Citation
Population genomics of a rare edaphic endemic

Material Information

Title:
Population genomics of a rare edaphic endemic
Creator:
Bard, Nicholas Wehby
Place of Publication:
Denver, CO
Publisher:
University of Colorado Denver
Publication Date:
Language:
English

Thesis/Dissertation Information

Degree:
Master's ( Master of science)
Degree Grantor:
University of Colorado Denver
Degree Divisions:
Department of Integrative Biology, CU Denver
Degree Disciplines:
Biology
Committee Chair:
Miller, Christopher S.
Committee Members:
Bruederle, Leo P.
Ramp-Neale, Jennifer

Notes

Abstract:
Range-limited rare endemic species are threatened by loss of genomic diversity, which can lead to extirpation and extinction. The focal taxon of this study, the edaphic endemic Carex scirpoidea subsp. convoluta (Cyperaceae), exhibits such risk. The narrowly distributed subsp. convoluta is primarily limited to alvar soils on the northeastern shores of Lake Huron in North America. In contrast, C. scirpoidea subsp. scirpoidea is widespread and found in diverse habitats and soil types across northern North America and Greenland, with few populations in Norway and Russia. Both taxa are of conservation concern in all or part of their ranges, and the characteristic alvar habitat of subsp. convoluta is considered to be imperiled. In this study, I measured genomic diversity in both taxa in order to provide insight about their continued survival and adaptation to future conditions. Finally, I assessed the differentiation of subsp. convoluta from subsp. scirpoidea, and how edaphic niche differs between taxa. As an edaphic endemic, I expected a narrow edaphic niche in subsp. convoluta that is distinct from subsp. scirpoidea. I compared niche and niche breadth for several soil parameters between taxa. When multiple parameters were considered, neither taxon occupied a distinct niche, while both taxa exhibited broad tolerance to key soil parameters, including cations. My results indicated that exchangeable sodium niche differed between taxa; subsp. convoluta soils consistently had a lower concentration than subsp. scirpoidea. The wide tolerance to edaphic conditions observed in C. scirpoidea has likely aided range expansion by occupying adverse soil conditions. Due to its limited distribution on fragmented alvar, subsp. convoluta was expected to have low genomic diversity, and be differentiated from subsp. scirpoidea. Double digest restriction-associated DNA (ddRAD) sequencing was used to compare genomic diversity and evaluate differentiation between subsp. convoluta and subsp. scirpoidea. My results suggest subsp. convoluta does not exhibit lower genomic diversity; nor does it exhibit signs of inbreeding depression — likely due to dioecy in C. scirpoidea. The results also suggest that differentiation of subsp. convoluta is at an early stage. The study supports earlier findings that North American subsp. scirpoidea populations are derived from three glacial refugia in Northwestern, Western, and Eastern North America. Carex scirpoidea subsp. convoluta arose following recolonization from Eastern North American populations of subsp. scirpoidea, rather than from other refugial populations. Finally, there is evidence for broad tolerance or local adaptation between populations, which may aid the apparent adaptability to extreme edaphic conditions. Though surveyed populations maintain genomic diversity, persistence of populations is contingent upon the preservation of suitable habitat. I recommend that land managers prioritize habitat protection for discrete populations, particularly at the margins of the range.
Restriction:
Embargo ended 06/03/2019

Record Information

Source Institution:
University of Colorado Denver
Holding Location:
Auraria Library
Rights Management:
Copyright Nicholas Wehby Bard. Permission granted to University of Colorado Denver to digitize and display this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.

Downloads

This item has the following downloads:


Full Text
POPULATION GENOMICS OF A RARE EDAPHIC ENDEMIC
by
NICHOLAS WEHBY BARD B.S., Portland State University, 2013
A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirements for the degree of Master of Science Biology Program
2018


©2018
NICHOLAS WEHBY BARD
ALL RIGHTS RESERVED


This thesis for the Master of Science degree by Nicholas Wehby Bard has been approved for the Biology Program by
Christopher S. Miller, Chair Leo P. Bruederle, Advisor Jennifer Ramp-Neale
Date: December 15, 2018


Bard, Nicholas Wehby. (M.S. Biology Program)
Population Genomics of a Rare Edaphic Endemic Thesis directed by Professor Leo P. Bruederle
Abstract
Range-limited rare endemic species are threatened by loss of genomic diversity, which can lead to extirpation and extinction. The focal taxon of this study, the edaphic endemic Carex scirpoidea subsp. convoluta (Cyperaceae), exhibits such risk. The narrowly distributed subsp. convoluta is primarily limited to alvar soils on the northeastern shores of Lake Huron in North America. In contrast, C. scirpoidea subsp. scirpoidea is widespread and found in diverse habitats and soil types across northern North America and Greenland, with few populations in Norway and Russia. Both taxa are of conservation concern in all or part of their ranges, and the characteristic alvar habitat of subsp. convoluta is considered to be imperiled. In this study, I measured genomic diversity in both taxa in order to provide insight about their continued survival and adaptation to future conditions. Finally, I assessed the differentiation of subsp. convoluta from subsp. scirpoidea, and how edaphic niche differs between taxa.
As an edaphic endemic, I expected a narrow edaphic niche in subsp. convoluta that is distinct from subsp. scirpoidea. I compared niche and niche breadth for several soil parameters between taxa. When multiple parameters were considered, neither taxon occupied a distinct niche, while both taxa exhibited broad tolerance to key soil parameters, including cations. My results indicated that exchangeable sodium niche differed between taxa; subsp. convoluta soils consistently had a lower concentration than subsp. scirpoidea. The wide tolerance to edaphic conditions observed in C. scirpoidea has likely aided range expansion by occupying adverse soil conditions.
IV


Due to its limited distribution on fragmented alvar, subsp. convoluta was expected to have low genomic diversity, and be differentiated from subsp. scirpoidea. Double digest restriction-associated DNA (ddRAD) sequencing was used to compare genomic diversity and evaluate differentiation between subsp. convoluta and subsp. scirpoidea. My results suggest subsp. convoluta does not exhibit lower genomic diversity; nor does it exhibit signs of inbreeding depression — likely due to dioecy in C. scirpoidea. The results also suggest that differentiation of subsp. convoluta is at an early stage. The study supports earlier findings that North American subsp. scirpoidea populations are derived from three glacial refugia in Northwestern, Western, and Eastern North America. Carex scirpoidea subsp. convoluta arose following recolonization from Eastern North American populations of subsp. scirpoidea, rather than from other refugial populations. Finally, there is evidence for broad tolerance or local adaptation between populations, which may aid the apparent adaptability to extreme edaphic conditions. Though surveyed populations maintain genomic diversity, persistence of populations is contingent upon the preservation of suitable habitat. I recommend that land managers prioritize habitat protection for discrete populations, particularly at the margins of the range.
The form and content of this abstract are approved. I recommend its publication.
Approved: Leo P. Bruederle
v


DEDICATION
To Erika, Mom, Dad, and Gammy, for your unending love and support. To my beloved friend Colin Ward, for being an extraterrestrial treepunk prophet. Afterlife on loop.
vi


ACKNOWLEDGEMENTS
Thank you to Dr. Leo P. Bruederle, for his invaluable expertise, inspiration, and mentorship. To Dr. Christopher S. Miller and Dr. Jennifer Ramp Neale, my deepest gratitude for your guidance and counsel. Writing my thesis would not have been possible without the aid provided by Dr. Erin Tripp, Dr. Michael Wunder, Dr. Gregory Ragland, Dr. Kristine Westergaard, and Dr. Brad Slaughter. Finally, thank you to my peers who provided support with laboratory methods and data analysis: Andrew Boddicker, McCall Calvert, Kelsie Faulds, Kathryn Kilpatrick, Jared Mastin, Elizabeth Pansing, Allison Pierce, Laura Sedivy, Aaron Wagner, and Scott Yanco.
Vll


TABLE OF CONTENTS
CHAPTER
I. A BIOLOGICAL OVERVIEW OF CAREXSCIRPOIDEA SUBSP. CONVOLUTA AND
SUBSP. SCIRPOIDEA (CYPERACEAE) AND THE CONSERVATION OF TAXON
AND HABITAT................................................................1
Taxonomy, Ecology, and Conservation Status of Carex scirpoidea subsp. convoluta and
subsp. scirpoidea..........................................................1
Conservation of Suitable Habitat...........................................2
Justification for Research.................................................4
Figure.....................................................................5
II. A COMPARISON OF THE SOIL CHEMICAL AND PHYSICAL CONDITONS FOR
POPULATIONS OF CAREX SCIRPOIDEA SUBSP. CONVOLUTA AND SUBSP.
SCIRPOIDEA (CYPERACEAE)................................................6
Abstract...............................................................6
Introduction...........................................................6
Methods................................................................8
Results...............................................................11
Discussion............................................................13
Tables and Figures....................................................18
III. POPULATION GENOMICS OF THE NARROW ENDEMIC CAREX SCIRPOIDEA
SUBSP. CONVOLUTA......................................................26
Abstract..............................................................26
Introduction..........................................................27
vm


Methods
31
Results.......................................................................32
Discussion....................................................................38
Tables and figures............................................................47
REFERENCES...........................................................................65
APPENDIX.............................................................................71
A. Descriptions of soil and leaf tissue sampling sites for Carex scirpoidea subsp. convoluta
(Cyperaceae)..................................................................72
B. Weights for all soil samples taken from Carex scirpoidea subsp. convoluta (Cyperaceae)
population sites..............................................................74
C. Soil chemical and physical parameters measured across the ranges of Carex scirpoidea
subsp. convoluta and subsp. scirpoidea (Cyperaceae)...........................75
D. Average daily soil temperature for all days determined to be in the growing season for all populations of Carex scirpoidea subsp. convoluta and subsp. scirpoidea (Cyperaceae). 85
E. Average number of alleles and associated standard deviation across 7384 measured loci
for 11 populations of Carex scirpoidea subsp. convoluta and subsp. scirpoidea (Cyperaceae)..................................................................92
F. Maximum likelihood unrooted phylogeny of all populations of Carex scirpoidea subsp.
convoluta and subsp. scirpoidea (Cyperaceae)..................................93
IX


LIST OF TABLES
TABLE
2.1 Soil sampling sites for Carex scirpoidea subsp. convoluta and subsp. scirpoidea
(Cyperaceae) populations..............................................................17
2.3 Descriptive statistics of the parameters differentiating between taxa.............18
2.2 Akaike Information Criterion (AIC) for models which predict taxon from discrete soil
chemistry parameters.........................................................
Tissue sampling sites for Carex scirpoidea subsp. convoluta and subsp. scirpoidea
(Cyperaceae)...................................................................47
3.2 Levels of genetic diversity for 11 populations of Carex scirpoidea subsp. convoluta and subsp. scirpoidea (Cyperaceae) based on 7384 genomic SNPs..................48
3.3. Locus-by4ocus AMOVA results by taxon as weighted average over 7384 genomic
loci of 11 populations of two taxa; 1) Carex scirpoidea subsp. convoluta and 2) subsp. scirpoidea (Cyperaceae).........................................................49
3.4. Locus-by-locus AMOVA results by region as weighted average over 7384 genomic
loci of 11 populations of three regions; 1) Northwestern, 2) Western, and 3) Eastern North American..................................................................50
3.5. Pairwise differentiation (Fst) measured across 7384 loci of Carex scirpoidea subsp.
convoluta and subsp. scirpoidea (Cyperaceae)....................................51
x


LIST OF FIGURES
FIGURE
1.1. Georeferenced occurrences for (a) Car ex scirpoidea subsp. convoluta (Cyperaceae)
and (b) subsp. scirpoidea........................................................5
2.1. Soil sampling sites representing populations of (a) Carex scirpoidea subsp. convoluta
and (b) subsp. scirpoidea (Cyperaceae); (c) soil core used for collection, and typical core sample in alvar locations.......................................................20
2.2. PC A biplot for 57 measured soil parameters at sites of Carex scirpoidea subsp.
convoluta and subsp. scirpoidea (Cyperaceae)...................................21
2.3. Clustered dendrogram of soil site similarity across Carex scirpoidea subsp.
convoluta and subsp. scirpoidea (Cyperaceae)....................................22
2.4. Plot of predicted probability of Carex scirpoidea subsp. convoluta and subsp.
scirpoidea (Cyperaceae) from a generalized logistic regression model, wherein a soil parameter is the predictor variable.............................................23
2.5. Coefficient of variation for (a) soil cumulative growing degree days and (b)
exchangeable Na measured in soil collected at 11 sites with populations of Carex scirpoidea subsp. convoluta and subsp. scirpoidea (Cyperaceae)..................24
3.1. Plant tissue sampling sites for (a) Carex scirpoidea subsp. convoluta and (b) subsp.
scirpoidea (Cyperaceae).........................................................52
3.2. Transect sampling plan for each Carex scirpoidea subsp. convoluta (Cyperaceae)
population site.................................................................53
3.3. Expected heterozygosity for 7384 genomic loci in 11 populations of Carex
scirpoidea subsp. convoluta (Cyperaceae) subsp. scirpoidea......................54
xi


3.4. Nucleotide diversity (0K), measured across 7384 genomic loci for 11 populations of
Car ex scirpoidea subsp. convoluta and subsp. scirpoidea (Cyperaceae).............55
3.5. Allele count for each locus across 7384 loci in 11 populations of Car ex scirpoidea
subsp. convoluta and subsp. scirpoidea (Cyperaceae)...............................56
3.6. fastStructure plots for strong structure and weak structure of all assayed populations
of Car ex scirpoidea subsp. convoluta and (b) subsp. scirpoidea (Cyperaceae)......57
3.7. BIC values for DAPC models with K clusters...................................58
3.8. DAPC cluster plots for (a) Car ex scirpoidea subsp. convoluta and (b) subsp.
scirpoidea (Cyperaceae)...........................................................59
3.9. Maximum likelihood rooted phylogeny of all populations of Carex scirpoidea subsp.
convoluta and subsp. scirpoidea (Cyperaceae)......................................60
3.10. Pairwise distance between loci using Nei’s distance, mean number of pairwise differences between populations, and within populations using 7384 genomic loci in 11 populations of Carex scirpoidea subsp. convoluta and subsp. scirpoidea (Cyperaceae). 61
3.11. Pairwise Fst values for 73 84 genomic loci in 11 populations of Carex scirpoidea
subsp. convoluta and subsp. scirpoidea (Cyperaceae)...............................62
3.12. Bayescan outlier loci (candidate loci under selection) for 11 populations of Carex
scirpoidea subsp. convoluta and subsp. scirpoidea (Cyperaceae) within Eastern North America...........................................................................63
Xll


LIST OF ABBREVIATIONS
1. AIC = Akaike Information Criterion — a score given to statistical models to rank their
predictive performance, relative to other models.
2. BIC = Bayesian Information Criterion — an alternative to AIC, with different parameter
penalization.
3. PC A = Principle Components Analysis — a dimension reduction analysis used to interpret
complex, multivariate relationships in data.
4. SNP = Single Nucleotide Polymorphism — a variant at one base at a locus in a genome.
xm


CHAPTERI
AN OVERVIEW OF CAREXSCIRPOIDEA SUBSP. CONVOLUTA AND SUBSP. SCIRPOIDEA (CYPERACEAE)
Taxonomy, Ecology, and Conservation Status of Carex scirpoidea subsp. convoluta and subsp. scirpoidea
Carex L. section Scirpinae Tuck. (Cyperaceae) comprises perennial sedges, several of which have been reported to be edaphic (soil) endemics adapted to specialized soil types. The section is largely dioecious and easily identified by its typically unispicate inflorescence and pubescent perigynia. Carex section Scirpinae includes Carex scirpoidea Michx. (northern single spike sedge), which comprises four interfertile subspecies: Carex scirpoidea subsp. scirpoidea,
C. scirpoidea subsp. convoluta (Kiik.) Dunlop, C. scirpoidea subsp. pseudoscirpoidea (Rydb.) Dunlop, Carex scirpoidea subsp. stenochlaena (Holm) A. Love and D. Love, and Carex curatorum Stacey. This thesis focuses on C. scirpoidea subsp. convoluta and subsp. scirpoidea (hereafter, subsp. convoluta and subsp. scirpoidea), which are described as differing based only on leaf morphology and ecology: subsp. convoluta has narrow, V-shaped leaves, with the widest leaves being narrower than its conspecific (i.e., less than 1.5 mm) (Dunlop and Crow, 1999). Phylogenetic relationships among taxa are largely unresolved (DePrenger-Levine, 2007; Pembrook, 2014).
Carex scirpoidea subsp. scirpoidea has a wide distribution, spanning alpine and boreal habitats from East Russia across Canada, the northern USA, and Greenland, to Norway (Fig. 1.1) (Dunlop and Crow, 1999; Carex Working Group, 2008; Flora of North America, 2003). In North America, it is found in the Northern Forests, Tundra, Taiga, and Hudson Plain ecoregions (Omernik, 1995). It tolerates a broad range of edaphic conditions, including dry to moist
1


substrates and calcareous, serpentine, and granodiorite soils. Populations occupy ridges, meadows, riparian stream areas, tundra, heathlands, and fellfields (Carex Working Group, 2008).
Carex scirpoidea subsp. convoluta has the most restricted range within the section. It is a rare habitat specialist and comprises approximately 50 variously sized, but often small populations in the Upper Great Lakes Region of Michigan, USA and Ontario, CAN (Fig 1.1) (Dunlop and Crow, 1999). The taxon typically occurs on soils that are highly calcareous with annually fluctuating wet and dry conditions, often reaching saturated to drought levels (Reschke et al., 1999; Shackleford, 2003). It dwells primarily on alvar, so-called limestone pavement, cobble beaches, and fens along or in close proximity to the shoreline of northern Lake Huron, including the North Channel and Georgian Bay (Fig. 1.1) (Shackleford, 2003).
Although subsp. scirpoidea is common throughout much of its range and globally ranked as secure (G5T5), it is possibly extirpated (SH) in New Brunswick, CAN, and imperiled (S2) or critically imperiled (SI) in certain states at the southern edge of its range in the USA (e.g., New York, Vermont, Wyoming, Washington), but globally ranked as secure (G5T5). In Michigan, C. scirpoidea has been designated a “threatened” species, although subsp. convoluta has been designated vulnerable by NatureServe in Ontario. Additionally, the USFS speculates that climate change could be a "significant threat to habitat of C. scirpoidea subsp. scirpoidea and subsp. convoluta in the contiguous United States" (Shackleford, 2003).
Conservation of Suitable Habitat
Worldwide, anthropogenic land-use directly causes degradation and fragmentation (e.g. wildemess-to-agriculture conversion, vegetation clearing), which may lead to drastic loss of suitable habitat (Fischer and Lindenmayer, 2007). Less common habitats (e.g., wetlands, alvar) are especially at risk, portending tenuous survival for rare endemic taxa, such as subsp.
2


convoluta (Reschke et al. 1999; Shackleford 2003). Alvar, the primary habitat for subsp. convoluta, is a rare community type found only in Europe (e.g., Sweden, Estonia) and in or adjacent to the Great Lakes region in North America (Schaefer and Larson, 1997; Brownell and Riley, 2000). Alvar is defined by its often treeless landscape that is almost exclusively composed of graminoids, forbs, and shrubs. However, the most definitive characteristic of alvar communities is its distinct edaphic conditions. Made up of marble or limestone pavement covered by a characteristically thin soil layer atop limestone or dolostone bedrock, alvar provides harsh conditions unsuitable for many plant species, while supporting many endemics and disjunct taxa (Reschke et al., 1999, Catling and Brownell, 1995). Soils are characterized by high alkalinity, silica levels, organic content, nitrogen, calcium, and magnesium levels, and low phosphorous content (Catling and Brownell, 1995; Brownell and Riley, 2000; Stark et al., 2004). Alvar is also subject to extreme hydrological conditions, often with seasonal fluctuations ranging from drought to flood conditions (Reschke et al. 1999) that alvar endemics can tolerate. The flora of Great Lakes alvar includes many disjunct relictual populations originating from boreal (Carex richardsonii R. Brown, Poa alpina L.), southern (Thaspium barbinode [Michx.] Nutt, Valerianella umbdicata Sulk Alph. Wood.), and prairie communities (Hymenoxys herbaceae Greene, Iris lacustris Nutt.) that established following Wisconsinan glaciation (Catling and Brownell, 1995; Reschke et al., 1999).
Alvar is threatened by anthropogenic activities (e.g., off-road vehicle activities, nonnative species), and broadly listed as “imperiled” at a global (G2) and statewide scale (S2 Michigan; Shackleford, 2003). Additionally, “open alvar” has been assigned as globally imperiled (Catling and Brownell, 1995). Other threats to alvar communities and soils include quarrying, logging, and rutting from vehicular travel (Reschke et al., 1999), direct pollution of
3


adjacent waters, hydrological changes, and overgrazing (Comer et al., 1997; Reschke et al., 1999). Like alvar, fens are also threatened by anthropogenic hydrological changes, as well as automotive and human trampling (Shackleford, 2003). Fens constitute one of several habitat types for subsp. scirpoidea, and a minor proportion of the total available habitat for Carex scirpoidea subsp. convoluta (Shackleford, 2003). Fens are low-oxygen endogenously-fed peatlands, in which Carex is known to comprise a significant portion of detrital peat (Gorham, 1991). Edaphic substrates, such as those characteristic of alvar and fens, place high selective pressures on plants; as such, the study of edaphic endemics can also provide valuable insight into the role of soil and other substrates on population differentiation (Rajakaruna, 2004).
Justification for Research
Adaptive ecosystem management stipulates that ecological research and scientific evidence be intertwined with management and conservation efforts (Lee, 1999). As anthropogenic climatic and habitat changes threaten plant populations, population decline may reduce genetic diversity and adaptive variation, in particular. Thus, conservation managers should prioritize assessments of genetic diversity, in order to elucidate threats to rare organisms, and estimate the likelihood and efficacy of adaptive change in small endemic populations under changing environmental conditions (Birchenko et al., 2009; Allendorf et al., 2010; Worch et al., 2011; Frankham et al., 2014; Modesto et al., 2014). Integrating genetic analyses with ecological assessment provides necessary information concerning rare endemic plants, including the range of acceptable conditions and extirpation risks.
4


#1 • • •
•J v?* .• I * , • V
* • • • n a* •
^ f f ^ Greenland •
^ 2 ^ # J» NT# # • • •
• YT A# _ f •
•?««»wr# . •••:• »,
• »• . “ Canady «• Hudson e * * ‘ * *• * \
. .'V- S 'if \
Norway
; Poland
Germany
Austria
F,,nCC Ron
> < Si
Italy
Greec
Figure 1.1: Georeferenced occurrences for: (a) Carex L. scirpoidea Michx. subsp. convoluta (Kiik) Dunlop and (b) C. scirpoidea subsp. scirpoidea (Cyperaceae). Locations were mapped in Fulcrum using data from SEINet, iDigBio, and expert knowledge (Leo P. Bruederle pers. comm., Anton Reznicek pers. comm., Brad Slaughter pers. comm.).
5


CHAPTER II
EDAPHIC NICHE OF CAREXSCIRPOIDEA SUBSP. CONVOLUTA AND SUBSP. SCIRPOIDEA Abstract
Edaphic endemism, wherein organisms are restricted to distinct soil types, has contributed to the evolution of plant diversity. Alvar is a habitat type with a unique edaphic substrate—often consisting of limestone or dolostone geological parent material, with calcareous, alkaline chemistry— and supports a number of endemic plant species, including Carex scirpoidea subsp. convoluta. However, it is unclear the extent to which soil chemistry differs between subsp. convoluta and C. scirpoidea subsp. scirpoidea. In this study, I characterized the edaphic niches of subsp. convoluta and subsp. scirpoidea by measuring 66 physical and chemical soil parameters across 14 sample sites. My analyses did not indicate that either taxon could be characterized by a distinct soil type, although subsp. convoluta site soils had consistently lower exchangeable sodium concentrations, and longer growing seasons with hotter soil temperatures. Additionally, the subsp. convoluta population located in an opening in a cedar swamp (Horseshoe Bay, MI, USA) had the most distinct conditions of all sites surveyed. These findings reveal a wider than expected edaphic niche for subsp. convoluta in comparison to subsp. scirpoidea. Future studies should determine the extent of local adaptation in subsp. convoluta.
Introduction
Edaphic endemism, wherein organisms are adapted and isolated to soils with unique physical and chemical properties, has resulted in high levels of plant diversity (Kruckeberg,
1951; Rajakaruna, 2018). Two distinct scenarios are thought to have contributed to the origin of
6


edaphic endemics. A widespread taxon may colonize a variety of habitats, eventually specializing on a unique soil type. Alternatively, edaphic endemic populations may be the only remaining populations of a once-generalist taxon (Kruckeberg, 1951; Anacker, 2014). Much edaphic endemism theory derives from research on serpentine soil endemics (Burge et al., 2017; Rajakaruna, 2018); soil types boasting high endemism (e.g., calcareous, metal mine tailings) have received less attention (Antonovics, 2006; Bagella and Urbani, 2006; Burge et al., 2017; Rajakaruna, 2018).
Dunlop (1990), as part of her monograph on Section Scirpinae, suggested that Car ex scirpoidea subsp. convoluta (hereafter, subsp. convoluta) is an edaphic ecotype of open limestone pavement. The taxon is strongly associated with Great Lakes alvar — a habitat type with limited distribution and moderate geographic disjunction, often comprised of patchy mosaics (Catling, 1995). Carex scirpoidea subsp. convoluta affiliates with several alvar community types, specifically, little bluestem alvar, creeping juniper-shrubby cinquefoil alvar, and white cedar-jack pine-shrubby cinquefoil alvar (Reschke et al. 1999). Each alvar community type varies with respect to species composition, exposed bedrock, and vegetative cover; however, these communities are typically characterized by limestone or dolostone bedrock, thin soil, and low to moderate canopy cover (Reschke et al., 1999). The soil chemistry of alvar is characterized as having high alkalinity, silica levels, organic content, nitrogen (N), calcium (Ca), and magnesium (Mg) levels, but low phosphorous (P) content (Stark et al., 2004, Brownell and Riley, 2000). Comparing alvar with adjacent forest communities, Schaefer (1997) found that physical properties (e.g., soil depth, exposed rock, sub-surface temperature) differed and that potassium (K) levels were lower, with no apparent difference in Mg, P, Ca, ammonia, or moisture content.
7


In contrast, Car ex scirpoidea subsp. scirpoidea (hereafter, subsp. scirpoidea) is broadly distributed and occupies a range of soil types, including wet, dry, calcareous, serpentine, and acidic soils. Common habitats include meadows, rock cracks, and fens at high latitudes (see Chapter I). Though habitat and soil types are highly variable, all have alpine-arctic elements (Given and Soper, 1981).
The affinity of subsp. convoluta to alvar, coupled with limited geographic range, suggests that this taxon occupies a narrow ecological niche. However, the physical and chemical properties of these soils have not been previously examined across the range of this taxon. Here,
I examine and compare the edaphic niche for subsp. convoluta, a narrow endemic, relative to a widespread conspecific taxon, subsp. scirpoidea. In this study, I hypothesize that: 1) due to the narrow range and strong association with alvar, subsp. convoluta has a narrower edaphic niche than its conspecific, and 2) subsp. convoluta has a distinct edaphic niche for several soil parameters, including Ca, Mg, P, K, and pH — those associated with limestone open pavement soils (Stark et al., 2004).
Methods
Sampling
Fourteen field sites were chosen representing much of the range of both taxa (Fig. 2.1, Table 2.1). Soil chemistry and soil temperature data were collected for nine sites supporting subsp. scirpoidea populations by Westergaard et al. (in press), while I collected the same data for five populations of subsp. convoluta (Appendix A). At each subsp. convoluta site, three 21-m transects were established 6m apart, with points marked every 3m. Soil samples and temperature measurements were obtained at one randomly-chosen point per transect (three per site), adjacent to the nearest subsp. convoluta individual. At each point, a soil temperature logger (Thermo
8


Button 22L 1199Y85) was placed -10-15 cm deep, and programmed to record four temperature measurements throughout the day (once per six hours). It was often necessary to combine multiple soil cores due to the thin layer of soil at most sites. A soil sample of approximately lOOg was obtained at each sample point by. Leaf litter and other debris, plants, roots, and stones were subsequently removed. Soil samples were kept on ice in the field and later frozen in the lab.
Soil samples were air dried (-24 hours) in the laboratory, with soil weights recorded both before and after drying (Appendix B). Dried soil samples were sent to the Norwegian Institute of Bioeconomy Research (https://www.nibio.no/en) for analysis. Estimates were obtained for 65 chemical and physical parameters; additional ammonium acetate tests were performed to measure exchangeable cation content (Castilho and Rix, 1993). Soil sampling occurred in 2014 (subsp. scirpoidea) and 2016 (subsp. convoluta), with subsequent removal of soil temperature loggers in 2016 and 2017, respectively. Soil sampling for subsp. scirpoidea was conducted by Kristine Westergaard and Leo P. Bruederle, with loggers deposited and soil cores taken at the edges and middle of each population.
Soil temperature analysis
Soil temperature magnitude and duration was characterized with cumulative growing degree days (CGDD). However, loggers failed to take readings at later dates of the 2017 growing season at subsp. convoluta sites. To extrapolate missing subsp. convoluta data, Intercept, Linear, and Quadratic generalized linear models (GLMs) were created using per-day averages of the four daily temperature measurements (Tavg). At each site, the best-fitting model was chosen based upon the optimal Akaike Information Criterion (AIC). The model was used in lieu of raw data points for CGDD estimation at subsp. convoluta sites (Appendix D). However, Tavg was calculated from raw data for the complete 2015 growing season for subsp. scirpoidea sites. The
9


base temperature (Tbase) was 0°C, the observed minimum temperature required for growth and development in temperate and arctic zones, specifically shortgrass steppe and alpine (Moore et al., 2015; White et al., 2015). However, a caveat was applied in order to limit errant measurements near 0°C due to Thermo Button Temperature Data Logger accuracy limits (±0.5°C). Measurements were included between the first point above 0°C — such that subsequent values showed an upward slope — and the last point above 0°C; although remaining values that increased up to 0.5 °C were discarded. The following equation for CGDD (Huflft et al., 2018) was implemented:
CGDD = Yik=1 Tavg ~ Tbase â– 
Soil chemical and physical parameters
Of the soil chemical and physical parameters (hereafter, parameters) measured, those with no variance or in which >50% were a single value were filtered out (e.g., selenium). Percent solid matter was also filtered out. A principle components analysis (PCA) on all remaining variables was conducted after scaling and centering, using prcomp in R (R Core Team, 2014). A dendrogram using the PCA results was also created with hclust in R using the “complete method,” which determines the maximum Euclidian distance between points in different clusters.
To determine which parameters varied most widely between subsp. convoluta and subsp. scirpoidea, several parameters were incorporated as the ‘predictor variables’ in a GLM binomial model to describe the ‘dependent’ taxon variable. Schaefer and Larson (1997), Stark et al. (2004) and Burge et al. (2017) — studies on alvar soil and edaphic endemism — were consulted to select relevant soil parameters to be tested in the model. All macronutrients and micronutrients were also included in the model. CGDD, loss of ignition, and carbon were used in lieu of subsurface temperature (Schaefer and Larson, 1997), soil organic matter, and inorganic carbon
10


