic variation; this is not unexpected, because other studies have shown that long‐lived, perennial plant species maintain the majority of their genetic diversity within populations (Hamrick and Godt 1996; Nybom 2004). In addition, these results are consistent with the genetic structure of two other rare species endemic to the Fall‐line Sandhills that maintain a large portion of the overall genetic diversity within populations. Pyxidanthera brevifolia occupies xeric upland habitats similar to those occupied by A. michauxii, with genetic evidence suggesting within‐population genetic variation accounted for 90.5% of the total genetic variation of the species (Douglas et al. 2011; Wall et al. 2010). Lilium pyrophilum occupying relatively more mesic habitats in the Fall‐line Sandhills may represent a peripheral isolate of the more widespread L. superbum, and demonstrates little population differentiation (N A. Douglas and W. A. Wall, unpublished data).

While within‐population genetic diversity of A. michauxii may be comparable to that of other plant species, there is evidence of a past genetic bottleneck based on the ratio of the number of alleles to the allelic range size (M ratio) across A. michauxii populations. Although the expected mutation‐drift heterozygosity (Heq) was not significantly less than He, indicating no evidence of a recent genetic bottleneck, this method of assessing genetic bottlenecks may not be as sensitive as evaluating M ratio values (Girod et al. 2011). In addition, M ratio values may be reduced for more generations relative to methods such as analyzing the expected mutation‐drift, which quantify rare

82 allelic deficits (Garza and Williamson 2001). Reductions in M ratio values can persist for 100 or more generations, and so it is not possible to date the bottleneck using M ratio values alone.

Since efforts to date the timing of the genetic bottleneck were unsuccessful (results not shown), it is unclear whether the genetic bottleneck experienced by A. michauxii occurred pre‐ or post‐European settlement. While it is easy to assume that A. michauxii population numbers were reduced following fragmentation, reduction, and degradation of the longleaf pine ecosystem, there are confounding factors that complicate the situation. As an plant species endemic to the Fall‐line

Sandhills region of the GACP, A. michauxii most likely was present in the region since the

Pleistocene. As such, it occupied an environment much different from the present longleaf pine ecosystem. The formation of Aeolian river dunes and braided river channels that have been documented from the Gulf and Atlantic Coastal Plain during the Late Pleistocene (Ivester et al. 2001;

Leigh 2008) suggest an environment with exposed soil and dry, windy climatic conditions (Leigh

2008). The colder, drier conditions of the Pleistocene most likely reduced plant productivity and biomass accumulation, and the region would have been characterized as an open savanna with scattered Picea and Pinus spp. and an herbaceous understory (Watts 1980). As climatic conditions became progressively warmer and wetter during the Holocene, it is likely that the biomass and canopy cover increased in many areas. Evidence from extant and historical local populations suggests that A. michauxii is sensitive to woody encroachment (North Carolina Natural Heritage

Program data): A. michauxii may have become isolated to habitats where competition was reduced through increasing fire frequency in combination with extreme xeric conditions presented by the deep sands that are characteristic of many upland sites in the Fall‐line Sandhills.

83 The distribution of the common alleles across populations and the low population differentiation suggests that A. michauxii had a more continuous range in the past relative to its present‐day fragmented distribution, and that interpopulation gene flow was more common in the past. While the two A. michauxii populations in Georgia appear somewhat different from the North

Carolina populations based on the NMDS ordination, the genetic differences between the two regions may best be explained by an isolation by distance model (Fig. 4.4) with little evidence of population subdivision (Fig. 4.3). Even though A. michauxii populations are currently geographically isolated from each other, the Fort Bragg populations demonstrate evidence of recent interpopulation gene flow (Table 4.6). Indeed, recent gene flow appears to have occurred broadly across the North Carolina populations, with no evidence of isolation by distance in these populations. This pattern has been observed in at least one other species in the Southeastern United

States. The genetic structure of the federally‐endangered plant species, Echinacea laevigata, may best be explained through a metapopulation model, with current range fragmented due to anthropogenic factors but evidence of past gene flow between populations (Peters et al. 2009).

Both genetic and demographic factors can affect the long‐term viability of plant populations. Our results indicate that, although A. michauxii may have undergone a reduction in rare alleles due to a past population bottleneck, gene flow is continuing between at least some of the NC populations and that ultimately demographic factors may be more critical for the persistence of the species. Most of the populations in NC are in habitats maintained with prescribed fire, and both of the GA populations are burned. Prescribed growing season fire at relatively frequent intervals (every 3‐4 years) should maintain open habitat by reducing woody encroachments. Even with prescribed fire, demographic analyses suggest that A. michauxii may not be maintaining stable

84 populations (Wall et al. 2012), mainly due to limited recruitment of seedlings. Maintaining connectivity between populations will be necessary from a demographic and genetic standpoint to ensure population persistence and reduce the probability of genetic drift and inbreeding.

85 4.6 References

Avise JC (1989) A role for molecular genetics in the recognition and conservation of endangered species. Trends in Ecology & Evolution 4:279–281.

Bonferroni CE (1935) Il calcolo delle assicurazioni su gruppi di teste. Studi in onore del Professore Salvatore Ortu Carboni. Tipografia del Senato, Rome, Italy, pp 13–60

Cornuet JM, Luikart G (1996) Description and power analysis of two tests for detecting recent population bottlenecks from allele frequency data. Genetics 144:2001–2014.

Cronn R, Liston A, Parks M, et al. (2008) Multiplex sequencing of plant chloroplast genomes using Solexa sequencing‐by‐synthesis technology. Nucleic Acids Research 36:e122.

Douglas NA, Wall WA, Xiang QY (Jenny), et al. (2011) Recent vicariance and the origin of the rare, edaphically specialized Sandhills lily, Lilium pyrophilum (): evidence from phylogenetic and coalescent analyses. Molecular Ecology 20:2901–2915.

Doyle JJ, Doyle JL (1987) A rapid DNA isolation procedure for small quantities of fresh leaf tissue. Phytochemical bulletin 19:11–15.

Ellstrand NC, Elam DR (1993) Population genetic consequences of small population size: Implications for plant conservation. Annual Review of Ecology and Systematics 24:217–242.

Evanno G, Regnaut S, Goudet J (2005) Detecting the number of clusters of individuals using the software structure: a simulation study. Molecular ecology 14:2611–2620.

Excoffier L, Laval G, Schneider S (2005) Arlequin (version 3.0): An integrated software package for population genetics data analysis. Evolutionary Bioinformatics 1:47–50.

Falush D, Stephens M, Pritchard JK (2007) Inference of population structure using multilocus genotype data: dominant markers and null alleles. Molecular Ecology Notes 7:574–578.

Frankham R, Ballou JD (Jonathan D), Briscoe DA (David A (2010) Introduction to conservation genetics, 2nd ed. Cambridge University Press,, Cambridge, UK

Franklin IR (1980) Evolutionary change in small populations. In: Soule ME, Wilcox BA (eds) Conservation biology: an evolutionary‐ecological perspective. Sinauer Associates, Sunderland, Massachussetts, pp 135–149

86 Frost C (2006) History and future of the longleaf pine ecosystem. In: Jose S, Jokela EJ, Miller DL (eds) The Longleaf Pine Ecosystem: Ecology, Silvaculture, and Restoration. Springer Science, New York, pp 9–42

Frost CC (1993) Four centuries of changing landscape patterns in the longleaf pine ecosystem. In: Hermann SM (ed) Proceedings of the Tall Timbers Fire Ecology Conference. Tall Timbers Research Station, Tallahassee, FL, pp 17–37

Garza J, Williamson E (2001) Detection of reduction in population size using data from microsatellite loci. Molecular Ecology 10:305–318.

Girod C, Vitalis R, Leblois R, Freville H (2011) Inferring Population Decline and Expansion From Microsatellite Data: A Simulation‐Based Evaluation of the Msvar Method. Genetics 188:165– 179.

Guo SW, Thompson EA (1992) Performing the exact test of Hardy‐Weinberg proportion for multiple alleles. Biometrics 48:361–372.

Hamrick JL, Godt MJW (1996) Effects of life history traits on genetic diversity in plant species. Philosophical Transactions of the Royal Society B: Biological Sciences 351:1291–1298.

Herrera CM, Bazaga P (2008) Population‐genomic approach reveals adaptive floral divergence in discrete populations of a hawk moth‐pollinated violet. Molecular Ecology 17:5378–5390.

Ivester AH, Leigh DS, Godfrey‐Smith DI (2001) Chronology of inland eolian dunes on the Coastal Plain of Georgia, USA. Quaternary Research 55:293–302.

James MMR (2000) Legumes in loamy soil communities of the Carolina sandhills: Their natural distributions and performance of seeds and seedlings along complex ecological gradients. Master of Science thesis, University of North Carolina

Jennings TN, Knaus BJ, Mullins TD, et al. (2011) Multiplexed microsatellite recovery using massively parallel sequencing. Molecular Ecology Resources 11:1060–1067.

Kalinowski ST (2004) Counting alleles with rarefaction: private alleles and hierarchical sampling designs. Conservation Genetics 5:539–543.

Kalinowski ST (2005) HP‐RARE 1.0: a computer program for performing rarefaction on measures of allelic richness. Molecular Ecology Notes 5:187–189.

Leigh DS (2008) Late Quaternary climates and river channels of the Atlantic Coastal Plain, Southeastern USA. Geomorphology 101:90–108. doi: 10.1016/j.geomorph.2008.05.024

87 Leimu R, Mutikainen PIA, Koricheva J, Fischer M (2006) How general are positive relationships between plant population size, fitness and genetic variation? Ecology 94:942–952.

Manel S, Gaggiotti O, Waples R (2005) Assignment methods: matching biological questions with appropriate techniques. Trends in Ecology & Evolution 20:136–142

Maruyama T, Fuerst PA (1985) Population bottlenecks and nonequilibrium models in population genetics. II. Number of alleles in a small population that was formed by a recent bottleneck. Genetics 111:675–689.

Matschiner M, Salzburger W (2009) TANDEM: integrating automated allele binning into genetics and genomics workflows. Bioinformatics 25:1982.

Moran MD (2003) Arguments for rejecting the sequential Bonferroni in ecological studies. Oikos 100:403–405.

NatureServe (2012) NatureServe Explorer: An online encyclopedia of life [web application]. In: NatureServe Explorer: An online encyclopedia of life [web application]. Version 7.1. http://www.natureserve.org/explorer. Accessed 28 May 2012

Nei M (1972) Genetic distance between populations. Am Nat 106:283.