(Stark et al., 2004), respectively. Where applicable, exchangeable cation measurements were used, as well. For each taxon, descriptive statistics were calculated for all predictor variables. To estimate niche breadth differences between taxa, differences in variation for each parameter with the optimal AIC scores were tested using the asymptotic test for the equality of the coefficient of variation (ATECV; Feltz and Miller, 1996).
Results
Estimates were retained for 65 soil parameters, of which seven (i.e., selenium (Se), exchangeable beryllium, exchangeable cobalt, exchangeable Se, solid matter, molybdenum) were filtered out (Appendix C), and CGDD was added. CGDD was calculated from a best-fitting soil temperature model for subsp. convoluta., for all subsp. convoluta sites, the quadratic GLM was the best-fitting model (Appendix B). The PCA and dendrogram (based on 59 parameters) did not reveal strict taxonomic differences based on soil type (Fig 2.2, 2.3). Instead, sites clustered geographically (e.g., Norway, Greenland, and western North America; Fig 2.2). Notably, the Upper Midwestern North American site harboring subsp. scirpoidea (i.e., Pembina) and subsp. convoluta sites exhibited similar soil chemistry, with the exception of Horseshoe Bay, MI. Horseshoe Bay, the site of a subsp. convoluta population, occupies an opening in a cedar swamp, which is an anomaly, both with respect to the community and soil chemistry, which is distinct from all other sites studied (Appendix C).
In general, estimates for soil parameters were highly variable in one or both taxa, making niche characterization difficult. For instance, in some cases, standard deviation for a parameter exceeded the average value (p) (e.g., Ca, Cu; Table 2.2). For nearly all parameters, the values for one taxon overlapped with those of the conspecific, although estimates for subsp. scirpoidea typically exhibited larger ranges (Table 2.2). Though unexpected, the average concentration and
11


range of total Ca (p = 129664.27 ± 127122 mg/kg in subsp. convoluta vs. 36234.59 ± 50148.31 mg/kg in subsp. scirpoidea, range = 36113 mg/kg vs. 211230 mg/kg) and Mg (24898.13 ± 20581.2 vs. 7111.04 ± 5141 mg/kg, range = 65598 mg/kg vs. 15913 mg/kg) was larger for subsp. convoluta, as a result of high Ca and low Mg content at the Horseshoe Bay, MI site (Table 2.2, Appendix C). However, the ranges for exchangeable Ca and Mg were higher in subsp. scirpoidea populations. Additionally, soils at subsp. convoluta sites were more alkaline (pH = 7.49 ± 0.16 vs. 6.73 ± 0.63, Table 2.2), had lower N (0.7 ± 0.54% vs. 1.11 ± 1.08%), and total P (579.80 ± 345.73 mg/kg vs. 843.19 ± 573.75 mg/kg). Sites for each taxon exhibited approximately similar concentrations of total K (2724.40 ± 2283.70 mg/kg vs. 2940.48 ± 1924.90 mg/kg).
Exchangeable sodium (Na) best predicted taxonomic outcome (subsp. convoluta vs. subsp. scirpoidea), providing a relative model weight of 100% (Table 2.3), with consistently higher levels at subsp. scirpoidea sites, even at the otherwise parametrically-similar Minnesota subsp. scirpoidea sites (Fig. 2.4, Appendix C). CGDD had a relative model weight of 0.8% (Table 2.3). Other predictors did not have a measurable effect on the model (Akaike weight <
0.1%). Estimates of exchangeable Na were especially high at the High Creek Fen, CO site, where the average was over three times greater than the next closest average (Eagle Creek, AK). As indicated by the ATECV results (p < 0.05, Fig. 2.5), subsp. scirpoidea also had a broader range of exchangeable Na and CGDD values.
Discussion
Soil parameter suite
The goal of this aspect of my thesis research was to characterize and compare edaphic niche between subsp. scirpoidea and subsp. convoluta, which have been presumed to have broad and narrow edaphic niches, respectively. Variation for the suite of soil parameters studied,
12


indicated that edaphic niche did not differ substantially between taxa (Fig 2.2). Instead, edaphic variation corresponded with geography; more variability occurred among subsp. convoluta sites than between sites of subsp. convoluta and subsp. scirpoidea sites in close proximity (e.g., Pembina, MN; Figs. 2.2, 2.3).
While alvar is a distinctive habitat, the soil parameter suite for the alvar endemic (Figs. 2.2, 2.3) was not unequivocally distinct. However, Schaefer and Larson (1997) found that physical, not chemical characterisics (e.g., soil depth, photosynthetically active radiation, woody debris cover) comprised the majority of parameters that differed between alvar and adjacent forest habitats. However, chemical parameters comprised the majority of soil measurements in my study (Table 2.2, Appendix C). It is possible that parameters not measured here, such as microclimate, radiation, and plant species composition (e.g., % lichen cover) better describe niche divergence between taxa. Finally, drought periodicity niche may differ between taxa; Dunlop (1990) speculated that drought tolerant morphology may in fact be a distinguishing feature of subsp. convoluta.
My findings show that both taxa surveyed in the section have broad niche with respect to nearly all soil parameters (Range > p in 20/24 parameters in subsp. convoluta; 22/24 parameters in subsp. scirpoidea). Strikingly, subsp. convoluta sites exhibited a higher total Ca and total Mg niche breadth (Table 2.2). The higher average pH, higher concentrations of total Ca and Mg, and lower concentrations of total P in subsp. convoluta align with the chemical characterization of alvar by Stark et al. (2004). However, many exceptions were noted in one or more subsp. scirpoidea population sites, indicating that the niche of subsp. convoluta did not unequivocally differ from subsp. scirpoidea with respect to pH, or total Ca, Mg, K, or P.
13


Horseshoe Bay, MI accounted for much of the edaphic niche breadth exhibited by subsp. convoluta (Figs. 2.2, 2.3). This sampling site, the only one in a forested cedar swamp, had a marl-like soil visibly distinct from the rest. Soils at this site also had the highest water content, and lowest CGDD for any subsp. convoluta site, possibly due to microclimate effects in this forested habitat (Appendix A, B). Additionally, many macro- and micronutrients (e.g., total K, P, Ni, Fe, and Zn) were at lower concentrations than other subsp. convoluta sites, though total Ca was much higher (Appendix C). However, like subsp. convoluta alvar habitats, Horseshoe Bay, MI is underlain by dolomite, albeit with greater soil depth (Ludwig et al., 1996; Schaefer and Larson, 1997). Evidently, subsp. convoluta is not confined to a narrow range of soil types, and can withstand saturated, nutrient-poor conditions.
Association with variant, adverse soil types may indicate poor interspecific competitive ability (Boisson et al., 2017; Kruckeberg and Rabinowitz, 2017; Rajakaruna, 2018). As such, colonization of alvar by competing plants (e.g., shrubs) is known to be prevented by grazing and long periods of drought and flood (Schaefer and Larson, 1997; Stark et al., 2004). Dunlop (1990) reported other edaphic ecotypes within C. scirpoidea, suggesting that the taxon is highly adaptive edaphically, which has likely contributed to diversification across its range. My results support this, although I suggest that broad niche is also exhibited at a finer taxonomic level—within the most restricted ecotype, subsp. convoluta. Thus, colonization of variably adverse soils has aided persistence of C. scirpoidea ecotypes.
Assessment of taxonomic predictor parameters
Soil exchangeable Na niche and niche breadth differed consistently between taxa (p = 6.99 for subsp. convoluta, u = 45.10 for subsp. scirpoidea, Table 2.2; ATECV, p < 0.05, Fig.
2.5; Appendix C). Carex scirpoidea subsp. convoluta occurs in sites with low exchangeable Na
14


— with little variation among sites (range = 719.03 mg/kg, Table 2.2) — while soils for subsp. scirpoidea exhibit highly variable exchangeable Na content among sites (range = 2516 mg/kg). This finding supports documented occurrences of subsp. scirpoidea on saline margins of prairies and fens in Minnesota and Colorado, respectively (Shackleford, 2003). While subsp. scirpoidea has a larger exchangeable Na niche breadth, it is unclear whether it has a broader tolerance than its conspecific, or if it is simply an artifact of its distribution.
Physiologically, it is unclear whether Na is a limiting factor for plant growth, like P or Ca (Fageria et al., 2011); however, it is known to contribute to plant metabolism, particularly as a substitute for K (Subbarao et al., 2003). However, at high levels, salinity is known to be an agent of selection. Na may stress plants by blocking water and nutrient uptake in the roots, leading to deficiency (Fageria et al., 2011). Persistence in saline soils is uncommon for non-halophytic plants, and supports the hypothesis that subsp. scirpoidea exhibits high plasticity with respect to edaphic conditions as an evolutionary strategy.
Due to its distribution at high latitudes, shorter, colder growing seasons were expected in subsp. scirpoidea compared to its conspecific. My CGDD results supported this, with the exception of the Pembina, MN site (Figs. 2.4, 2.5; Appendix D). The seasonal dynamics of subsp. scirpoidea are quite distinct, adopting a bi/trimodal parabola-like shape with several subsp. scirpoidea sites oscillating above and below 0°C (Appendix D). These seasonal patterns differ somewhat from the standard parabolic shape as seen in the Upper Midwestern United States populations of both taxa. Additionally, higher soil temperature variation was noted in subsp. scirpoidea than subsp. convoluta. Soil CGDD has been found to moderately influence sedge biomass (Brooker and Van Der Wal, 2003). Further, low soil temperatures are known to
15


limit nutrient uptake by roots, decrease nitrogen availability, and slow or prevent plant growth (Brooker and Van Der Wal, 2003; Moore et al., 2015; Krab et al., 2018).
Future Directions
The findings suggest that neither taxon is restricted to a single soil type based on my analyses of the parameters assessed here. While Dunlop & Crow (1999) described subsp. convoluta as an edaphic ecotype, the chemical composition of the soil type remains ambiguous. Furthermore, the uniquely divergent soil types of the Horseshoe Bay, MI site for subsp. convoluta and the High Creek Fen, CO site for subsp. scirpoidea may suggest novel ecotypes. While genetic differentiation was detected with other populations, it is unclear whether local genomic adaptation has influenced specialization to unique soil conditions (Pembrook, 2014; Westergaard et al., in press; Chapter 3, this thesis). Future genomic studies will determine whether local genomic adaptation to low-nutrient, calcareous soil, or sodic soil has occurred; and whether edaphic “speciation genes” may be detected (Rajakaruna, 2018).
Although beyond the scope of this research, greenhouse studies and reciprocal transplant experiments should be employed to test whether populations of subsp. convoluta and subsp. scirpoidea grow equally well in soils with high salinity, low CGDD, and low nutrient soils, to indicate any differences in edaphic tolerance between taxa. Reciprocal transplant experiments — alvar-associated subsp. convoluta individuals at Horseshoe Bay, MI and vice versa — may indicate niche differences due to local adaptation, or broad tolerance due to plasticity. Such studies may also be useful to understand how adaptation may progress as climate change and habitat degradation occurs.
16


Table 2.1: Soil sampling sites for Carex scirpoidea Michx. subsp. convoluta (Kiik.) Dunlop and subsp. scirpoidea (Cyperaceae) populations.
Taxon Sample Site Country Latitude Longitude
Thompson’s Harbor, MI USA 45.35 -83.58
d C/3 3 C/3 Horseshoe Bay, MI USA 45.99 -84.745
2* Maxton Plains, MI USA 46.08 -83.66
o •K: ?3 y d Manitoulin Island, ON CAN 45.60 -82.12
X O b S Cabot Head, ON CAN 45.24 -81.30
Eagle Summit, AK USA 65.48 -145.41
12 Mile Summit, AK USA 65.40 -145.96
High Creek Fen, CO USA 39.10 -105.97
Pembina, MN USA 48.08 -96.45
o Mestersvig GRL 72.23 -23.98
d C/3 3 C/3 Traill GRL 72.52 -23.98
Kjelvatn NOR 68.18 17.21
o Kjerringa NOR 66.78 14.21
X a Solvagtind NOR 66.83 15.42
17


Table 2.2: Akaike Information Criterion (AIC) for models that predict taxon from discrete soil chemistry parameters. The taxa being predicted are Carex scirpoidea Michx. subsp. convoluta (Kiik.) Dunlop and subsp. scirpoidea (Cyperaceae). The Intercept model is a null model. Soil parameters are measured in mg/kg, except where stated otherwise. Asterisks (*) imply concentrations of exchangeable cations. Akaike weights indicate an estimate for probabilities of each model. Dashes (-) indicate probabilities under 0.1%.
Model AIC Akaike Weight
Na* 13.59124 100%
CGDD 23.07332 0.8%
Intercept 56.74755 -
pH 40.45050 -
Ca 48.28243 -
Ca* 58.50410 -
P 58.63474 -
p* 57.76991 -
Mg 42.41074 -
Mg* 58.28331 -
K 58.63474 -
K* 57.76991 -
N % 56.80382 -
C% 58.32251 -
Loss on Ignition 57.70272 -
Zn 57.50691 -
Zn* 57.00345 -
S 58.74152 -
S* 58.63804 -
Na 56.59971 -
Fe 50.85687 -
Fe* 51.46165 -
Mn 57.54802 -
Mn* 57.16004 -
Cu 57.65553 -
Cu* 56.92506 -
Cation Exchange Capacity 58.01994 -
18


Table 2.3: Descriptive statistics of the parameters differentiating between Carex scirpoidea Michx. Subsp. convoluta (Kiik.) Dunlop and subsp. scirpoidea (Cyperaceae). All chemicals were measured in mg/kg, except where noted with %, and cation exchange capacity, measured in in milliequivalents per lOOg. Asterisks (*) imply concentrations of exchangeable cations.
Parameter Carex scirpoidea subsp. convoluta Carex scirpoidea subsp. scirpoidea
M SD Range M SD Range
Na* 6.99 1.30 719.03 45.10 59.97 2909.13
CGDD 3168.80 273.44 4.8 972.76 977.44 293
pH 7.49 0.16 0.64 6.73 0.67 2.13
Ca 129664.27 127122 361113 36234.59 50148.31 211230
Ca* 8923.40 4613.81 13156 10089.48 8732.92 27191
P 579.80 345.37 1658 843.19 573.75 1726
p* 9.91 6.94 25.93 8.35 8.20 26.02
Mg 24898.13 20581.21 65598 7111.04 5141.00 15913
Mg* 967.94 845.52 2881.54 735.19 1167.55 4797.65
K 2724.40 2283.70 7029 2940.48 1924.09 7791
K* 107.63 87.81 305.9 141.08 118.52 364.3
N % 0.70 0.54 1.952 1.11 1.08 2.78
C% 13.58 5.54 20.616 16.48 17.25 44.61
Loss on Ignition % 22.99 10.92 38.46 31.11 30.79 80.25
Zn 53.22 35.74 125.8 66.11 38.24 186.7
Zn* 1269.80 900.47 2933 2202.37 2880.37 10978
S 1994.35 2104.27 5627.6 2048.64 2293.01 9416.5
S* 56.67 83.44 236.6 67.14 110.94 512.9
Na 229.67 101.34 308 462.41 712.64 2516
Fe 11216.47 9219.87 26891 20977.78 11595.428 34595
Fe* 2.87 0.83 2.51 6.94 7.47 27.52
Mn 533.13 679.86 1988 831.41 1022.64 4719
Mn* 7.13 2.82 8.1 15.41 29.14 120.82
Cu 20.95 26.61 79.3 30.50 31.30 104.9
Cu* 277 0 0 296 71.19 327
Cation Exchange 528.23 238.55 709.5 654.67 561.91 1766.2
Capacity
19


a
b

•L2MS, ES

NT
Northw—tern nu -v Passages
Canada
Greenland
Iceland
K,
K2,«
S
United
Kingdom
Denmark
Germany
Polan
Figure 2.1: Soil sampling sites representing populations of (a) Carex scirpoidea Michx. subsp. convolutci (Kiik.) Dunlop and (b) subsp. scirpoidea (Cyperaceae); (c) soil core used for collection, and typical core sample in alvar locations. Plant site abbreviations are as follows: TH = Thompson’s Harbor, MI, USA; HB = Horseshoe Bay, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CAN; CH = Cabot Head, ON, CAN; 12MS = 12 Mile Summit, AK, USA; ES= Eagle Summit,, AK, USA; HCF = High Creek Fen, CO, USA; P = Pembina, MN, USA; M = Mestersvig, GRL; T = Traill, GRL; K = Kjelvatn, NOR; K2 = Kjerringa, NOR; S = Solvagtind, NOR.
20


a
b
A
Alaska Colorado Great Lakes Greenland Norway
Figure 2.2. PCA biplot for 59 measured soil parameters at sites of Carex scirpoidea Michx. subsp. convoluta (Kiik.) Dunlop and subsp. scirpoidea (Cyperaceae). In a), taxon is circumscribed, in b) population is circumscribed, and in c) region is circumscribed. Plant site abbreviations are as follows: TH = Thompson’s Harbor, MI, USA; HB = Horseshoe Bay, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CAN; CH = Cabot Head, ON, CAN; 12MS = 12 Mile Summit, AK, USA; ES= Eagle Summit,, AK, USA; HCF = High Creek Fen, CO, USA; P = Pembina, MN, USA; M = Mestersvig, GRL; T = Traill, GRL; K = Kjelvatn, NOR; K2 = Kjerringa, NOR; S = Solvagtind, NOR.
21


Figure 2.3: Clustered dendrogram of soil site similarity across Carex scirpoidea Michx. subsp. convolutci (Kiik.) Dunlop (in blue, abbreviated as ‘c’) and subsp. scirpoidea (Cyperaceae) (in red, abbreviated as ‘s’) (from PC A). Plant site abbreviations are as follows: TH = Thompson’s Harbor, MI, USA; HB = Horseshoe Bay, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CAN; CH = Cabot Head, ON, CAN; 12MS = 12 Mile Summit, AK, USA; ES= Eagle Summit,, AK, USA; HCF = High Creek Fen, CO, USA; P = Pembina, MN, USA; M = Mestersvig, GRL; T = Traill, GRL; K = Kjelvatn, NOR; K2 = Kjerringa, NOR; S = Solvagtind, NOR.
22


CGDD
Figure 2.4: Plot of predicted probability of Carex scirpoidea Michx. subsp. convoluta (Kiik.) Dunlop (abbreviated as ‘convoluta’), (set as 0) and subsp. scirpoidea (‘scirpoidea’) Cyperaceae, (set as 1) from generalized logistic regression model, wherein a soil parameter is the predictor variable. In a), exchangeable Na (Na*) is the predictor variable, in b), soil cumulative growing degree days (CGDD) is the variable.
23


taxon
Figure 2.5: Coefficient of variation for (a) soil cumulative growing degree days (CGDD) and (b) exchangeable Na (Na*) measured in soil collected at 11 sites with populations of Carex scirpoidea Michx. subsp. convolutci (Kiik.) Dunlop (abbreviated as ‘convoluta’), and C. scirpoidea subsp. scirpoidea (‘scirpoidea’). Statistical significance of test for equality of variation between populations (p) included.
24


CHAPTER III
POPULATION GENOMICS OF THE NARROW EDAPHIC ENDEMIC CAREXSCIRPOIDEA SUBSP. CONVOLUTA Abstract
Rare, narrow endemics are expected to harbor low genomic diversity due to small number of populations and limited range. In this study, I compared genomic diversity between Carex scirpoidea subsp. convoluta and subsp. scirpoidea, and analyzed genomic differentiation and structuring between taxa and populations of these taxa. I used double digest restriction site-associated DNA (ddRAD) sequencing to identify genome-wide SNPs from individuals representing 11 populations across the North American range of both taxa. Unexpectedly, subsp. convoluta exhibited relatively high genomic diversity compared to its conspecific. Further, there was little evidence for evolutionary differentiation between taxa, although regional differentiation was observed among populations of subsp. scirpoidea. Phylogenomic results and Bayesian/multivariate cluster analyses suggest that subsp. convoluta arose from periglacial refugial populations of subsp. scirpoidea in Eastern North America as glaciation receded and C. scirpoidea recolonized northern North America. Multiple statistical analyses revealed that subsp. scirpoidea comprises three distinct genomic clusters, corresponding to three probable North American glacial refugia. It is likely that a nearly obligate dioecious mode of reproduction has contributed to the maintenance of genomic diversity in subsp. convoluta, preventing inbreeding depression, and contributing to the evolutionary potential of populations of this taxon. I recommend that populations of subsp. convoluta be monitored, particularly along the margins of its range, due to reduced gene flow between distant populations.
25


Introduction
The high diversity of flowering plants worldwide can be attributed to population divergence over evolutionary time, to which population dynamics — including size, age structure, and the ecological factors that influence them — contribute. Limited, fragmented habitat may lead to bottlenecks and reduced effective size of a population, which in turn may reduce heterozygosity (Kruckeberg and Rabinowitz, 1985; Gomez-Femandez et al., 2016). Loss of heterozygosity may be exacerbated in habitat-limited populations by increases in inbreeding rates and genetic drift, which may lead to differentiation from the metapopulation (Kruckeberg and Rabinowitz, 1985; Ellstrand and Elam, 1993; Hamrick and Godt, 1996). Similarly, selection in response to variable ecological conditions across a species’ range, such as soil chemistry, may also lead to population differentiation and, ultimately speciation (De La Torre et al., 2014; Zhou et al., 2014; Rajakaruna, 2018). However, adaptation to novel conditions must precede such differentiation (Richter-Boix et al., 2011; Zhou et al., 2014; Sandoval-Castillo et al., 2018).
Adaptation is enabled by phenotypic plasticity, standing genetic variation, or the introduction of novel alleles through genetic mutation or gene flow (Olson et al., 2013). Often, adaptive alleles are selected for in response to ecological conditions, such as annual temperature and soil chemistry (Coop et al., 2010; Roda et al., 2013). As such, adaptive alleles may influence a taxon’s geographic distribution, sometimes resulting in endemism (Richter-Boix et al., 2011; Worch et al., 2011). Endemics may form when a species’ ability to adapt to a novel niche coincides with reduced competition from other less well-adapted species (Kruckeberg, 1951; Stebbins and Major, 1965). Edaphic endemism specifically arises when a species can tolerate a broad range of soil types or has the means to adapt to unusual soil types. To explain the origin of some edaphic endemics, Rajakaruna (2018) hypothesized that habitats with novel soil types
26


formed glacial refugia during the Last Glacial Maximum (LGM). As such, the spatially reconfigured populations altered evolutionary processes, such as gene flow, and adaptation to new habitat types followed (Catling and Brownell, 1995; Hewitt, 2000; Rajakaruna, 2018). Dillenberger and Kadereit (2013) reported a niche shift in subspecies of Adenostyles alpina (L.) Bluff. & Fingerh from calcareous to non-calcareous soils following dispersal to glacial refugia.
Differentiation can occur between populations, regardless of levels of genetic diversity. Genetic drift, selection, and restricted gene flow may increase rapid fixation of alleles and reduce diversity, resulting in greater genetic structure between populations (Frankham, 2005). Genetic diversity provides the novel variation on which selection acts, particularly as an organism colonizes new environments and or faces forthcoming environmental conditions, thereby contributing to its “evolutionary potential” (Rajakaruna, 2018; Frankham, 2005). Widely distributed species comprising large populations tend to maintain the highest levels of genetic diversity, while narrowly distributed species with small populations typically maintain less (Ellstrand and Elam, 1993; Hamrick and Godt, 1996; Frankham, 2005). In small populations, alleles that confer adaptation to shifting environmental conditions and resilience to unanticipated events may be lost to genetic drift, while they are expected to persist in larger populations. Furthermore, loss of genetic diversity in a population can hamper evolution, resulting in ‘genostasis’ (Bradshaw, 1991; Rajakaruna, 2018) and extinction vortices, wherein anthropogenic threats combine with stochastic demographic and genetic events resulting in extinction (Gilpin and Soule, 1986).
Reliable methods of measuring genetic diversity aid in forecasting viability and extirpation risk in small populations, especially amidst emerging ecological threats, like habitat loss (Allendorf et al., 2010; Frankham, 2010). Monitoring genetic diversity enables proper
27


estimation of evolutionary potential, which may ensure plant survival under stochastic conditions, determine whether rare alleles persist in small populations, or indicate if a population is threatened by inbreeding depression (Allendorf et al., 2010). However, genomic techniques may also be used to identify how populations and taxa have adapted and diversified in response to historical ecological conditions (Roda et al., 2013; Sandoval-Castillo et al., 2018; Westergaard et al., in press), providing useful information to predict future scenarios. Thus, studying species’ population genetic diversity and structure is paramount for determining conservation status and implementing management strategies (Frankham, 2010).
Car ex scirpoidea subsp. convoluta (hereafter, subsp. convoluta) is endemic to alvar and related habitats (e.g., cobble beach) along the Niagara Escarpment in close proximity to northern Lake Huron (see Chapter 2) (Dunlop & Crow 1999). Although species associated with Great Lakes alvar are thought to be descendants of ancestral parkland, tundra, and prairie species prior to glaciation (Hamilton & Eckert 2007; Catling & Brownell 1995), little is known about the evolutionary history of subsp. convoluta. It has been hypothesized that subsp. convoluta arose from C. scirpoidea subsp. scirpoidea (hereafter, subsp. scirpoidea) as an edaphic alvar-associated ecotype during the Pleistocene (Dunlop, 1990; Dunlop and Crow, 1999; Pembrook, 2014). Dunlop (1990) hypothesized that subsp. scirpoidea persisted in three North American glacial refugia during the Pleistocene: Beringian Alaska, the Cordilleran region in Western North America, and periglacially in Eastern North America. Westergaard et al., (in press) found genomic evidence for these refugia, as well as additional glacial refugia in Norway and Greenland. However, the ancestral populations of subsp. convoluta remain undocumented.
Predicting genetic diversity based upon life history yields conflicting results. Shackleford (2003) speculated that subsp. convoluta would harbor low levels of genetic diversity, especially
28


compared to its widespread conspecific, due to its restricted distribution. However, C. scirpoidea is dioecious, a trait known to preserve genetic diversity (Hamrick and Godt, 1996). Mechanisms and effectiveness of pollen and diaspore dispersal are not well understood, further complicating predictions regarding genetic diversity and structure in this species. DePrenger-Levin (2007) and Pembrook (2014), using allozyme analysis, reported levels of genetic diversity in subsp. convoluta that were not significantly lower than in the widespread subsp. scirpoidea in all populations except the Beringian Alaska population. However, it should be noted that population samples for subsp. scirpoidea were primarily obtained from the taxon’s southernmost populations, and allozymes have a coarser resolution than most contemporary genomics methods.
In order to better characterize genetic diversity and structure in this C. scirpoidea, including subsp. convoluta, a relatively novel molecular genetic technique with higher resolution (greater number of data points) was employed. Double digest restriction-site associated DNA (ddRAD) sequencing is a contemporary method of reduced-representation genome sequencing wherein only genomic loci flanked by two specific restriction enzyme cut sites are sequenced. The technique has gained popularity due to its ability to partially sequence multiple individuals per sequencing run, and assemble a partial genome de novo for organisms without publicly available reference genomes (Baird et al., 2008; Peterson et al., 2012).
In this study, ddRAD sequencing was conducted to assess genomic diversity for the rare edaphic endemic subsp. convoluta and the widespread, presumed generalist, subsp. scirpoidea. Additionally, population genetic structure and phylogenomic relationships were inferred using ddRAD genomic data in order to detect divergence among populations of both taxa. This paper seeks to quantify genomic diversity and explain evolutionary diversification within a
29


geographical context. The objectives of this study are to 1) quantify genomic diversity in subsp. convoluta relative to its widespread conspecific, 2) assess structure within and among populations of the subspecies, 3) test evolutionary hypotheses regarding the origin of subsp. convoluta, and 4) estimate risks of extirpation and evolutionary potential of subsp. convoluta. Notably, this study marks the first population genomic study of Carex — and among the first for plant taxa — using ddRAD sequencing.
Methods
Tissue Collection
Fieldwork was conducted at five sites for subsp. convoluta from across its range (Dunlop and Crow, 1999; Fig. 3.1a; Table 3.1; Appendix A). At each site, leaf tissue was collected from 24 individuals per site following a transect-based sampling protocol for a total of 120 individuals (Fig. lb). Leaf tissue of subsp. scirpoidea individuals was previously collected using the methods described by Pembrook (2014) (Fig. 3.1b; Table 3.1). Leaf tissue was silica dried and stored at room temperature.
DNA Extraction
Silica-dried leaf tissue (20mg) was cut into pieces with sterile scissors, then frozen. Three 2.8 mm ceramic beads (OMNI International SKU 19-646) were added to each Eppendorf tube, and pulverized at 1.5 minutes with a reciprocating saw modified from Alexander et al. (2007). DNA was extracted following a CTAB protocol using NucleoSpin® Plant II kits (Macherey-Nagel 740440.50) with the following modifications. Lysis incubation was increased to 30 minutes, crude lysate was centrifuged for 5 minutes at 11,000 x g prior to filtration, and the column containing DNA was dried in an oven at 40°C for 30 minutes prior to elution. Warm milliQ® water (40 pL) was used in place of the provided elution buffer. DNA concentrations
30


were quantified with a Nanodrop 2000 and later with Qubit (dsDNA HS Assay Kit Q32851). Samples with signs of carbohydrate contamination or low DNA concentration were cleaned and concentrated with ethanol precipitation, re-extraction of DNA from additional leaf material, and/or vacuum concentration, as needed. Ethanol precipitation was performed using 1/10th volume of 3M sodium acetate (pH 5.5) and 2X the volume of 100% EtOH, washed with 70% EtOH, and resuspended in milliQ® water. After DNA extraction, cleanup, and concentration, samples with the highest concentration were selected for sequencing: 10-11 individuals from each subsp. convolutci population, and 5-8 individuals from each subsp. scirpoidea population.
ddRAD library preparation was conducted using a modified version of Parchman et al. (2012). MSEI and EcoRI-HF were used as restriction enzymes to digest the DNA, and their respective sequencing adaptors with complementary overhangs were ligated onto digested DNA with DNA ligase. PCR amplification was used to amplify correctly ligated fragments for DNA sequencing. The Biofrontiers Sequencing Facility (University of Colorado, Boulder) conducted fragment size-selection, cleanup, and conducted sequencing. BluePippin (Sage Science) was used for size-selection at the 300-600 fragment range and cleanup using Agencourt AMPure XP beads (A63881). Sequencing was performed to generate single-ended reads using the Illumina NextSeq V2 High Output (75-cycle kit) platform.
SNP calling and filtering
Pooled Illumina output was demultiplexed using the first step of the iPyrad pipeline (https://ipyrad.readthedocs.io/). A barcodes file matching those used during ddRAD library preparation was supplied. Restriction overhang sequences CAATTC (protector base [C] + EcoRI) and TAA (Msel) were used. During demultiplexing, barcode mismatches of up to two nucleotides were permitted. Once demultiplexed, adapter trimming was employed by dDocent
31


(Puritz et al., 2014). De novo assembly was conducted using Rainbow with read clustering at 80% minimum similarity (Chong et al., 2012). Read mapping was conducted using BWA with default scores and penalties (Li and Durbin, 2009). For reference assembly, K1 (coverage within each individual) and K2 (minimum number of individuals) parameters were chosen based upon manual inspection of the asymptotic curve inflection. K1 was set to a minimum depth of five, while K2 was set to nine. freeBayes was invoked for SNP calling using default parameters supplied by dDocent (Garrison and Marth, 2012).
The raw SNP file was filtered with vcflib (Garrison, 2012) and vcftools (Danecek et al., 2011), following the dDocent SNP filtering process, with some exceptions: minor alleles with a count less than four were filtered, individuals with >40% missing data were discarded. Next, a 95% genotype call rate across all populations was specified, followed by a 95% call rate within each population. Allele balances between 20% and 80% were retained. Mapping quality between reference and alternate alleles were kept if between 0.8 and 1.2. Loci that did not have a minimum quality score within 'A of the depth were removed. Next, loci above the mean depth + three standard deviations, with a quality score above twice the depth, were filtered. To filter potential paralogous sites, loci with a high mean depth were filtered. To choose a cutoff for high mean depth, the upper bound value of the larger histogram (mean depth = 160) was chosen. Loci not at Hardy Weinberg Equilibrium (p > 0.05) were discarded. To minimize linkage between loci, rad_haplotyper (Willis et al., 2017) was employed, and loci were kept that successfully haplotyped in 90% of individuals and saved in variant call format (VCF). PGDspider was employed to convert data to the correct format for each subsequent software program (Lischer and Excoffier, 2012).
32