Nei M (1987) Molecular evolutionary genetics. Columbia University Press,, New York

Nybom H (2004) Comparison of different nuclear DNA markers for estimating intraspecific genetic diversity in plants. Molecular Ecology 13:1143–1155.

Ohta T (1982) Linkage disequilibrium due to random genetic drift in finite subdivided populations. Proceedings of the National Academy of Sciences 79:1940.

Oksanen J, Kindt R, Legendre P, et al. (2009) vegan: Community Ecology Package.

Paetkau D, Slade R, Burden M, Estoup A (2004) Genetic assignment methods for the direct, real‐time estimation of migration rate: a simulation‐based exploration of accuracy and power. Molecular Ecology 13:55–65.

Peakall R, Smouse PE (2006) GENALEX 6: genetic analysis in Excel. Population genetic software for teaching and research. Molecular Ecology Notes 6:288–295.

Peet RK (2006) Ecological Classification of Longleaf Pine Woodlands. In: Jose S, Jokela EJ, Miller DL (eds) The Longleaf Pine Ecosystem: Ecology, Silviculture, and Restoration. Springer, New York, pp 51–93

88 Peters MD, Xiang QY, Thomas DT, et al. (2009) Genetic analyses of the federally endangered Echinacea laevigata using amplified fragment length polymorphisms (AFLP)—Inferences in population genetic structure and mating system. Conservation Genetics 10:1–14.

Piry S, Alapetite A, Cornuet JM, et al. (2004) GENECLASS2: a software for genetic assignment and first‐generation migrant detection. Journal of Heredity 95:536.

Piry S, Luikart G, Cornuet J‐M (1999) BOTTLENECK: A computer program for detecting recent reductions in the effective population size using allele frequency data. Journal of Heredity 90:502–503.

Pritchard JK, Stephens M, Donnelly P (2000) Inference of population structure using multilocus genotype data. Genetics 155:945–959.

R Development Core Team (2012) R: a Language and environment for statistical computing. Vienna, Austria

Rannala B, Mountain JL (1997) Detecting immigration by using multilocus genotypes. Proceedings of the National Academy of Sciences 94:9197.

Roberts DW (2010) labdsv: Ordination and Multivariate Analysis for Ecology.

Schafale MP, Weakley AS (1990) Classification of the natural communities of North Carolina: Third approximation. 321.

Sorrie BA, Weakley AS (2001) Coastal plain vascular plant endemics: Phytogeographic patterns. Castanea 66:50–82.

Spielman D, Brook BW, Frankham R (2004) Most species are not driven to extinction before genetic factors impact them. Proceedings of the National Academy of Sciences of the United States of America 101:15261.

Swatdipong A, Primmer CR, Vasemägi A (2009) Historical and recent genetic bottlenecks in European grayling, Thymallus thymallus. Conservation Genetics 11:279–292.

Walker J, Peet RK (1983) Composition and species diversity in pine‐wiregrass savannas of the Green Swamp, North Carolina. Vegetatio 55:166–179.

Wall W, Hoffmann W, Wentworth T, et al. (2012) Demographic effects of fire on two endemic plant species in the longleaf pine‐wiregrass ecosystem. Plant Ecology 213:1093–1104.

Wall WA, Douglas NA, Xiang QY (Jenny), et al. (2010) Evidence for range stasis during the latter Pleistocene for the Atlantic Coastal Plain endemic genus, Pyxidanthera Michaux. Molecular Ecology 19:4302–4314.

89 Watts WA (1980) Late‐Quaternary vegetation history at White Pond on the inner Coastal Plain of South Carolina. Quaternary Research 13:187–199.

You F, Huo N, Gu Y, et al. (2008) BatchPrimer3: a high throughput web application for PCR and sequencing primer design. BMC bioinformatics 9:253.

90 Table 4.1: Eight polymorphic loci identified and developed for Astragalus michauxii. Column are primer pair name, sequence, repeat, microsatellite range (with fragment size in parentheses), and number of alleles observed.

Primer Sequence Repeat Size (BP) Alleles

AM_15 F: GTTTCACACTGAGACACAGTTC GA 24‐34 (124‐134) 6

R: AATTCCCAAGTGTAAAAGCTC

AM_18 F: GAAAACACAAACAAATTCTGG GA 8‐38 (165‐195) 13

R: AGAAAGTCTGTGCTCTCTCATT

AM_25 F: CAATCCCTAACCTTGAGTTCT GA 8‐36(107‐135) 14

R: AGCAACGTGGGATAAAAATA

AM_29 F: AACGGTGTCTGTGTCTATGTC GT 32‐42 (160‐170) 6

R: ATGAAGCGTTTCACATTTTT

AM_34 F: TGACATACATGCTGAAAGTTG AG 20‐26 (155‐161) 4

R: TTTGGATTCATATAACCACCA

AM_46 F: GAAAATGGTGGAAAAGGAAT AG 18‐64 (102‐148) 22

R: GTGTAAAAATCGTGCACTTCT

AM_71 F: AAGATTGTCTAACGATCACCA GT 187‐203 (201 missing) 7

R:AAAGCCCATGTTTCACTAAAT

AM_91 F: GGACAAAAGAAGAGGAGAGAG AG(TACTGG)TG 22‐40 (107‐125) 10

R: TAAGTCGAGTTGTTCCAAAGT

91 Table 4.2: Proportion of linkage disequilibrium (LD) tests that detected significant LD for eight polymorphic microsatellite loci.

AM15 AM18 AM25 AM29 AM34 AM46 AM71 AM91 Mean AM15 ‐ 0.14 0.14 0.09 0.14 0.18 0.18 0.23 0.16 AM18 0.14 ‐ 0.18 0.14 0.27 0.23 0.14 0.27 0.19 AM25 0.14 0.18 ‐ 0.09 0.23 0.09 0.09 0.14 0.14 AM29 0.09 0.14 0.09 ‐ 0.09 0.09 0.09 0.05 0.09 AM34 0.14 0.27 0.23 0.09 ‐ 0.05 0.14 0.18 0.16 AM46 0.18 0.23 0.09 0.09 0.05 ‐ 0.18 0.18 0.14 AM71 0.18 0.14 0.09 0.09 0.14 0.18 ‐ 0.14 0.14 AM91 0.23 0.27 0.14 0.05 0.18 0.18 0.14 ‐ 0.17

92 Table 4.3: Genetic variation in Astragalus michauxii populations from North Carolina and Georgia based on eight polymorphic microsatellite loci. Column headings are: N, number of individuals; A, average number of alleles across loci; AR, average allelic richness; P, number of private alleles; PR, private allelic richness; HO, observed heterozygosity; HE, expected heterozygosity; H‐W disequilibrium, loci identified as not in Hardy‐Weinberg equilibrium.

Population Ultra population N A AR P PR HO HE H‐W disequilibrium ASMI053 Camp Mackall 7 3.429 2.61 1 0.11 0.551 0.531 None ASMI054 Camp Mackall 14 4.875 3.23 0 0.07 0.587 0.622 AM29 ASMI088 Camp Mackall 5 3.125 2.91 1 0.14 0.506 0.548 None ASMI020 Fort Bragg ‐ North 5 3.125 2.95 0 0.12 0.550 0.548 None ASMI022 Fort Bragg ‐ North 7 4 3.42 1 0.14 0.601 0.690 AM29,AM46 ASMI023 Fort Bragg ‐ North 27 5.5 3.38 1 0.04 0.581 0.665 AM29 ASMI032 Fort Bragg ‐ North 30 6.125 3.48 0 0.03 0.617 0.654 None AM18, AM25, AM29, ASMI034 Fort Bragg ‐ North 24 6.625 3.8 2 0.11 0.557 0.714 AM91 AM18, AM25, AM29, ASMI035 Fort Bragg ‐ North 8 4.25 3.36 0 0.05 0.578 0.657 AM91 ASMI049 Fort Bragg ‐ North 13 4.5 3.36 0 0 0.635 0.637 None ASMI056 Fort Bragg ‐ North 68 7.75 3.63 2 0.1 0.623 0.690 AM29, AM34, AM71 ASMI057 Fort Bragg ‐ North 33 5 2.93 1 0.02 0.474 0.559 AM18, AM46 ASMI061 Fort Bragg ‐ North 6 2.857 2.43 0 0.01 0.619 0.554 None ASMI091 Fort Bragg ‐ North 12 5 3.56 0 0.07 0.667 0.692 None ASMI096 Fort Bragg ‐ North 4 3.5 3.5 0 0.12 0.688 0.621 None ASMI030 Fort Bragg ‐ South 9 4.25 3.29 0 0.07 0.569 0.618 None ASMI046 Fort Bragg ‐ South 11 3.875 3.04 0 0.02 0.557 0.579 None ASMI050 Fort Bragg ‐ South 34 5.75 3.19 3 0.08 0.581 0.611 AM29 ASMI051 Fort Bragg ‐ South 24 6 3.32 1 0.1 0.542 0.591 AM29 ASMI097 Fort Bragg ‐ South 5 3 2.81 0 0.09 0.475 0.528 none AMBUGA Georgia 4 3.25 3.25 0 0.03 0.563 0.652 none AMMEGA Georgia 5 3.625 3.34 0 0.12 0.550 0.661 none

Camp Mackall 26 5.625 3.16 2 0.52 0.546 0.588 AM29 Fort Bragg ‐ North 237 9.625 3.67 13 0.63 0.592 0.693 all Fort Bragg ‐ South 83 8.125 3.4 5 0.53 0.558 0.626 AM29 Georgia 9 4.5 3.34 1 0.75 0.556 0.662 None

All 355 10.88 3.67 NA NA 0.579 0.683 NA

93 Table 4.4: Analysis of molecular variance (AMOVA) results for Astragalus michauxii populations from North Carolina and Georgia (USA). Populations were defined by Natural Heritage Program protocols. Grouping of populations into regions follows Table 2.

sum of variance percentage Source of variation d.f. squares components of variation P‐value

Among regions 3 46.058 0.07028 2.60 <0.001

Among populations 18 132.731 0.15762 5.83 <0.001 within regions

Within populations 688 1704.2 2.47703 91.57 <0.001

Total 709 1882.989 2.70493 100.00

94 Table 4.5: Tests for genetic bottlenecks in Astragalus michauxii using Bottleneck version 1.2 in populations with > 20 gene copies and M ratio for all populations as calculated in Arlequin 3.1. Results indicate no recent genetic bottleneck events, but evidence of a severe genetic bottleneck in the more distant past.