Data Analysis
The following population genomic statistics were obtained using Arlequin: expected heterozygosity (HE), observed heterozygosity (H0), percent of polymorphic loci (P), nucleotide diversity (0K), pairwise differentiation between populations (Fst), and Nei’s distance between populations (d) (Excoffier and Lischer, 2010). Two locus by locus AMOVA tests were conducted with 1000 permutations in order to analyze hierarchical patterns of genomic differentiation across the dataset. Individuals were categorized in their respective populations, and each population was categorized into broad groups. In the first AMOVA, group was defined as ‘taxon’ (subsp. convoluta and subsp. scirpoidea); in the second, group was defined as ‘region:’ Northwestern represented by Anvil Mountain, AK; Western represented by Butte, MT and High Creek Fen, CO; and Eastern comprising all remaining populations. Parameters measured included inbreeding coefficient (Fis), variance among individual populations within groups (Fsc), variance among taxa compared to total variance (FCt), and overall fixation index (Fit). FSt was calculated with 100 permutations at the 0.05 significance level, while QK was calculated by taking the average number of pairwise differences under the infinite site model. In order to assess candidate loci under selection, an outlier loci analysis was calculated with Bayescan (Foil and Gaggiotti, 2008) for the eight Eastern North American populations (all but Anvil Mountain, AK; Butte, MT; and High Creek Fen, CO). BayeScan uses a reverse-jump Markov Chain Monte Carlo to estimate locus-population-specific genomic differentiation (Fst). I ran 10,000 iterations of the Bayesian model, using the default settings. Western North American populations were excluded in order to limit false positives from isolation by distance.
Structure in the data was assessed using fastStructure (Raj et al., 2014) and discriminant analysis of principle components (DAPC) function of adegenet (Jombart and Ahmed, 2011)
33


using R Version 3.4.4 (R Core Team, 2014). fastStructure uses a model-based approach, while DAPC uses a discriminant analysis to maximize variability between groups designated by PCA. In fastStructure (Raj et al., 2014), K values between 1 and 11 (total number of populations) were tested using a simple model with default settings. K values indicating strong and subtle population genetic structure were chosen using the chooseK python script. Structure at both scales were plotted in Distruct (Rosenberg, 2004). Using adegenet (Jombart and Ahmed, 2011), BIC scores were obtained for K groups (K = 1-40). Models with optimal BIC scores (ABIC < 1) were assessed for optimal a-scores. DAPC, using the number of principle components (PCs) and discriminant functions specified, were used to determine structure.
SNP loci with ambiguous sites in an individual were filtered using the raxmlascbias Python script (Martin, 2018) to avoid ascertainment bias during alignment. jModelTest2 (Guindon and Gascuel, 2003; Darriba et al., 2015) was used to test 88 different models of nucleotide substitution. Phylogenies were built using RAxML (Stamatakis, 2014), using a ASC GTRGAMMA model for 500 bootstrapping iterations, and the resulting tree (Appendix E) was rooted by employing the simple rooting algorithm. All trees were plotted using Dendroscope (Huson and Scomavacca, 2012).
Results
NextSeq 1 x 75 sequencing produced 472,350,095 reads. After demultiplexing, 344,139,306 reads remained. After filtering and SNP calling, 719,059 SNPs were “called” by dDocent (Puritz et al., 2014). Of the 96 individual samples that were sequenced, 77 remained after filtering. Of the 19 individuals that were filtered out, all were Carex scirpoidea subsp. convoluta representing Horseshoe Bay, MI (5); Manitoulin Island, ON (7); Maxton Plains, MI (4); Thompson’s Harbor, MI (3). After filtering loci, 7384 SNPs remained, which were used in
34


each analysis, with the exception of the phylogenomic analysis.
Population Genomics
All measures of genomic diversity in subsp. convoluta exceeded those for subsp. scirpoidea, including proportion of polymorphic loci (P), nucleotide diversity (0K), and expected heterozygosity (He) (subsp. convoluta: HE = 0.39, P = 51.688, 0n = 1490.33; subsp. scirpoidea: He = 0.38, P = 46.62, 0^ = 1307.79). Observed heterozygosity was higher than expected heterozygosity in every assayed population (Table 3.2, Fig. 3.3). The Anvil Mountain, AK population of subsp. scirpoidea exhibited the lowest genomic diversity (P = 34.41, 0K= 995.05; Table 3.2; Figs. 3.3, 3.4). Manitoulin Island, ON subsp. convoluta had the highest rate of nucleotide diversity (QK = 1552.53), while Cabot Head, ON had the highest rate of polymorphic loci (P = 58.85%). Allele count across all surveyed loci ranged between 2.003 and 2.005 (Fig. 3.5 Appendix E).
The AMOVA takes into consideration analysis of variance for mutations among loci to determine where in the sample variation occurs. In the taxon-partitioned AMOVA, only 5.42% of the variation was among taxa (p < 0.001), while nearly one third of the variation observed among populations within taxa (29.10%, p < 0.001, Table 3.3). In the region-partitioned AMOVA, the reverse was observed: one third of the variation was observed among regions (33.95%, p < 0.001, Table 3.4), with only 9.74% (p < 0.001) of the variation occurring among populations within regions. Geographically distant populations were highly differentiated; the western populations (Anvil Mountain, AK; Butte, MT; and High Creek Fen, CO) were highly differentiated from the eastern populations (FSt= 0.41-0.50; Table 3.5; Figs. 3.10, 3.11), while the greatest differentiation was found between the High Creek Fen, CO and Anvil Mountain, AK populations (FSt= 0.50, p < 0.005).
35


Of the 7384 assayed loci, 11 loci were found to be outlier loci that exceed the false discovery rate (FDR; q-value = 0.05) (Fig. 3.12). Fsiwas bimodal, with all intermediate values falling under the FDR. Of the 11 loci, three loci produced FSt values below 0.06; eight loci produced FSt values exceeding 0.36, suggesting population differentiation due to selection (Foil and Gaggiotti, 2008).
Structure
Optimal model complexity calculated using fastStructure (Raj et al., 2014) resulted in three groups (K = 3) (Fig. 3.6). No clustering of populations occurred along taxonomic lines (e.g., subsp. scirpoidea vs. subsp. convoluta). Instead, populations clustered over three broad geographical regions, which were also identified by Westergaard et al. (pers. comm.); these are: 1) a Beringian group, as represented here by the Anvil Mountain, AK population; 2) a Western North American group represented by the Butte, MT and High Creek Fen, CO populations; and 3) an Eastern North American group as represented by the remaining populations, including each of the five subsp. convoluta populations. fastStructure (Raj et al., 2014) model selection indicated that five clusters or groups (K = 5) account for 99.9% cumulative ancestry contribution. Therein, evidence of admixture between populations was exhibited in eight of the eleven populations surveyed, all representing the Eastern North American group (Fig. 3.6).
DAPC analyses revealed that five groups (K =5) had the highest model support (BIC,
Fig. 3.7), with several notable similarities to fastStructure (K=5) results (Fig. 3.8). For instance, the Brig Bay, NL subsp. scirpoidea population clustered with populations of subsp. convoluta at all K, using both methods. Interestingly, each analysis showed that one population on the range margin of subsp. convoluta (Thompson’s Harbor [MI], K=5, Fig. 3.6; Cabot Head [ON]; Fig.
3.8) formed a distinct group. Further, remaining central populations formed a group (Figs. 3.6,
36


3.8), although there was also very high model support for three and four groups (K = 3-4; ABIC < 1; Fig. 3.7). Notably, atK=3, fastStructure and DAPC produced identical results.
Phytogeny
After filtering ambiguous sites to be used to create the phylogeny, 5985 sites remained. Of the 88 nucleotide substitution models tested, a General Time Reversible model with gamma distribution and invariable sites (GTR+I+G) model produced optimal BIC and AIC scores, though the transversion gamma model (TVM+G) was best supported by the decision-theoretic performance-based metric (DT). Thus, the GTR-Gamma model with ascertainment bias correction (ASC GTRGAMMA) was selected to for the Maximum Likelihood cladogram.
With the exception of the Maxton Plains, MI subsp. convoluta population, which received weak bootstrap support (50%), each population was recovered as monophyletic with moderate to very strong bootstrap support (> 80; Fig. 3.9). There was no support for a monophyletic subsp. convoluta. Instead, the western High Creek Fen, CO, Butte, MT, and Anvil Mountain, AK subsp. scirpoidea populations segregated with strong bootstrap support (100%), while the eastern group — recovered in fastStructure and DAPC — was unresolved and polyphyletic. Within the western clade, the Butte, MT and High Creek Fen, CO populations segregated together with strong bootstrap support (100%). Within the Eastern clade, the Escanaba River, MI and Pembina, MN populations (subsp. scirpoidea) and Horseshoe Bay, MI population (subsp. convoluta) form a clade, albeit with weak (64%) bootstrap support. Within subsp. convoluta, only the Maxton Plains, MI and Thompson’s Harbor, MI populations formed a well-supported clade (91%, Fig. 3.9).
37


Discussion
Population Genomic Diversity
In contrast to expectations based upon restricted range, fragmented habitat, and relatively small population size (Bard and Bruederle, personal observation), population genomic diversity in subsp. convoluta is, in fact, higher than all but one population of the widespread conspecific subsp. scirpoidea (e.g., P = 51.69± 4.9 vs. P = 46.62 ± 6.70; Table 3.2). Although some inbreeding depression was expected, all populations of C. scirpoidea exhibit heterozygote excess (Fis= -0.18, Table 3.4, 3.5), with observed heterozygosity exceeding expected heterozygosity in all populations (Table 3.2). Obligate outcrossing due to dioecy has likely contributed to the preservation of genomic variation by preventing selfing, the most severe type of inbreeding (Ellstrand and Elam, 1993; Hamrick and Godt, 1996; Dunlop and Crow, 1999). These findings generally support Pembrook (2014) and DePrenger-Levin (2007), neither of whom found lower genetic diversity in populations of subsp. convoluta compared to subsp. scirpoidea.
Population Genomic Structure and Phylogenomics
Rather than exhibiting genomic differentiation along taxonomic lines, as was expected, strong structure was instead observed among three geographic regions: Northwestern North America, Western North America, and Eastern North America (fastStructure/DAPC K=3, Figs. 3.6, 3.8). This region-based structure was also revealed from the phylogenomic analyses, with the Northwestern and Western regions forming monophyletic groups (clades) with 100% bootstrap support (Fig. 3.9). Eastern North American populations of subsp. scirpoidea formed a group with subsp. convoluta (fastStructure/DAPC K = 3, Figs. 3.6, 3.8), although the phylogenomic analysis revealed the Eastern clade to be polyphyletic (Fig. 3.9). An AMOVA showed that 33.95% of all genomic variation was partitioned among the aforementioned regions,
38


while only 5.42% was between taxa (Table 3.4). Finally, genomic differentiation between populations among regions (Fst) ranged from 0.41 to 0.50 (p < 0.05 in 88.5% comparisons,
Table 3.5, Fig. 3.11).
The observed regional structuring pattern provides support for Dunlop's (1990) proposed refugia for Carex scirpoidea during Wisconsin glaciation: Beringian Russia, Alaska, and Yukon; Cordilleran (Rocky Mountain) western North America, south of the last glacial maximum (LGM); and periglacially in Eastern North America. Among the populations sampled in my study, the Northwestern North American populations represent a Beringian cluster also identified by Westergaard et al. (in press), while the Western North American populations represent a well-supported Cordilleran cluster that differentiated during the Pleistocene. In contrast, Eastern populations of subsp. scirpoidea are clearly the product of recolonization from periglacial North American refugia, south of the LGM, rather than recolonization from the Beringian and Cordilleran refugia (Figs. 3.7-3.9).
One such glacial refugia for Eastern North American subsp. scirpoidea populations may be the Driftless Area. Westergaard et al. (in press) found evidence for distinct and unadmixed populations in Pembina, MN and Escanaba River, MI. Their results also suggested that Western Greenland populations admixed with North American and Eastern Greenland/Norway influence, suggesting East-West range colonization during glacial retreat. Evidently, C. scirpoidea populations in the Great Lakes area of North America expanded Northeastward as the Laurentide Ice Sheet receded. The Brig Bay, NL population in my study supports this expansion, as it forms a group with the Pembina, MN population and is only moderately differentiated from subsp. scirpoidea populations from the Upper Midwest (K=3-5, Fig. 3.8; FSt=0.17, Table 3.5, Figs.
3.10, 3.11).
39


My results suggest that extant populations of subsp. convoluta also descended from periglacial Eastern North American refugial populations. The furthest northeastern population of the Eastern group (Brig Bay, NL) exhibited no apparent substructure when compared to subsp. convoluta (All K, Figs. 3.6, 3.8), and low differentiation (Fst= 0.10-0.18, Table 3.5, Fig. 3.11) compared to populations of subsp. scirpoidea (FSt= 0.44-0.45). Interestingly, there is a closer affiliation between subsp. convoluta and Eastern Canada subsp. scirpoidea populations than with Great Lakes subsp. scirpoidea populations — where in fact subtle structure is exhibited (K=5,
Fig 3.6; K=4-5, Fig. 3.8).
Although my results do not resolve the pattern of postglacial recolonization, multiple origins for subsp. convoluta are suggested. Previous studies have suggested that Great Lakes disjuncts represent elements of the Atlantic and Gulf Coastal Plains flora (Reznicek, 1994; Fant et al., 2014). Several species are hypothesized to have originated from the southern Gulf Coastal Plain and migrated to the Great Lakes along the Mississippi, while others originated from the North Atlantic Coastal Plain (McLaughlin, 1932). Fant et al. (2014) proposed northeastern Michigan, in the vicinity of the Straits of Mackinac, as a postglacial entry point for many plant species following the Atlantic Coastal route, as evidenced by patterns of genetic diversity in Cirsium pitcheri (Torr. Ex Eaton) Torr & A. Gray, another Great Lakes endemic.
Thus, three scenarios seem likely for the origin of subsp. convoluta, though none are mutually exclusive. First, founding populations of extant subsp. convoluta may have established following westward expansion of Eastern North American C. scirpoidea populations, as has been reported for other Great Lakes disjuncts (Reznicek, 1994). Alternatively, subsp. convoluta populations and Eastern Canada (e.g., Brig Bay, NL) populations may be sister groups, having independently established following range expansion of ancestral populations near the Driftless
40


Area. Finally, populations may have recolonized the extant range of subsp. convoluta from other, unknown periglacial refugia south of the LGM.
Speciation, if occurring in this complex, is apparently at an early stage, with little evidence for coalescence. The two taxa harbor no reproductive isolation (Dunlop and Crow, 1999), and subsp. convoluta exhibits high affinity for Eastern populations of subsp. scirpoidea (i.e. Brig Bay, NL). Slight admixture between Escanaba River, Ml/Pembina, MN and subsp. convoluta populations was observed using fastStructure, indicating that gene flow between taxa has occurred or is ongoing (Fig. 3.6). Further, the taxonomical boundaries of subsp. convoluta remain ambiguous, at best (Figs. 3.6-3.9). As such, the taxon may more appropriately be categorized as biotypes (Clausen, 1951; Lowry, 2012).
Extant alvar populations are thought to be relicts of prairie and parkland habitats that were spatially and, ultimately, genetically isolated from conspecific populations during glacial recession. North American alvar is restricted to the Niagara Escarpment and supports small, fragmented, and discontinuous communities comprising variously small populations between which restricted gene flow is expected for endemics (Catling, 1995; Hamilton and Eckert, 2007). As a result, alvar endemics are expected to harbor low genetic diversity and be differentiated from ancestral populations (Hamilton and Eckert, 2007). One alvar ecotype, the monoecious Geum trijlorum Pursh, was found to exhibit lower genetic diversity and greater differentiation among its alvar populations than prairie populations across its range; this was attributed to the disjunction of alvar habitats from prairielands, as well as changes in habitat availability during glaciation and recession (Hamilton and Eckert, 2007).
In contrast to the findings of Hamilton and Eckert (2007), subsp. convoluta populations exhibit similar, and in some cases, higher levels of genomic diversity compared to subsp.
41


scirpoidea (e.g. 0K= 1442.38-1552.52 vs. 995.05-1455.08, Table 3.2; Figs. 3.3-3.5). Further, there was very little evidence of genetic differentiation among populations of subsp. convoluta and subsp. scirpoidea from Eastern North America (Fst = 0.11-0.18, Table 3.5; Fig 3.10, 3.11). Differences in breeding system between C. scirpoidea and G. triflorum — dioecy vs. monoecy — likely contribute to these differences in population genomic diversity, which otherwise have similar recent evolutionary histories. Obligate outcrossing in subsp. convoluta is likely preserving genomic diversity, and halting loss of allelic variation resulting from genetic drift.
Structure (Fig. 3.6, 3.8) and significant differentiation (p < 0.05 in 7/10 pairwise comparisons, Table 3.5) of the two marginal populations, Cabot Head, ON and Thompson’s Harbor, MI, indicate restricted gene flow in peripheral populations of subsp. convoluta (Fant et al., 2014). However, the diaspore dispersal distance, a critical determinant of gene flow, is unknown. Limited discontinuous alvar habitats may be aiding the enforcement of structure within all sampled populations of subsp. convoluta — populations are monophyletic (Fig. 3.9), with genomic differentiation occurring between them (FSt 0.05 > 0.13, p < 0.05 in 7/10 pairwise comparisons, Table 3.5, Fig. 3.11). Geomorphological features (e.g., Lake Huron) may further obstruct gene flow restriction.
Selection and isolation by distance
Population differentiation implies reduced gene flow among populations, local adaptation within populations, or both. The isolated outlier loci identified herein (Fig. 3.12) are likely to be under selection, suggesting that local adaptation may be contributing to population differentiation (Foil and Gaggiotti, 2008). Carex scirpoidea subsp. scirpoidea and subsp. convoluta occupy a wide range of habitats with varying ecological conditions (Dunlop and Crow, 1999; Shackleford, 2003, Chapter 2). As discussed in Chapter 2, soil chemistry correlates, at
42


least to some degree, with the Northwestern, Western, and Eastern North American genetic clusters (Chapter 2; Figs. 3.6, 3.8). Although regional adaptation may be expected, genomic differentiation at the regional level may also be attributed to reduced gene flow among regions over many generations.
Although most populations of subsp. convoluta occur along the Niagara Escarpment, the soil differed more than was expected between sites. Furthermore, I found 11 loci among the Eastern North American populations that are putatively under selection; however, this is likely a conservative estimate, as BayeScan limits false positives in small population samples (Foil and Gaggiotti, 2008; Fig. 3.12). Thus, it is possible that adaptive loci may be facilitating local adaptation among populations with different environmental conditions. Alternatively, wide tolerance across subsp. convoluta may facilitate occupation of different soil types (Chapter 2, this thesis). It is possible that the diversification of subsp. convoluta is itself due to edaphic adaptation; Dunlop (1990) suggests that the phenotypic leaf properties that is diagnostic for subsp. convoluta may be adaptation to long periods of drought.
My study also provides some evidence for geographic isolation-caused differentiation. The highest levels of population differentiation were detected among the Cabot Head, ON); Horseshoe Bay, MI; and Thompson’s Harbor, MI populations (FSt =0.11-0.13, Table 3.5, Figs. 3.10, 3.11). Cabot Head, ON and Thompson’s Harbor, MI are located on extreme ends of the range — the east and west sides of Lake Huron, respectively. Thus, geographical distance, as well as environmental conditions acting as selective pressures, may be contributing to the maintenance of structure in subsp. convoluta.
43


Conservation of Carex scirpoidea subsp. convoluta
There has been no apparent loss of genomic diversity in Carex scirpoidea subsp. convoluta, either at the level of the taxon or population, compared to its widespread conspecific. Although my results suggest that genomic diversity is being maintained by populations of subsp. convoluta at the time, habitat fragmentation and loss remain as risks limiting gene flow and outbreeding among isolated populations of this taxon. Restricted gene flow, as well as local adaptation may contribute to differentiation between populations of subsp. convoluta (FSt = 0.07-0.13, Table 3.5). Furthermore, isolation of marginal populations, such as Thompson’s Harbor, MI and Cabot Head, ON, may increase their sensitivity to habitat changes or stochastic events. Finally, genomic adaptation may be isolated to local populations; subsp. convoluta inhabits soils with a wide range of chemical and physical conditions. As such, populations, rather than taxa, should be treated as conservation units, particularly those at the range margin (e.g., Thompson’s Harbor, MI and Cabot Head, ON) may be prioritized, due to their differentiation from other subsp. convoluta populations (FSt=0.07-0.13, Table 3.5, Fig. 3.11; Allendorf et al., 2010; Frankham, 2010). Preventing genomic diversity loss and locally adapted populations are measures to retain evolutionary potential.
Carex scirpoidea subsp. convoluta is a seemingly resilient taxon, retaining genomic diversity despite fragmented habitats and limited range. Thus, the taxon may be considered for replanting, particularly for restoring endemic-rich habitats with unusual soil types (Reschke et al., 1999; Hamilton and Eckert, 2007). Such efforts may have ecological benefits; detrital sedge tissue contributes to the unique edaphic conditions of wetland (i.e., sedge fen) habitats (Gorham, 1991), and is thought to be a food source for grazing fauna (Ludwig et al., 1996; Shackleford, 2003).
44


Ongoing conservation and management practices should be informed by monitoring habitat degradation, recreational use, invasion by non-native species, and climate change.
Edaphic endemics likely colonize adverse soils due to their low competitive ability (Kruckeberg and Rabinowitz, 1985; Dillenberger and Kadereit, 2013). With alvar soil, influx of nutrients allows colonization of non-native shrubs (Catling and Brownell, 1995). Therefore, the known threat from native and non-native plant species (e.g., Thuja occidentalis L.) may destabilize populations and lead to local extinction (Comer et al., 1997). Similarly, habitat degradation (e.g., quarrying, recreational use) could destabilize suitable habitat for subsp. convoluta and other alvar endemics (Shackleford, 2003). The survival of subsp. convoluta may be contingent upon protecting rare habitat types, especially those inhospitable to competing plant species.
Future Directions
I intend to elucidate patterns of edaphic selection and adaptive divergence using the ddRAD genomic data in tandem with my soil data. Although the data presented in Chapter 2 substantiates expectations of a broad edaphic niche for Carex scirpoidea subsp. scirpoidea, a broader than expected niche for subsp. convoluta was also revealed. Habitats for the two taxa range from nutrient-poor, calcareous swamps (Horseshoe Bay, MI) to saline fen margins (High Creek Fen, CO). Bayesian analysis of locus-population specific differentiation (Foil and Gaggiotti, 2008) revealed 11 loci putatively under selection (Fig. 3.12). Statistical association of genomic SNPs with environmental variables will be assessed following Rellstab et al. (2015) in order to determine whether there is evidence for genomic adaptation to local environmental conditions.
Secondly, due to the fragmented nature of habitat for subsp. convoluta, a better understanding of the mechanisms of pollen and seed (diaspore) dispersal is mandated to
45


understand gene flow among populations. Disjunct, marginal populations of subsp. convoluta, in particular (e.g., Thompson’s Harbor, MI and Cabot Head, ON) show evidence of genomic differentiation and may be threatened by restricted gene flow from central populations. However, long distance dispersal — which has been well documented for the genus — can be expected to lead to gene flow among populations and fragmented habitat, as well as the establishment of new populations. In subsp. convoluta, this may be facilitated by water dispersal, dispersal across snow and ice, and proximity to a shoreline.
Finally, further evidence is needed to elucidate the evolutionary divergence of subsp. convoluta from Eastern North American populations of subsp. scirpoidea. Currently, data are limited to the partial genomes of only seven individuals between the Great Lakes and Western Greenland (this study; Westergaard et al., in press). I recommend future phylogeographic studies that integrate a greater number of Eastern North American populations in order to resolve the postglacial expansion of Eastern North American C. scirpoidea, including the origin of subsp. convoluta, as well as the state of speciation in this taxon.
46


Table 3.1: Tissue sampling sites for Carex scirpoidea Michx. subsp. convoluta (Kiik.) Dunlop (Cyperaceae) and subsp. scirpoidea (Cyperaceae).
Taxon Sample Site Country Latitude Longitude
Thompson’s Harbor, MI USA 45.35 -83.58
d C/3 3 C/3 Horseshoe Bay, MI USA 45.99 -84.745
2* Maxton Plains, MI USA 46.08 -83.66
o •K: ?3 y d Manitoulin Island, ON CAN 45.60 -82.12
X O b S Cabot Head, ON CAN 45.24 -81.30
Anvil Mountain, AK USA 65.48 -145.41
Butte, MT USA 46.19 -112.77
d C/3 3 C/3 High Creek Fen, CO USA 39.10 -105.97
2* Escanaba River, MI USA 45.90 -87.21
o Cl •d Q % ^ Pembina, MN USA 48.08 -96.45
x 3 b .gr r« U \J to Brig Bay, NL CAN 51.06 -56.91
47


Table 3.2: Levels of genetic diversity for 11 populations of Carex scirpoidea Michx. subsp. convoluta (Kiik.) Dunlop and subsp. scirpoidea (Cyperaceae) based on 7384 genomic SNPs. Site abbreviations are as follows: HB = Horseshoe Bay, MI, USA; TH = Thompson’s Harbor, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CAN; CH = Cabot Head, ON, CAN; AM = Anvil Mountain, AK, USA; B = Butte, MT, USA; HCF = High Creek Fen,
CO, USA; P = Pembina, MN, USA; ER = Escanaba River, MI, USA; BB = Brig Bay, NL, CAN.
Site He Ho
Taxon N P (%) bn
Mean + SD Mean + SD
AM 8 0.39 + 0.15 0.46 + 0.29 34.41 995.05
B 8 0.37 + 0.15 0.42 + 0.28 47.84 1295.55
s HCF 8 0.38 + 0.15 0.43 + 0.28 47.45 1315.18
> s: o o P 8 0.36 + 0.15 0.46 + 0.28 52.99 1417.97
ci
g ER 8 0.38 + 0.14 0.45 + 0.27 52.08 1455.08
Q BB 5 0.41 + 0.14 0.51+0.32 44.96 1367.89
o
c 3 >3 Mean + SD 0.39 + 0.04 0.44 + 0.04 51.69 + 4.91 1490.33 + 46.62
2i Q o Range 0.11 0.11 13.18 110.15
TH 7 0.39 + 0.14 0.46 + 0.26 50.56 1442.38
Q is HB 5 0.39 + 0.14 0.43 + 0.26 49.71 1445.16
O § 3 >3 MP 6 0.38 + 0.14 0.42 + 0.26 53.65 1503.97
ci t/j g MI 3 0.46 + 0.12 0.50 + 0.28 45.67 1552.53
^2 3 Cl CH 11 0.35 + 0.15 0.39 + 0.26 58.85 1507.61
3 >3 Mean + SD 0.38 + 0.02 0.46 + 0.03 46.62 + 6.70 1307.79+ 164.59
2i Q u Range 0.05 0.09 18.58 460.03
48


Table 3.3: Locus-by-locus AMOVA results by taxon as weighted average over 7384 genomic loci of 11 populations of 2 taxa; 1) Carex scirpoidea Michx. subsp. convoluta (Kiik.) Dunlop and 2) subsp. scirpoidea (Cyperaceae). F-statistics, including inbreeding coefficient (Fis), variance among individual populations within taxa (FSc), variance among taxa compared to total variance (FCt), and overall fixation index (Fit). Significant (p<0.01) variance components and fixation indices in bold.
Source of variation Sum of squares Variance Percentage
components variation
Among taxa 9337.940 56.26525 5.41862
Among populations 42360.383 302.11838 29.09549
within taxa
Among individuals 36655.009 -124.60599 -12.0017
within populations
Within individuals 61953.500 804.59091 77.48606
Total 150306.831 1038.36854
Fis -0.18325
Fsc 0.30762
Fct 0.05419
Fit 0.22514
49


Table 3.4: Locus-by-locus AMOVA results by region as weighted average over 7384 genomic loci of 11 populations of Carex scirpoidea Michx. subsp. convoluta (Kiik.) Dunlop and subsp. scirpoidea (Cyperaceae) at 3 regions: 1) Northwestern, 2) Western, and 3) Eastern North American. F-statistics, including inbreeding coefficient (Fis), variance among individual populations within taxa (FSc), variance among regions compared to total variance (FCt), and overall fixation index (Fit). Significant (p<0.01) variance components and fixation indices in bold.
Source of variation Sum of squares Variance components Percentage variation
Among regions 34650.637 410.01290 33.95238
Among populations 17047.685 117.61385 9.73938
within regions
Among individuals 36655.009 -124.60599 -10.31838
within populations
Within individuals 61953.500 804.59091 66.62663
Total 150306.831 1207.61166
Fis -0.18325
Fsc 0.14746
Fct 0.33952
Fit 0.33373
50


Table 3.5: Pairwise differentiation (Fst) measured across 7384 loci of Carex scirpoidea Michx. subsp. convolutci (Kiik.) Dunlop and subsp. scirpoidea (Cyperaceae). Significant values (p < 0.05) in bold. Comparison between different conspecifics in shaded area. Site abbreviations are as follows: HB = Horseshoe Bay, MI, USA; TH = Thompson’s Harbor, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CAN; CH = Cabot Head, ON, CAN; AM = Anvil Mountain, AK, USA; B = Butte, MT, USA; HCF = High Creek Fen, CO, USA; P = Pembina, MN, USA; ER = Escanaba River, MI, USA; BB = Brig Bay, NL, CAN.
Carex scirpoidea subsp. scirpoidea Carex scirpoidea subsp. convoluta
AM B HCF P ER BB TH HB MP MI c H
AM -
o B 0.49
cf C/3 3 C/3 ^3 HCF 0.50 0.10 -
O £ P 0.44 0.44 0.44 -
X a ER 0.44 0.44 0.45 0.16 -
BB 0.45 0.45 0.45 0.17 0.17 -
i TH 0.44 0.45 0.45 0.16 0.18 0.16 -
o p- s o o HB 0.43 0.43 0.43 0.13 0.13 0.13 0.13 -
cf C/3 3 C/3 ^3 MP 0.43 0.43 0.43 0.13 0.14 0.12 0.09 0.09 -
O MI 0.43 0.42 0.42 0.11 0.12 0.10 0.08 0.07 0.05 -
X a CH 0.41 0.43 0.43 0.15 0.15 0.14 0.13 0.11 0.10 0.07 -
51


b
'AM
Canada

WA
OR
CA
Northwestern
Passages
Hudson Bay
QC
& or
BB
ANB
-f MEN |
NY
PE
7^" MEN ns NH MA
“HC

:ed States
KS MO
IN OH PA
’ MD
-•WV-- ,ut KY VA
Figure 3.1: Plant tissue sampling sites for (a) Carex scirpoidea Michx. subsp. convoluta (Kiik.) Dunlop (Cyperaceae) and (b) subsp. scirpoidea (Cyperaceae). Site abbreviations are as follows: HB = Horseshoe Bay, MI, USA; TH = Thompson’s Harbor, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CAN; CH = Cabot Head, ON, CAN; AM = Anvil Mountain, AK, USA; B = Butte, MT, USA; HCF = High Creek Fen, CO, USA; P = Pembina, MN, USA; ER = Escanaba River, MI, USA; BB = Brig Bay, NL, CAN.
52


21 rnetGt.s
Figure 3.2: Transect sampling plan for each Carex scirpoidea Michx. subsp. convoluta (Kuk.) Dunlop (Cyperaceae) population site. At each site, 24 tissue samples were obtained, 3 meters
apart.
53


Expected heterozygosity
sAM
sB
sHCF
sP
00 00 00 00
-S' CO o O o £? co o o o ^ o o <*> -S' CO 0 0 ° O 0 0 CO
(/) o o 05 o o O) (/> o o 05 o o c (/) o o O)
fiw _ o o . o o N ^ _ o o _ o o
0) o a> m .. a> a> aj ..
Id - Hi 0.2 Hi 0.2 H. 0.2
o d T 00 o d T 00
1 1174 2787 4400 6013 Locus
sER
1 1174 2787 4400 6013 Locus
sBB
1 1174 2787 4400 6013 Locus
1 1174 2787 4400 6013 Locus
•■= ^ - o
00 00 o 00
Z? CO oooo ° ° _ O O- & CO O O OOO ODD -Z? CO
cn o cn o (/) o
05 Q Q O) 05
o o O o O d
Is- Is- Is-
1 ) 00 T ) 00 r ) 00
1174 2787 4400 6013 Locus
cTH
cCH
1 1174 2787 4400 6013 Locus
1174 2787 4400 6013 Locus
cHB
Figure 3.3: Expected heterozygosity for 7384 genomic loci in 11 populations of Carex scirpoidea Michx. subsp. convolata (Kiik.) Dunlop (preceded by c) and subsp. scirpoidea (s) (Cyperaceae). Site abbreviations are as follows: HB = Horseshoe Bay, MI, USA; TH = Thompson’s Harbor, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CAN; CH = Cabot Head, ON, CAN; AM = Anvil Mountain, AK, USA; B = Butte, MT, USA; HCF = High Creek Fen, CO, USA; P = Pembina, MN, USA; ER = Escanaba River, MI, USA; BB = Brig Bay, NL, CAN.
54


Molecular diversity indices
Figure 3.4: Nucleotide diversity (9*), measured across 7384 genomic loci for 11 populations of Carex scirpoidea Michx. subsp. convoluta (Kiik.) Dunlop (preceded by c) and subsp. scirpoidea (s) (Cyperaceae). Site abbreviations are as follows: HB = Horseshoe Bay, MI, USA; TH = Thompson’s Harbor, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CAN; CH = Cabot Head, ON, CAN; AM = Anvil Mountain, AK, USA; B = Butte, MT, USA; HCF = High Creek Fen, CO, USA; P = Pembina, MN, USA; ER = Escanaba River, MI, USA; BB = Brig Bay, NL, CAN. Dashed lines indicating 1 standard deviation.
55


Number of alleles at different loci
cCH
cHB
cMI
CMP
Locus
cTH
sER
Locus
sAM
E
sHCF
Locus
sB
sP
Locus
sBB
E ^ "
Locus
Locus
Locus
Figure 3.5: Allele count for each locus across 7384 loci in 11 populations of Car ex scirpoidea Michx. subsp. convoluta (Kiik.) Dunlop (preceded by c) and subsp. scirpoidea (s) (Cyperaceae). Site abbreviations are as follows: HB = Horseshoe Bay, MI, USA; TH = Thompson’s Harbor, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CAN; CH = Cabot Head, ON, CAN; AM = Anvil Mountain, AK, USA; B = Butte, MT, USA; HCF = High Creek Fen, CO, USA; P = Pembina, MN, USA; ER = Escanaba River, MI, USA; BB = Brig Bay, NL, CAN.
56


K=3
K=5
Fig 3.6: fastStructure plots for strong structure (K=3) and weak structure (K=5) of all assayed populations of Carex scirpoidea Michx. subsp. convoluta (Kiik.) Dunlop (preceded by c) and (b) subsp. scirpoidea (s) (Cyperaceae). Site abbreviations are as follows: HB = Horseshoe Bay, MI, USA; TH = Thompson’s Harbor, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CAN; CH = Cabot Head, ON, CAN; AM = Anvil Mountain, AK, USA; B = Butte, MT, USA; HCF = High Creek Fen, CO, USA; P = Pembina, MN, USA; ER = Escanaba River, MI, USA; BB = Brig Bay, NL, CAN.
57


520 540 560 580
Value of BIC
versus number of clusters
Number of clusters
Fig. 3.7. BIC values for DAPC models with K clusters.