Wilcoxon's test Wilcoxon's test Copies (SMM) (TPM) Mode‐shift test M ratio ASMI020 9.25 ‐ ‐ ‐ 0.43205 ASMI022 13.25 ‐ ‐ ‐ 0.40397 ASMI023 53 0.62891 0.47266 L‐shape 0.49321 ASMI030 18 ‐ ‐ ‐ 0.4812 ASMI032 59.5 0.875 0.76953 L‐shape 0.50957 ASMI034 46 0.84375 0.80859 L‐shape 0.49201 ASMI035 16 ‐ ‐ 0.49813 ASMI046 22 0.27344 0.27344 L‐shape 0.4777 ASMI049 26 0.15625 0.125 shifted 0.48117 ASMI050 67 0.875 0.875 L‐shape 0.456 ASMI051 48 0.99414 0.99414 L‐shape 0.45295 ASMI053 14 ‐ ‐ ‐ 0.50025 ASMI054 27.75 0.76953 0.72656 L‐shape 0.47976 ASMI056 141.5 0.99023 0.98047 L‐shape 0.5195 ASMI057 57.25 0.97266 0.67969 L‐shape 0.46245 ASMI061 12 ‐ ‐ ‐ 0.44602 ASMI088 9.75 ‐ ‐ ‐ 0.43187 ASMI091 24 0.67969 0.47266 L‐shape 0.48862 ASMI096 8 ‐ ‐ ‐ 0.47949 ASMI097 10 ‐ ‐ ‐ 0.50159 AMBUGA 8 ‐ ‐ ‐ 0.5181 AMMEGA 10 ‐ ‐ ‐ 0.49401

95 Table 4.6: A. michauxii individuals identified by GeneClass 2 as being the result of possible interpopulation gene flow (p‐value >= 0.001). Fourteen individuals across eleven populations were identified.

resident assigned distance population population individual probability (m) ASMI022 ASMI032 /Ind‐36 0.005 5098 ASMI023 ASMI054 /Ind‐45 0.006 40415 ASMI035 ASMI054 /Ind‐124 0.001 39628 ASMI049 ASMI034 /Ind‐137 0.009 6986 ASMI051 ASMI034 /Ind‐323 0.003 10304 ASMI051 ASMI030 /Ind‐327 0.005 5379 ASMI053 ASMI050 /Ind‐1 0.005 26853 ASMI054 ASMI032 /Ind‐9 0.01 32460 ASMI056 ASMI034 /Ind‐198 0.003 1956 ASMI056 ASMI046 /Ind‐188 0.007 13891 ASMI057 ASMI096 /Ind‐219 0.001 8783 ASMI057 ASMI097 /Ind‐223 0.006 17144 ASMI057 ASMI035 /Ind‐238 0.01 2184 ASMI096 ASMI049 /Ind‐263 0.01 9159 ASMI097 ASMI057 /Ind‐342 0 17144

96

Fig. 4.1. Historic range and collection sites of Astragalus michauxii. Historical range determined based on voucher specimens (UNC Herbarium Flora of the Southeast; http://www.herbarium.unc.edu/seflora). Current range is greatly restricted, with most sites in North Carolina. Survey of Georgia populations located 13 individuals and there are only two known sites in South Carolina and Alabama.

97

Fig. 4.2. Non‐metric multidimensional scaling ordination of Astragalus michauxii population genetic distances based on eight polymorphic microsatellite loci. The Georgia populations appear separate from the North Carolina populations, while little separation appears between the North Carolina populations from Fort Bragg and Camp Mackall.

98

Fig. 4.3. Population genetic structure for Astragalus michauxii as determined by the program STRUCTURE (K = 2‐9). Individual columns represent genetic samples and colors represent proportion of ancestry assigned to different ancestral populations. Results reflect low genetic population differentiation in A. michauxii.

99

Fig. 4.4. Pairwise population isolation by distance for sampled Astragalus michauxii populations based on eight polymorphic microsatellite loci. Genetic distances demonstrate significant (R = 0.43, P<0.001) isolation by distance. When GA populations are removed isolation by distance is not evident (R = 0.05, P=0.36).

100

Fig. 4.5. M ratio values (black circles) estimated for 22 Astragalus michauxii populations across North Carolina and Georgia (USA). Horizontal lines represent 95% CIs, the vertical line is the threshold indicative of a past genetic bottleneck, the open triangles are the critical Mc (90% SSM), and the gray triangles are the critical Mc (80% SMM) (Garza and Williamson 2001). Population numbers are in parentheses.

101 CHAPTER 5 DEMOGRAPHIC EFFECTS OF FIRE ON TWO ENDEMIC PLANT SPECIES IN THE LONGLEAF PINE‐WIREGRASS ECOSYSTEM.

*Previously published in Plant Ecology, 213, 1093‐1104, 2012. Used by permission.

5.1 Abstract

Fire can have dramatic effects on the vital rates of plant species and has been used successfully for management in a number of ecosystems. However, the demographic response of species to fire in fire‐dependent ecosystems is variable, making it important to study the effects of fire on rare and threatened species. We quantified the effects of fire on Astragalus michauxii and Pyxidanthera brevifolia, two rare endemics of the longleaf pine‐wiregrass ecosystem of the southeastern USA, using periodic matrix models to project the effect of fire frequency on population growth. In contrast to many species in the longleaf pine‐wiregrass ecosystem, fire had short term negative effects on both species, causing reductions in survival, size, flowering, and fruit production. Relative to the three‐year fire intervals the study populations are currently exposed, more frequent burning is projected to cause population decline, with the most dramatic effects under annual burning.

Although the current longleaf pine‐wiregrass ecosystem is fire dependent and has experienced frequent fire for at least several thousand years, we propose that the two endemic species may be remnants from a past vegetation assemblage that experienced less frequent fire and thus may be adapted to longer fire‐return intervals compared to other species currently in the ecosystem.

Despite the short‐term negative effects of fire on the vital rates of these species, longer‐term

102 benefits such as reduction of woody encroachment and litter removal may be important for the ultimate success of the species.

5.2 Introduction

Fire is a natural disturbance in many ecosystems and can have a profound effect on the population dynamics of plant species (Glitzenstein et al. 1995), which directly influences plant community composition (Morrison et al. 1995) and vegetation structure (Moreira 2000). Under natural fire frequencies, many regions with precipitation and nutrient availability adequate for supporting closed canopy forest are instead maintained as grasslands or savannas (Staver et al.

2011). Due to the recurring destruction of aboveground biomass, frequent fires tend to favor plant species of smaller size with less carbon investment in aboveground biomass (Hoffmann 1999), and many areas with high fire frequencies tend to be dominated by herbaceous species. Fire suppression during the 20th century has led to changes in vegetation structure in many ecosystems, including increased density of woody vegetation (Bond et al. 2004), decreased biomass of herbaceous plants, and loss of species diversity, making it one of the primary threats to rare plants in the United States

(Schemske et al. 1994).

Prescribed fire has become one of the primary tools for restoring and maintaining fire‐ dependent plant communities and managing rare plants in the longleaf pine ecosystem of the southeastern United States (Brockway and Lewis 1997). Both representative rare and common plant species in the region have been shown to respond positively to fire, exhibiting increased flowering

(Seamon et al. 1989; Gowe and Brewer 2005; Brewer et al. 2009), fecundity (Hartnett and

Richardson 1989; Brewer 2001), and seedling establishment (Hartnett and Richardson 1989; Brewer

103 and Platt 1994; Brewer 2001). Based partly on the observed positive responses of many plant species to fire in this system, over the last few decades biologists and land managers have advocated for restoring historical fire regimes in many natural areas as a critical component of rare species management.

Although re‐establishing historical fire regimes is a well‐intentioned goal, local and regional differences in historical fire regimes are expected due to spatio‐temporal variation in fuels, topographic position and climate. It is also difficult to identify fire frequency over the timeframes within which most plant species are likely to have evolved. As a surrogate for other information

(Power et al. 2007), some have used the demographic responses of native plants to infer past fire regimes (Liu et al. 2005; Menges 2007). This is appropriate in areas that have been climatically stable for an extended period of time, but in the northern extent of the longleaf pine ecosystem the situation is complicated because of substantial climatic change since the Pleistocene (Jackson et al.

2000). Vegetation assemblages in these areas are likely a mixture of species with different evolutionary relationships with fire, for example endemic species that were most likely present through the latter Pleistocene (Wall et al. 2010) and species that migrated following the end of the

Pleistocene (Soltis et al. 2006). Consequently, it is important to specifically understand the demographic response of rare or endangered species to variable fire regimes in order to make better informed management decisions (Kirkman et al. 1998; Slapcinsky et al. 2010).

It is within this context that we studied the fire ecology of two rare endemic species restricted to the Fall‐line Sandhills, an ancient dune system at the eastern edge of the Gulf and

Atlantic Coastal Plain physiographic province (GACP) that was historically dominated by the longleaf pine ecosystem. We examined the effects of prescribed fire on the mortality, growth, flowering, and

104 seed production of Astragalus michauxii (Kuntze) F.J. Herm. and Pyxidanthera brevifolia Wells. We monitored individuals across multiple populations and estimated the effects of fire frequency on the species population growth rates and long‐term population viability using matrix modeling and interpreted the results in the context of the climatic and vegetation history of the GACP.

5.3 Methods

5.3.1 Study area and species

Astragalus michauxii (Fabaceae) is a perennial, herbaceous legume with populations in

North Carolina (NC), South Carolina (SC), Georgia (GA) and Alabama (AL), USA. Flowering occurs in early May with fruit maturation by mid‐July. Pyxidanthera brevifolia (Diapensiaceae) is an evergreen, woody cushion plant with a range limited to four counties in NC and two counties in SC, with over

80% of extant populations found on Fort Bragg Military Reservation, NC. Flowering occurs from late

December through early April, with the majority of flowers opening in mid‐ to late‐March. Fruit maturation occurs between the last week of April and the first week of May. Anecdotal evidence suggests that both species are relatively long lived perennials that experience low natural recruitment rates, as prior to this study no seedling establishment for either species had been documented. Both species have been designated by the Department of Defense as Species‐at‐Risk and are considered vulnerable to local extinction in NC (S3).

Fort Bragg Military Reservation is located at the northern limit of the Fall‐line Sandhills. The landscape is characterized by rolling hills dissected by numerous streams that create a matrix of wetland and xeric habitats. For the duration of the study, Fort Bragg was managed with prescribed fire on a fixed, three‐year burn interval. Growing season burns are defined as occurring in April

105 through June. During the course of the study, all growing season fires were completed by 1 July.

Dormant season fires occur from December through March. The flora and vegetation communities of Fort Bragg have been described separately (Sorrie et al. 2006). The study species occur on excessively well drained upland sandy soils with low nutrient availability, and a priori habitat characterization of the two species suggests that both require a relatively open forest canopy and relatively low ground‐level biomass.

5.3.2 Field Methods

In summer 2007 we surveyed all known A. michauxii populations on Fort Bragg, locating individuals in 39 of the 87 surveyed populations. Population sizes averaged 13 individuals and ranged from 1 to 116 individuals. We placed aluminum tags next to all identified individuals; over the course of the study we identified and tagged 496 individuals. During winter 2008 we randomly selected 24 of the 277 known P. brevifolia populations which were evenly divided into one‐, two‐, and three‐years post burn. Before the 2008 growing season burn period (April‐June), we surveyed and demarcated the areal extent of the 24 populations and established a transect through the area with the highest population density. We placed aluminum tags next to all individuals within one meter of the established transect. In total, we marked 1042 individuals, with sample size averaging

42 individuals per population and ranging from 13 to 71 individuals.

We measured the height (± 1 cm) and number of fruits for the five tallest stems of each tagged A. michauxii individual annually during the last two weeks in June from 2007 to 2010 (). We estimated the number of seeds per fruit by counting the number of viable seeds in 466 fruits produced between 2007 and 2010. We established twenty‐one 2.25 m2 seed addition plots (mean

106 number of seeds = 98.4) during 2007 and 2008 and monitored for recruitment for four years (2008 through 2011). Size of P. brevifolia plants was quantified by measuring the major and minor axis (± 1 cm) of each tagged individual and estimating the percentage cover within the cushion to the nearest

10%. Area was calculated as follows: . We recorded occurrence of flowering and revisited each population to estimate the number of fruits of ~10 randomly selected individuals per population. We recensused all populations in 2009 and 2010 before the beginning of the burn season (Fig. 5.1). To estimate the number of seeds produced per fruit, we collected and counted seeds of 111 fruits from 12 of the populations in 2008. We established thirty 2.25 m2 seed addition plots by distributing 100 seeds per plot and monitored for seedling recruitment from 2008 to 2010. No seedlings were observed in any of the A. michauxii or P. brevifolia seed addition plots, but one A. michauxii seedling was identified in the field during the course of the study and 41 P. brevifolia seedlings were identified across four populations during the

2010 census.

5.3.3 Vital rates data analysis

We modeled the effects of fire and plant size on the vital rates using generalized linear mixed effects models (Pinheiro and Bates 2000), with year and population as random factors for P. brevifolia and year as a random factor in the case of A. michauxii, as population variance was small and causing convergence issues. We tested for the effects of fire on plant size using a linear regression model. Survivorship for both species was modeled using logistic regression. We quantified recovery rates following fire for both species by calculating the ratio of the pre‐burn size

107 to post‐burn size one and two years post‐fire and performed an ANOVA to test for differences in recovery rates by pre‐burn size. We modeled the effects of fire and plant size on number of fruits produced using a Poisson distribution with a log link function. While all of the P. brevifolia populations were subjected to spring burns, some A. michauxii populations were burned in winter, so we compared the effects of seasonality of burn (growing (April ‐July) versus dormant season burns) on the survivorship, post‐fire recovery, and fruit production of A. michauxii.

5.3.4 Matrix construction and analysis

We classified A. michauxii into four height classes based on the tallest stem: small (0.1 – 20 cm), small‐medium (>20‐40 cm), medium (>40‐80 cm), and large (> 80 cm). Because of low numbers,

A. michauxii individuals were pooled across populations and years to estimate three separate transition matrices based on years post‐burn: burned during current growing season, one year post‐ burn, and two or more years post‐burn (Table 5.1 and Table 5.2). We classified P. brevifolia into ten size classes using area (cm2): >0‐10 cm2, >10‐25 cm2, >25‐50 cm2, >50‐100 cm2, >100‐200 cm2, >200‐

400 cm2, >400‐800 cm2, >800‐1600 cm2, >1600‐3200 cm2, and >3200 cm2; an additional age class was recognized for first‐year seedlings. We calculated mean seed production by multiplying the mean number of seeds per fruit by per capita fruit production. Seedling recruitment per seed was estimated by dividing the observed number of seedlings at time t by total seed production at time t‐

1. For P. brevifolia, seedling survivorship and growth were estimated by tagging and revisiting in spring 2011 the 41 seedlings located during the 2010 census. For P. brevifolia, we constructed six separate transition matrices grouped according to years post‐burn (one, two, or three years post‐ burn) and the two times steps (2008‐2009 and 2009‐2010) by pooling individuals across populations

108 (Table 5.3 and Table 5.4). Because the census of P. brevifolia populations was conducted prior to burn season, response data collected one year post‐fire, represents the first census after fire. We estimated size class survivorship at 0.972 (average survivorship of two largest size classes across all years) for A. michauxii and 0.998 (average survivorship of four largest size classes across all years) for P. brevifolia when no mortality was observed within a size class during a time step. We did not include a seed bank in the modeling for either species. While we did not directly parameterize the role of the seed bank in the population dynamics, prior work on A. michauxii suggests that the size of the seed bank is small (Weeks 2005), and the low overall recruitment rates for both species indicate that the contribution of the seed bank is minimal.

We simulated the stochastic population growth rate (λs) of A. michauxii and P. brevifolia under one to four year fire‐return intervals using a matrix selection approach, which preserves the correlations between transition elements. The population growth rate provides information on whether a population is likely to increase, decrease, or remain the same through time. To increase the variation of possible λs, we constructed nine new transition matrices from the A. michauxii data set divided into three time since fire categories (burned in current year, one year post‐burn, and two years post‐burn) and three time steps (2007‐2010). For P. brevifolia we utilized the six matrices described above. We calculated the average annual λs over 200 time steps and 5000 iterations using

the following formula: (Lewontin and Cohen 1969). At each time step of the simulation, we first randomly selected one “year” (e.g. either 2007‐2008, 2008‐2009, or 2009‐2010 for A. michauxii or 2008‐2009, 2009‐2010 for P. brevifolia). So, under a simulated two‐year fire‐ return interval we randomly selected a year and then multiplied the population vector Nt by the

109 burned matrix and the unburned1 matrix for the randomly selected year; under a three‐year fire‐ return interval the simulation also included the two‐year post‐fire matrix from the selected year. To simulate a four‐year fire‐return interval, we used the two‐year post‐fire matrix for the fourth year.

Elasticities (de Kroon et al. 1986) were calculated for the estimated A. michauxii and P. brevifolia matrices based on the deterministic population growth rate (λ). Elasticities provide information on the proportional contribution of the individual matrix elements to proportional changes in the population growth rate (Caswell 1996). We calculated the reproductive value of each size class and the stable stage distribution, which are mathematically defined as the left and right eigenvectors of a matrix, respectively. The reproductive value of each size class is defined as the expected number of offspring for an individual of that size class, relative to the smallest size class

(defined as 1). For P. brevifolia, the reproductive values were relativized to the seedling size class, not the seed class (Goodman 1968). The stable stage distribution refers to the proportional representation of each size class under a stable population structure scaled to sum to 1. Stable stage distributions represent long‐term population dynamics and can mask underlying transient dynamics, especially in frequently disturbed systems (Stott et al. 2011). We estimated the proportional difference between the observed size distribution at each time since fire and stable stage distribution of the two year post fire matrix for both species using Keyfitz’s ∆ (Keyfitz 1968). This index is a value between 1 and 0, with 1 representing the largest proportional differences and 0 no difference between the observed and stable size distributions.

Fire influences λ through its multiple effects on reproduction, mortality and growth of individuals. To quantify the contributions of these vital rates to the overall change in λ caused by fire, we performed a life table response experiment, or LTRE (Caswell 1996). We utilized a fixed one‐

110 way design to compare the burn matrix with the second year post‐fire matrix. The difference in λ between these two matricies (∆) was decomposed using the linear approximation: ∆

∑ ∆ . The overall difference in the population growth rate (∆) is approximated as the sum of

the differences of the individual contributions (∆ of the underlying parameters () of the transition matrices. This formula is appropriate for any set of parameters that can be used to reconstruct the transition matrices (Caswell 1996); here we utilized an alternate parameterization

(Hoffmann 1999).

All statistical analyses were performed using R version 2.10.1 (R Development Core Team

2009); R script is available from the authors on request. Point estimates include standard errors.

5.4 Results

5.4.1 Effects of plant size and fire on vital rates

Mortality was generally low for A. michauxii and P. brevifolia, averaging 4% and 2% per year,

respectively. Mortality was highest among small individuals for both A. michauxii ( =88.1,

P<0.01; Fig. 5.2) and P. brevifolia ( = 198.9, P<0.001; Fig. 5.2). Burning increased mortality in A.

michauxii ( =89.7, P<0.01) and P. brevifolia ( = 2579.5, P<0.001; Fig. 5.2). Burning had a

greater effect on mortality of small P. brevifolia individuals relative to the larger size classes ( =

2303.1, P<0.001), but in A. michauxii all size classes were affected ( = 5.83, P=0.05).

Fruit production was positively correlated with plant size in both A. michauxii ( = 89.3,

P<0.001) and P. brevifolia ( = 1769.2, P<0.001; Fig. 5.3). Burning reduced flowering (not shown) and fruit production (Fig. 5.3) in both species. Growing‐season burns nearly eliminated fruit

111 production in A. michauxii across all size classes, with mean per capita fruit production of 0.2 in

years when burned and 9.2 in unburned years ( = 16.6, P<0.001). Burning reduced flowering

in P. brevifolia by 48% in the first year post‐burn ( = 639.5, P<0.001, not shown), and reduced

fruit production ( = 4816.7, P<0.001).