2 3 1 2 3 4 1 2 3 4 5
Ousters Clusters Clusters
Fig 3.8: DAPC cluster plots of K=3-5 for Carex scirpoidea Michx. subsp. convoluta (Kiik.) Dunlop (preceded by c) and (b) subsp. scirpoidea (s) (Cyperaceae). Site abbreviations are as follows: HB = Horseshoe Bay, MI, USA; TH = Thompson’s Harbor, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CAN; CH = Cabot Head, ON, CAN; AM = Anvil Mountain, AK, USA; B = Butte, MT, USA; HCF = High Creek Fen, CO, USA; P = Pembina, MN, USA; ER = Escanaba River, MI, USA; BB = Brig Bay, NL, CAN.
so


iff
«-
•to-
€#-
!»■
100-
•»
100------------
V*
1R-
Cf
2ft-
100-
3ft-
45"
25-
■»5-
1<»-
1'fT

95-
I
3t-

9
g
95-
*B_2 *B_1 sB_9 jB.10 sB_6 sB_4 *B_3 jB.11 lHCF.18 iHCF.33 $HCf_24 *HCF_28 $HCF_23 SHCF_16 $HCf_21 JHCF.34 J/UL22 JAM.19 JAMJ7 jML24 J/UL18 sAM_28 SM4.16 j«C20 cHB.12 cHB_13 iEa.ii
jer.13
•cC»t21
cCH_4
eCH>11
â– cCH.9
«CH_15
cOU4
cCH.8
cCH_7
eCH.23
«CH_13
cCH.17
'568.4 1 jES.i 588.5 sB8_3 sB8_2 el*_17 eML15 cM_22 cMP_24 cfcr.e cMP.19 «**_? cfcT.22 cwr_5 •cTH_1 eTH_12 eTH.15 «TH_8 cTH.85 cTK-20 eTH_2”l
Fig 3.9: Maximum likelihood rooted phylogeny of all populations of Carex scirpoidea Michx. subsp. convolutci (Kiik.) Dunlop (blue, preceded by c) and subsp. scirpoidea (red, preceded by s) (Cyperaceae). Site abbreviations are as follows: HB = Horseshoe Bay, MI, USA; TH = Thompson’s Harbor, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CAN; CH = Cabot Head, ON, CAN; AM = Anvil Mountain, AK, USA; B = Butte, MT, USA; HCF = High Creek Fen, CO, USA; P = Pembina, MN, USA; ER = Escanaba River, MI, USA; BB = Brig Bay, NL, CAN. ASC GTRGAMMA nucleotide substitution model was used, and plotted with RAxML (Stamatakis, 2014). Bootstrap support indicated at nodes; 500 iterations.
60


Figure 3.10: Pairwise distance between loci using Nei’s distance (d), mean number of pairwise differences between populations, and within populations (9*) of using 7384 genomic loci in 11 populations of Carex scirpoidea Michx. subsp. convoluta (Kiik.) Dunlop (preceded by c) and subsp. scirpoidea (s) (Cyperaceae). Site abbreviations are as follows: HB = Horseshoe Bay, MI, USA; TH = Thompson’s Harbor, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CAN; CH = Cabot Head, ON, CAN; AM = Anvil Mountain, AK, USA; B = Butte, MT, USA; HCF = High Creek Fen, CO, USA; P = Pembina, MN, USA; ER = Escanaba River, MI, USA; BB = Brig Bay, NL, CAN.
61
Nei's distance (d) within populations between populations


Matrix of pairwise FSt
Figure 3.11: Pairwise Fst values for 7384 genomic loci in 11 populations of Carex scirpoidea Michx. subsp. convolutci (Kiik.) Dunlop (preceded by c) and subsp. scirpoidea (s) (Cyperaceae). Site abbreviations are as follows: HB = Horseshoe Bay, MI, USA; TH = Thompson’s Harbor, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CAN; CH = Cabot Head, ON, CAN; AM = Anvil Mountain, AK, USA; B = Butte, MT, USA; HCF = High Creek Fen, CO, USA; P = Pembina, MN, USA; ER = Escanaba River, MI, USA; BB = Brig Bay, NL, CAN.
62


Figure 3.12: Bayescan outlier loci for 11 populations of Carex scirpoidea Michx. subsp. convolata (Kiik.) Dunlop and subsp. scirpoidea (Cyperaceae) within Eastern North America. Square indicates loci above false discovery rate (FDR) of q=0.05. 11 loci were determined to be candidate loci under selection (above FDR). Populations include Horseshoe Bay, MI, USA; Thompson’s Harbor, MI, USA; Maxton Plains, MI, USA; Manitoulin Island, ON, CAN; Cabot Head, ON, CAN; Pembina, MN, USA; Escanaba River, MI, USA; Brig Bay, NL, CAN.
63


REFERENCES
Alexander, P.J., G. Rajanikanth, C.D. Bacon, and C.D. Bailey. 2007. Recovery of plant DNA using a reciprocating saw and silica-based columns. Molecular Ecology Notes 7: 5-9.
Allendorf, F.W., P.A. Hohenlohe, and G. Luikart. 2010. Genomics and the future of conservation genetics. Nature Reviews Genetics 11: 697-709.
Anacker, B.L. 2014. The nature of serpentine endemism. American Journal of Botany 101: 219-224.
Antonovics, J. 2006. Evolution in closely adjacent plant populations X: Long-term persistence of prereproductive isolation at a mine boundary. Heredity 97: 33-37.
Bagella, S., and M. Urbane 2006. Vascular flora of calcareous outcrops in North-Western Sardinia (Italy). Webbia 61: 95-132.
Baird, N.A., P.D. Etter, T.S. Atwood, M.C. Currey, A.L. Shiver, Z.A. Lewis, E.U. Selker, et AL. 2008. Rapid SNP discovery and genetic mapping using sequenced RAD markers. PLoSONE 3: 1-7.
Ball, P.W., and A.A. Reznicek. 2003. Carex scirpoidea Michaux subsp. scirpoidea (Kiikenthal) F. of N. A. E. Committee [ed.],. Flora of North America North of Mexico 23: 551. Available at: http://www.efloras.org.
Birchenko, I., Y. Feng, and J. Romero-Severson. 2009. Biogeographical Distribution of Chloroplast Diversity in Northern Red Oak (Quercus rubra L.). The American Midland Naturalist 161: 134-145.
Boisson, S., M.P. Faucon, S. Le Stradic, B. Lange, N. Verbruggen, O. Garin, A. Tshomba Wetshy, et al. 2017. Specialized edaphic niches of threatened copper endemic plant species in the D.R. Congo: implications for ex situ conservation. Plant and Soil 413: 261-273.
Bradshaw, A.D. 1991. The Croonian Lecture, 1991. Genostasis and the limits to evolution.
Philosophical Transactions of the Royal Society of London B Biological Sciences 333: 289-305.
Brooker, R., and R. Van Der Wal. 2003. Can soil temperature direct the composition of high arctic plant communities? Journal of Vegetation Science 14: 535-542.
Brownell, V.R., and J.R. Riley. 2000. The alvars of Ontario: Significant alvar natural areas in the Ontario Great Lakes region. Federation of Ontario Naturalists, Don Mills, Ontario.
Burge, D., T. Stoughton, and D. Jolles. 2017. Soil Chemistry Patterns In An Edaphic Endemism Hotspot: The Pebble Plains of the San Bernardino Mountains , California. Madrono 64: 9-21.
64


Carex Working Group. 2008. Field guide to sedges of the Pacific Northwest. Oregon State University Press, Corvallis, OR.
Castilho, P. Del, and I. Rix. 1993. Ammonium Acetate Extraction for Soil Heavy Metal
Speciation; Model Aided Soil Test Interpretation. International Journal of Environmental Analytical Chemistry 51: 59-64.
Catling, P.M. 1995. The extent of confinement of Vascular Plants to Alvars in Southern Ontario. Canadian Field Naturalist 109: 172-181.
Catling, P.M., and V.R. Brownell. 1995. Alvars of the Great Lakes Region. The Canadian Field-Naturalist 109: 143-171.
Chong, Z., J. Ruan, and C. Wu. 2012. Sequence analysis Rainbow : an integrated tool for efficient clustering and assembling RAD-seq reads. 28: 2732-2737.
Clausen, J. 1951. Stages in the evolution of plant species. Cornell University Press, Ithaca, NY.
Comer, P.J., D.L. Cuthrell, D.A. Albert, and M.R. and Penskar. 1997. Natural community abstract for limestone bedrock lakeshore. Michigan Natural Features Inventory, Lansing, MI.
Coop, G., D. Witonsky, A. Di Rienzo, and J.K. Pritchard. 2010. Using environmental correlations to identify loci underlying local adaptation. Genetics 185: 1411-1423.
Danecek, P., A. Auton, G. Abecasis, C.A. Albers, E. Banks, M.A. DePristo, R.E.
Handsaker, et al. 2011. The variant call format and VCFtools. Bioinformatics 27: 2156-2158.
Darriba, D., G.L. Taboada, R. Doallo, and D. Posada. 2015. Europe PMC Funders Group jModelTest 2 : more models , new heuristics and high- performance computing. 9: 6-9.
DePrenger-Levine, M.E. 2007. Genetic diversity in a rare North American endemic Carex scirpoidea ssp. convoluta. University of Colorado Denver.
Dillenberger, M.S., and J.W. Kadereit. 2013. The phylogeny of the European high mountain genus Adenostyles (Asteraceae-Senecioneae) reveals that edaphic shifts coincide with dispersal events. American Journal of Botany 100: 1171-1183.
Dunlop, D.A. 1990. The biosystematics of Carex section Scirpinae ( Cyperaceae ) The
biosystematics of Carex section Scirpinae ( Cyperaceae ). University of New Hampshire.
Dunlop, D.A., and G.E. Crow. 1999. The taxonomy of Carex section Scirpinae (Cyperaceae). Rhodora 101: 163-199.
Ellstrand, N.C., and D.R. Elam. 1993. Population Genetic Consequences of Small Population Size : Implications for Plant Conservation. Annual Review of Ecology and Systematics 24: 217-242.
65


Excoffier, L., and H.E.L. Lischer. 2010. Arlequin suite ver 3.5: A new series of programs to perform population genetics analyses under Linux and Windows. Molecular Ecology Resources 10: 564-567.
Fageria, N.K., H.R. Gheyi, and A. Moreira. 2011. Nutrient bioavailability in salt affected soils. Journal of Plant Nutrition 34: 945-962.
Fant, J.B., K. Havens, J.M. Keller, A. Radosavljevic, and E.D. Yates. 2014. The influence of contemporary and historic landscape features on the genetic structure of the sand dune endemic, Cirsium pitched (Asteraceae). Heredity 112: 519-530.
Feltz, C.J., and G.E. Miller. 1996. An asymptotic test for the equality of coefficients of variation from k populations. Statistics in Medicine 15: 647-658.
Fischer, J., and D.B. Lindenmayer. 2007. Landscape modification and habitat fragmentation: a synthesis. Global Ecology and Biogeography 16: 265-280.
Foll, M., and O. Gaggiotti. 2008. A genome-scan method to identify selected loci appropriate for both dominant and codominant markers: A Bayesian perspective. Genetics 180: 977-993.
Frankham, R. 2005. Genetics and extinction. Biological Conservation 126: 131-140.
Frankham, R. 2010. Where are we in conservation genetics and where do we need to go ? 661-663.
Frankham, R., C.J.A. Bradshaw, and B.W. Brook. 2014. Genetics in conservation management: Revised recommendations for the 50/500 rules, Red List criteria and population viability analyses. Biological Conservation 170: 56-63.
Garrison, E. 2012. Vcflib: A C++ library for parsing and manipulating VCF files. Available at: https ://github. com/vcflib/vcflib.
Garrison, E., and G. Marth. 2012. Haplotype-based variant detection from short-read sequencing. 1-9. Available at: http://arxiv.org/abs/1207.3907.
Gilpin, M.E., and M.E. Soule. 1986. Minimum viable populations: Processes of species extinction. In Conservation Biology: The Science of Scarcity and Diversity, 19-34.
Given, D.R., and J.H. Soper. 1981. The arctic-apine element of the vascular flora at Lake Superior. National Museums of Canada, Ottawa.
Gomez-FernAndez, A., I. Alcocer, and S. Matesanz. 2016. Does higher connectivity lead to higher genetic diversity? Effects of habitat fragmentation on genetic variation and population structure in a gypsophile. Conservation Genetics 17: 631-641.
Gorham, E. 1991. Northern peatlands: role in the carbon cycle and probable responses to climatic warming. Ecological Applications 1: 182-195.
66


Guindon, S., and O. Gascuel. 2003. A simple, fast and accurate method to estimate large phylogenies by maximum-likelihood. Systematic Biology 52: 696-704.
Hamilton, J.A., and C.G. Eckert. 2007. Population genetic consequences of geographic disjunction: A prairie plant isolated on Great Lakes alvars. Molecular Ecology 16: 1649-1660.
Hamrick, A.J.L., and M.J.W. Godt. 1996. Effects of life history traits on genetic diversity in plant species. Philosophical Transactions of the Royal Society of London B Biological Sciences 351: 1291-1298.
Hewitt, G.M. 2000. The genetic legacy of the Quarternary ice ages. Nature 405: 907-913.
Hufft, R.A., M.E. DePrenger-Levin, R.A. Levy, and M.B. Islam. 2018. Using herbarium specimens to select indicator species for climate change monitoring. Biodiversity and Conservation 27: 1487-1501.
Huson, D.H., and C. Scornavacca. 2012. Dendroscope 3: An interactive tool for rooted phylogenetic trees and networks. Systematic Biology 61: 1061-1067.
Jombart, T., and I. Ahmed. 2011. adegenet 1.3-1: new tools for the analysis of genome-wide SNP data. Bioinformatics 27: 3070-3071.
Krab, E.J., J. Roennefarth, M. Becher, G. Blume-Werry, F. Keuper, J. Klaminder, J. Kreyling, et al. 2018. Winter warming effects on tundra shrub performance are species-specific and dependent on spring conditions. Journal of Ecology 106: 599-612.
Kruckeberg, A.R. 1951. Intraspecific Variability in the Response of Certain Native Plant Species to Serpentine Soil. American Journal of Botany 38: 408-419.
Kruckeberg, A.R., and D. Rabinowitz. 1985. Biological Aspects of Endemism in Higher Plants Publis. Annual Review of Ecology and Systematics 16: 447-479.
DeLa Torre, A.R., D.R. Roberts, and S.N. Aitken. 2014. Genome-wide admixture and ecological niche modelling reveal the maintenance of species boundaries despite long history of interspecific gene flow. Molecular Ecology 23: 2046-2059.
Lee, K.N. 1999. Appraising Adaptive Management. Conservation Ecology 3: 1-18.
Li, H., and R. Durbin. 2009. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25: 1754-1760.
Lischer, H.E.L., and L. Excoffier. 2012. PGDSpider: An automated data conversion tool for connecting population genetics and genomics programs. Bioinformatics 28: 298-299.
Lowry, D.B. 2012. Ecotypes and the controversy over stages in the formation of new species. Biological Journal of the Linnaean Society 106: 241-257.
67


Ludwig, B., S. Chadde, and L. Prout. 1996. Establishment Record for Horseshoe Bay: Research Natural Area within Hiawatha National Forest Mackinac County, Michigan.
Martin, B. 2018. raxml_ascbias. Available at: https://github.com/btmartin721/raxml_ascbias/.
McLaughlin, W.T. 1932. Atlantic Coastal Plain Plants in the Sand Barrens of Northwestern Wisconsin Author. Ecological Monographs 2: 335-383.
Modesto, I.S., C. Miguel, F. Pina-Martins, M. Glushkova, M. Veloso, O.S. Paulo, and D. Batista. 2014. Identifying signatures of natural selection in cork oak (Quercus suber L.) genes through SNP analysis. Tree Genetics and Genomes 10: 1645-1660.
Moore, L., W.K. Lauenroth, D.M. Bell, and D.R. Schlaepfer. 2015. Soil, water, and
temperature explain canopy phenology and greeness on the Shortgrass Steppe. Great Plains Research 25: 121-138.
Olson, M.S., N. Levsen, R.Y. Soolanayakanahally, R.D. Guy, W.R. Schroeder, S.R.
Keller, and P. Tiffin. 2013. The adaptive potential of Populus balsamifera L. to phenology requirements in a warmer global climate. Molecular Ecology 22: 1214-1230.
Omernik, J.M. 1995. Ecoregions : A Framework for Managing Ecosystems. The George Wright Forum 12: 35-50.
Parchman, T.L., Z. Gompert, J. Mudge, F.D. Schilkey, C.W. Benkman, and C.A. Buerkle. 2012. Genome-wide association genetics of an adaptive trait in lodgepole pine. Molecular Ecology 21: 2991-3005.
Pembrook, J.W. 2014. Systematics of Carex section Scirpinae Tuck. (Cyperaceae), with insight to the origin of edaphic endemics. University of Colorado Denver.
Peterson, B.K., J.N. Weber, E.H. Kay, H.S. Fisher, and H.E. Hoekstra. 2012. Double digest RADseq: An inexpensive method for de novo SNP discovery and genotyping in model and non-model species. PLoS ONE 7: 1-11.
Puritz, J.B., C.M. Hollenbeck, and J.R. Gold. 2014. dDocent: a RADseq, variant-calling pipeline designed for population genomics of non-model organisms. PeerJ2: 1-14.
R Core Team. 2014. R: A language and environment for statistical computing. Available at: http://www.r-project.org/.
Raj, A., M. Stephens, and J.K. Pritchard. 2014. FastSTRUCTURE: Variational inference of population structure in large SNP data sets. Genetics 197: 573-589.
Rajakaruna, N. 2018. Lessons on Evolution from the Study of Edaphic Specialization. Botanical Review 84: 39-78.
Rajakaruna, N. 2004. The Edaphic Factor in the Origin of Plant Species. International Geology Review 46: 471-478.
68


Rellstab, C., F. Gugerli, A.J. Eckert, A.M. Hancock, and R. Holderegger. 2015. A practical guide to environmental association analysis in landscape genomics. Molecular Ecology 24: 4348-4370.
Reschke, C., R. Reid, J. Jones, T. Feeney, H. Potter, A.W. Group, T.N. Conservancy, and G.L. Program. 1999. Conserving Great Lakes Alvars. The Nature Conservancey Great Lakes Program, Chicago, IL.
Reznicek, A. A. 1994. The disjunct coastal plain flora in the great lakes region. Biological Conservation 68: 203-215.
Richter-Boix, A., M. Quintela, G. Segelbacher, and A. Laurila. 2011. Genetic analysis of differentiation among breeding ponds reveals a candidate gene for local adaptation in Rana arvalis. Molecular Ecology 20: 1582-1600.
Roda, F., L. Ambrose, G.M. Walter, H.L. Liu, A. Schaul, A. Lowe, P.B. Pelser, et al. 2013. Genomic evidence for the parallel evolution of coastal forms in the Senecio lautus complex. Molecular Ecology 22: 2941-2952.
Rosenberg, N. A. 2004. DISTRUCT : a program for the graphical display of population structure. 137-138.
Sandoval-Castillo, J., N.A. Robinson, A.M. Hart, L.W.S. Strain, and L.B. Beheregaray. 2018. Seascape genomics reveals adaptive divergence in a connected and commercially important mollusc , the greenlip abalone (Haliotis laevigata), along a longitudinal environmental gradient. Molecular Ecology 27: 1603-1620.
Schaefer, C., and D. Larson. 1997. Vegetation, Environmental Characteristics and Ideas on the Maintenance of Alvars on the Bruce Peninsula, Canada. Journal of Vegetation Science 8: 797-810.
Shackleford, R. 2003. Conservation assessment for Bulrush Sedge (Carex scirpoidea Michx.): Subspecies scirpoidea and convoluta (Kiikenth.) Dunlop. Milwaukee, Wisconsin.
Stamatakis, A. 2014. RAxML version 8 : a tool for phylogenetic analysis and post-analysis of large phylogenies. 30: 1312-1313.
Stark, K.E., J.T. Lundholm, and D.W. Larson. 2004. Arrested development of soil on alvars of Ontario, Canada: Implications for conservation and restoration. Natural Areas Journal 24: 95-100.
Stebbins, G.L., and J. Major. 1965. Endemism and Speciation in the California Flora. Ecological monographs 35: 1-35.
Subbarao, G. V., O. Ito, W.L. Berry, and R.M. Wheeler. 2003. Sodium - A Functional Plant Nutrient. Critical Reviews in Plant Sciences 22: 391-416.
69


Westergaard, K.B., S. Fior, L.P. Bruederle, H.K. Stenoien, N. Zemp, and A. Widmer.
2018. Population genomic evidence for plant glacial survival in Scandinavia. Molecular Ecology 1-45.
White, S.N., N.S. Boyd, and R.C. Van Acker. 2015. Temperature Thresholds and Growing-Degree-Day Models for Red Sorrel (Rumex acetosella) Ramet Sprouting, Emergence, and Flowering in Wild Blueberry. Weed Science 63: 254-263.
Willis, S.C., C.M. Hollenbeck, J.B. Puritz, J.R. Gold, and D.S. Portnoy. 2017. Haplotyping RAD loci: an efficient method to filter paralogs and account for physical linkage. Molecular Ecology Resources 17: 955-965.
Worch, S., K. Rajesh, V.T. Harshavardhan, C. Pietsch, V. Korzun, L. Kuntze, A. Borner, et al. 2011. Haplotyping, linkage mapping and expression analysis of barley genes regulated by terminal drought stress influencing seed quality. BMC Plant Biology 11: 1-14.
Zhou, Y., L. Zhang, J. Liu, G. Wu, and O. Savolainen. 2014. Climatic adaptation and ecological divergence between two closely related pine species in Southeast China. Molecular Ecology 23: 3504-3522.
70


Appendix:
A. Descriptions of soil and leaf tissue sampling sites for Carex scirpoidea Michx. subsp. convoluta (Kiik) Dunlop. Habitat descriptions and associated species observations made on site by Leo Bruederle and Nick Bard.
B. Weights for all soil samples taken from Carex scirpoidea Michx. subsp. convoluta (Kiik.) Dunlop (Cyperaceae) population sites before and after drying out and removal of rocks, if applicable. Numbers following the site name are the transect number, and the number of the sampling point along that transect.
C. Soil chemical and physical parameters measured across the ranges of Carex scirpoidea Michx. subsp. convoluta (Kuk.) Dunlop and subsp. scirpoidea (Cyperaceae).
D. Average daily soil temperature for all days determined to be in the growing season (>0°C) for all populations of Carex scirpoidea Michx. subsp. convoluta (Kiik.) Dunlop and subsp. scirpoidea (Cyperaceae). Due to missing data, extrapolations were generated for subsp. convoluta using quadratic GLMs, which comprise the first 5 plots.
E. Average number of alleles, and associated standard deviation across 7384 measured loci for 11 populations of Carex scirpoidea Michx. subsp. convoluta (Kiik.) Dunlop and subsp. scirpoidea (Cyperaceae).
F. Maximum likelihood unrooted phylogeny of all populations of Carex scirpoidea subsp. convoluta (Kiik.) Dunlop and subsp. scirpoidea (Cyperaceae).
71


Appendix A: Descriptions of soil and leaf tissue sampling sites for Carex scirpoidea Michx. subsp. convoluta (Kiik.) Dunlop (Cyperaceae). Habitat descriptions and associated species observations made on site by Leo Bruederle and Nick Bard.
Site State/Province Country Latitude & Longitude Elevation (in m) Habitat Description Associated Species
Cabot 45.244, 183.79 Open meadow Potentilla fruticosa,
Head, Ontario, CAN -81.304 with limestone pavement with abundant dense clumps of C. scirpoidea subsp. convoluta. Aster sp., Soli dago sp., Poa sp., Rosa sp., Pimis sp.
Surrounded by Pimis sp., little cedar. Grazed by pigs
possibly._____
Horseshoe Bay, Michigan, USA 45.987, -84.745 199.34 Cedar glade swamps, marl substrate, mostly open canopy, lots of dead cedars, grey soil even at top layer. Schizacryium scoparium, Picea glauca, Thuja occidentalis, Juniperus horizontalis, Arctostaphylos uva-ursi, Andromeda sp., Gentiana sp., Parnassea palustris, Potentilla fruticosa
Manitoulin 45.596, 181.66 Right on lake Viburnum triloba,
Island, -82.121 shoreline. Swale Schizacryium
Ontario, CAN between white scoparium,
spruce, cedar Juniperus
spruce glade and horizontalis,
limestone Potentilla fruticosa,
pavements. Open Fragaria
canopy. Like virginiana, Thuja
others, an occidentalis,
intermediate zone. Thalictrum sp.,
Arctostaphylos uva-
ursi.
72


Maxton 46.075, 174.65 Limestone Jimiperus
Plains, -83.6617 pavement with horizontalis, Rhus
Michigan, scattered grasses, radacans, Primus
USA exposed rocks, sp., Bryophyta sp.,
early colonizing Potentilla fruticosa,
mosses. Schizacryium scoparium
Thompson’s 45.351, 178.43 Open swale with Jimiperus
Harbor, -83.580 cedar. Intermediate horizontalis,
Michigan, zone, between Potentilla fruticosa,
USA cedar glade and Arctostaphylos uva-
lakeshore, between ursi, Carex flayva,
cedar glade and Shephardia
lake shore. Near canadensis,
trail. Common C. Gentiana sp., Thuja
scirpoidea subsp. occidentalis,
convoluta. Lots of Schizacryium
Great lakes scoparium, Aster
endemics. sp., Picea glauca or P. mariana, Rudbeckia sp.
73


Appendix B: Weights for all soil samples taken from Carex scirpoidea Michx. subsp. convoluta (Kiik.) Dunlop (Cyperaceae) population sites before and after drying out and removal of rocks, if applicable. Numbers following the site name are the transect number, and the number of the sampling point along that transect. Abbreviations for sites are as follows: MP = Maxton Plains, MI, USA; TH = Thompson’s Harbor, MI, USA; HB = Horseshoe Bay, MI, USA; CH = Cabot Head, ON, CAN; MI = Manitoulin Island, ON, CAN.
Site Before (g) After (g) Rock wt. (g) Water wt (g) h2o %
MP 1-3 146.4 108.92 0 37.48 26
MP 2-8 79.8 47.27 0 32.53 41
MP 3-8 88.59 51.6 0 36.99 42
TH 1-5 142.69 104.46 12.35 38.23 29
TH 2-7 175.69 129.27 14.48 46.42 29
TH 3-5 142.6 103.85 5.25 38.75 28
HB 1-1 261.53 151.41 0 110.12 42
HB 2-2 201.52 107.23 0 94.29 47
HB 3-7 288.28 176 0 112.28 39
CH 1-4 212.61 146.67 0 65.94 31
CH 2-5 185.75 125.61 0 60.14 32
CH 3-8 177.22 118.6 0 58.62 33
MI 1-5 185.91 123.03 0 62.88 34
MI 2-2 204.39 164.37 0 40.02 20
MI 3-4 225.9 169.29 0 56.61 25
74


Appendix C: All soil chemical and physical parameters measured across the ranges of Carex scirpoidea Michx. subsp. convoluta (Kuk.) Dunlop and subsp. scirpoidea (Cyperaceae). Asterisks (*) imply concentrations of exchangeable cations. The following sites were sampled: Horseshoe Bay, MI, USA; Maxton Plains, MI, USA; Thompson’s Harbor, MI, USA; Cabot Head, ON, CAN; Manitoulin Island, ON, CAN; Solvagtind, GRL; Kjelvatn, NOR; Pembina, MN, USA; Kjerringa, NOR; High Creek Fen, CO, USA; Eagle Summit, AK, USA; Twelve Mile Summit, AK, USA; Traill, GRL; Mestervig, GRL.
ID Solid matter % Loss on ignition % PH A1 mg/kg As mg/kg B mg/kg Ba mg/kg Be mg/kg C a mg/kg Cd mg/kg Co mg/kg Cr mg/kg Cu mg/kg
Thompson’s Harbor 97.3 19.29 7.36 5472 3.6 13.7 25 0.3 113647 0.5 1.6 5472 3.6
Thompson’s Harbor 96.5 22.51 7.43 5024 2.2 13.9 26.6 0.3 117583 0.5 1.4 5024 2.2
Thompson’s Harbor 96.4 23.27 7.43 5247 2.9 13 27.2 0.3 110451 0.6 1.5 5247 2.9
5 Horseshoe Bay 93.4 16.2 7.65 1166 2.4 12.4 43.3 0.3 362647 0.2 0.9 1166 2.4
*5 Horseshoe Bay 95.1 18.45 7.49 1452 2.8 13.2 41.8 0.3 345156 0.4 1.1 1452 2.8
s: o Horseshoe Bay 89.8 15.48 7.89 1285 3.5 10.8 45.6 0.3 385123 0.3 0.8 1285 3.5
d. & Maxton Plains 96.8 18.23 7.44 16062 2.8 24.6 71 0.4 34366 1.5 4.8 16062 2.8
pO s Maxton Plains 92.4 43.14 7.33 15695 5.8 38.4 91.9 0.5 48969 2.3 4 15695 5.8
Maxton Plains 90.1 51.34 7.25 15752 8.7 39.5 98.6 0.5 35557 2.8 5.1 15752 8.7
.£• Manitoulin Island 97.7 14.22 7.47 6779 2.5 9 43.7 0.3 24010 0.9 3.2 6779 2.5
S'} H Manitoulin Island 98 13.25 7.42 6667 1.3 9.1 37.9 0.3 25869 0.9 2.9 6667 1.3
a Manitoulin Island 98.1 12.88 7.38 7272 2.6 8.8 48.3 0.3 25297 0.8 2.9 7272 2.6
Cabot Head 95.5 23.58 7.53 25155 16.4 71.5 108.1 1 97238 2 9.5 25155 16.4
Cabot Head 95.8 25.66 7.64 16809 14.6 52.9 89.9 0.6 122812 1.7 7 16809 14.6
Cabot Head 95 27.35 7.59 24401 16 63.6 117.1 1 96239 2.3 10 24401 16
C/i