Plant size was reduced in the current year by burning in A. michauxii ( = 243.2,

P<0.001). The mean height in 2007 for plants burned three years previously was 44.2 ± 2.1 cm. At the end of the growing season following fire, the mean height was 26.6 ± 1.4 cm. However, mean stem height had recovered to pre‐burn heights by the end of the second growing season after fire

(43.2 ± 2.5 cm vs. 44.2 ± 2.1 cm, P>0.05). Plant size was also still reduced in P. brevifolia roughly nine

months following fire ( = 50.6, P<0.001). The area of P. brevifolia individuals across all years averaged 908 ± 72.6 cm2 prior to burning, declining to 627 ± 48.1 cm2 after one year and recovered to 854 ± 49.8 cm2 two years after burning. Post‐burn recovery (ratio of post‐burn size to pre‐burn size) varied across size classes with small individuals exhibiting greater recovery than large

individuals after one year (F = 9.5, P<0.001, Fig. 5.4) and two years post‐burn (F = 26.5,

P<0.001, Fig. 5.4).

Burn season (dormant versus growing) did not significantly increase mortality in A. michauxii

across size classes ( = 2.4, P>0.05, Fig. 5.2), but growing season fires decreased the size of

individuals, while dormant season burns did not (F = 134.8, P<0.001). Growing season burns

also decreased A. michauxii fruit production relative to dormant season burns ( = 3521.0,

P<0.001, Fig. 5.3); individuals burned during the dormant season produced an average of 17.2 ± 3.8 fruits, while individuals burned during the growing season produced an average of 0.03 ± 0.01 fruits.

112 5.4.2 Population growth rates, elasticities, and LTRE in relation to fire

Simulations of the long‐term population dynamics of A. michauxii and P. brevifolia predicted that λs would decline significantly under annual burning, and increase with increasing fire‐return interval (Fig. 5.5), with the greatest increase between one‐ and two‐year fire‐return intervals. Two‐, three‐, and four‐year fire‐return intervals were similar, but for both species, λS was greatest when the fire‐return interval was four years. Under fire‐return intervals longer than one year, λS averaged

0.970 (0.964‐0.978 95% CI) for A. michauxii and 1.00 (0.990‐1.015 95% CI) for P. brevifolia.

Elasticities under annual burning for A. michauxii populations indicated that λ was most sensitive to changes in the survivorship of small individuals. Under longer fire‐return intervals, λ was most sensitive to changes in the survivorship of medium‐ and large‐sized individuals (Table 5.5).

Elasticities for P. brevifolia demonstrated similar patterns (Table 5.6). Stable stage distributions of A. michauxii and P. brevifolia indicated that annual burning would result in a larger proportion of individuals in smaller size classes (Table 5.7 and Table 5.8), but the two species had different growth responses following fire. For A. michauxii, Keyfitz’s ∆ decreased from 0.68 in the year burned to 0.09 the first year post‐burning and 0.05 in the second year, indicating that burning altered the stage distribution, but by the second year there was essentially no difference in the observed stage distribution and the stable stage distribution. Recovery to the projected stable stage distribution was slower in P. brevifolia, with ∆ = 0.38 one year post‐burning and ∆ = 0.31 three years post‐ burning, indicating that P. brevifolia may take a longer time than the current three‐year fire‐return interval to reach a stable stage distribution.

LTRE results indicated that decreased mortality (0.025 ∆λ) and increased growth (0.027 ∆λ) of A. michauxii individuals in the “small‐medium” size class accounted for the 51% of the difference

113 in λ between burned and unburned projection matrices (Fig. 5.6). Overall, individual growth made the greatest contribution to the difference in λ, with fecundity making only a minor contribution. For

P. brevifolia, increased survivorship of the smallest individuals (0.038 ∆λ) and increased growth of medium‐sized individuals (0.040 ∆λ; Fig. 5.6) made the greatest contribution to the difference in λ.

As with A. michauxii, fecundity contributed very little to the overall difference in λ (0.003 ∆λ).

5.5 Discussion

In the short‐term, fire caused increased mortality and reduced size and seed production of

A. michauxii and P. brevifolia, two rare, endemic species of the Fall‐line Sandhills, USA. This is in contrast to many common and rare plant species in the Gulf and Atlantic Coastal Plain (GACP) that demonstrate short‐term positive responses to burning (Brewer 2001; Kesler et al. 2008; Kirkman et al. 1998; Spier and Snyder 1998). The reduced flowering in A. michauxii following fire contrasts with other members of Fabaceae in the longleaf pine ecosystem that generally flower in the same season following fire (Hiers et al. 2000). While species in fire‐dependent ecosystems may demonstrate a trade‐off in vital rate responses to fire (Menges and Quintana‐Ascencio 2004), such trade‐offs were not evident in the study species. Recruitment was low for both species throughout the fire cycle and did not compensate for the increased mortality and reduced reproduction following fire.

The reduction in plant size caused by burning increased the proportion of smaller individuals in the stable stage distribution for both species relative to unburned populations (Table 5.7 and

Table 5.8) and increased the elasticities of smaller size classes. This shift indicates that under annual burning the population growth rate would be more sensitive to the vital rates of the smaller size classes and that survivorship and growth of smaller individuals would make the greatest

114 contribution to the differences in λ between burned and unburned populations. Although individuals from both species experienced size reduction following fire, recovery to pre‐burn size was faster in A. michauxii relative to P. brevifolia. By the first growing season after fire, A. michauxii had roughly the same observed size distribution as the projected stable stage distribution in the absence of burning. The recovery of P. brevifolia was much slower; by the third year post fire the proportional difference between the observed size distribution and the stable stage distribution was still quite different. These results support anecdotal observations that the species is slow‐growing and that frequent fire leads to a greater proportion of small sized individuals, relative to populations that are burned less frequently.

The reduced stochastic population growth rates (λs) under simulated annual burning are not surprising, as fire‐dependent ecosystems are most likely composed of species adapted to different fire frequencies and other species in fire‐dependent ecosystems have demonstrated similar results

(Gross et al. 1998; Kaye et al. 2001; Menges and Quintana‐Ascencio 2004). However, even under the three year fire‐return interval to which these populations are currently exposed, simulations predict a gradual population decline for both species and only under a regular four year fire‐return interval was λs greater than one for either species (Fig. 5.5). We strongly caution against using this particular result to inform a change in management policy, owing to uncertainty inherent in short‐term demographic studies. The vital rates of P. brevifolia and A. michauxii most likely vary extensively through time (Hairston et al. 1996), with episodic recruitment and aperiodic bonanza years that could greatly influence long‐term population viability (Ludwig 1999). During the course of the study, the area was under a drought for most of the study and annual rainfall averaged 85% of the 30‐year normal rainfall of 120.7 cm (State Climate Office of North Carolina, http://www.nc‐

115 climate.ncsu.edu). Additionally, seedling detection probabilities are most likely less than one for P. brevifolia, and we may have missed a number of seedlings that established near maternal plants, as they are difficult to distinguish from the below ground stems of existing individuals (W. Wall, pers. obs.). Finally, the seed bank for both species may play a role in the long‐term population dynamics of both species, especially during bonanza years, but the low recruitment rates did not allow us to parameterize the seed bank over the course of the study as we did not observe any recruitment in the seed addition plots. Thus, recruitment is most likely higher than estimated in P. brevifolia and possibly A. michauxii, and λs < 1 are not necessarily a cause for immediate concern. Despite this, our results provide valuable information on fire effects because population growth rates in these species are most responsive to changes in growth and survival, rather than reproduction.

While fire has negative short‐term effects on the vital rates of these two species, fire is necessary under present climatic conditions to reduce competition and allow for population persistence. Long‐term fire exclusion in the longleaf pine ecosystem has been shown to have deleterious effects on population dynamics and plant biodiversity (Brockway and Lewis 1997). Fort

Bragg is burned on a fixed, three‐year fire‐return interval, so we were unable to observe the effects of fire suppression after three years. By the second year post‐burn, A. michauxii fruit production was decreasing for small and small‐medium sized individuals. Observations of P. brevifolia populations on adjacent lands that are under long‐term fire suppression suggest that fruit production is substantially reduced, relative to the number of fruits produced in populations that are regularly burned (W. Wall, pers. obs.). As with other studies of the responses of plant species to fire in the longleaf pine ecosystem (Hiers et al. 2000), a fire regime that includes variation in frequency and seasonality may be the optimal management strategy.

116 It is likely that A. michauxii and P. brevifolia have persisted in the Fall‐line Sandhills in dry, open habitats that have certain characteristics, such as sparse tree cover and low productivity, in common with Pleistocene landscapes (Watts 1980a; Webb III et al. 1998). Climatic reconstructions for the region suggest a Pleistocene environment with exposed soil and a dry, windy environment

(Ivester et al. 2001; Leigh 2008); these conditions most likely reduced plant productivity and biomass accumulation. As climatic conditions became progressively warmer and wetter, A. michauxii and P. brevifolia may have been isolated to deep sand habitats in the Fall‐line Sandhills where competition was reduced because of the extreme xeric conditions. The two endemic species could be viewed as remnants from a past vegetation assemblage that is not currently present. The end of the Pleistocene also lead to the northward migration of many plant species out of one or more refugia in the southeastern United States (Soltis et al. 2006), resulting in a mixture of species adapted to variable fire‐return intervals. Thus, it is not surprising to discover that two species endemic to a currently fire‐prone landscape do not appear adapted to one year fire‐return intervals.

Climatic and vegetation changes over the last 20,000 years in the GACP are complex, and more research is warranted to understand the interplay between fire, climate, and the demographic responses of species. While many species in fire‐dependent ecosystems may respond positively to increased fire frequency, we should not expect that all species will do so. A thorough investigation of the effects of fire on rare plant species in fire‐dependent ecosystems and a better understanding of the evolutionary history of targeted species are critical to both understanding the role of fire in determining the population dynamics of rare species and the development of successful management and recovery plans.

117 5.6 References

Bond WJ, Woodward FI, Midgley GF (2004) The global distribution of ecosystems in a world without fire. New Phytologist 165:525‐538.

Brewer JS (2001) A demographic analysis of fire‐stimulated seedling establishment of Sarracenia alata (Sarraceniaceae). American Journal of Botany 88:1250‐1257.

Brockway DG, Lewis CE (1997) Long‐term effects of dormant‐season prescribed fire on plant community diversity, structure and productivity in a longleaf pine wiregrass ecosystem. Forest Ecology and Management 96:167–183.

Caswell H (1996) Analysis of life table response experiments II. Alternative parameterizations for size‐ and stage‐structured models. Ecological Modelling 88:73‐82.

Goodman LA (1968) An elementary approach to the population projection‐matrix, to the population reproductive value, and to related topics in the mathematical theory of population growth. Demography 5:382‐409.

Hairston NG, Ellner SP, Kearns CM (1996) Overlapping generations: the storage effect and the maintenance of biotic diversity. In: Rhodes OE, Chesser RK, Smith MH (eds) Population dynamics in ecological space and time. University of Chicago Press, Chicago, IL, pp 109‐145

Hartnett DC, Richardson DR (1989) Population biology of Bonamia grandiflora (Convolulaceae): Effects of fire on plant and seed bank dynamics. American Journal of Botany 76:361–369.

Hiers JK, Wyatt R, Mitchell RJ (2000) The effects of fire regime on legume reproduction in longleaf pine savannas: is a season selective? Oecologia 125:521–530

Hoffmann WA (1999) Fire and Population Dynamics of Woody Plants in a Neotropical Savanna: Matrix Model Projections. Ecology 80:1354‐1369.

Ivester AH, Leigh DS, Godfrey‐Smith DI (2001) Chronology of inland eolian dunes on the Coastal Plain of Georgia, USA. Quaternary Research 55:293–302.

Jackson ST, Webb RS, Anderson KH, et al. (2000) Vegetation and environment in eastern North America during the last glacial maximum. Quaternary Science Reviews 19:489–508.

Jackson ST, Williams JW (2004) Modern analogs in Quaternary paleoecology: Here today, gone yesterday, gone tomorrow? Annual Review of Earth and Planetary Sciences 32:495‐537.

118 Kaye TN, Pendergrass KL, Finley K, Kauffman JB (2001) The effect of fire on the population viability of an endangered prairie plant. Ecological Applications 11:1366–1380

Kesler HC, Trusty JL, Hermann SM, Guyer C (2008) Demographic responses of Pinguicula ionantha to prescribed fire: a regression‐design LTRE approach. Oecologia 156:545‐557.

Keyfitz N (1968) Introduction to the mathematics of population. Addison‐Wesley, Reading, MA

Kirkman LK, Drew MB, Edwards D (1998) Effects of experimental fire regimes on the population dynamics of Schwalbea americana L. Plant Ecology 137:115‐137. de Kroon H, Plaiser A, van Groenendael J, Caswell H (1986) The relative contribution of demographic parameters to population growth rate. Ecology 67:1427–1431.

Leigh DS (2008) Late Quaternary climates and river channels of the Atlantic Coastal Plain, Southeastern USA. Geomorphology 101:90‐108.

Lewontin RC, Cohen D (1969) On population growth in a randomly varying environment. Proceedings of the National Academy of Sciences of the United States of America 62:1056‐ 1060.

Liu H, Menges ES, Quintana‐Ascencio PF (2005) Population viability analyses of Chamaecrista keyensis: effects of fire season and frequency. Ecological Applications 15:210–221.

Ludwig D (1999) Is it meaningful to estimate a probability of extinction? Ecology 80:298–310.

Menges ES (2007) Integrating demography and fire management: an example from Florida scrub. Australian Journal of Botany 55:261–272.

Menges ES, Quintana‐Ascencio PF (2004) Population viability with fire in Eryngium cuneifolium: Deciphering a decade of demographic data. Ecological Monographs 74:79‐99.

Moreira AG (2000) Effects of fire protection on savanna structure in Central Brazil. Journal of Biogeography 27:1021–1029.

Morrison DA, Gary GJ, Pengelly SM, et al. (1995) Effects of fire frequency on plant species composition of sandstone communities in the Sydney region: Inter‐fire interval and time‐ since‐fire. Australian Journal of Ecology 20:239–247.

Pinheiro JC, Bates DM (2000) Mixed‐effects models in S and S‐PLUS. Springer Verlag, New York

Power MJ, Marlon J, Ortiz N, et al. (2007) Changes in fire regimes since the Last Glacial Maximum: an assessment based on a global synthesis and analysis of charcoal data. Climate Dynamics 30:887–907.

119 Quintana‐Ascencio PF, Menges ES (2000) Competitive abilities of three narrowly endemic plant species in experimental neighborhoods along a fire gradient. American Journal of Botany 87:690‐699.

R Development Core Team (2009) R: a Language and environment for statistical computing. Vienna, Austria

Schemske DW, Husband BC, Ruckelshaus MH, et al. (1994) Evaluating approaches to the conservation of rare and endangered plants. Ecology 75:584–606.

Seamon P, Myers R, Robbins L, Seamon G (1989) Wiregrass reproduction and community restoration. Natural Areas Journal 9:264‐265.

Slapcinsky JL, Gordon DR, Menges ES (2010) Responses of rare plant species to fire in Florida’s pyrogenic communities. Natural Areas Journal 30:4‐19.

Soltis DE, Morris AB, McLachlan JS, Manos PS, Soltis PS (2006) Comparative phylogeography of unglaciated eastern North America. Molecular Ecology 15:4261–4293

Sorrie BA, Weakley AS (2001) Coastal plain vascular plant endemics: Phytogeographic patterns. Castanea 66:50‐82.

Spier L, Snyder J (1998) Effects of wet‐ and dry‐season fires on Jacquemontia curtisii, a South Florida pine forest endemic. Natural Areas Journal 18:350–357.

Staver AC, Archibald S, Levin SA (2011) The global extent and determinants of savanna and forest as alternative biome states. Science 334:230–232.

Wall WA, Douglas NA, Xiang QY (Jenny), Hofmann WA, Wentworth TR, Hohmann MG. (2010) Evidence for range stasis during the latter Pleistocene for the Atlantic Coastal Plain endemic genus, Pyxidanthera Michaux. Molecular Ecology 19:4302–4314

Watts WA (1980a) The late Quaternary vegetation history of the southeastern United States. Annual Review of Ecology and Systematics 11:387–409.

Watts WA (1980b) Late‐Quaternary vegetation history at White Pond on the inner Coastal Plain of South Carolina. Quaternary Research 13:187–199.

Webb III T, Anderson KH, Bartlein PJ, Webb RS (1998) Late quaternary climate change in eastern North America a comparison of pollen‐derived estimates with climate model results. Quaternary Science Reviews 17:587–606.

120 Weeks, S (2005) Factors limiting growth in Astragalus michauxii (Sandhills milk‐vetch). Doctoral dissertation, North Carolina State University

121 Table 5.1: Number of A. michauxii individuals used to estimate survivorship, transition probabilities, and reproductions for each size class. TSB refers to time since burn, with “0” indicating year that population was burned.

TSB Year Size Class

small small-medium medium large 0 2007 10 31 46 2 0 2008 7 19 41 9 0 2009 6 37 82 27

1 2007 50 81 27 2 1 2008 40 61 12 0 1 2009 45 41 47 5

2 2007 13 36 64 9 2 2008 8 24 108 27 2 2009 11 26 58 16

122 Table 5.2: Transition matrices for A. michauxii individuals that were recently burned, 1 year post‐ fire, and 2 years post‐fire. Stages delineated based on the tallest stem: small (0.01 – 20 cm), small‐ medium (>20‐40 cm), medium (>40‐80 cm), and large (> 80 cm).

Size Class burned small small‐medium medium large small 0.65217 0.47127 0.23078 0.07895 small‐medium 0.21739 0.40230 0.62130 0.50000 medium 0.00000 0.04598 0.10059 0.34211 large 0.00000 0.00000 0.00000 0.00000

1 year post‐fire small small‐medium medium large small 0.12653 0.03411 0.02429 0.00323 small‐medium 0.39259 0.15847 0.10465 0.00000 medium 0.37778 0.66667 0.53488 0.27771 large 0.02963 0.12568 0.33721 0.69428

2 years post‐fire small small‐medium medium large small 0.34391 0.07022 0.01024 0.00317 small‐medium 0.25000 0.43023 0.12609 0.05882 medium 0.28125 0.47674 0.67391 0.35294 large 0.00000 0.01163 0.17391 0.56863

123 Table 5.3: Number of P. brevifolia individuals used to estimate survivorship, transition probabilities, and reproductions for each size class. Populations were sampled before the burning season, so TSB = “1” indicates first measurements after population was burned.

size classes (cm2)

1- 10- 25- 50- 100- 200- 400- 800- 1600- TSB Year 10 25 50 100 200 400 800 1600 3200 >3200 1 2009 8 28 26 38 65 64 72 60 38 20 1 2010 4 3 14 25 48 59 47 42 19 8

2 2009 17 28 35 44 50 61 42 20 13 6 2 2010 18 14 33 51 68 56 65 53 21 20

3 2009 6 14 25 42 60 49 47 21 13 14 3 2010 5 4 12 31 38 62 65 33 17 8

124 Table 5.4: Transition matrices for P. brevifolia during the 2008‐2009 and 2009‐2010 transition intervals. Seedling class is an age class, while the other 10 size classes are defined by plant area (cm2).

2008‐2009 Size Class 1 Year Post‐fire 50‐ 100‐ 200‐ 400‐ 800‐ 1600‐ seed seedling 1‐10 10‐25 25‐50 100 200 400 800 1600 3200 >3200 seed 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 seedling 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1‐10 0.0000 0.0000 0.1250 0.0714 0.1538 0.0263 0.0154 0.0000 0.0000 0.0167 0.0000 0.0000 10‐25 0.0000 0.0000 0.1250 0.2143 0.1538 0.0789 0.0000 0.0156 0.0000 0.0000 0.0000 0.0000 25‐50 0.0000 0.0000 0.3750 0.2500 0.2672 0.1316 0.1538 0.0313 0.0278 0.0167 0.0000 0.0000 50‐100 0.0000 0.0000 0.1250 0.2857 0.3077 0.2895 0.2000 0.0938 0.0694 0.0000 0.0000 0.0000 100‐200 0.0000 0.0000 0.0000 0.0357 0.0769 0.3421 0.3385 0.2188 0.1806 0.0500 0.0000 0.0000 200‐400 0.0000 0.0000 0.0000 0.0357 0.0385 0.0526 0.1692 0.4043 0.1806 0.0500 0.0000 0.0000 400‐800 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0923 0.2344 0.4167 0.2000 0.1053 0.0000 800‐1600 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0154 0.0000 0.1111 0.5813 0.3684 0.0500 1600‐3200 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0833 0.3401 0.2000 >3200 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.1842 0.7480