ID Solid Loss on PH A1 As B Ba Be Ca Cd Co Cr Cu
matter % ignition % mg/kg mg/kg mg/kg mg/kg mg/kg mg/kg mg/kg mg/kg mg/kg mg/kg
Twelve Mile Summit 86.3 80.18 5.86 7554 47.7 3 218.5 0.3 18376 4.4 29.9 11.9 107.9
Twelve Mile Summit 86.5 79.81 5.55 10154 23.1 4.9 182.6 0.3 18658 1.9 8.4 16.2 103.8
Twelve Mile Summit 86 80.25 5.92 10432 7.2 5.4 178.3 0.3 23194 0.9 4.6 14.4 101.2
Eagle Summit 84.7 83.02 5.88 8337 8.2 5.9 221.6 0.3 26133 1.8 4.6 10.4 52.3
Eagle Summit 85.1 79.92 6.32 9196 8.3 7.7 210.6 0.3 28669 1.6 4.2 11.8 57.6
Eagle Summit 83.9 82.68 6.33 8564 18.2 7.1 293.1 0.3 30040 2.4 12.8 9.1 46
High Creek Fen 88.6 68.65 7.51 8406 19.6 42.6 462.4 0.8 66410 3.1 4.2 11.7 8.9
High Creek Fen 92.8 35.92 7.59 9797 7.6 29 467.7 0.3 73106 1.7 2.5 8.1 4.9
High Creek Fen 92.4 35.65 7.66 2757 10.1 27.6 942.2 0.3 213495 2.6 2.2 1.9 3
Pembina 93.2 29.43 7.68 10916 2.1 34.2 190.3 0.3 125418 0.9 3.2 13.1 11
Pembina 93.9 26.12 7.62 11643 2.3 26.7 193.8 0.3 128043 0.9 3.4 13.9 9.4
Pembina 93.1 28.34 7.6 12180 2.4 36.2 176.4 0.3 82306 1 3.5 14.7 12.7
Mestervig 97.9 7.74 6.89 20494 1.8 13.7 252.1 0.9 4809 1.5 8.1 33.3 14
Mestervig 98.7 3.98 7.13 23389 1.9 18.9 255.1 1 3340 1.6 9.4 36.2 15.6
Mestervig 99.1 2.8 7.05 17098 1.6 12.8 180.7 0.7 2265 1.2 6.8 27 11.4
Traill 0 99.3 2.77 6.96 9813 1.6 4.3 57.3 0.4 3125 0.7 4.1 18.2 8.3
Traill 0 98.6 5.12 6.9 13573 1.7 5.8 89.6 0.6 4483 0.9 5.3 20.1 11.4
Traill 0 99 3.77 7.09 11529 1.7 4.8 67.5 0.5 3950 0.8 4.7 17.1 8.7
Solvagtind 95.4 15.07 6.37 16636 6.6 2.2 85.1 0.5 8267 1.5 10.9 27.8 38.7
Solvagtind 92.8 29.37 6.28 21895 3.1 2.8 118.3 0.3 11932 2 11.8 80.8 43.3
Solvagtind 95.1 16.14 6.42 24589 4.4 1.7 143.7 0.3 10922 2.2 13 91.4 58.8
Kjelvatn 97.5 5.51 6.14 23265 2.6 0.6 73.7 0.5 4244 2.7 15.9 37 18.9
Kjelvatn 97.3 7.78 6.18 18215 5.7 0.6 55.3 0.5 4499 2.4 14.5 29.7 21
Kjelvatn 97.7 6.74 5.88 17118 1.1 0.7 62 0.5 3411 2.2 11.8 27.5 19.5
Kjerringa 97.9 4.9 7.24 37238 0.7 1.4 57 1.1 20384 2.7 10.6 50.5 13.2
Kjerringa 97.4 8.54 6.35 34910 0.7 1.1 36.2 1 16037 2.9 11.1 44.3 11.7
Kjerringa 97.2 9.88 7.38 30731 1.6 2.8 33.4 0.9 42818 2.4 9.2 35.1 10.2
^1
On


ID Fe mg/kg Ga mg/kg K mg/kg Li mg/kg Mg mg/kg Mn mg/k Mo mg/kg Na mg/kg Ni mg/kg P mg/kg Pb mg/kg S mg/kg Sc mg/kg
Thompson’s Harbor 5066 2.2 1340 8.7 33656 103 0.8 226 6.8 274 12.1 787.7 1.2
Thompson’s Harbor 4772 2.2 1263 8.3 31546 115 0.7 179 5.7 302 14.2 732.3 1.2
Thompson’s Harbor 4966 2.2 1241 8.6 32016 109 0.8 172 5.7 300 16.4 714.5 1.1
5 a Horseshoe Bay 942 2.2 301 7.2 1664 111 0.9 62 2.9 142 9.4 6027 0.2
sj Horseshoe Bay 1105 2.2 338 6.5 1587 117 0.9 60 4.1 162 11.9 5698 0.3
s o Horseshoe Bay 1009 2.2 283 7.4 1616 120 0.8 59 3.2 152 11.6 5927 0.2
d. & pfi Maxton Plains 15913 5.2 3750 13.2 18887 346 0.7 313 12.1 1049 32.6 1074 3
Maxton Plains 16752 5.4 4275 13.5 14969 588 0.9 263 15.9 1800 80.7 2651 3
q Manitoulin Island 8553 2.7 1447 8.5 11999 193 0.5 291 9.9 284 19 466.9 2
3 Manitoulin Island 7877 3.5 1643 9.4 12731 146 0.6 311 8.7 255 13.6 457.1 2
i Manitoulin Island 8207 3.1 1874 9.1 12810 193 0.4 367 8.6 243 12.1 399.4 2.1
Cabot Head 27833 6.7 7312 23.3 55105 1519 0.8 320 17.6 646 20.5 903.1 5
Cabot Head 22416 4.4 5187 16 67185 1792 0.9 283 13.7 611 18.4 878.3 3.6
Cabot Head 27333 6.5 6258 20.9 53819 2091 0.8 283 19.3 750 27.2 1055 4.9
^1
^1


ID Fe Ga K Li Mg Mn Mo Na Ni P Pb S Sc
mg/kg mg/kg mg/kg mg/kg mg/kg mg/k mg/kg mg/kg mg/kg mg/kg mg/kg mg/kg mg/kg
Twelve Mile Summit 41762 3.5 936 2.6 1151 2146 1.2 106 32.9 1581 3.8 4066 3.1
Twelve Mile Summit 11271 2.2 1662 4.3 1771 2430 1.1 149 31.6 1754 5.6 3735 4.8
Twelve Mile Summit 7428 2.2 1488 4.6 1632 1208 0.4 152 32.5 1604 3.5 4274 5
Eagle Summit 8977 2.2 1299 2.8 1588 1877 0.9 139 23.3 1984 4.8 4704 3.1
Eagle Summit 7167 2.2 1664 3.2 1670 1776 0.6 126 20.2 2003 4.7 5276 4
Eagle Summit 11205 2.2 1494 3.1 1596 4822 0.9 114 22.1 1847 6 4274 2.8
High Creek Fen 25392 3.1 1782 7.9 10456 103 1.6 395 9 670 41.9 9560 1.1
High Creek Fen 21138 3.6 2196 8.6 7031 108 0.4 240 3.3 418 21 4198 1.2
High Creek Fen 39452 2.8 677 7.7 14506 205 0.4 309 1.5 459 18.1 3408 0.3
Pembina 9688 2.8 1717 11.3 14841 512 0.4 182 7.5 972 5.6 1387 2
Pembina 10219 3.2 1873 11.9 14622 420 0.4 194 7.8 883 5.8 1149 2.1
Pembina 10831 3.2 2030 11.5 17064 580 0.4 173 7.8 1070 6.2 1419 2.3
Mestervig 17583 6.7 6290 10.7 3423 357 0.4 212 19.5 500 9.1 460.9 4.7
Mestervig 20040 7.3 8468 10 3291 376 0.4 232 24.4 411 9.7 295.8 5
Mestervig 15418 5.8 5874 7.8 2312 278 0.4 179 16.7 303 8 174.8 3.7
Traill 0 9528 3.5 2296 6.9 2633 147 0.4 184 8.5 277 4.7 143.5 2.2
Traill 0 11887 4.4 2915 9.9 3532 218 0.4 237 11.8 311 6 232.2 3
Traill 0 10515 4 2571 8.1 3012 199 0.4 194 10.1 316 5.3 152.8 2.7
Solvagtind 18136 6.2 2865 15.3 4859 622 0.4 289 23.9 733 14.1 880.7 7.3
Solvagtind 24433 7.7 4759 25.3 12092 478 0.4 362 47.1 1114 9.3 1948 9
Solvagtind 27810 9.8 6809 28.7 14341 678 0.4 469 56.8 742 6.6 955.3 9.5
Kjelvatn 33448 9.8 3549 38.6 10979 582 0.4 226 27.6 580 7.6 443.4 6.4
Kjelvatn 29386 8.4 2448 34.3 7767 781 0.4 189 19.6 671 7.2 651.6 5.5
Kjelvatn 28142 7.9 2821 25.7 7207 547 0.4 178 18.5 543 8.4 506.3 4.9
Kjerringa 38752 11.2 4051 28.2 11190 353 0.4 2622 23.2 328 28.1 229 6.4
Kjerringa 41162 10.4 2411 26.3 9507 337 0.4 2168 21.1 372 83.8 380.8 5.9
Kjerringa 35630 8.4 2448 21.1 7925 308 0.4 2465 18.6 320 32 408.1 4.9
^1
00


ID Se mg/kg Sr mg/kg Ti mg/kg V mg/kg Y mg/kg Zn mg/kg C % N % C/N Vol. wt. g/1 Al* mg/kg B* ug/kg Ba* mg/kg
Thompson’s Harbor 2.2 54.3 113 11 3.4 29.3 11.08 0.453 24.49 971 2.2 257 4
Thompson’s Harbor 2.2 54.5 112 10 3.3 27.6 12.86 0.485 26.52 851 3.1 381 5.3
Thompson’s Harbor 2.2 53.7 124 9.9 3.3 34.4 12.74 0.493 25.83 858 2.8 216 5.5
Horseshoe Bay 2.2 0.1 31 1.8 0.6 20 15.73 0.488 32.25 640 1.8 996 4.3
Horseshoe Bay 2.2 829.4 36 2.3 1 23.8 16.26 0.537 30.27 629 0.7 1404 4.6
Horseshoe Bay 2.2 0.1 31 1.8 0.6 22 15.84 0.436 36.37 632 0.2 748 4.5
Maxton Plains 2.2 31.6 350 28.3 5.9 60.3 9.005 0.772 11.67 930 6 619 6.7
Maxton Plains 2.2 29.5 215 26.1 5.7 117.8 21.32 1.72 12.39 593 5.9 690 11.1
Maxton Plains 2.2 25 172 28.9 6.4 145.8 26.46 2.194 12.06 574 4.7 550 12.3
Manitoulin Island 2.2 32.9 236 18.2 5 62.5 6.884 0.315 21.88 959 3 366 8.6
Manitoulin Island 2.2 32 221 18.2 4.1 45.5 6.467 0.272 23.74 940 2.8 298 5.1
Manitoulin Island 2.2 33 202 19.2 4.1 47.3 5.844 0.242 24.19 922 2.7 326 8
Cabot Head 2.2 34.2 137 36.7 17 54.1 13.38 0.709 18.85 813 5.2 415 8.7
Cabot Head 2.2 36.4 111 26 13.3 40.1 15.06 0.631 23.88 768 5.1 537 10.1
Cabot Head 2.2 33.1 145 35.3 16 67.8 14.75 0.807 18.27 742 7.2 606 11.5
^1
VO


ID Se Ti V Y Zn Vol. Al* B* Ba*
mg/kg Sr mg/kg mg/kg mg/kg mg/kg mg/kg C % N % C/N wt g/1 mg/kg ug/kg mg/kg
Twelve Mile Summit 2.2 93.1 149 16.2 26.2 98.6 44.18 2.85 15.5 177 6.1 2790 77.5
Twelve Mile Summit 2.2 97 282 18.5 23 104.2 43.58 2.61 16.7 197 5.5 2394 49
Twelve Mile Summit 2.2 119.1 261 13.7 18.1 57.7 43.9 2.83 15.5 185 2.3 1923 56.5
Eagle Summit 2.2 156.7 172 12.4 26 58.4 45.54 2.83 16.1 207 4.7 1909 83.3
Eagle Summit 2.2 165.8 189 11.7 22.6 85.6 44.23 2.86 15.4 193 2.4 1688 70.1
Eagle Summit 3.1 176.2 149 12.5 21.1 90.2 45.27 2.77 16.4 202 5.5 1737 98.6
High Creek Fen 4.9 199.6 131 24.3 7.9 59.6 37.48 2.72 13.8 199 4.4 4491 194.9
High Creek Fen 2.2 156.7 138 10.9 4.5 25.8 18.32 1.38 13.3 416 1.7 1824 170.2
High Creek Fen 2.2 405.3 46 4.3 1 22.8 22.12 1.31 16.9 382 0.9 1614 194.4
Pembina 2.2 98.5 256 21.2 4.8 51.4 14.97 1.08 13.8 695 2 2105 37.5
Pembina 2.2 104.3 270 22.7 5.2 41 13.62 0.88 15.4 738 1.4 1757 34
Pembina 2.2 81.8 293 23.7 5.4 65.2 14.07 1.12 12.6 721 1.8 2095 39.5
Mestervig 2.2 13.3 386 31.2 10.4 45.9 3.24 0.26 12.6 985 6.4 576 118.3
Mestervig 2.2 11.7 328 34.7 10.3 48.6 1.42 0.12 11.5 1176 2 399 106.9
Mestervig 2.2 9.2 262 26.6 7.2 38.4 0.93 0.08 11.2 1308 3.6 527 102.3
Traill 0 2.2 14 367 20.9 8.1 22.5 1.1 0.1 10.9 1189 3.9 417 19.6
Traill 0 2.2 19.2 398 22.3 9.6 29.9 2.1 0.17 12.3 988 3.5 487 33.6
Traill 0 2.2 17.3 379 20.6 10.4 28.2 1.58 0.12 13.2 1084 4 468 24.1
Solvagtind 2.2 33.6 564 33.1 58.5 41.7 7.19 0.63 11.4 547 6.4 1803 16
Solvagtind 2.2 26.5 1014 49.2 49 61.7 13.01 1.12 11.6 510 4.1 1898 21.3
Solvagtind 2.2 28.8 717 57.8 63.7 58 7.75 0.58 13.3 670 3.8 1420 19.3
Kjelvatn 2.2 21.3 610 46.9 17.1 85.9 1.97 0.19 10.2 928 4.1 1188 11.2
Kjelvatn 2.2 22.8 714 41.6 18 66.4 3.29 0.3 11.1 910 3 1240 6.5
Kjelvatn 2.2 17.3 799 38.4 13.2 62.7 2.88 0.26 11.2 948 3.1 1339 8.2
Kjerringa 2.2 341.3 630 24.9 18.2 115.1 2.1 0.13 16.5 1016 4.9 1241 11.3
Kjerringa 2.2 293.2 636 24.5 15.7 209.2 3.89 0.26 14.8 822 4.8 1640 4.8
Kjerringa 2.2 434.1 904 16.4 16.4 110.4 5.21 0.28 18.4 818 2.7 2.2 434.1
00
o


ID Be* ug/kg Ca* mg/kg Cd* ug/kg Co* ug/kg Cr* ug/kg Cu* ug/kg Fe* mg/kg K* mg/kg Li* ug/kg Mg* mg/kg Mn* mg/kg Mo* ug/kg Na* mg/kg
Thompson’s Harbor 12 7458 42 14 20 277 1.74 56.4 140 312.3 4.14 193 6.2
Thompson’s Harbor 12 9499 40 14 20 277 2.25 80.8 149 433.8 6.66 189 6.6
Thompson’s Harbor 12 8748 36 14 60 277 2.15 77.3 163 404 5.19 202 6.5
5 a Horseshoe Bay 12 16008 38 14 96 277 2.28 37.9 397 221.9 4.43 339 10.3
sj Horseshoe Bay 12 16367 59 14 54 277 1.87 40.7 305 236.9 4.29 359 7.1
s o Horseshoe Bay 12 15627 44 14 20 277 1.66 32.7 343 83.46 4.91 409 6
d. & pfi Maxton Plains 12 5594 69 14 20 277 3.73 127.7 158 1209 2.98 233 5.8
Maxton Plains 12 13473 97 14 20 277 3.43 338.6 232 2965 10.23 295 7.6
q Manitoulin Island 12 3887 104 14 45 277 3.29 43.6 82 733 9.35 149 5.5
Manitoulin Island 12 3336 114 14 20 277 2.98 48.9 82 690.2 5.06 146 5.5
i Manitoulin Island 12 3211 77 14 20 277 2.91 59.3 82 652.5 9.03 127 7.3
Cabot Head 12 6157 35 14 93 277 3.1 145.6 200 1345 9.56 180 8.2
Cabot Head 12 6004 41 14 20 277 3.46 117.7 129 1240 10.45 180 7.1
Cabot Head 12 7292 50 14 128 277 4.17 152.9 202 1499 11.08 199 8.6
00


ID Be* Ca* Cd* Co* Cr* Cu* Fe* K* Li* Mg* Mn* Mo* Na*
ug/kg mg/kg ug/kg ug/kg ug/kg ug/kg mg/kg mg/kg ug/kg mg/kg mg/kg ug/kg mg/kg
Twelve Mile Summit 12 13724 279 148 20 277 28.65 263.4 1132 342.6 45.49 19 39.7
Twelve Mile Summit 12 13005 397 88 20 277 6.84 388.9 925 372 121.4 19 40.8
Twelve Mile Summit 12 16442 126 14 20 277 3.69 282.6 1103 374.4 25.85 19 50.3
Eagle Summit 12 18495 385 14 20 277 7.01 319.3 1282 626.4 96.06 19 55
Eagle Summit 12 20619 198 14 20 277 3.6 347 1247 630 17.7 19 42.7
Eagle Summit 12 22259 250 14 20 277 9.14 339.5 1678 663.3 31.08 19 36.3
High Creek Fen 397 28377 113 128 20 277 26.79 292 2380 4848 10.5 312 301.5
High Creek Fen 12 21925 26 14 20 277 17.7 96.8 1611 2681 6.43 19 119.9
High Creek Fen 12 26431 7 42 20 277 17.96 128.7 1933 3137 11 19 146
Pembina 12 15339 7 14 50 277 1.16 151.7 255 1560 2.69 19 20
.§■ Pembina 12 14175 7 14 20 277 1.14 156 220 1723 3.78 19 18.5
6-5 Pembina 12 14882 7 14 20 277 1.13 186.2 220 1523 2.08 19 17.4
Cm 3 â–  sa Mestervig 12 1223 6 14 20 277 2.76 31.2 147 55.32 0.88 19 13.4
B- â–  sa Traill 12 1186 7 14 20 277 2.53 24.6 106 57.9 0.84 19 8.9
65 8 Traill 12 2226 7 14 20 277 2.42 29.8 216 103.3 0.74 19 13.2
K o Traill 12 1630 7 14 20 277 3.21 30.2 163 69.68 0.58 19 8.5
Solvagtind 12 3705 34 14 20 604 1.87 122.6 82 143.4 4.56 19 34.2
Solvagtind 12 7432 48 14 20 277 2.28 161.8 82 218.2 9.91 19 35.4
Solvagtind 12 6046 16 14 20 463 2.41 102.4 82 161.4 4.57 19 26.6
Kjelvatn 12 2010 11 14 20 277 4.02 41.6 82 64.33 1.54 19 17.2
Kjelvatn 12 2428 17 14 37 277 4.07 38.4 82 50.35 1.55 19 14.3
Kjelvatn 12 1553 16 14 20 277 3.78 40.7 82 56.13 1.46 19 13.2
Kjerringa 12 3839 15 14 20 277 8.72 61 90 51.41 2.82 19 41.6
Kjerringa 12 2617 28 14 20 277 6.09 49.7 82 90.51 3.28 19 33.8
Kjerringa 12 6025 31 14 20 277 11.55 31.2 119 76.2 6.07 19 18
00
to


ID Ni* ug/kg p* mg/kg Pb* ug/kg S* mg/kg Se* ug/kg Si* mg/kg Sr* ug/kg Ti* ug/kg V* ug/kg Zn* ug/kg H* mg/kg Ionic exchange capacity* Degree of base saturation*
Thompson’s Harbor 20 9.87 349 15.2 239 15.7 7313 34 56 537 0 399.7 100
Thompson’s Harbor 20 13.12 506 15.2 239 20.1 8219 104 52 719 0 512.3 100
Thompson’s Harbor 27 11.5 487 16 239 17.6 8803 26 130 604 0 472.2 100
5 a Horseshoe Bay 20 8.36 519 247.1 239 10.8 77678 26 52 1343 0 818.7 100
s; Horseshoe Bay 46 10.96 527 230.8 239 11.2 81378 26 52 1889 0 837.7 100
s o Horseshoe Bay 48 7.55 519 164.9 239 7.8 53263 26 52 1307 0 787.9 100
d. & pfi Maxton Plains 20 8.17 400 15.4 239 25.4 3374 266 52 616 0 382.3 100
•a • Maxton Plains 20 27.05 758 33.2 239 24 5925 26 52 1729 0 925.6 100
Q Manitoulin Island 27 9.07 483 10.5 239 16.6 3436 163 52 3123 0 255.9 99.9
3 Manitoulin Island 20 8.32 323 13 239 18.2 2473 49 106 2026 0 224.9 99.9
i Manitoulin Island 25 8.91 232 10.9 239 17.9 2371 64 52 2534 0 216.1 99.8
Cabot Head 20 1.67 232 18.7 239 25.8 5272 180 106 190 0 422.3 99.9
Cabot Head 20 1.5 217 12.3 239 27.3 4308 156 52 299 0 405.3 99.9
Cabot Head 20 1.12 192 17.3 239 30.9 5791 291 124 277 0 491.9 99.9
00
u>


ID Ni* p* Pb* S* Se* Si* Sr* Ti* V* Zn* H* Ion exch. Deg. of
ug/kg mg/kg ug/kg mg/kg ug/kg mg/kg ug/kg ug/kg ug/kg ug/kg mg/kg Cap* base sat*
Twelve Mile Summit 689 16.28 476 91 239 71.7 73123 26 52 8353 326.3 1046 69
Twelve Mile Summit 573 22.21 426 89.4 239 86.5 69466 26 52 11176 404.8 1096 63
Twelve Mile Summit 536 21.75 363 81.2 239 60.2 90516 26 52 5596 335.4 1194 72.1
Eagle Summit 328 19.52 356 121.1 239 44.7 119673 26 52 6141 316.9 1302 75.6
Eagle Summit 167 26.35 426 115.7 239 60.2 129965 26 52 5246 202.9 1293 84.4
Eagle Summit 147 22.33 412 103.2 239 50.9 143503 309 52 5933 233.1 1407 83.5
High Creek Fen 20 13.1 893 516.1 239 155.1 108295 26 52 2763 0 1836 100
High Creek Fen 20 5.66 311 194.9 239 108.8 71746 26 52 1106 0 1323 100
High Creek Fen 20 4.13 312 265.7 239 158 90274 26 52 574 0 1587 100
â–  sa Pembina 20 11.41 243 17.2 239 94.8 19587 324 182 243 0 898.6 100
© Pembina 20 10.72 227 15.2 239 75.5 18557 26 52 256 0 854 100
6-5 Pembina 20 13.86 253 14.6 239 117.3 22243 26 52 198 0 873.5 100
Mestervig 20 0.33 69 3.2 239 11.9 2286 166 52 475 7.9 74.8 89.5
B- â–  sa O Traill 0 20 0.81 66 4.2 239 19 2961 227 52 968 4.8 69.8 93.1
s* 8 Traill 0 20 1.15 80 5.7 239 19.7 5400 198 52 654 17 137.7 87.8
K O Traill 0 20 0.93 70 4.1 239 22 3895 214 52 878 5.2 93.4 94.4
Solvagtind 273 5.53 432 29.8 239 45.9 5390 26 52 700 76 276.7 72.7
Solvagtind 132 8.8 402 48 239 40.3 9071 539 52 930 81.9 476 82.9
Solvagtind 231 3.87 235 22.6 239 23.5 7113 78 52 863 57.7 376 84.8
Kjelvatn 66 2.39 133 10.1 239 16.3 7480 235 52 873 53.8 160.7 66.8
Kjelvatn 57 2.45 121 12.7 239 8.8 9713 254 68 1028 54.2 180.6 70.3
Kjelvatn 54 2.77 127 11.3 239 9.3 5945 178 52 1728 59 142.2 58.9
Kjerringa 55 2.04 196 4.4 239 13.8 29120 166 52 445 5.6 204.8 97.2
Kjerringa 78 2.89 812 10.2 239 11.8 15495 166 75 753 44.5 184.9 76.1
Kjerringa 41 2.94 230 8 239 12.9 28438 41 52 670 0 308.7 99.9
00


Average temperature Average temperature
Appendix D: Average daily soil temperature for all days determined to be in the growing season (>0°C) for all populations of Carex scirpoidea Michx. subsp. convolata (Kiik.) Dunlop and subsp. scirpoidea (Cyperaceae). Due to missing data, extrapolations were generated for subsp. convoluta using quadratic GLMs, which comprise the first 5 plots.
Thompson’s Harbor, ON, CAN
Horseshoe Bay, Ml, USA
85


Average temperature Average temperature
Maxton Plains, Ml, USA
Day
Manitoulin Island ON, CAN
Day


Average Temperature Average temperature
Cabot Head, Ml, USA
o
CM
LO
in
0
50
100
200
150 Day
12 Mile Summit, AK, USA
250
300
87


Full Text

PAGE 1

POPULATION GENOMICS OF A RARE EDAPHIC ENDEMIC b y NICHOLAS WEHBY BARD B.S., Portland State University, 2013 A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirements for the degree of Master of Science Biology Program 2018

PAGE 2

ii © 2018 NICHOLAS WEHBY BARD ALL RIGHTS RESERVED

PAGE 3

iii This thesis for the Master of Science degree by Nicholas Wehby Bard has been approved for the Biology Program by Christopher S. Miller , Chair Leo P. Bruederle, Advisor Jennifer Ramp Neale Date: December 15, 2018

PAGE 4

iv Bard, Nicholas Wehby. (M.S. Biology Program) Population Genomics of a Rare Edaphic Endemic Thesis directed by Professor Leo P. Bruederle Abstract Range limited rare endemic species are threatened by loss of gen om ic diversity, which can lead to extirpation and extinction. The focal taxon of this stud y , the edaphic endemic Carex scirpoidea subsp. convoluta ( Cyperaceae) , exhibits such risk . The narrowly distributed subsp. convoluta is primarily limited to alvar soils on the n ortheastern shores of Lake Huron in North America. In contrast, C. scirpoidea subsp. scirpoidea is widespread and found in diverse habitats and soil types across n orthern North America and Greenland, with few populations in Norway and Russia. Both taxa are of conservation concern in all or part of their ranges, and the characteristi c alvar habitat of subsp. convoluta is considered to be imperiled. In this study, I measured genomic diversity in both taxa in order to provide insight about their continued survival and adaptation to future conditions. Finally, I assess ed the di fferentiation of subsp. convoluta from subsp. scirpoidea, and how edaphic niche differs between taxa. As an edaphic endemic, I expected a narrow edaphic niche in subsp. convoluta that is distinct from subsp. scirpoidea. I compare d niche and niche breadth f or several soil parameters between taxa . W hen multiple parameters were considered, neither taxon occupie d a distinct niche, while both taxa exhibit ed broad tolerance to key soil parameters, including cations. My results indicate d that exchangeable sodium niche differed between taxa ; subsp. convoluta soils consistently ha d a lower concentration than subsp. scirpoidea . The wide tolerance to edaphic conditions observed in C. scirpoidea has likely aided range expansion by occupying adverse soil conditions .

PAGE 5

v D ue to its limited distribution on fragmented alvar, subsp. convoluta was expected to have low gen om ic diversity, and be differentiated from subsp. scirpoidea. D ouble digest restriction associated DNA (d dRAD ) sequencing was used to compa re gen om ic diversity and evaluate di fferentiation between subsp. convoluta and subsp. scirpoidea. My results suggest subsp. convoluta does not exhibit lower gen om ic diversi ty; nor does it exhibit signs of inbreeding depression Ñ likely due to dioecy in C . scirpoidea. The results also suggest that differentiation of subsp. convoluta is at an early stage. The study supports e arlier findings that North American subsp. scirpoidea populations are derived from three glacial refugia in N orthwestern , W estern, and E astern North America . Carex scirpoidea subsp. convoluta arose following recolonization from E astern North American populations of subsp. scirpoidea , rather than from other refugia l populations . Finally, there is evidence for broad tolerance or local adapta tion between populations, which may aid the apparent adaptability to extreme edaphic conditions. Though surveyed populations maintain genomic diversity, persistence of populations is contingent upon the preservation of suitable habitat. I recommend that la nd managers prioritize habitat protection for discrete populations , particularly at the margins of the range . The fo rm and content of this abstract are approved . I recommend its publication. Ap proved: Leo P. Bruederle

PAGE 6

vi DEDICATION To Erika , Mom, Dad, and Gammy, for your unending love and support. To my beloved friend Colin Ward , for being an extraterrestrial treepunk prophet. Afterlife on loop.

PAGE 7

vii ACKNOWLEDGEMENTS Thank you to Dr. Leo P. Bruederle, for his invaluable expertis e, inspiration, and mentorship. To Dr. Christopher S. Miller and Dr. Jennifer Ramp Neale, my deepest gratitude for your guidance and counsel. Writing my thesis would not have been possible without the aid provided by Dr. Erin Tripp , Dr. Michael Wunder , Dr. Gregory Ragland , Dr. Kristine Westergaard , and Dr. Brad Slaughter. Finally, thank you to my peers who provid ed support with laboratory methods and data analysis : Andrew Boddicker , McCall Calvert, Kelsie Faulds, Kathryn Kilpatrick, Jared Mastin, Elizabeth Pansing , Allison Pierce, Laura Sedivy , Aaron Wagner , and Scott Yanco .