2 Years Post‐fire 50‐ 100‐ 200‐ 400‐ 800‐ 1600‐ seed seedling 1‐10 10‐25 25‐50 100 200 400 800 1600 3200 >3200 seed 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 seedling 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1‐10 0.0000 0.0000 0.1765 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 10‐25 0.0000 0.0000 0.1765 0.0000 0.0000 0.0227 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 25‐50 0.0000 0.0000 0.1765 0.2143 0.0551 0.0455 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 50‐100 0.0000 0.0000 0.1765 0.5714 0.3143 0.0682 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 100‐200 0.0000 0.0000 0.1765 0.1429 0.3429 0.3636 0.1180 0.0492 0.0238 0.0000 0.0000 0.0000 200‐400 0.0000 0.0000 0.0000 0.0357 0.2857 0.3864 0.4000 0.2931 0.0000 0.0000 0.0000 0.0000 400‐800 0.0000 0.0000 0.0000 0.0000 0.0000 0.0909 0.4800 0.5738 0.2123 0.0500 0.0000 0.0000 800‐1600 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0820 0.7143 0.2480 0.0769 0.0000 1600‐3200 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0476 0.7000 0.3826 0.0000 >3200 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.5385 0.9980

3 Years Post‐fire 50‐ 100‐ 200‐ 400‐ 800‐ 1600‐ seed seedling 1‐10 10‐25 25‐50 100 200 400 800 1600 3200 >3200 seed 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 seedling 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1‐10 0.0000 0.0000 0.1647 0.0714 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 10‐25 0.0000 0.0000 0.3333 0.0000 0.0000 0.0000 0.0167 0.0204 0.0000 0.0000 0.0000 0.0000 25‐50 0.0000 0.0000 0.3333 0.2857 0.1580 0.0000 0.0500 0.0204 0.0000 0.0000 0.0000 0.0000 50‐100 0.0000 0.0000 0.1667 0.4286 0.2800 0.1885 0.0333 0.0000 0.0213 0.0000 0.0000 0.0000 100‐200 0.0000 0.0000 0.0000 0.0714 0.4000 0.5000 0.2647 0.0000 0.0000 0.0000 0.0769 0.0000 200‐400 0.0000 0.0000 0.0000 0.0000 0.1600 0.2619 0.4167 0.3858 0.0426 0.0000 0.0769 0.0000 400‐800 0.0000 0.0000 0.0000 0.0000 0.0000 0.0476 0.2167 0.4694 0.3171 0.0000 0.0000 0.0000 800‐1600 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.1020 0.5957 0.4266 0.1538 0.0000 1600‐3200 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0213 0.5714 0.4595 0.1429 >3200 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.2308 0.8551

125 Table 5.4, continued

2009‐2010 Size Class 1 Year Post‐fire 50‐ 100‐ 200‐ 400‐ 800‐ 1600‐ seed seedling 1‐10 10‐25 25‐50 100 200 400 800 1600 3200 >3200 seed 0.0000 0.0000 0.0 1.1 37.9 63.7 232.6 398.3 1236.9 1463.8 2431.0 2292.6 seedling 0.00003 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1‐10 0.0000 0.1250 0.4980 0.3333 0.0714 0.0800 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 10‐25 0.0000 0.0000 0.0000 0.3333 0.0714 0.0800 0.1042 0.0339 0.0638 0.0238 0.0000 0.0000 25‐50 0.0000 0.0000 0.5000 0.0000 0.3571 0.1600 0.1458 0.0339 0.0213 0.0000 0.0526 0.0000 50‐100 0.0000 0.0000 0.0000 0.0000 0.0714 0.4800 0.2708 0.1525 0.0426 0.0238 0.0000 0.0000 100‐200 0.0000 0.0000 0.0000 0.0000 0.2857 0.0400 0.2500 0.2712 0.1702 0.0238 0.0000 0.0000 200‐400 0.0000 0.0000 0.0000 0.0000 0.0714 0.0800 0.1667 0.3559 0.3191 0.1667 0.0000 0.0000 400‐800 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.1356 0.3617 0.1905 0.1053 0.0000 800‐1600 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0213 0.4524 0.3158 0.0000 1600‐3200 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0417 0.0000 0.0000 0.1190 0.4211 0.0000 >3200 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.1053 0.8750

2 Years Post‐fire 50‐ 100‐ 200‐ 400‐ 800‐ 1600‐ seed seedling 1‐10 10‐25 25‐50 100 200 400 800 1600 3200 >3200 seed 0.0000 0.0000 6.9 57.5 187.8 182.7 568.7 1247.7 2269.2 4321.4 5750.7 9056.4 seedling 0.00003 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1‐10 0.0000 0.2414 0.0980 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 10‐25 0.0000 0.0690 0.7000 0.0000 0.0606 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 25‐50 0.0000 0.1034 0.1000 0.2143 0.0909 0.0784 0.0000 0.0000 0.0154 0.0000 0.0000 0.0000 50‐100 0.0000 0.0690 0.0000 0.6429 0.2727 0.0960 0.0147 0.0179 0.0308 0.0000 0.0000 0.0000 100‐200 0.0000 0.0345 0.1000 0.0714 0.4848 0.4314 0.2039 0.0357 0.0308 0.0000 0.0476 0.0000 200‐400 0.0000 0.0000 0.0000 0.0000 0.0606 0.3725 0.4853 0.3373 0.0308 0.0189 0.0000 0.0000 400‐800 0.0000 0.0000 0.0000 0.0000 0.0000 0.0196 0.2647 0.5357 0.3057 0.0377 0.0476 0.1000 800‐1600 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0294 0.0714 0.5385 0.4508 0.0476 0.0000 1600‐3200 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0462 0.4528 0.7123 0.0500 >3200 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0377 0.1429 0.8480

3 Years Post‐fire 50‐ 100‐ 200‐ 400‐ 800‐ 1600‐ seed seedling 1‐10 10‐25 25‐50 100 200 400 800 1600 3200 >3200 seed 0.0000 0.0000 0.0 11.9 82.4 321.5 437.3 705.7 1234.1 2663.1 2763.3 3898.8 seedling 0.00003 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1‐10 0.0000 0.2414 0.3313 0.0000 0.0000 0.0000 0.0263 0.0000 0.0000 0.0000 0.0000 0.0000 10‐25 0.0000 0.0690 0.3333 0.2480 0.0000 0.0000 0.0000 0.0161 0.0000 0.0000 0.0000 0.0000 25‐50 0.0000 0.1034 0.0000 0.5000 0.2480 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 50‐100 0.0000 0.0690 0.0000 0.2500 0.5833 0.4496 0.0789 0.0323 0.0000 0.0000 0.0000 0.0000 100‐200 0.0000 0.0345 0.3333 0.0000 0.1667 0.4839 0.3664 0.0645 0.0000 0.0000 0.0588 0.0000 200‐400 0.0000 0.0000 0.0000 0.0000 0.0000 0.0645 0.4474 0.3710 0.1231 0.0303 0.0000 0.0000 400‐800 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0263 0.4677 0.6288 0.0909 0.0588 0.0000 800‐1600 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0263 0.0323 0.2462 0.5738 0.1176 0.0000 1600‐3200 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0263 0.0000 0.0000 0.3030 0.7039 0.0000 >3200 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0588 0.9980

126 Table 5.5: Elasticity values for A. michauxii transition matrices. Size classes are based on tallest stem and are small (0.1 – 20 cm), small‐medium (>20‐40 cm), medium (>40‐80 cm), and large (> 80 cm).

Size Class burned small small‐medium medium large small 0.47235 0.16544 0.00474 0.00000 small‐medium 0.17017 0.15264 0.01378 0.00000 medium 0.00000 0.01852 0.00237 0.00000 large 0.00000 0.00000 0.00000 0.00000

1 year post‐fire small small‐medium medium large small 0.00195 0.00196 0.00967 0.00157 small‐medium 0.00646 0.00978 0.04460 0 medium 0.00629 0.04155 0.23018 0.14620 large 0.00047 0.00754 0.13975 0.35200

2 years post‐fire small small‐medium medium large small 0.00838 0.00999 0.00490 0.00065 small‐medium 0.00728 0.07314 0.07204 0.01443 medium 0.00826 0.08176 0.38839 0.08735 large 0.00000 0.00200 0.10043 0.14102

127 Table 5.6: Elasticity estimates for P. brevifolia during the 2008‐2009 and 2009‐2010 transition intervals. Seedling class is an age class, while the other 10 size classes are defined by plant area (cm2).

2008‐2009 Size Class 1 Year Post‐fire 50‐ 100‐ 200‐ 400‐ 800‐ 1600‐ seed seedling 1‐10 10‐25 25‐50 100 200 400 800 1600 3200 >3200 seed 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 seedling 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1‐10 0.0000 0.0000 0.0037 0.0030 0.0152 0.0036 0.0023 0.0000 0.0000 0.0006 0.0000 0.0000 10‐25 0.0000 0.0000 0.0045 0.0110 0.0187 0.0132 0.0000 0.0019 0.0000 0.0000 0.0000 0.0000 25‐50 0.0000 0.0000 0.0151 0.0142 0.0360 0.0243 0.0316 0.0042 0.0032 0.0009 0.0000 0.0000 50‐100 0.0000 0.0000 0.0050 0.0163 0.0416 0.0537 0.0413 0.0126 0.0079 0.0000 0.0000 0.0000 100‐200 0.0000 0.0000 0.0000 0.0023 0.0118 0.0718 0.0790 0.0334 0.0233 0.0029 0.0000 0.0000 200‐400 0.0000 0.0000 0.0000 0.0025 0.0063 0.0118 0.0424 0.0661 0.0250 0.0032 0.0000 0.0000 400‐800 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0235 0.0390 0.0586 0.0129 0.0013 0.0000 800‐1600 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0044 0.0000 0.0174 0.0416 0.0049 0.0006 1600‐3200 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0067 0.0051 0.0026 >3200 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0032 0.0111

2 Years Post‐fire 50‐ 100‐ 200‐ 400‐ 800‐ 1600‐ seed seedling 1‐10 10‐25 25‐50 100 200 400 800 1600 3200 >3200 seed 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 seedling 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1‐10 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 10‐25 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 25‐50 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 50‐100 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 100‐200 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 200‐400 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 400‐800 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 800‐1600 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1600‐3200 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 >3200 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000