PAGE 8

viii TABLE OF CONTENTS CHAPTER I. ! A BIOLOGICAL OVERVIEW OF CAREX SCIRPOIDEA SUBSP. CONVOLUTA AND SUBSP. SCIRPOIDEA (CYPERACEAE) AND THE CONSERVATION OF TAXON AND HABITAT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Taxonomy, Ecology, and Conservation Status of Carex scirpoidea subsp. convolut a and subsp. scirpoidea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Conservation of Suitable Habitat. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Justification for Research. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Figure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 II. ! A COMPARISON OF THE SOIL CHEMICAL AND PHYSICAL CONDITONS FOR POPULATIONS OF CAREX SCIRPOIDEA SUBSP. CONVOLUTA AND SUBSP. SCIRPOIDEA (CYPERACEAE). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 Tables and Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 8 III. ! POPULATION GENOMIC S OF THE NARROW ENDEMIC CAREX SCIRPOIDEA SUBSP. CONVOLUTA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 6 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 7

PAGE 9

ix Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 8 Tables and figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 7 ! REFERENCES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 APPENDI X . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 A. ! Descriptions of soil and leaf tissue sampling sites for Carex scirpoidea subsp. convoluta (Cyperaceae). "!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"!"! "!"!"!! 72 B. ! Weights for all soil samples taken from Carex scirpoidea subsp. convoluta ( Cyperaceae ) population sites. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74 C. ! Soil chemical and physical parameters measured across the ranges of Carex scirpoidea subsp. convoluta and subsp. scirpoidea (Cyperaceae) . . . . . . . . . . . . . . . "!"!"!"!"!"!"!"!"!"!"!"!"! 75 ! D. ! Average daily soil temperature for all days determined to be in the growing season for all populations of Carex scirpoidea subsp. convoluta and subsp. scirpoidea (Cyperaceae) . 85 E. ! Average number of alleles and associated standar d deviation across 7384 measured loci for 11 populations of Carex scirpoidea subsp. convoluta and subsp. scirpoidea (Cyperaceae) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 ! F. ! Maximum likelihood unrooted phylogeny of all populations of Carex scirpoidea subsp. convoluta and subsp. scirpoidea (Cyperaceae) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

PAGE 10

x LIST OF TABLES TABLE 2.1 Soil sampling sites for Carex scirpoidea subsp. convoluta and subsp. scirpoidea (Cyperaceae) populations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .17 2. 3 Descriptive statistics of the parameters differentiating between taxa. . . . . . . . . . . . . 1 8 2. 2 Akaike Information Criterion (AIC) for models which predict taxon from discrete soil chemistry parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Tissue sampling sites for Carex scirpoidea subsp. convoluta and subsp. scirpoidea (Cyperaceae) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 3.2 Levels of genetic diversity for 11 populations of Carex scirpoidea subsp. convoluta and subsp. scirpoidea (Cyperaceae) based on 7384 genomic SNPs . . . . . . . . . . . . . . . . . 4 8 3. 3 . Locus by locus AMOVA results by taxon as weighted average over 7384 genomic loci of 11 populations of two taxa; 1) Carex scirpoidea subsp. convoluta and 2) subsp. scirpoidea ( Cyperaceae) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 3. 4 . Locus by locus AMOVA results by region as weighted average over 7384 genomic loci of 11 populations of three regions; 1) Northwestern, 2) Western, and 3) E astern North American . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 3. 5 . Pairwise differentiation (F st ) measured across 7384 loci of Carex scirpoidea subsp. convoluta and subsp. scirpoidea (Cyperaceae). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

PAGE 11

xi LIST OF FIGURES FIGURE 1. 1. Georeferenced occurrences for (a) Carex scirpoidea subsp. convoluta (Cyperaceae) and (b) subsp. scirpoidea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2. 1. Soil sampling sites representing populations of (a) Carex scirpoidea subsp. convoluta and (b) subsp. scirpoidea (Cyperaceae) ; (c) soil core used for collection, and typical core sample in alvar locations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 2. 2. PCA biplot for 57 measured soil parameters at sites of Carex scirpoidea subsp. convoluta and subsp. scirpoidea (Cyperaceae). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1 2. 3. Clustered dendrogram of soil site similarity across Carex scirpoidea subsp. convoluta and subsp. scirpoidea (Cyperaceae) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 2 2. 4. Plot of predicted probability of Carex scirpoidea subsp. convoluta and subsp . scirpoidea (Cyperaceae) from a generalized logistic regression model, wherein a soil parameter is the predictor variable. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 3 2. 5. Coefficient of variation for (a) soil cumulative growing degree days and (b) exchangeable Na measured in soil collected at 11 sites with populations of Carex scirpoidea subsp. convoluta and subsp. scirpoidea (Cyperaceae) . . . . . . . . . . . . . . . . . . 2 4 3.1 . Plant tissue sampling sites for (a) Carex scirpoidea subsp. convoluta and (b) subsp. scirpoidea (Cyperaceae) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 3. 2 . Transect sampling plan for each Carex scirpoidea subsp. convoluta (Cyperaceae) population site. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 3. 3 . Expected heterozygosity for 7384 genomic loci in 11 populations of Carex scirpoidea subsp. convoluta (Cyperaceae) subsp. scirpoidea . . . . . . . . . . . . . . . . . . . . . 54

PAGE 12

xii 3. 4 . Nucleotide diversity ( ! " ), measured across 7384 genomic loci for 11 popu lations of Carex scirpoidea subsp. convoluta and subsp. scirpoidea (Cyperaceae) . . . . . . . . . . . . . 55 3. 5 . Allele count for each locus across 7384 loci in 11 populations of Carex scirpoidea subsp. convoluta and subsp. scirpoidea (Cyperaceae) . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 6 3. 6 . fastStructure plots for strong structure and weak structure of all assayed populations of Carex scirpoidea subsp. convoluta and (b) subsp. scirpoidea (Cyperaceae) . . . . . . . . 57 3. 7 . BIC values for DAPC models with K clusters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 3. 8 . DAPC cluster plots for (a) Carex scirpoidea subsp. convoluta and (b) subsp. scirpoidea (Cyperaceae) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 3. 9 . Maximum likelihood rooted phylogeny of all populations of Carex scirpoidea subsp. convoluta and subsp. scirpoidea (Cyperaceae) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 3. 10 . Pairwise distance between loci using Nei's distance , mean number of pairwise differences between populations, and within populations using 7384 genomic loci in 11 populations of Carex scirpoidea subsp. convoluta and subsp. scirpoidea (Cyperaceae) . 61 3. 11 . Pairwise F ST values for 7384 genomic loci in 11 populations of Carex scirpoidea subsp. convoluta and subsp. scirpoidea (Cyperaceae) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 3. 1 2 . Bayescan outlier loci (candidate loci under selection) for 11 populations of Carex scirpoidea subsp. convoluta and subsp. scirpoidea (Cyperaceae) within Eastern North America. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

PAGE 13

xiii LIST OF ABBREVIATIONS 1 . AIC = Akaike Information Criterion Ñ a score given to statistical models to rank their predictive performance , relative to other models . 2 . BIC = Bayesian Information Criterion Ñ an alternative to AIC, with different parameter penalization. 3 . PCA = Principle Components Analysis Ñ a dimension reduction analysis used to interpret complex, multivariate relationships in data. 4 . SNP = Single Nucleotide Polymorphism Ñ a variant at one base at a locus in a genome.

PAGE 14

1 C HAPTER I A N OVERVIEW OF CAREX SCIRPOIDEA SUBSP. CONVOLUTA AND SUBSP. SCIRPOIDEA (CYPERACEAE) Taxonomy, Ecology, and Conservation Status of Carex scirpoidea subsp. convoluta and subsp. scirpoidea Carex L. section Scirpinae Tuck. (Cyperaceae) comprises perennial sedges, several of which have been reported to be edaphic (soil) endemics adapted to specialized soil types. The section is largely dioecious and easily identified by its typically unispicate inflorescence and pubescent perigyni a . Carex section S cirpinae includes Carex scirpoidea Michx . ( northern single spike sedge), which comprises four interfertile subspecies : Carex scirpoidea subsp. scirpoidea, C. scirpoidea subsp. convoluta ( KŸk . ) Dunlop, C. scirpoidea subsp. pseudoscirpoidea (Rydb . ) Dunlop , Carex scirpoidea subsp. stenochlaena (Holm) A. Lšve and D. Lšve, and Carex curatorum Stacey. This thesis focuses on C . scirpoidea subsp. convoluta and subsp. scirpoidea (hereafter, subsp. convoluta and subsp. scirpoidea ) , which are described as differing b ased only on leaf morphology and ecology: subsp. convoluta has narrow , V shaped leaves, with the widest leaves being narrower than its conspecific ( i.e., less than 1.5 mm) (Dunlop and Crow, 1999) . Phylogenetic relationships among taxa are largely unresolved (DePrenger Levine, 2007; Pembrook, 2014) . Carex scirpoidea subsp. scirpoidea has a wide distribution, spanning alpine and boreal habitats from East Russia across Canada, the northern USA, and Greenland, to Norway (Fig. 1. 1) (Dunlop and Crow, 1999; Carex Working Group , 2008; Flor a of North America , 2003) . In North America, it is found in the Northern Forests, Tundra, Taiga, and Hudson Plain ecoregions (Omernik, 1995) . It tolerates a broad range of edaphic conditions , including dry to moist

PAGE 15

2 substrates and calcareous, serpentine, and granodiorite soils. Populations occupy ridges, meadows, riparian stream areas, tundra, heathlands, and fellfields (Carex Working Group, 2008) . Carex scirpoidea subsp. convoluta has the most restricted range within the section . It is a rare habitat specialist and comprise s approximately 5 0 variously sized, but often small populations in the Upper Great Lakes Region of Michigan, USA and Ontario, CA N (Fig 1 .1 ) (Dunlop and Crow, 1999) . The taxon typically occurs on soils that are highly calcareous with annually fluctuating wet and dry conditions, often reaching saturated to dr ought levels (Reschke et al., 1999; Shackleford, 2003) . It dwells pr imarily on alvar, so called limestone pavement, cobble beaches, and fens along or in close proximity to the shoreline of northern Lake Huron , including the North Channel and Georgian Bay (Fig. 1 .1 ) ( Shackleford, 2003) . Although subsp. scirpoidea is common throughout much of its range and globally ranked as secure (G5T5) , it is possibly extirpated (SH) in New Brunsw ick, CAN, and imperiled (S2) or critically imperiled ( S1 ) in certain states at the southern edge of its range in the USA (e.g., New York, Vermont , Wyoming , Washington ) , but globally ranked as secure ( G5T5 ). In Michigan, C. scirpoidea has been designated a "threatened" species, although subsp. convoluta has been designated vulnerable by NatureServe in Ontario. Additionally, the USFS speculates that climate change could be a "significant threat to habitat of C. scirpoidea subsp. scirpoidea and subsp. convolut a in the contiguous United States" (Shackleford, 2003) . Conservation of Suitable Habitat Worldwide, anthropogenic land use directly causes degradation and fragmentation (e.g. wilderness to agriculture conversion, vegetati on clearing), which ma y lead to drastic loss of suitable habitat (Fischer and Lindenmayer, 2007) . Less common habitats (e.g., wetlands, alvar) a re especially at risk, portending tenuous survival for rare endemic taxa, such as subsp.

PAGE 16

3 convoluta (Reschke et al. 1999; Shackleford 2003) . Alvar, the primary habitat for subsp. convoluta , is a rare community ty pe found only in Europe (e.g., Sweden , Estoni a) and in or adjacent to the Great Lakes region in North America (Schaefer and Larson, 1997; Brownell and Riley, 2000) . Alvar is defined by its often treeless landscape that is almost exclusively composed of graminoids , forbs, and shrubs. However, the most definitive characteristic of alvar communities is its distinct edaphic conditions. Made up of marble or limestone pavement covered by a characteristically thin soil layer atop limestone or dolostone bedrock, alvar pro vides harsh conditions unsuitable for many plant species, while supporting many endemics and disjunct taxa (Reschke et al. , 1999 , Catling and Brownell , 1995). Soils are characterized by high alkalinity, silica levels, organic content, nitrogen, calcium, an d magnesium levels, and low phosphorous content (Catling and Brownell, 1995; Brownell and Riley, 2000; Stark et al., 2004) . Alvar is also subject to extreme hydrological conditions, often with seasonal fluctuations ranging from drought to flood conditions (Reschke et al. 1999) that alvar endemics can tolerate. The flora of Great Lakes alvar includes many disjunct relictual populations originating from boreal ( Carex richardsonii R. Brown , Poa alpina L. ) , southern ( Thaspium barbinode [ Michx. ] Nutt , Valerianella umbilicata Sull. Alph. Wood. ) , and prairie communities ( Hymenoxys herbaceae Greene , Iris lacustris Nutt. ) that established following Wisconsinan glaciation (Catling and Brownell, 1995; Reschke et al., 1999) . Alvar is threatened by anthropogenic activities (e.g., off road vehicle activities, non native species), and broadly listed as " imperiled" at a global (G2) and statewide scale (S2 Michigan ; Shackleford , 2003). Additionally, "open alvar" has been assigned as globally imperiled (Catling and Brownell , 1995). Other threats to alvar communities and soils include quarrying, logging, and rutting from vehicular travel (Reschke et al. , 1999) , direct pollution of

PAGE 17

4 adjacent waters, hydrological changes, and overgrazing (Comer et al., 1997; Reschke et al., 1999) . Like alvar, fens are also threatened by anthropogenic hydrological changes, as well as automotive and human trampling (Shackleford , 2003). Fens constitute one of several habitat types for subsp. scirpoidea, and a minor proportion of the total available habitat for Carex scirpoidea subsp. convoluta (Shackleford , 2003). Fens are low ox ygen endogenously fed peatlands, in which Carex is known to comprise a significant portion of detrital peat (Gorham, 1991) . Edaphic substrates, such as those characteristic of alvar and fens, place high selective pre ssures on plants ; as such, the study of edaphic endemics can also provide valuable insight into the role of soil and other substrates on population differentiation (Rajakaruna, 2004) . Justification for Research Adaptive ecosystem management stipulates that ecological research and scientific evidence be intertwined with management and conservation efforts (Lee, 1999) . As a nthropogenic climatic and habitat changes threaten plant populations , population decline may reduce genetic diversity and adaptive variation, in particular . Thus , conservation managers should prioritize assessments of genetic diversity , in order to e lucidate threats to rare organisms, and estimate the likelihood and efficacy of adaptive change in small endemic populations under changing environmental conditions (Birchenko et al., 2009; Allendorf et al., 2010; Worch et al., 2011; Frankham et al., 2014; Modesto et al., 2014) . Integrating genetic analyses with ecological assessment provides necessary information concerning rare endemic plants , including the range of acceptable conditions and extirpation risks .

PAGE 18

5 Figure 1 .1 : Georeferenced o ccurrences for: (a) Carex L. scirpoidea Michx. subsp. convoluta (KŸk) Dunlop and (b) C. scirpoidea subsp. scirpoidea (Cyperaceae) . Locations were m apped in Fulcrum using data from SEINet, iDigBio, and expert knowledge (Leo P. Bruederle pers. comm., Anton Reznicek pers. comm., Brad Slaughter pers. comm. ) . a b

PAGE 19

6 CHAPTER II EDAPHIC NICHE OF CAREX SCIRPOIDEA SUBSP. CONVOLUTA AND SUBSP. SCIRPOIDEA Abstract Edaphic endemism, wherein organisms are restricted to distinct soil types, has contributed to the evolution of plant diversity. Alvar is a habitat type with a unique edaphic substrate Ñ often consisting of limestone or dolostone geological parent material, with calcareous, alkaline chemistry Ñ and supports a number of endemic plant species , including Carex scirpoidea subsp . convoluta . However, it is unclear the extent to which soil chemistry differs between subsp. convoluta and C. scirpoidea subsp. scirpoidea . In this study, I characterized the edaphic niches of subsp. convoluta and subsp. scirpoidea by measuring 66 physica l and chemical soil parameters across 14 sample sites. My analyses did not indicate that either taxon could be characterized by a distinct soil type, although subsp. convoluta site soils had consistently lower exchangeable sodium concentrations , and longer growing seasons with hotter soil temperatures. Additionally, the subsp. convoluta population located in an opening in a cedar swamp ( Horseshoe Bay, MI, USA ) had the most distinct conditions of all sites surveyed. These findings reveal a wide r than expected edaphic niche for subsp. convoluta in comparison to subsp. scirpoidea . F uture studies should determine the extent of local adaptation in subsp. convoluta . Introduction Edaphic endemism, wherein organisms are adapted and isolated to soils with unique physical and chemical properties, has resulted in high levels of plant diversity (Kruckeberg, 1951; Rajakaruna, 2018) . Two distinct scenarios are thought to have cont ributed to the origin of

PAGE 20

7 edaphic endemics. A widespread taxon may colonize a variety of habitats, eventually specializing on a unique soil type. Alternatively, edaphic endemic populations may be the only remaining populations of a once generalist taxon (Kruckeberg, 1951; Anacker, 2014) . Much edaphic endemism theory derives from research on serpentine soil endemics (Burge et al., 2017; Rajakaruna, 2018) ; soil types boasting high endemism (e.g., calcareous, metal mine tailings ) have received less attention (Antonovics, 2006; Bagella and Urbani, 2006; Burge et al., 2017; Rajakaruna, 2018) . Dunlop (1990 ), as part of her monograph on Section Scirpinae , suggested that Carex scirpoidea subsp. convoluta (hereafter, subsp. convoluta ) is an edaphic ecotype of open limestone pavement. The taxon is strongly associated with G reat Lakes alvar Ñ a habitat type with limited distribution and moderate geographic disjunction, often comprised of patchy mosaics (Catling, 1995) . Carex scirpoidea subsp. convoluta affiliates with several alvar communit y types , specifically, little bluestem alvar, creeping juniper shrubby cinquefoil alvar, and white cedar Ð jack pine shrubby cinquefoil alvar (Reschke et al. 1999 ) . Each alvar community type varies with respect to species composition, exposed bedrock, and vegetative cover; however, these communities are typically characterized by limestone or dolostone bedrock, thin soil, and low to moderate canopy cover (Reschke et al., 1999) . The soil chemistry of alvar is characterized as having high alkalinity , silica levels, organic content, nitrogen (N) , calcium (Ca) , and magnesium (Mg) levels, but low phosphorous (P) content (Stark et al. , 2004 , Brownell and Riley , 2000). Comparing alvar with adjacent forest communities, Schaefer (1997) found that ph ysical properties (e.g. , soil depth, exposed rock, sub surface temperature) differed and that potassium (K) levels were lower, with no apparent difference in Mg, P , Ca , ammonia, or moisture content.

PAGE 21

8 In contrast, Carex scirpoidea subsp. scirpoidea ( hereaft er, subsp. scirpoidea ) is broadly distributed and occupies a range of soil types, including wet, dry, calcareous, serpentine, and acidic soils. Common habitats include meadows, rock cracks, and fens at high latitudes (see Chapter I) . Though habitat and soi l types are highly variable, all have alpine arctic elements (Given and Soper, 1981) . The affinit y of subsp. convoluta to alvar, coupled with limited geographic range, suggests that this taxon occupies a narrow ecological niche . However, the physical and chemical properties of these soils have not been previously examined across the range of this taxon. Here, I examine and compare the edaphic niche for subsp. convoluta , a narrow endemic, relative to a widespread conspecific taxon, subsp. scirpoidea. In this study, I hypothesize that: 1) due to the narrow range and strong association with alvar, subsp. convoluta has a narrower edaphic niche than its conspecific, and 2) subsp. convoluta has a distinct edaphic niche for several soil parameters, includin g Ca, Mg, P, K, and pH Ñ those associated with limestone open pavement soils (Stark et al., 2004) . Methods Sampling Fourteen field sites were chosen representing much of the rang e of both taxa (Fig. 2. 1, Table 2. 1). Soil chemistry and soil temperature data were collected for nine sites supporting subsp. scirpoidea populations by Westergaard et al . (in press), while I collected the same data for five populations of subsp. convoluta (Appendix A). At each subsp. convoluta site, three 21 m transects were established 6m apart, with points marked every 3m. Soil samples and temperature measurements were obtained at one randomly chosen point per transect (three per site), adjacent to the n earest subsp. convoluta individual. At each point, a soil temperature logger (Thermo

PAGE 22

9 Button 22L 1199Y85) was placed ~10 15 cm deep, and programmed to record four temperature measurements throughout the day (once per six hours). It was often necessary to co mbine multiple soil cores d ue to the thin layer of soil at most sites . A soil sample of approximately 100g was obtained at each sample point by. Leaf litter and other debris, plants, roots, and stones were subsequently removed. Soil samples were kept on ice in the field and later frozen in the lab. Soil samples were air dried (~24 hours) in the laboratory, with soil weights recorded both before and after drying (Appendix B). Dried soil samples were sent to the Norwegian Institute of Bioeconomy Research ( https://www.nibio.no/en) for analysis. Estimates were obtained for 65 chemical and physical parameters; additional ammonium acetate tests were performed to measure exchangeable cation content (Castilho and Rix, 1993) . Soil sampling occurred in 2014 (subsp. scirpoidea ) and 2016 (subsp. convoluta ) , with subsequent removal of soil temperature loggers in 2016 and 2017, respectively. Soil sampling for subsp. scirpoidea was conducted by Kristine Westergaard and Leo P. Bruederle , with loggers deposited and soil cores taken at the edges and middle of eac h population. Soil temperature analysis Soil temperature magnitude and duration was characterized with cumulative growing degree days (CGDD). However, loggers failed to take readings at later dates of the 2017 growing season at subsp. convoluta sites. To e xtrapolate missing subsp. convoluta data, Intercept, Linear, and Quadratic generalized linear models (GLMs) were created using per day averages of the four daily temperature measurements ( ! "#$ % . At each site, the best fitting model was chosen based upo n the optimal Akaike Information Criterion (AIC). The model was used in lieu of raw data points for CGDD estimation at subsp. convoluta sites (Appendix D). However, ! "#$ & was calculated from raw data for the complete 2015 growing season for subsp. scirp oidea sites. The

PAGE 23

10 base temperature ' ! (")* % & was 0¼C, the observed minimum temperature required for growth and development in temperate and arctic zones, specifically shortgrass steppe and alpine (Moore et al., 2015; White et al., 2015) . However, a caveat was applied in order t o limit errant measurements near 0¼C due to Thermo Button Temperature Data Logger accuracy limits (±0.5¼C ) . M easurements were included between the first point above 0¼C Ñ such that subsequent values showed an upward slope Ñ and the last point above 0¼C ; although remaining va lues that increased up to 0.5 ¼C were discarded . The following equation for CGDD (Hufft et al., 2018) was implemented : +,-. & ! "#$ / 0 1 2 3 ! (")* . Soil chemical and physical parameters Of the soil chemical and physical parameter s (hereafter, parameters) measured, those with no variance or in which >50% were a single value were filtered out (e.g., selenium). Percent so li d matter was also filtered out. A principle components analysis (PCA) on all remaining variables was conducted a fter scaling and centering, using prcomp in R (R Core Team, 2014) . A dendrogram using the PCA results was also created with hclust in R using the "complete method," which determines the maximum Euclidian distance between points in different clusters. To determine which parameters varied most widely between subsp. convoluta and subsp. scirpoidea, several parameters were incorporated as the Ôpredictor variables' in a GLM binomial model to describe the Ôdependent' taxon varia ble. Schaefer and Larson (1997), Stark et al. (2004) and Burge et al. (2017) Ñ studies on alvar soil and edaphic endemism Ñ were consulted to select relevant soil parameters to be tested in the model. All macronutri ents and micronutrients were also included in the model. CGDD, loss of ignition, and carbon were used in lieu of subsurface temperature (Schaefer and Larson, 1997) , soil organic matter , and inorganic carbon

PAGE 24

11 (Stark et al., 2004) , respectively. Where applicable, exchangeable cation measurements were used, as well. For each taxon, descriptive statistics were calculated for all predictor variables. To estimate niche breadth differences between taxa, differences in variation for each parameter with the optimal AIC scores were tested using the asymptotic test for the equality of the coefficient of variation (ATECV ; Feltz and Miller, 1996) . Results Estimates were retained for 65 soil parameters, of which seven (i.e., selenium (Se) , exchangeable beryllium , exchangeable c o balt , exchangeable S e, solid matter, molybdenum ) were filtered out (Appendix C ), and CGDD was added. CGDD was calculated from a best fitting soil temperature model for subsp. convoluta ; for all subsp. convoluta sites , t he quadratic GLM was the best fitting model (Appendix B). The PCA and dendrogram (based on 59 parameters) did not reveal strict taxonomic differences based on soil type (Fig 2 .2 , 2.3 ). Instead, sites clustered geographically (e.g., Norway , Greenland, and western North America; Fig 2.2 ). Notably, the Upper Midwest ern North American site harboring subsp. scirpoidea (i.e., Pembina) and subsp. convoluta sites exhibited similar soil chemistry , with the exception of Horseshoe Bay, MI. Horseshoe Bay , t he site of a subsp. convoluta population , occupies an op ening in a cedar swamp , which is an anomaly, both with respect to the community and soil chemistry, which is distinct from all other sites studied (Appendix C) . In general, estimates for soil parameters were highly variable in one or both taxa, making nich e characterization difficult. For instance, in some cases, standard deviation for a parameter exceeded the average value ( # ) (e.g., Ca, Cu; Table 2 .2 ). For nearly all parameters, the values for one taxon overlapped with those of the conspecific, although e stimates for subsp. scirpoidea typically exhibited larger ranges (Table 2 .2 ). Though unexpected, the average concentration and

PAGE 25

12 range of total Ca ( # = 129664.27 ± 127122 mg/kg in subsp. convoluta vs. 36234.59 ± 50148.31 mg/kg in subsp. scirpoidea , range = 36113 mg/kg vs. 211230 mg/kg ) and Mg ( 24898.13 ± 20581.2 vs. 7111.04 ± 5141 mg/kg, range = 65598 mg/kg vs. 15913 mg/kg ) was larger for subsp. convoluta, as a result of high Ca and low Mg content at the Horseshoe Bay, MI site (Table 2 .2 , Appendix C). Howeve r, the range s for exchangeable Ca and Mg w ere higher in subsp. scirpoidea populations. Additionally, s oils at subsp. convoluta sites were more alkali ne ( pH = 7.49 ± 0.16 vs. 6.73 ± 0.63 , Table 2 .2 ) , had lower N (0.7 ± 0.54% vs. 1.11 ± 1.08%) , and total P ( 579.80 ± 345.73 mg/kg vs. 843.19 ± 573.75 mg/kg) . Sites for each taxon exhibited approximately similar concentrations of total K (2724.40 ± 2283.70 mg/kg vs. 2940.48 ± 1924.90 mg/kg). Exchangeable sodium (Na) best predicted taxonomic outcome (subsp. convol uta vs. subsp. scirpoidea ), providing a relative model weight of 100% (Table 2.3 ), with consistently higher levels at subsp. scirpoidea sites, even at the otherwise parametrically similar Minnesota subsp. scirpoidea sites (Fig. 2. 4, Appendix C ). CGDD had a relative model weight of 0.8% (Table 2.3 ). Other predictors did not have a measurable effect on the model (Akaike weight < 0.1%). Estimates of exchangeable Na were especially high at the High Creek Fen, CO site, where the average was over thr ee times greater than the next closest average (Eagle Creek, AK). As indicated by the ATECV results (p < 0.05, Fig. 2. 5), subsp. scirpoidea also had a broader range of exchangeable N a and CGDD values. Discussion Soil parameter suite The goal of this aspec t of my thesis research was to characterize and compare edaphic niche between subsp. scirpoidea and subsp. convoluta , which have been presumed to have broad and narrow edaphic niches, respectively. Variation for the suite of soil parameters studied ,

PAGE 26

13 indica t ed that edaphic niche did not differ substantially between taxa (Fig 2 .2 ). Instead, edaphic variation corresponded with geography ; more variability occurr ed among subsp. convoluta sites than between sites of subsp. convoluta and subsp. scirpoidea sites in close proximity (e.g., Pembina, MN; Fig s . 2 .2 , 2. 3). While alvar is a distinctive habitat, the soil parameter suite for the alvar endemic (Figs. 2 .2, 2. 3) was not unequivocally distinct. However, Schaefer and Larson (1997) found that physical, not chemical characterisics (e.g., soil depth, photosynthetically active radiation, wood y debris cover) comprised the majority of parameters that differed between alvar and adjacent forest habitats. However, chemical parameters comprised the majority of soil measurements in my study (Table 2.2 , Appendix C). It is possible that parameters not measured here, such as microclimate, radiation, and plant species composition (e.g., % lichen cover) better describe nic he divergence between taxa . Finally, drought periodicity niche may differ between tax a ; Dunlop (1990) speculated that drought tolerant morphology may in fact be a distinguishing feature of subsp. convoluta. My findings show that both taxa surveyed in the s ection have broad niche with respect to nearly all soil parameters (Range > # in 20/24 parameters in subsp. convoluta; 22/24 parameters in subsp. scirpoidea ). Strikingly, subsp. convoluta sites exhibited a higher total Ca and total Mg niche breadth ( Table 2. 2). The higher average pH, higher concentrations of total Ca and Mg , and low er concentrations of total P in subsp. convoluta align with the chemical characterization of alvar by Stark et al . ( 2004) . However, many exceptions were noted in one or more subsp. scirpoidea population sites, indicating that the niche of subsp. convoluta did not unequivocally differ from subsp. scirpoidea with respect to pH, or total Ca, Mg, K, or P.

PAGE 27

14 Horseshoe Bay, MI accounted for much of the edaphic niche breadth exhibited by subsp. convoluta (Figs. 2. 2 , 2. 3) . Th is sampling site, the only one in a forested cedar swamp, had a marl like soil visibly distinct from the rest. Soils at this site also had the highest water content, and lowest CGDD for any subsp. convoluta site, possibly due to microclimate e ffects in this forested habitat (Appendix A, B). Additionally, many macro and micronutrients ( e.g., total K, P, Ni, Fe, and Zn) were at lower concentrations than other subsp. convoluta sites, though total Ca was much higher (Appendix C). However, like sub sp. convoluta alvar habitats, Horseshoe Bay , MI is underlain by dolomite, albeit with greater soil depth (Ludwig et al., 1996; Schaefer and Larson, 1997) . Evidently, subsp. convoluta is not confined to a narrow range of soil types, and can withstand saturated, nutrient poor conditions. Association with variant, adverse soil types may indicate poor interspecific competitive ability (Boisson et al., 2017; Kruckeberg and Rabinowitz, 2017; Rajakarun a, 2018) . As such, c olonization of alvar by competing plants (e.g., shrubs) is known to be prevented by grazing and long periods of drought and flood (Schaefer and Larson, 1997; Stark et al., 2004) . Dunlop (1990) reported other edaphic ecotypes within C. scirpoidea , suggesting that the taxon is highly adaptive edaphically, which has likely contributed to diversification across it s range. My results support this, although I suggest that broad niche is also exhibited at a finer taxonomic level Ñ within the most restricted ecotype, subsp. convoluta . Thus, colonization of variably adverse soils has aided pers istence of C. scirpoidea ecotypes. Assessment of taxonomic predictor parameters Soil exchangeable Na niche and niche breadth differed consistently between taxa ( ! = 6.99 for subsp. convoluta, ! = 45.10 for subsp. scirpoidea , Table 2 .2 ; ATECV, p < 0.05 , Fig. 2. 5 ; Appendix C). Carex scirpoidea subsp. convoluta occurs in sites with low exchangeable N a

PAGE 28

15 Ñ with little variation among sites (r ange = 719.03 mg/kg, Table 2. 2 ) Ñ while soils for subsp. scirpoidea exhibit highly variable exchangeable N a content amon g sites ( r ange = 2516 mg/kg). This finding supports documented occurrences of subsp. scirpoidea on saline margins of prairies and fens in Minnesota and Colorado , respectively (Shackleford, 2003 ) . While subsp. scirpoidea has a larger exch a ngeable Na niche breadth, it is unclear whether it has a broader tolerance than its con specific, or if it is simply an artifact of its distribution. Physiologically, it is unclear whether Na is a limiting fac tor for plant growth, like P or Ca (Fageria et al., 2011) ; however, it is known to contribute to plant meta bolism, particularly as a substitute for K (Subbarao et al., 2003) . However, at high levels, salinity is known to be an agent of selection. Na may stress plants by blocking water and nutrient uptake in the roots, lead ing to deficiency (Fageria et al., 2011) . Persistence in saline soils is uncommon for non halophytic plants, and supports the hypothesis that subsp. scirpoidea exhibits high plasticity with respect to edaphic cond itions as an evolutionary strategy. Due to its distribution at high latitudes, shorter, colder growing seasons were expected in subsp. scirpoidea compared to its conspecific. My CGDD results supported this, with the exception of the Pembina, MN site (Fig s . 2. 4 , 2. 5; Appendix D). The seasonal dynamics of subsp. scirpoidea are quite distinct, adopting a bi/trimodal parabola like shape with several subsp. scirpoidea sites oscillating above and below 0¼C (Appendix D). These seasonal patterns differ somewhat fr om the standard parabolic shape as seen in the Upper Midwestern United States populations of both taxa. Additionally, higher soil temperature variation was noted in subsp. scirpoidea than subsp. convoluta . Soil CGDD has been found to moderately influence s edge biomass (Brooker and Van Der Wal, 2003) . Further, low soil temperatures are known to

PAGE 29

16 limit nutrient uptake by roots, decrease nitrogen availability, and slow or prevent plant growth (Brooker and Van Der Wal, 2003; Moore et al., 2015; Krab et al., 2018) . Future Directions The findings suggest that neither taxon is restricted to a single soil type based on my analyses of the parameters assessed here. While Dunlop & Crow (1999) described subsp. convoluta as an edaphic ecotype, the chemical composition of the soil type remains ambiguous. Furthermore, the uniquely divergent soil types of the Horseshoe Bay, MI site for subsp. convoluta and the High Creek Fen, CO site for subsp. scirpoidea may suggest n ovel ecotypes. While genetic differentiation was detected with other populations, it is unclear whether local genomi c adaptation has influenced specialization to unique soil conditions (Pembrook, 2014; Westergaard et al., in press; Chapter 3, this thesis) . Future genomic studies will determine whether local gen om ic adaptation to low nutrient, calcareous soil, or sodic soil has occurred; and whether edaphic "speciation genes" may be detected (Rajakaruna, 2018) . Although beyond the scope of this research, greenhouse stu dies and reciprocal transplant experiments should be employed to test whether populations of subsp. convoluta and subsp. scirpoidea grow equally well in soils with high salinity, low CGDD, and low nutrient soils, to indicate any differences in edaphic tole rance between taxa . Reciprocal transplant experiments Ñ alvar associated subsp. convoluta individuals at Horseshoe Bay, MI and vice versa Ñ may indicate niche differences due to local adaptation , or broad tolerance due to plasticity. Such studies may also be useful to understand how adaptation may progress as climate change and habitat degradation occurs.

PAGE 30

17 Table 2. 1: Soil sampling sites for Carex scirpoidea Michx. s ubsp. convoluta (KŸk.) Dunlop and subsp. scirpoidea (Cyperaceae) populations . Taxon Sample Site Country Latitude Longitude Carex scirpoidea subsp. convoluta Thompson's Harbor, MI USA 45.35 83.58 Horseshoe Bay, MI USA 45.99 84.745 Maxton Plains, MI USA 46.08 83.66 Manitoulin Island, ON CAN 45.60 82.12 Cabot Head, ON CAN 45.24 81.30 Carex scirpoidea subsp. scirpoidea Eagle Summit, AK USA 65.48 145.41 12 Mile Summit, AK USA 65.40 145.96 High Creek Fen, CO USA 39.10 105.97 Pembina, MN USA 48.08 96.45 Mestersvig GRL 72.23 23.98 Traill GRL 72.52 23.98 Kjelvatn NOR 68.18 17.21 Kjerringa NOR 66.78 14.21 SolvŒgtind NOR 66.83 15.42

PAGE 31

18 Table 2.2 : Akaike Information Criterion (AIC) for models that predict taxon from discrete soil chemistry parameter s. The taxa being predicted are Carex scirpoidea Michx. s ubsp. convoluta (K Ÿk.) Dunlop and subsp. scirpoidea (Cyperaceae) . The Intercept model is a null model. Soil parameters are measured in mg/kg, except where stated otherwise. Asterisks (*) imply concentrations of exchangeable cations. Akaike weights indicate an es timate for probabilities of each model. Dashes ( ) indicate probabilities under 0.1%. Model AIC Akaike Weight Na * 13.59124 100% CGDD 23.07332 0.8% Intercept 56.74755 pH 40.45050 Ca 48.28243 Ca* 58.50410 P 58.63474 P* 57.76991 Mg 42.41074 Mg* 58.28331 K 58.63474 K * 57.76991 N % 56.80382 C % 58.32251 Loss on Ignition 57.70272 Zn 57.50691 Zn* 57.00345 S 58.74152 S* 58.63804 Na 56.59971 Fe 50.85687 Fe* 51.46165 Mn 57.54802 Mn* 57.16004 Cu 57.65553 Cu* 56.92506 Cation Exchange Capacity 58.01994

PAGE 32

19 Table 2 .3 : Descriptive statistics of the parameters differentiating between Carex scirpoidea Michx. S ubsp. convoluta (KŸk.) Dunlop and subsp. scirpoidea (Cyperaceae) . All chemicals were measured in mg/kg , except where noted with %, and cation exchange capacity, measured in in milliequivalents per 100g. Asterisks (*) imply concentrations of exchangeable cations. Parameter Carex scirpoidea subsp. convoluta Carex scirpoidea subsp. scirpoidea ! SD Range ! SD Range Na * 6.99 1.30 719.03 45.10 59.97 2909.13 CGDD 3168.80 273.44 4.8 972.76 977.44 293 pH 7.49 0.16 0.64 6.73 0.67 2.13 Ca 129664.27 127122 361113 36234.59 50148.31 211230 Ca * 8923.40 4613.81 13156 10089.48 8732.92 27191 P 579.80 345.37 1658 843.19 573.75 1726 P * 9.91 6.94 25.93 8.35 8.20 26.02 Mg 24898.13 20581.21 65598 7111.04 5141.00 15913 Mg * 967.94 845.52 2881.54 735.19 1167.55 4797.65 K 2724.40 2283.70 7029 2940.48 1924.09 7791 K * 107.63 87.81 305.9 141.08 118.52 364.3 N % 0.70 0.54 1.952 1.11 1.08 2.78 C % 13.58 5.54 20.616 16.48 17.25 44.61 Loss on Ignition % 22.99 10.92 38.46 31.11 30.79 80.25 Zn 53.22 35.74 125.8 66.11 38.24 186.7 Zn * 1269.80 900.47 2933 2202.37 2880.37 10978 S 1994.35 2104.27 5627.6 2048.64 2293.01 9416.5 S * 56.67 83.44 236.6 67.14 110.94 512.9 Na 229.67 101.34 308 462.41 712.64 2516 Fe 11216.47 9219.87 26891 20977.78 11595.428 34595 Fe * 2.87 0.83 2.51 6.94 7.47 27.52 Mn 533.13 679.86 1988 831.41 1022.64 4719 Mn * 7.13 2.82 8.1 15.41 29.14 120.82 Cu 20.95 26.61 79.3 30.50 31.30 104.9 Cu * 277 0 0 296 71.19 327 Cation Exchange Capacity 528.23 238.55 709.5 654.67 561.91 1766.2

PAGE 33

20 Figure 2. 1: Soil sampling sites representing populations of (a) Carex scirpoidea Michx. subsp. convoluta (KŸk.) Dunlop and (b) subsp. scirpoidea (Cyperaceae) ; (c) soil core used for collection, and typical core sample in alvar locations. Plant site abbreviations are as follows: TH = Thompson's Harbor , MI , USA; HB = Horseshoe Bay , MI , USA; MP = Maxton Plains , MI , USA; MI = Manitoulin Island , ON , CA N ; CH = Cabot Head , ON , CA N ; 12MS = 12 Mile Summit , AK , USA; ES= Eagle Summit, , AK , USA; HCF = High Creek Fen , CO , USA; P = Pembina , MN , USA; M = Mestersvig, G R L; T = Traill , G R L; K = Kjelvatn, NO R ; K2 = Kjerringa, NO R ; S = Solvagtind, NO R . c a b

PAGE 34

21 a b c Fig ure 2. 2. PCA biplot for 5 9 measured soil parameters at sites of Carex scirpoidea Michx. subsp. convoluta (KŸk.) Dunlop and subsp. scirpoidea (Cyperaceae) . In a), taxon is circumscribed, in b) population is circumscribed , and in c) region is circumscribed . Plant site abbreviations are as follows: TH = Thompson's Harbor , MI , USA; HB = Horseshoe Bay , MI , USA; MP = Maxton Plains , MI , USA; MI = Manitoulin Island , ON , CA N ; CH = Cabot Head , ON , CA N ; 12MS = 12 Mile Summit , AK , USA; ES= Eagle Summit, , AK , USA; HCF = High Creek Fen , CO , USA; P = Pembina , MN , USA; M = Mestersvig, G R L; T = Traill, G R L; K = Kjelvatn, NO R ; K2 = Kjerringa, NO R ; S = Solvagtind, NO R .

PAGE 35

22 Fig ure 2. 3: Clustered dendrogram of soil site similarity across Carex scirpoidea Michx. subsp. convoluta (KŸk.) Dunlop (in blue, abbreviated as Ôc') and subsp. scirpoidea (Cyperaceae) (in red, abbreviated as Ôs') (from PCA). Plant site abbreviations are as follows: TH = Thompson's Harbor , MI , USA; HB = Horseshoe Bay , MI , USA; MP = Maxton Plains , MI , USA; MI = Manitoulin Island , ON , CA N ; CH = Cabot Head , ON , CA N ; 12MS = 12 M ile Summit , AK , USA; ES= Eagle Summit, , AK , USA; HCF = High Creek Fen , CO , USA; P = Pembina , MN , USA; M = Mestersvig, G R L; T = Traill, G R L; K = Kjelvatn, NO R ; K2 = Kjerringa, NO R ; S = Solvagtind, NO R .

PAGE 36

23 Fig ure 2. 4: Plot of predicted probability of Carex scirpoidea Michx. subsp. convoluta (KŸk.) Dunlop (abbreviated as Ôconvoluta'), (set as 0) and subsp. scirpoidea (Ôscirpoidea') Cyperaceae , (set as 1) from generalized logistic regression model, wh erein a soil parameter is the predictor variable. In a), exchangeable Na (Na * ) is the predictor variable, in b), soil cumulative growing degree days (CGDD) is the variable. a b CGDD Na* mg/kg

PAGE 37

24 Fig ure 2. 5: Coefficient of variation for (a) soil cumulative growing degree days (CGDD) and (b) exchangeable Na ( Na* ) measured in soil collected at 11 sites with populations of Carex scirpoidea Michx. subsp. convoluta (KŸk.) Dunlop (abbreviated as Ôconvoluta'), a nd C. scirpoidea subsp. scirpoidea (Ôscirpoidea'). Statistical significance of test for equality of variation between populations (p) included. a b Na* mg/kg CGDD

PAGE 38

25 C HAPTER III POPULATION GENOMICS OF THE NARROW EDAPHIC ENDEMIC CAREX SCIRPOIDEA SUBSP. CONVOLUTA Abstract Rare, narrow endemics are expected to harbor low genomic diversity due to small number of populations and limited range . In this study, I compared genomic diversity between Carex scirpoidea subsp. convoluta and subsp. scirpoidea, and analyzed genomic diffe rentiation and structuring between taxa and populations of these taxa. I used double digest restriction site associated DNA (ddRAD) sequencing to identify genome wide SNPs from individuals representing 11 populations across the North American range of both taxa. Unexpectedly, subsp. convoluta exhibit ed relatively high genomic diversity compared to its conspecific. Further, there was little evidence for evolutionary di fferentiation between taxa, although regional differentiation was observed among populations of subsp. scirpoidea . Phylogen omic results and Bayesian /multivariate cluster analyses suggest that subsp. convoluta arose from periglacial refugial populations of subsp. scirpoi dea in Eastern North America as glaciation receded and C. scirpoidea recolonized northern North America. Multiple statistical analyses revealed that subsp. scirpoidea comprises three distinct genomic clusters, corresponding to three probable North American glacial refugia. It is likely that a nearly obligate dioecious mode of reproduction has contributed to the maintenance of gen om ic diversity in subsp. convoluta, preventing inbreeding depression, and contributing to the evolutionary potential of population s of this tax on . I recommend that populations of subsp. convolut a be monitored , particularly along the margins of its range, due to red uced gene flow between distant populations.

PAGE 39

26 Introduction The high diversity of flowering plants worldwide can be attributed to population divergence over evolutionary time, to which population dynamics Ñ including size, age structure , and the ecological factors that influence them Ñ contribute. Limited, fragmented habitat may lead to bottleneck s and reduced effective size of a population , which in turn may reduce heterozygosity (Kruckeberg and Rabinowitz, 1985; G—mez Fern‡ndez et al., 2016) . Loss of heterozygosity may be exacerbated in habitat li mited populations by increases in inbreeding rates and genetic drift, which may lead to differentiation from the metapopulation (Kruck eberg and Rabinowitz, 1985; Ellstrand and Elam, 1993; Hamrick and Godt, 1996) . Similarly, selection in response to variable ecological conditions across a species ' range, such as soil chemistry, may also lead to population differentiation and, ultimatel y speciation (De La Torre et al., 2014; Zhou et al., 2014; Rajakaruna, 2018) . However, adaptation to novel conditions must precede such differentiation (Richter Boix et al., 2011; Zhou et al., 2014; Sandoval Castillo et al., 2018) . Adaptation is enabled by phenotypic plasticity, standing genetic variation, or the introduction of novel alleles through genetic mutation or gene flow (Olson et al., 2013) . Often, adaptive alleles are selected for in response to ecological conditions, such as annual temperature and soil chemistry (Coop et al., 2010; Roda et al., 2013) . As such, adaptive alleles may influence a taxon's geogra phic distribution, sometimes resulting in endemism (Richter Boix et al., 2011; Worch et al., 2011) . Endemics may form when a species' ability to adapt to a novel niche coincides with reduced competition from other less well adapted species (Kruckeberg, 1951; Stebbins and Major, 1965 ) . Edaphic endemism specifically arises when a species can tolerate a broad range of soil types or has the means to adapt to unusual soil types. To explain the origin of some edaphic endemics, Rajakaruna (2018) hypothesized that habitats with novel soil types

PAGE 40

27 formed glacial refugia during the Last Glacial Maximum (LGM). As such, the spatially reconfigured populations altered evolutionary processes, such as gene flow, and adaptation to new habitat types followed (Catling and Brownell, 1995; Hewitt, 2000; Rajakaruna, 2018) . Dillenberger and Kadereit (2013) reported a niche shift in subspecies of Adenostyl e s alpin a (L.) Bluff. & Fingerh from calcareous to non calcareou s soils following dispersal to glacial refugia. D ifferentiation can occur between populations , regardless of levels of genetic diversity. Genetic dri ft, selection, and restricted gene flow may increase rapid fixation of alleles and reduce diversity , result ing in greater genetic structure between populations (Frankham, 20 05) . G enetic diversity provide s the novel variation on which selection acts, particularly as an organism colonizes new environments and or faces forthcoming environmental conditions, thereby contributing to its "evolutionary potential" (Rajakaruna, 2018; Frankham, 2005) . Widely distributed species comprising large populations tend to maintain t he highest levels of genetic diversity, while narrowly distributed species with small populations typically maintain less (Ellstrand and Elam , 1993; Hamrick and Godt, 1996; Frankham, 2005) . In small populations, alleles that confer adaptation to shifting environmental conditions and resilience to unanticipated events may be lost to genetic drift, while they are expected to persist in larger populations. Furthermore, loss of genetic diversity in a population can hamper evolution, resulting in Ôgenostasis' (Bradshaw, 1991; Rajakaruna, 2018) and extinction vortices, wherein anthropogenic threats combine with stochastic demog raphic and genetic events resulting in extinction (Gilpin and SoulŽ, 1986) . Reliable methods of measuring genetic diversity aid in forecasting viability and extirpation risk in small populations, especially amidst emerging ecological threats, like habitat loss (Allendorf et al., 2010; Frankham, 2010) . Monitoring genetic diversity enables proper

PAGE 41

28 estimation of evolutionary p otential, which may ensure plant survival under stochastic conditions, determine whether rare alleles persist in small populations, or indicate if a population is threatened by inbreeding depression (Allendorf et al., 2010) . However, genomic techniques may also be used to identify how populations and taxa have adapted and dive rsified in response to historical ecological conditions (Roda et al., 2013; Sandoval Castillo et al., 2018; Westergaard et al., in press ) , providing useful information to predict future scenarios. Thus, studying species' population genetic diversity and structure is paramount for determining conservation status and implementing management strategies (Frankham, 2010) . Carex scirpoidea subsp. convoluta (hereafter, subsp. convoluta ) is endemic to alvar and related habitats (e.g., cobble beach) along the Niagara Escarpment in close proximity to northern Lake Huron (see Chapter 2) (Dunlop & Crow 1999). Although species associated with Great Lakes alvar are thought to be descendants of ancestral parkland, tundra, and prairie species prior to glaciation (Hamilton & Eckert 2007 ; Catling & Brownell 1995), little is known about the evolutionary history of subsp. convoluta . It has been hypothesized that subsp. convoluta arose from C. scirpoidea subsp. scirpoidea (hereafter, subsp. scirpoidea ) as an edaphic alvar associated ecotype during the Pleistocene (Dunlop, 1990; Dunlop and Crow, 1999; Pembrook, 2014) . Dunlop (1990) hypothesized that subsp. scirpoidea persisted in three North American glacial refugia during the Pleistocene: Beringian Alaska, the Cordilleran region in Western North America, and periglacially in Eastern North America. Westergaard et al., ( in press) found genomic evidence for these refugia, as well as additional glacial refugia in Norway and Greenland. However, the ancestral populations of subsp. convoluta remain undocumented. Predicting genetic diversity based upon life history yields confl icting results. Shackleford (2003) speculated that subsp. convoluta would harbor low levels of genetic diversity, especially

PAGE 42

29 compared to its widesprea d conspecific, due to its restricted distribution. However, C. scirpoidea is dioecious, a trait known to preserve genetic diversity (Hamrick and Godt, 1996) . Mechanisms and effectiveness of pollen and diaspore dispersal are not well understood, further complicating predictions regarding genetic diversity and structure in this species. DePrenger Levin (2007) and Pembrook (2014) , using allozyme analysis, reported levels of genetic diversity in subsp. convoluta that were not significantly lower than in the widespread subsp. scirpoidea in all populations except the Beringian Alaska population. However, it should be noted that popul ation samples for subsp. scirpoidea were primarily obtained from the taxon's southernmost populations, and allozymes have a coarser resolution than most contemporary genomics methods. In order to better characterize genetic diversity and structure in this C. scirpoidea , including subsp. convoluta , a relatively novel molecular genetic technique with higher resolution (greater number of data points) was employed. Double digest restriction site associated DNA (ddRAD) sequencing is a contemporary method of red uced representation genome sequencing wherein only genomic loci flanked by two specific restriction enzyme cut sites are sequenced. The technique has gained popularity due to its ability to partially sequence multiple individuals per sequencing run, and as semble a partial genome de novo for organisms without publicly available reference genomes (Baird et al., 2008; Peterson et al., 2012) . In this study, ddRAD sequencing was conducted to assess genomic diversity for the rare edaphic endemic subsp. convoluta and the widespread, presumed generalist, subsp. scirpoidea . Additionally, population genetic structure and phylogen om ic relationships were inferred us ing ddRAD genomic data in order to detect divergence among populations of both taxa. This paper seeks to quantify genomic diversity and explain evolutionary diversification within a

PAGE 43

30 geographical context. The objectives of this study are to 1) quantify gen o m ic diversity in subsp. convoluta relative to its widespread conspecific , 2 ) assess structure within and among populations of the subspecies , 3) test evolutionary hypotheses regarding the origin of subsp. convoluta, and 4 ) estimate risks of extirpation and evolutionary potential of subsp. convoluta . Notably, this study marks the first population genomic study of Carex Ñ and among the first for plant taxa Ñ using ddRAD sequencing. Methods Tissue Collection Fieldwork was conducted at five sites for subsp. convoluta from across its range (Dunlop and Crow, 1999; Fig. 3. 1a; Table 3.1; Appendix A ). At each site, leaf tissue was collected from 24 individuals per site following a transect based sampling protocol for a total of 120 individuals (Fig. 1b). Leaf tissue of subsp. scirpoidea individuals was previously collected using the methods described by Pembrook (2014) (Fig. 3.1b; Table 3.1) . Leaf tissue was silica dried and stored at room temperature. DNA Extraction Silica dried leaf tissue (20mg) was cut into pieces with sterile scissors, then frozen. Three 2.8 mm ceramic beads (OMNI International SKU 19 646) were added to each Eppendorf tube, and pulverized at 1.5 minutes with a reciprocating saw modified from Alexander et al. (2007) . DNA was extracted following a CTAB protocol using NucleoSpin¨ Plant II kits (Macherey Nagel 740440.50) with the following modifications. Lysis incubation was increased to 30 minutes, crude lysate was centrifug ed for 5 minutes at 11,000 x g prior to filtration, and the column containing DNA was dried in an oven at 40¼C for 30 minutes prior to elution. Warm milliQ¨ water (40 # L) was used in place of the provided elution buffer. DNA concentrations

PAGE 44

31 were quantified with a Nanodrop 2000 and later with Qubit (dsDNA HS Assay Kit Q32851). Samples with signs of carbohydrate contamination or low DNA concentration were cleaned and concentrated with ethanol precipitation, re extraction of DNA from additional leaf material, a nd/or vacuum concentration, as needed. Ethanol precipitation was performed using 1/10 th volume of 3M sodium acetate (pH 5.5) and 2X the volume of 100% EtOH, washed with 70% EtOH, and resuspended in milliQ¨ water. After DNA extraction, cleanup, and concentr ation, samples with the highest concentration were selected for sequencing: 10 11 individuals from each subsp. convoluta population, and 5 8 individuals from each subsp. scirpoidea population. ddRAD library preparation was conducted using a modified version of Parchman et al. ( 2012) . MSEI and EcoRI HF were used as r estriction enzymes to digest the DNA, and their respective sequencing adaptors with complementary overhangs were ligated onto digested DNA with DNA ligase. PCR amplification was used to amplify correctly ligated fragments for DNA sequencing . The Biofrontiers Sequencing Facility (University of Colorado, Boulder) conducted fragment size selection, cleanup, and conducted sequencing. BluePippin (Sage Science) was used for size selection at the 300 600 fragment range and cleanup using Agencourt AMPure XP beads (A63881). Sequencing was performed to generate single ended reads using the Illumina NextSeq V2 High Output (75 cycle kit) platform. SNP calling and filtering Pooled Illumina output was demultiplexed using the first step of the iPyrad pipeline ( https://ipyrad.readthedocs.io/) . A barcodes file matching those used during ddRAD library preparation was supplied. Restriction overhang sequences CAATTC (protector base [C] + EcoRI) and TAA (MseI) were used. D uring demultiplexing, barcode mismatches of up to two nucleotides were permitted. Once demultiplexed, adapter trimming was employed by dDocent

PAGE 45

32 (Puritz et al., 2014) . De novo assembly was conducted using Rainbow with read clustering at 80% minimum similarity (Chong et al., 2012) . Read mapping was conducted using BWA with default scores and penalties (Li and Durbin, 2009) . For reference assembly, K1 (coverage within each individual) and K2 (minimum number of indivi duals) parameters were chosen based upon manual inspection of the asymptotic curve inflection. K1 was set to a minimum depth of five, while K2 was set to nine. freeBayes was invoked for SNP calling using default parameters supplied by dDocent (Garrison and Marth, 2012) . The raw SNP file was filtered with vcflib (Garrison, 2012) and vcftools (Danecek et al., 2011) , following the dDocent SNP filtering p rocess, with some exceptions: minor alleles with a count less than four were filtered, individuals with >40% missing data were discarded. Next, a 95% genotype call rate across all populations was specified, followed by a 95% call rate within each populatio n. Allele balances between 20% and 80% were retained. Mapping quality between reference and alternate alleles were kept if between 0.8 and 1.2. Loci that did not have a minimum quality score within $ of the depth were removed. Next, loci above the mean dep th + three standard deviations , with a quality score above twice the depth , were filtered. To filter potential paralogous sites, loci with a high mean depth were filtered. To choose a cutoff for high mean depth, the upper bound value of the larger histogra m (mean depth = 160) was chosen. Loci not at Hardy Weinberg Equilibrium (p > 0.05) were discarded. To minimize linkage between loci, rad_haplotyper (Willis et al., 2017) was employed, and loci were kept that successfully haplotyped in 90% of individuals and saved in variant call format (VCF). PGDspider was employed to convert data to the correct format for each subsequent software program (Lischer and Excoffier, 2012) .

PAGE 46

33 Data Analysis The following population genomic statistics were obtained using Arlequin: expected heterozygosity (H E ), observed heterozygosity (H O ), percent of polymorphic loci (P), nucleotide diversity ( ! " ), pairwise differentiation between populations (F ST ), and Nei's distance between populations (d) (Excoffi er and Lischer, 2010) . Two locus by locus AMOVA tests were conducted with 1000 permutations in order to analyze hierarchical patterns of genomic differentiation across the dataset. Individuals were categorized in their respective populations, and each p opulation was categorized into broad groups. In the first AMOVA, group was defined as Ôtaxon' (subsp. convoluta and subsp. scirpoidea ); in the second, group was defined as Ôregion:' Northwestern represented by Anvil Mountain, AK; Western represented by But te, MT and High Creek Fen, CO; and E astern comprising all remaining populations. Parameters measured included inbreeding coefficient (F IS ), variance among individual populations within groups (F SC ), variance among taxa compared to total variance (F CT ), and overall fixation index (F IT ) . F ST was calculated with 100 permutations at the 0.05 significance level, while ! " was calculated by taking the average number of pairwise differences under the infinite site model. In order to assess candidate loci under sele ction, an outlier loci analysis was calculated with Bayescan (Foll and Gaggiotti, 2008) for the eight E astern North American populations (all but Anvil Mountain, AK; Butt e, MT; and High Creek Fen, CO). BayeScan uses a reverse jump Markov Chain Monte Carlo to estimate locus population specific genomic differentiation (F ST ). I ran 10,000 iterations of the Bayesian model, using the default settings. Western North American pop ulations were excluded in order to limit false positives from isolation by distance. Structure in the data was assessed using fastStructure (Raj et al., 2014) and discriminant analysis of principle components (DAPC) function of adegenet (Jombart and Ahmed, 2011)

PAGE 47

34 using R Version 3.4.4 (R Core Team, 2014) . fastStructure uses a model based approach, while DAPC uses a discriminant analysis to maximize variability between groups des ignated by PCA. In fastStructure (Raj et al., 2014) , K values between 1 and 11 (total number of populations) were tested using a simple model with default settings . K values indicating strong and subtle population gene tic structure were chosen using the chooseK python script. Structure at both scales were plotted in Distruct (Rosenberg, 2004) . Using adegenet (Jombart and Ahmed, 2011) , BIC scores were obtained for K groups (K = 1 40). Models with optimal BIC scores ( % BIC < 1) were assessed for optimal a scores. DA PC, using the number of principle components (PCs) and discriminant functions specified, were used to determine structure. SNP loci with ambiguous sites in an individual were filtered using the raxml_ascbias Python script (Martin, 2018) to avoid ascertainment bias during alignment. jModelTest2 (Guindo n and Gascuel, 2003; Darriba et al., 2015) was used to test 88 different models of nucleotide substitution. Phylogenies were built using RAxML (Stamatakis, 2014) , using a ASC_GTRGAMMA model for 500 bootstrapping iterations, and the resulting tree (Appendix E) was rooted by employing the simple rooting algorithm. All trees were plotted using Dendroscope (Huson and Scornavacca, 2012) . Results NextSeq 1 x 75 sequencing pro duced 472,350,095 reads. After demultiplexing, 344,139,306 reads remained. After filtering and SNP calling, 719,059 SNPs were "called" by dDocent (Puritz et al., 2014) . Of the 96 individual samples that were sequenced, 77 remained after filtering. Of the 19 individuals that were filtered out, all were Carex scirpoidea subsp. convoluta representing Horseshoe Bay, MI (5); Manitoulin Island, ON (7); Maxton Plains, MI (4); Thompson's Harbor, MI (3). After filtering loci, 7384 SNPs remained, which were used in

PAGE 48

35 each analysis , with the exception of the phylogenomic analysis . Population Genomi cs All measures of genomic diversity in subsp. convoluta exceeded those for subsp. scirpoidea, including proportion of polymorphic loci (P), nucleotide diversity ( ! " ), and expected heterozygosity (H E ) ( subsp. convoluta : 4 5 = 0.39, 6 = 51.688, 7 8 = 1490. 33; subsp. scirpoidea : 4 5 = 0.38, 6 = 46.62, 7 8 = 1307.79) . Observed heterozygosity was higher than expected heterozygosity in every assayed population (Table 3.2 , Fig. 3.3 ). The Anvil Mountain , AK population of subsp. scirpoidea exhibited the lowest genomic diversity (P = 34.41, ! " = 995.05 ; Table 3.2 ; Figs. 3. 3 , 3.4 ) . Manitoulin Island, ON subsp. convoluta had the highest rate of nucleotide diversity ( ! " = 1552.53) , while Cabot Head, ON had the highest rate of polymorphic loci (P = 58.85%). Allele count across all surveyed loci ranged between 2.003 and 2.005 ( Fig. 3.5 Appendix E). The AMOVA takes into consideration analysis of variance for mutations among loci to determine where in the sample variation occurs. In the taxon partitioned AMOVA, only 5.42% of the variation was among taxa (p < 0.001), while nearly one third of the varia tion observed among populations within taxa ( 29.10 %, p < 0.001, Table 3 . 3 ). In the region partitioned AMOVA, the reverse was observed: one third of the variation was observed among regions (33.95%, p < 0.001, Table 3 .4 ), with only 9.74% (p < 0.001) of the variation occurring among populations within regions. Geographically distant populations were highly differentiated; t he western populations (Anvil Mountain, AK; Butte, MT; and High Creek Fen, CO) were highly differentiated from the eastern populations ( F S T = 0.41 0.50; Table 3.5 ; Figs. 3.10 , 3.11 ), while the greatest differentiation was found between the High Creek Fen, CO and Anvil Mountain, AK populations (F ST = 0.50, p & 0.005).

PAGE 49

36 Of the 7384 assayed loci, 11 loci were found to be outlier loci that excee d the false discovery rate (FDR; q value = 0.05) (Fig. 3.12 ). F ST was bimodal, with all intermediate values falling under the FDR. Of the 11 loci, three loci produced F ST values below 0.0 6 ; eight loci produced F ST values exceeding 0.36, suggesting populati on differentiation due to selection (Foll and Gaggiotti, 2008) . Structure Optimal model complexity calculated using fastStructure (Raj et al., 2014) resulted in three groups (K = 3) (Fig . 3.6 ). No clustering of populations occurred along taxonomic lines (e.g., subsp. scirpoidea vs. subsp. convoluta ). Instead, populations clustered over three broad geographical regions, which were also identified by Westergaard et al. (pers . comm . ); these are: 1) a Beringian group, as represented here by the Anvil Mountain, AK population; 2) a Western North American group represented by the Butte, MT and High Creek Fen, CO popula tions; and 3) an Eastern North American group as represented by the remaining populations, including each of the five subsp. convoluta populations. fastStructure (Raj et al., 2014) model selection indicated that five c lusters or groups (K = 5) account for 99.9% cumulative ancestry contribution. Therein, evidence of admixture between populations was exhibited in eight of the eleven populations surveyed, all representing the Eastern North American group (Fig. 3.6 ). DAPC a nalyses revealed that five groups (K =5) had the highest model support (BIC, Fig. 3 .7 ), with several notable similarities to fastStructure (K=5) results (Fig. 3.8) . For instance, the Brig Bay, NL subsp. scirpoidea population clustered with populations of s ubsp. convoluta at all K, using both methods. Interestingly, each analysis showed that one population on the range margin of subsp. convoluta (Thompson's Harbor [MI], K=5, Fig. 3.6 ; Cabot Head [ON]; Fig. 3 .8 ) formed a distinct group. Further, remaining cen tral populations formed a group (Figs. 3.6,

PAGE 50

37 3.8 ), although there was also very high model support for three and four groups (K = 3 4; % BIC < 1; Fig. 3.7 ). Notably, at K=3, fastStructure and DAPC produced identical results. Phylogeny After filtering ambiguo us sites to be used to create the phylogeny, 5985 sites remained. Of the 88 nucleotide substitution models tested, a General Time Reversible model with gamma distribution and invariable sites (GTR+I+G) model produced optimal BIC and AIC scores, though the transversion gamma model (TVM+G) was best supported by the decision theoretic performance based metric (DT). Thus, the GTR Gamma model with ascertainment bias correction (ASC_GTRGAMMA) was selected to for the Maximum Likelihood cladogram. With the exceptio n of the Maxton Plains, MI subsp. convoluta population, which received weak bootstrap support (50%), each population was recovered as monophyletic with moderate to very strong bootstrap support ( ' 80; Fig. 3.9 ). There was no support for a monophyletic subs p. convoluta. Instead, the western High Creek Fen, CO, Butte, MT, and Anvil Mountain, AK subsp. scirpoidea populations segregated with strong bootstrap support (100%), while the eastern group Ñ recovered in fastStructure and DAPC Ñ was unresolved and polyp hyletic. Within the western clade, the Butte, MT and High Creek Fen, CO populations segregated together with strong bootstrap support (100%). Within the Eastern clade, the Escanaba River, MI and Pembina, MN populations (subsp. scirpoidea ) and Horseshoe Bay , MI population (subsp. convoluta ) form a clade, albeit with weak (64%) bootstrap support. Within subsp. convoluta , only the Maxton Plains, MI and Thompson's Harbor, MI populations formed a well supported clade (91%, Fig. 3.9 ).

PAGE 51

38 Discussion Population Genomic Diversity In contrast to expectations based upon restricted range, fragmented habitat, and relatively small population size (Bard and Bruederle, personal observation) , population genomic diversity in subsp. convoluta is, in fact, higher than all bu t one population of the widespread conspecific subsp. scirpoidea (e.g., 9 = 51.69± 4.9 vs. 9 = 46.62 ± 6.70; Table 3.2 ) . Although some inbreeding depression was expected, all populations of C. scirpoidea exhibit heterozygote excess (F IS = 0.18, Table 3.4, 3.5 ), with observed heterozygosity exceeding expected heterozygosity in all populations (Table 3.2 ). Obligate outcrossing due to dioecy has likely contributed to the preservation of genomic variation by preventing selfing, the most severe type of inbreedi ng (Ellstrand and Elam, 1993; Hamrick and Godt, 1996; Dunlop and Crow, 1999) . These findings generally support Pembrook (2014) and DePrenger Levin (2007), neither of whom found lower genet ic diversity in populations of subsp. convoluta compared to subsp. scirpoidea . Population Genomic Structure and Phylogenomics Rather than exhibiting genomic differentiation along taxonomic lines, as was expecte d, strong structure was instead observed among three geographic regions: Northwestern North America, Western North America, and Eastern North America ( fastStructure/DAPC K=3, Figs. 3.6 , 3 .8 ). This region based structure was also revealed from the phylogen om ic analyses, with the Northwestern and W estern regions forming monophyletic groups (clades) with 100% bootstrap support (Fig. 3.9 ). Eastern North American populations of subsp. scirpoidea formed a group with subsp. convoluta ( fastStructure/DAPC K = 3, Figs. 3.6, 3 .8 ), although the phylogen om ic a nalys i s revealed the Eastern clade to be polyphyletic (Fig. 3.9 ). An AMOVA showed that 33.95% of all genomic variation was partitioned among the aforementioned regions,

PAGE 52

39 while only 5.42% was between taxa (Table 3 .4 ). Finally, genomic differentiation between populations among regions (F ST ) ranged from 0.41 to 0.50 (p & 0.05 in 88.5% comparisons , Table 3.5 , Fig. 3.11 ). The observed regional structuring pattern provides support for Dunlop's (1990) proposed refugia for Carex scirpoidea during Wisconsin glaciation: Beringian Russia, Alaska, and Yukon; Cordilleran (Rocky Mountain) western North America, south of the last glacial maximum (LGM); and periglacially in E astern North America. Among the populations sampled in my study, the Northwestern North American populations represent a Beringian cluster also identified by Westergaard et al . (in press), while the Western North American populations represent a well supported Co rdilleran cluster that differentiated during the Pleistocene. In contrast, Eastern populations of subsp. scirpoidea are clearly the product of recolonization from periglacial North American refugia, south of the LGM, rather than recolonization from the Ber ingian and Cordilleran refugia (Figs. 3.7 3.9 ). One such glacial refugia for E astern North American subsp. scirpoidea populations may be the Driftless Area. Westergaard et al . (in press) found evidence for distinct and unadmixed populations in Pembina , MN and Escanaba River , MI . Their results also suggested that Western Greenland populations admixed with North American and Eastern Greenland/Norway influence, suggesting East West range colonization during glacial retreat. Evidently, C. scirpoidea population s in the Great Lakes area of North America expanded Northeastward as the Laurentide Ice Sheet receded. The Brig Bay , NL population in my study supports this expansion, as it forms a group with the Pembina , MN population and is only moderately differentiate d from subsp. scirpoidea populations from the Upper Midwest (K=3 5, Fig. 3 .8 ; F ST =0.17, Table 3.5 , Fig s . 3.10 , 3.11 ).