3 Years Post‐fire 50‐ 100‐ 200‐ 400‐ 800‐ 1600‐ seed seedling 1‐10 10‐25 25‐50 100 200 400 800 1600 3200 >3200 seed 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 seedling 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1‐10 0.0000 0.0000 0.0000 0.0001 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 10‐25 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0005 0.0011 0.0000 0.0000 0.0000 0.0000 25‐50 0.0000 0.0000 0.0001 0.0005 0.0007 0.0000 0.0017 0.0013 0.0000 0.0000 0.0000 0.0000 50‐100 0.0000 0.0000 0.0000 0.0008 0.0012 0.0010 0.0011 0.0000 0.0012 0.0000 0.0000 0.0000 100‐200 0.0000 0.0000 0.0000 0.0001 0.0017 0.0027 0.0090 0.0000 0.0000 0.0000 0.0203 0.0000 200‐400 0.0000 0.0000 0.0000 0.0000 0.0007 0.0014 0.0142 0.0246 0.0023 0.0000 0.0204 0.0000 400‐800 0.0000 0.0000 0.0000 0.0000 0.0000 0.0003 0.0074 0.0300 0.0176 0.0000 0.0000 0.0000 800‐1600 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0065 0.0330 0.0601 0.0409 0.0000 1600‐3200 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0012 0.0804 0.1222 0.0615 >3200 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0615 0.3687

128 Table 5.6, continued

2009‐2010 Size Class 1 Year Post‐fire 50‐ 100‐ 200‐ 400‐ 800‐ 1600‐ seed seedling 1‐10 10‐25 25‐50 100 200 400 800 1600 3200 >3200 seed 0.0000 0.0000 0.0000 0.0000 0.0000 0.0001 0.0002 0.0003 0.0002 0.0001 0.0002 0.0004 seedling 0.0014 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1‐10 0.0000 0.0014 0.0861 0.0367 0.0195 0.0174 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 10‐25 0.0000 0.0000 0.0000 0.0204 0.0109 0.0097 0.0115 0.0028 0.0016 0.0002 0.0000 0.0000 25‐50 0.0000 0.0000 0.0751 0.0000 0.0848 0.0303 0.0252 0.0044 0.0008 0.0000 0.0009 0.0000 50‐100 0.0000 0.0000 0.0000 0.0000 0.0163 0.0874 0.0450 0.0192 0.0016 0.0003 0.0000 0.0000 100‐200 0.0000 0.0000 0.0000 0.0000 0.0708 0.0079 0.0450 0.0370 0.0068 0.0003 0.0000 0.0000 200‐400 0.0000 0.0000 0.0000 0.0000 0.0191 0.0170 0.0323 0.0522 0.0138 0.0024 0.0000 0.0000 400‐800 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0209 0.0164 0.0029 0.0021 0.0000 800‐1600 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0011 0.0076 0.0070 0.0000 1600‐3200 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0087 0.0000 0.0000 0.0019 0.0087 0.0000 >3200 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0004 0.0054

2 Years Post‐fire 50‐ 100‐ 200‐ 400‐ 800‐ 1600‐ seed seedling 1‐10 10‐25 25‐50 100 200 400 800 1600 3200 >3200 seed 0.0000 0.0000 0.0000 0.0000 0.0000 0.0001 0.0004 0.0009 0.0027 0.0058 0.0127 0.0160 seedling 0.0386 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1‐10 0.0000 0.0169 0.0017 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 10‐25 0.0000 0.0050 0.0127 0.0000 0.0010 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 25‐50 0.0000 0.0080 0.0020 0.0041 0.0017 0.0022 0.0000 0.0000 0.0014 0.0000 0.0000 0.0000 50‐100 0.0000 0.0057 0.0000 0.0131 0.0053 0.0029 0.0007 0.0010 0.0029 0.0000 0.0000 0.0000 100‐200 0.0000 0.0030 0.0022 0.0015 0.0100 0.0137 0.0107 0.0022 0.0031 0.0000 0.0089 0.0000 200‐400 0.0000 0.0000 0.0000 0.0000 0.0013 0.0122 0.0263 0.0213 0.0032 0.0022 0.0000 0.0000 400‐800 0.0000 0.0000 0.0000 0.0000 0.0000 0.0007 0.0153 0.0360 0.0339 0.0048 0.0098 0.0166 800‐1600 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0018 0.0052 0.0641 0.0610 0.0105 0.0000 1600‐3200 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0057 0.0630 0.1620 0.0091 >3200 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0058 0.0359 0.1714

3 Years Post‐fire 50‐ 100‐ 200‐ 400‐ 800‐ 1600‐ seed seedling 1‐10 10‐25 25‐50 100 200 400 800 1600 3200 >3200 seed 0.0000 0.0000 0.0000 0.0000 0.0000 0.0002 0.0003 0.0007 0.0020 0.0036 0.0037 0.0103 seedling 0.0207 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1‐10 0.0000 0.0095 0.0052 0.0000 0.0000 0.0000 0.0014 0.0000 0.0000 0.0000 0.0000 0.0000 10‐25 0.0000 0.0027 0.0052 0.0028 0.0000 0.0000 0.0000 0.0010 0.0000 0.0000 0.0000 0.0000 25‐50 0.0000 0.0042 0.0000 0.0059 0.0032 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 50‐100 0.0000 0.0029 0.0000 0.0030 0.0077 0.0156 0.0043 0.0022 0.0000 0.0000 0.0000 0.0000 100‐200 0.0000 0.0015 0.0057 0.0000 0.0023 0.0175 0.0206 0.0045 0.0000 0.0000 0.0058 0.0000 200‐400 0.0000 0.0000 0.0000 0.0000 0.0000 0.0024 0.0260 0.0266 0.0156 0.0031 0.0000 0.0000 400‐800 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0016 0.0361 0.0855 0.0101 0.0064 0.0000 800‐1600 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0018 0.0027 0.0366 0.0698 0.0140 0.0000 1600‐3200 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0019 0.0000 0.0000 0.0383 0.0875 0.0000 >3200 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0103 0.3476

129 Table 5.7: Population growth rates, stable stage distributions, and reproductive values for the three transition matrices (burned, one year post‐fire, and two or more years post‐fire) estimated for A. michauxii. Stable stage distribution values represent proportions and sum to 1; reproductive values are scaled to the smallest size class, which is equal to 1.

One year post‐ Two years post‐ size class burned fire fire

lambda 0.88713 0.98575 0.98167

small 0.66093 0.01604 0.02866 stage

small‐medium 0.32035 0.06017 0.16736

medium 0.01873 0.41549 0.56245 distribution stable

large 0.00000 0.50827 0.24153

small 1.00000 1.00000 1.00000

small‐medium 1.08080 1.07018 1.19493 value medium 1.14716 1.08065 1.20540 reproductive

large 1.14053 1.040736 1.20786

130 Table 5.8: Population growth rates, stable stage distributions, and reproductive values for the 6 transition matrices (2 time steps and burned, 1 year post‐fire, and 2 years post‐fire) estimated for P. brevifolia. Stable stage distribution values represent proportions and sum to 1; reproductive values are scaled to the smallest size class, which is equal to 1. Stable stage distribution does not include the seed stage and the reproductive values are scaled to seedling stage. 2008‐2009 2009‐2010

burned unburned1 unburned2 burned unburned1 unburned2

lambda 0.962 0.998 0.998 0.932 1.054 1.028

seed 0.00000 0.00000 0.00000 NA NA NA

seedling 0.00000 0.00000 0.00000 0.00919 0.10267 0.05637

1‐10 0.04282 0.00000 0.00017 0.13832 0.02591 0.02232

10‐25 0.06043 0.00000 0.00193 0.08807 0.02531 0.01641

25‐50 0.14323 0.00000 0.00430 0.21885 0.02428 0.01801

distribution 50‐100 0.19675 0.00000 0.00540 0.17427 0.03737 0.04711

100‐200 0.21877 0.00000 0.03405 0.15895 0.06177 0.07351

200‐400 0.14286 0.00000 0.06381 0.12035 0.07190 0.09077 stage

400‐800 0.12078 0.00000 0.05523 0.03548 0.11870 0.16012

800‐1600 0.05526 0.00000 0.14046 0.01192 0.13478 0.13090

stable 1600‐3200 0.01026 0.00000 0.26529 0.01573 0.22025 0.12853

>3200 0.00885 1.00000 0.42936 0.02886 0.17703 0.25594

seed NA NA NA <0.0001 <0.0001 <0.0001

seedling NA NA NA 1.0 1.0 1.0

1‐10 0 0 0 7.5 1.9 2.0

10‐25 0 0 0 4.2 2.0 1.9 value 25‐50 0 0 0 6.5 2.1 2.0

50‐100 0 0 0 6.2 2.3 2.1

100‐200 0 0 0 6.8 2.4 2.1

200‐400 0 0 0 7.3 2.5 2.2

400‐800 0 0 0 7.6 2.6 2.4

reproductive 800‐1600 0 0 0 8.5 2.8 2.6

1600‐3200 0 0 0 7.9 2.9 2.7

>3200 0 0 0 1.3 3.2 3.8

131

Fig. 5.1. time line for collection of demographic data for Astragalus michauxii (top) and Pyxidanthera brevifolia (bottom). Grey boxes represent prescribed burn season, long dashes represent data collection, and letters along x‐axis represent seasons. Data collection for A. michauxii occurred during the burn season, with burned populations measured at the end of the growing season (short dashed line). Data collection for P. brevifolia occurred before the burning season; measurements were roughly 9 months post‐fire in burned populations.

132

Fig. 5.2. Mortality of Astragalus michauxii (top) and Pyxidanthera brevifolia (bottom) as a function of time since last burn and size class. Error bars represent standard error of the mean.

133

Fig. 5.3. Mean number of fruits produced as a function of time since last burn and size class for Astragalus michauxii (top) and Pyxidanthera brevifolia (bottom). Error bars represent standard error of the mean.

134

Fig. 5.4. Modified box plot of post‐fire recovery rates for Astragalus michauxii (top) and Pyxidanthera brevifolia (bottom) individuals as a function of pre‐burn size. Dotted line represents recovery to pre‐burn size. Smaller individuals recover at a faster rate relative to larger individuals for both species.

135

Fig. 5.5. Projected stochastic population growth rates under different fire‐return intervals (1‐4 years) for Astragalus michauxii (left) and Pyxidanthera brevifolia (right) using a matrix selection approach. Error bars represent bootstrapped 95% confidence intervals.

136

Fig. 5.6. Contributions of growth, survivorship, and fecundity by size class to the difference in the population growth rate between unburned and burned Astragalus michauxii (top) and Pyxidanthera brevifolia (bottom) populations. Size classes for A. michauxii are small = 1‐20 cm, small‐medium = >20‐40 cm, medium = >40‐80 cm, and large > 80 cm. Size classes for P. brevifolia are small = 1‐50 cm2, medium = >50‐400 cm2, and large > 400 cm2.

137