PAGE 53

40 My results suggest that extant populations of subsp. convoluta also descended from perigl a cial Eastern North American refugial populations. The furthest n ortheastern population of the Eastern group (Brig Bay , NL) exhibited no apparent substructure when compared to subsp. convoluta (All K, Figs. 3.6 , 3 .8 ), and low differentiation (F ST = 0.10 0.18 , Table 3.5 , Fig. 3.11 ) compared to populations of subsp. scirpoidea (F ST = 0.44 0.45) . Interestingly, there is a closer affiliation between subsp . convoluta and Eastern Canada subsp. scirpoidea populations than with Great Lakes subsp. scirpoidea populatio ns Ñ where in fact subtle structure is exhibited (K=5, Fig 3.6 ; K=4 5, Fig. 3 .8 ) . Although my results do not resolve the pattern of postglacial recolonization, multiple origins for subsp. convoluta are suggested. Previous studies have suggest ed that Great Lakes disjuncts represent elements of the Atlantic and Gulf Coastal Plain s flora (Reznicek, 1994; Fant et al., 2014) . Several species are hypothesized to have originated from the s outhern Gulf Coastal Plain and migrated to the Great Lakes along the Mississippi, while others originated from the North Atlantic Coastal Plain (McLaughlin, 1932) . Fant et al. (2014) proposed n ortheastern Michigan, in the vicinity of the Straits of Mackinac, as a postglacial entry point for many plant species following the Atlantic Coastal route , as evidenced by patterns of genetic diversity in Cirsium pi tcheri (Torr. Ex Eaton) Torr & A. Gray , another Great Lakes endemic . T hus, t hree scenarios seem likely for the origin of subsp. convoluta , though none are mutually exclusive. First, founding populations of extant subsp. convoluta may have established following westward expansion of Eastern North American C. scirpoidea populations, as has been reported for other Great Lakes disjuncts (Reznicek, 1994). Alternatively, subsp. convoluta populations and Eastern Canada (e.g., Brig Bay, NL ) populations may be sister groups, having independently established following range expansion of ancestral populations ne ar the Driftless

PAGE 54

41 Area. Finally, populations may have recolonized the extant range of subsp. convoluta from other, unknown periglacial refugia south of the LGM. S peciation, if occurring in this complex, is apparently at an early stage, with little evidence for coalescence. The two taxa harbor no reproductive isolation (Dunlop and Crow, 1999) , and subsp. convoluta exhibits high affinity for Eastern populations of subsp. scirpoidea (i.e. Brig Bay, NL). S light admixture between Escanaba River , MI /Pembina , MN and subsp. convoluta populations was observed using fastStructure, indicating that gene flow between taxa has occurred or is ongoing (Fig. 3.6) . Further, t he taxonomical boundaries of subsp. convoluta remain ambiguous, at best (Figs. 3.6 3.9 ) . As such, the taxon may more appro priately be categorized as biotypes (Clausen, 1951; Lowry, 2012) . Extant alvar populations are thought to be relicts of prairie and parkland habitats that were spatially and, ultimately, genetically isolated from conspecific populations during glacial recession. North American alvar is restricted to the Niagara Escarpmen t and supports small, fragmented, and discontinuous communities comprising variously small populations between which restricted gene flow is expected for endemics (Catling, 1995; Hamilton and Eckert, 2007) . As a result, alvar endemics are expected to harbor low genetic diversity and be differentiated from ancestral populations (Hamilton and Eckert, 2007) . One alvar ecotype, the monoecious Geum triflorum Pursh, was found to exhibit lower genetic diversity and greater differentiation among its alvar populations than prairie populations across its range; this was attribute d to the disjunction of alvar habitats from prairielands, as well as changes in habitat availability during glaciation and recession (Hamilton and Eckert, 2007) . In contrast to the findin gs of Hamilton and Eckert (2007) , subsp. convoluta populations exhibit similar, and in some cases, higher levels of genomic diversity compared to subsp.

PAGE 55

42 scirpoidea (e.g. ! " = 1442.38 1552.52 vs. 995.05 1455.08, Table 3.2 ; Figs. 3.3 3.5 ) . Further, there was very little evidence of genetic differentiation a mong populations of subsp. convoluta and subsp. scirpoidea from E astern North America (F ST = 0.11 0.18, Table 3.5 ; Fig 3.10, 3.11 ). Differences in breeding system between C. scirpoidea and G. triflorum Ñ dioecy vs. monoecy Ñ likely contribute to these diff erences in population genomic diversity, which otherwise have similar recent evolutionary histories. Obligate outcrossing in subsp. convoluta is likely preserving genomic diversity, and halting loss of allelic variation resulting from genetic drift . Struct ure (Fig. 3.6 , 3 .8 ) and significant differentiation ( p & 0.05 in 7/10 pairwise comparisons, Table 3.5 ) of the two marginal populations, Cabot Head, ON and Thompson's Harbor, MI, indicate restricted gene flow in peripheral populations of subsp. convoluta (Fant et al., 2014). However, the diaspore dispersal distance, a critical determinant of gene flow, is unknown . Limited discontinuous alvar habitats may be aiding the enforcement of structure within all sampled populations of subsp. convoluta Ñ population s are monophyletic (Fig. 3.9 ), with genomic dif ferentiation occurring between them (F ST 0.05 ' 0.13, p & 0.05 in 7/10 pairwise comparisons, Table 3.5 , Fig. 3.11 ). G eomorphological features (e.g., Lake Huron) may further obstruct gene flow restriction. Se lection and isolation by distance Population differentiation implies reduced gene flow among populations, local adaptation within populations, or both. The isolated outlier loci identified herein (Fig. 3.12 ) are likely to be under selection , suggesting th at local adaptation may be contributing to population differentiation ( Foll and Gaggiotti, 2008) . Carex scirpoidea subsp. scirpoidea and subsp. convoluta occupy a wide range of habitats with varying ecological conditions (Dunlop and Crow, 1999; Shackleford, 2003, Chapter 2) . As discussed in Chapter 2, soil chemistry correlates, at

PAGE 56

43 least to some degree, with the Northwestern, Western, and E astern North American genetic clusters (Chapter 2; Figs. 3.6 , 3 .8 ). Although regional adaptation may be expected, genomic differentiation at the regional level may also be attributed to reduced gene flow among regions o ver many generations. Although most populations of subsp. convoluta occur along the Niagara Escarpment, the soil differed more than was expected between sites. Furthermore, I found 11 loci among the E astern North American populations that are putatively un der selection; however, this is likely a conservative estimate, as BayeScan limits false positives in small population samples (Foll and Gaggiotti, 2008; Fig. 3.12 ). Thus, it is possible that adaptive loci may be facilitating local adaptation among populations with different environmental conditions. Alternatively, wide tolerance across subsp. convoluta may facilitate occupa tion of different soil types (C hapter 2 , this thesis) . It is possible that the diversification of subsp. convoluta is itself due to edaphic adaptation; Dunlop (1990) suggests that the phenotypic leaf properties that is diagnostic for subsp. convoluta may b e adaptation to long periods of drought. My study also provides some evidence for geographic isolation caused differentiation. The highest levels of population differentiation were detected among the Cabot Head, ON); Horseshoe Bay, MI; and Thompson's Harbo r, MI populations (F ST =0.11 0.13, Table 3.5 , Figs. 3.10, 3.11 ). Cabot Head, ON and Thompson's Harbor, MI are located on extreme ends of the range Ñ the east and west sides of Lake Huron, respectively. Thus, geographical distance, as well as environmental conditions acting as selective pressures, may be contributing to the maintenance of structure in subsp. convoluta.

PAGE 57

44 Conservation of Carex scirpoidea subsp. convoluta There has been no apparent loss of gen om ic diversity in Carex scirpoidea subsp. convoluta, either at the level of the taxon or population, compared to i ts widespread conspecific. Although my results suggest that genomic diversity is being maintained by populations of subsp. convoluta at the time, habitat fragmentation and loss remain as risks limiting gene flow and outbreeding among isolated populations o f this taxon. Restricted gene flow, as well as local adaptation may contribute to differentiation between populations of subsp. convoluta (F ST = 0.07 0.13, Table 3.5 ). Furthermore, isolation of marginal populations, such as Thompson's Harbor, MI and Cabot Head, ON, may increase their sensitivity to habitat changes or stochastic events. Finally, genomic adaptation may be isolated to local populations; subsp. convoluta inhabits soils with a wide range of chemical and physical conditions. As such, populations, rather than taxa, should be treated as conservation units, particularly those at the range margin (e.g., Thompson's Harbor, MI and Cabot Head, ON) may be prioritized, due to their differentiation from other subsp. convoluta populations (F ST =0.07 0.13, Tab le 3.5 , Fig. 3.11 ; Allendo rf et al., 2010; Frankham, 2010) . Preventing genomic diversity loss and locally adapted populations are measures to retain evolutionary potential. Carex scirpoidea subsp. convoluta is a seemingly resilient taxon, retaining genomic diversity despite frag mented habitats and limited range. Thus, the taxon may be considered for replanting, particularly for restoring endemic rich habitats with unusual soil types (Reschke et al., 1999; Hamilton and Eckert, 2007) . Such efforts may have ecological benefits; detrital sedge tissue contributes to the unique edaphic conditions of wetland (i.e., sedge fen) habitats (Gorham, 1991) , and is thought to be a food source for grazing fauna (Ludwig et al., 1996; Shackleford, 2003) .

PAGE 58

45 Ongoing conservation and management practices should be informed by monitoring habitat degradation, recreational use, invasion by non native species, and climate change. Edaphic endemics likely colonize adverse soils due to their low competitive ability (Kruckeberg and Rabinowitz, 1985; Dillenberger and Kader eit, 2013) . With alvar soil, influx of nutrients allows colonization of non native shrubs (Catling and Brownell, 1995) . Therefore, the known threat from native and non native plant species (e.g., Thuja occidentalis L.) may destabilize populations and lead to local extinction (Comer et al., 1997) . Similarly, habitat degradation (e.g., quarrying, recreational use) could destabilize suitable habitat for subsp. convoluta and other alvar endemics (Shackleford, 2003) . The survival of subsp. convoluta may be contingent upon protecting rare habitat types, especially those inhospitable to competing plant species. Future Directions I intend to elucidate patterns of edaphi c selection and adaptive divergence using the ddRAD genomic data in tandem with my soil data. Although the data presented in Chapter 2 substantiates expectations of a broad edaphic niche for Carex scirpoidea subsp. scirpo i d e a , a broader than expected niche for subsp. convoluta was also revealed. Habitats for the two taxa range from nutrient poor, calcareous swamps (Horseshoe Bay, MI) to saline fen margins (High Creek Fen, CO). Bayesian analysis of locus population specific dif ferentiation (Foll and Gaggiotti, 2008) revealed 11 loci putatively under selection (Fig. 3.12 ) . Statistical association of genomic SNPs with environmental variables will be assessed following Rellstab et al. (2015) in order to determine whether there is evidence for genomic adaptation to local environmental con ditions. Secondly, due to the fragmented nature of habitat for subsp. convoluta , a better understanding of the mechanisms of pollen and seed (diaspore) dispersal is mandated to

PAGE 59

46 understand gene flow among populations. Disjunct, marginal populations of subs p. convoluta, in particular (e.g., Thompson's Harbor, MI and Cabot Head, ON) show evidence of gen om ic differentiation and may be threatened by restricted gene flow from central populations. However, long distance dispersal Ñ which has been well documented for the genus Ñ can be expected to lead to gene flow among populations and fragmented habitat, as well as the establishment of new populations. In subsp. convoluta , this may be facilitated by water dispersal, dispersal across snow and ice, and proximity to a shoreline. Finally, further evidence is needed to elucidate the evolutionary divergence of subsp. convoluta from E astern North American populations of subsp. scirpoidea . Currently, data are limited to the partial genomes of only seven individuals betwe en the Great Lakes and Western Greenland (this study ; Westergaard et al ., in press ). I recommend future phylogeographic studies that integrate a greater number of E astern North American populations in order to resolve the postglacial expansion of E astern North American C. scirpoidea , including the origin of subsp. convoluta , as well as the state of speciation in this taxon.

PAGE 60

47 Table 3. 1: Tissue sampling sites for Carex scirpoidea Michx . subsp. convoluta (KŸk . ) Dunlop (Cyperaceae) and subsp. scirpoidea (Cype raceae) . Taxon Sample Site Country Latitude Longitude Carex scirpoidea subsp. convoluta Thompson's Harbor, MI USA 45.35 83.58 Horseshoe Bay, MI USA 45.99 84.745 Maxton Plains, MI USA 46.08 83.66 Manitoulin Island, ON CAN 45.60 82.12 Cabot Head, ON CAN 45.24 81.30 Carex scirpoidea subsp. scirpoidea Anvil Mountain, AK USA 65.48 145.41 Butte, MT USA 46.19 112.77 High Creek Fen, CO USA 39.10 105.97 Escanaba River, MI USA 45.90 87.21 Pembina, MN USA 48.08 96.45 Brig Bay, NL CAN 51.06 56.91

PAGE 61

48 Table 3.2: Levels of genetic diversity for 11 populations of Carex scirpoidea Michx . subsp. convoluta (KŸk . ) Dunlop and subsp. scirpoidea (Cyperaceae) based on 7384 genomic SNPs. Site abbreviations are as follows: HB = Horseshoe Bay, MI, USA; TH = Thompson's Harbor, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CA N ; CH = Cabot Head, ON, CA N ; AM = Anvil Mountain, AK, USA; B = Butte, MT, USA; HCF = High Creek Fen, CO, USA; P = Pembina, MN, USA; ER = Escanaba River, MI, USA; BB = Brig Bay, NL, CA N . Taxon Site N He Ho P (%) ! " Mean + SD Mean + SD Carex scirpoidea subsp. convoluta AM 8 0.39 + 0.15 0.46 + 0.29 34.41 995.05 B 8 0.37 + 0.15 0.42 + 0.28 47.84 1295.55 HCF 8 0.38 + 0.15 0.43 + 0.28 47.45 1315.18 P 8 0.36 + 0.15 0.46 + 0.28 52.99 1417.97 E R 8 0.38 + 0.14 0.45 + 0.27 52.08 1455.08 BB 5 0.41 + 0.14 0.51 + 0.32 44.96 1367.89 Mean + SD 0.39 ± 0.04 0.44 ± 0.04 51.69 ± 4.91 1490.33 ± 46.62 Range 0.11 0.11 13.18 110.15 Carex scirpoidea subsp. scirpoidea TH 7 0.39 + 0.14 0.46 + 0.26 50.56 1442.38 HB 5 0.39 + 0.14 0.43 + 0.26 49.71 1445.16 MP 6 0.38 + 0.14 0.42 + 0.26 53.65 1503.97 MI 3 0.46 + 0.12 0.50 + 0.28 45.67 1552.53 CH 11 0.35 + 0.15 0.39 + 0.26 58.85 1507.61 Mean + SD 0.38 ± 0.02 0.46 ± 0.03 46.62 ± 6.70 1307.79 ± 164.59 Range 0.05 0.09 18.58 460.03

PAGE 62

49 Table 3 .3 : Locus by locus AMOVA results by taxon as weighted average over 7384 genomic loci of 11 populations of 2 taxa; 1) Carex scirpoidea Michx. subsp. convoluta (KŸk . ) Dunlop and 2) subsp. scirpoidea (Cyperaceae) . F statistics, including inbreeding coefficient (F IS ), variance among individual popula tions within taxa (F SC ), variance among taxa compared to total variance (F CT ), and overall fixation index (F IT ). Significant (p<0.01) variance components and fixation indices in bold. Source of variation Sum of squares Variance components Percentage vari ation Among taxa 9337.940 56.26525 5.41862 Among populations within taxa 42360.383 302.1 1 838 29.09549 Among individuals within populations 36655.009 124.60599 12.0017 Within individuals 61953.500 804.59091 77.48606 Total 150306.831 1038.36854 F IS 0.18325 F SC 0.30762 F CT 0.05419 F IT 0.22514

PAGE 63

50 Table 3 .4 : Locus by locus AMOVA results by region as weighted average over 7384 genomic loci of 11 populations of Carex scirpoidea Michx. subsp. convoluta (KŸk .) Dunlop and subsp. scirpoidea (Cyperaceae) at 3 regions : 1) Northwestern, 2) Western, and 3) E astern North American . F statistics, including inbreeding coefficient (F IS ), variance among individual populations within taxa (F SC ), variance among regions compared to total variance (F CT ), a nd overall fixation index (F IT ). Significant (p<0.01) variance components and fixation indices in bold. Source of variation Sum of squares Variance components Percentage variation Among regions 34650.637 410.01290 33.95238 Among populations within regions 17047.685 117.61385 9.73938 Among individuals within populations 36655.009 124.60599 10.31838 Within individuals 61953.500 804.59091 66.62663 Total 150306.831 1207.61166 F IS 0.18325 F SC 0.14746 F CT 0.33952 F IT 0.33373

PAGE 64

51 Table 3.5: Pairwise differentiation (F st ) measured across 7384 loci of Carex scirpoidea Michx. subsp. convoluta (KŸk.) Dunlop and subsp. scirpoidea (Cyperaceae) . Significant values (p < 0.05) in bold. Comparison between different co nspecifics in shaded area. Sit e abbreviations are as follows: HB = Horseshoe Bay, MI, USA; TH = Thompson's Harbor, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CA N ; CH = Cabot Head, ON, CA N ; AM = Anvil Mountain, AK, USA; B = Butte, MT, USA; HCF = High Creek Fen, CO , USA; P = Pembina, MN, USA; ER = Escanaba River, MI, USA; BB = Brig Bay, NL, CA N . Carex scirpoidea subsp. scirpoidea Carex scirpoidea subsp. convoluta AM B HCF P ER BB TH HB MP MI C H Carex scirpoidea subsp. scirpoidea AM B 0.49 HCF 0.50 0.10 P 0.44 0.44 0.44 ER 0.44 0.44 0.45 0.16 Carex scirpoidea subsp. convoluta BB 0.45 0.45 0.45 0.17 0.17 TH 0.44 0.45 0.45 0.16 0.18 0.16 HB 0.43 0.43 0.43 0.13 0.13 0.13 0.13 MP 0.43 0.43 0.43 0.13 0.14 0.12 0.09 0.09 MI 0.43 0.42 0.42 0.11 0.12 0.10 0.08 0.07 0.05 CH 0.41 0.43 0.43 0.15 0.15 0.14 0.13 0.11 0.10 0.07

PAGE 65

52 Figure 3. 1: Plant tissue sampling sites for (a) Carex scirpoidea Michx . subsp. convoluta (KŸk . ) Dunlop (Cyperaceae) and (b) subsp. scirpoidea (Cyperaceae) . Site abbreviations are as follows: HB = Horseshoe Bay, MI, USA; TH = Thompson's Harbor, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CA N ; CH = Cabot Head, ON, CA N ; AM = Anvil Mountain, AK, USA; B = Butte, MT, USA; HCF = High Creek Fen, CO, USA; P = Pembina, MN, USA; ER = Escanaba River, MI, USA; BB = Brig Bay, NL, CA N . a b

PAGE 66

53 Figure 3. 2 : Transect sampling plan for each Carex scirpoidea Michx. subsp. convoluta ( KŸk . ) Dunlop (Cyperaceae) population site. At each site, 24 tissue samples were obtained, 3 meters apart.

PAGE 67

54 Figure 3.3 : Expected heterozygosity for 7384 genomic loci in 11 populations of Carex scirpoidea Michx. subsp. convoluta (KŸk . ) Dunlop (preceded by c) and subsp. scirpoidea (s) (Cyperaceae) . Site abbreviations are as follows: HB = Horseshoe Bay, MI, USA; TH = Thompson's Harbor, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CA N ; CH = Cabot Head, ON, CA N ; AM = Anvil Mountain, AK, USA; B = Butte, MT, USA; HCF = High Creek Fen, CO, USA; P = Pembina, MN, USA; ER = Escanaba River, MI, USA; BB = Brig Bay, NL, CA N .

PAGE 68

55 Figure 3.4 : Nucleotide diversity ( ! " ), measured across 7384 genomic loci for 11 populations of Carex scirpoidea Michx. subsp. convoluta (KŸk .) Dunlop (preceded by c) and subsp. scirpoidea (s) (Cyperaceae) . Site abbreviations are as follows: HB = Horseshoe Bay, MI, USA; TH = Thompson's Harbor, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CA N ; CH = Cabot Head, ON, CA N ; AM = Anvil Mountain, AK, USA; B = Butte, MT, USA; HCF = High Creek Fen, CO, USA; P = Pembina, MN, USA; ER = Escanaba River, MI, USA; BB = Brig Bay, NL, CA N . Dashed lines indicating 1 standard deviation.

PAGE 69

56 Figure 3.5 : Allele count for each locus a cross 7384 loci in 11 populations of Carex scirpoidea Michx. subsp. convoluta (KŸk .) Dunlop (preceded by c) and subsp. scirpoidea (s) (Cyperaceae) . Site abbreviations are as follows: HB = Horseshoe Bay, MI, USA; TH = Thompson's Harbor, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CA N ; CH = Cabot Head, ON, CA N ; AM = Anvil Mountain, AK, USA; B = Butte, MT, USA; HCF = High Creek Fen, CO, USA; P = Pembina, MN, USA; ER = Escanaba River, MI, USA; BB = Brig Bay, NL, CA N .

PAGE 70

57 Fig 3. 6 : fastStructure plots for strong structure (K=3) and weak structure (K=5) of all assayed populations of Carex scirpoidea Michx . subsp. convoluta (KŸk . ) Dunlop (pr eceded by c) and (b) subsp. scirpoidea (s) (Cyperaceae) . Site abbreviations are as follows: HB = Horseshoe Bay, MI, USA; TH = Thompson's Harbor, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CA N ; CH = Cabot Head, ON, CA N ; AM = Anvil Mou ntain, AK, USA; B = Butte, MT, USA; HCF = High Creek Fen, CO, USA; P = Pembina, MN, USA; ER = Escanaba River, MI, USA; BB = Brig Bay, NL, CA N .

PAGE 71

58 Fig. 3 .7 . BIC values for DAPC models with K clusters.

PAGE 72

Fig 3 .8 : DAPC cluster plots of K=3 5 for Carex scirpoidea Michx . subsp. convoluta (KŸk . ) Dunlop (preceded by c) and (b) subsp. scirpoidea (s) (Cyperaceae) . Site abbreviations are as follows: HB = Horseshoe Bay, MI, USA; TH = Thompson's Harbor, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CA N ; CH = Cabot Head, ON, CA N ; AM = Anvil Mountain, AK, USA; B = Butte, MT, USA; HCF = High Creek Fen, CO, USA; P = Pembina, MN, USA; ER = Escanaba River, MI, USA; BB = Brig Bay, NL, CA N . 5 9

PAGE 73

60 Fig 3.9 : Maximum likelihood rooted phylogeny of all populations of Carex scirpoidea Michx. subsp. convoluta ( KŸk . ) Dunlop ( blue, preceded by c) and subsp. scirpoidea ( red, preceded by s) (Cyperaceae) . Site abbreviations are as follows: HB = Horseshoe Bay, MI, USA; TH = Thompson's Harbor, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CA N ; CH = Cabot Head, ON, CA N ; AM = Anvil Mountain, AK, USA; B = Butte, MT, USA; HCF = High Creek Fen, CO, USA; P = Pembina, MN, USA; ER = Escanaba River, MI, USA; BB = Brig Bay, NL, CA N . ASC_GTRGAMMA nucleotide substitution model was used, and plotted with RAxML (Stamatakis, 2014) . Bootstrap support indicated at nodes; 500 iterations.

PAGE 74

61 Figure 3.10 : Pairwise distance between loci using Nei's distance (d), mean number of pairwise differences between populations, and within populations ( ! " ) of using 7384 genomic loci in 11 populations of Carex scirpoidea Michx. subsp. convoluta ( KŸk . ) Dunlop (preceded by c) and subsp. scirpoidea (s) (Cyperaceae) . Site abbreviations are as follows: HB = Horseshoe Bay, MI, USA; TH = Thompson's Harbor, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CA N ; CH = Cabot Head, ON, CA N ; AM = Anvil Mountain, AK, USA; B = Butte, MT, USA; HCF = High Creek Fen, CO, USA; P = Pembina, MN, USA; ER = Escanaba River, MI, USA; BB = Brig Bay, NL, CA N .

PAGE 75

62 Figure 3.11 : Pairwise F ST values for 7384 genomic loci in 11 populations of Carex scirpoidea Michx. subsp. convoluta ( KŸk . ) Dunlop (preceded by c) and subsp. scirpoidea (s) (Cyperaceae) . Site abbreviations are as follows: HB = Horseshoe Bay, MI, USA; TH = Thompson's Harbor, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CA N ; CH = Cabot Head, ON, CA N ; AM = Anvil Mountain, AK, USA; B = Butte, MT, USA; HCF = High Creek Fen, CO, USA; P = Pembina, MN, USA; ER = Escanaba River, MI, USA; BB = Brig Bay, NL, CA N .

PAGE 76

6 3 Figure 3.12 : Bayescan outlier loci for 11 populations of Carex scirpoidea Michx. subsp. convoluta ( KŸk . ) Dunlop and subsp. scirpoidea (Cyperaceae) within Eastern North America. Square indicates loci above false discovery rate (FDR) of q=0.05. 11 loci were determined to be candidate loci under selection (above FDR). Populations include Horseshoe Bay , MI , USA; Thompso n's Harbor , MI , USA; Maxton Plains , MI , USA; Manitoulin Island , ON , CA N ; Cabot Head , ON , CA N ; Pembina , MN , USA; Escanaba River , MI , USA; Brig Bay , NL , CA N .

PAGE 77

64 REFERENCES A LEXANDER , P.J. , G. R AJANIKANTH , C.D. B ACON , and C.D. B AILEY . 2007. Recovery of plant DNA using a reciprocating saw and silica based columns. Molecular Ecology Notes 7: 5 Ð 9. A LLENDORF , F.W. , P.A. H OHENLOHE , and G. L UIKART . 2010. Genomics and the future of conservation genetics. Nature Reviews Genetics 11: 697 Ð 709. A NACKER , B.L. 2014. The nature of serpentine endemism. American Journal of Botany 101: 219 Ð 224. A NTONOVICS , J. 2006. Evolution in closely adjacent plant populations X: Long term persistence of prereproductive isolation at a mine bound ary. Heredity 97: 33 Ð 37. B AGELLA , S. , and M. U RBANI . 2006. Vascular flora of calcareous outcrops in North Western Sardinia (Italy). Webbia 61: 95 Ð 132. B AIRD , N.A., P.D. E TTER , T.S. A TWOOD , M.C. C URREY , A.L. S HIVER , Z.A. L EWIS , E.U. S ELKER , ET AL . 2008. Rap id SNP discovery and genetic mapping using sequenced RAD markers. PLoS ONE 3: 1 Ð 7. B ALL , P.W. , and A.A. R EZNICEK . 2003. Carex scirpoidea Michaux subsp. scirpoidea (KŸkenthal) F. of N. A. E. Committee [ed.],. Flora of North America North of Mexico 23: 551. Available at: http://www.efloras.org. B IRCHENKO , I. , Y. F ENG , and J. R OMERO S EVERSON . 2009. Biogeographical Distribution of Chloroplast Diversity in Northern Red Oak (Quercus rubra L.). The American Midland Naturalist 161: 134 Ð 145. B OISSON , S., M .P. F AUCON , S. L E S TRADIC , B. L ANGE , N. V ERBRUGGEN , O. G ARIN , A. T SHOMBA W ETSHY , ET AL . 2017. Specialized edaphic niches of threatened copper endemic plant species in the D.R. Congo: implications for ex situ conservation. Plant and Soil 413: 261 Ð 273. B RADS HAW , A.D. 1991. The Croonian Lecture, 1991. Genostasis and the limits to evolution. Philosophical Transactions of the Royal Society of London B Biological Sciences 333: 289 Ð 305. B ROOKER , R. , and R. V AN D ER W AL . 2003. Can soil temperature direct the composi tion of high arctic plant communities? Journal of Vegetation Science 14: 535 Ð 542. B ROWNELL , V.R. , and J.R. R ILEY . 2000. The alvars of Ontario: Significant alvar natural areas in the Ontario Great Lakes region. Federation of Ontario Naturalists, Don Mills, Ontario. B URGE , D. , T. S TOUGHTON , and D. J OLLES . 2017. Soil Chemistry Patterns In An Edaphic Endemism Hotspot # : The Pebble Plains of the San Bernardino Mountains , California. Madro–o 64: 9 Ð 21.

PAGE 78

65 C AREX W ORKING G ROUP . 2008. Field guide to sedges of the Pacifi c Northwest. Oregon State University Press, Corvallis, OR. C ASTILHO , P. D EL , and I. R IX . 1993. Ammonium Acetate Extraction for Soil Heavy Metal Speciation; Model Aided Soil Test Interpretation. International Journal of Environmental Analytical Chemistry 51 : 59 Ð 64. C ATLING , P.M. 1995. The extent of confinement of Vascular Plants to Alvars in Southern Ontario. Canadian Field Naturalist 109: 172 Ð 181. C ATLING , P.M. , and V.R. B ROWNELL . 1995. Alvars of the Great Lakes Region. The Canadian Field Naturalist 109: 14 3 Ð 171. C HONG , Z. , J. R UAN , and C. W U . 2012. Sequence analysis Rainbow # : an integrated tool for efficient clustering and assembling RAD seq reads. 28: 2732 Ð 2737. C LAUSEN , J. 1951. Stages in the evolution of plant species. Cornell University Press, Ithaca, N Y. C OMER , P.J. , D.L. C UTHRELL , D.A. A LBERT , and M.R. AND P ENSKAR . 1997. Natural community abstract for limestone bedrock lakeshore. Michigan Natural Features Inventory, Lansing, MI. C OOP , G. , D. W ITONSKY , A. D I R IENZO , and J.K. P RITCHARD . 2010. Using environmental correlations to identify loci underlying local adaptation. Genetics 185: 1411 Ð 1423. D ANECEK , P., A. A UTON , G. A BECASIS , C.A. A LBERS , E. B ANKS , M.A. D E P RISTO , R.E. H ANDSAKER , ET AL . 2011. The variant call format and VCFtools. Bio informatics 27: 2156 Ð 2158. D ARRIBA , D. , G.L. T ABOADA , R. D OALLO , and D. P OSADA . 2015. Europe PMC Funders Group jModelTest 2 # : more models , new heuristics and high performance computing. 9: 6 Ð 9. D E P RENGER L EVINE , M.E. 2007. Genetic diversity in a rare Nor th American endemic Carex scirpoidea ssp. convoluta. University of Colorado Denver. D ILLENBERGER , M.S. , and J.W. K ADEREIT . 2013. The phylogeny of the European high mountain genus Adenostyles (Asteraceae Senecioneae) reveals that edaphic shifts coincide with dispersal events. American Journal of Botany 100: 1171 Ð 1183. D UNLOP , D.A. 1990. The biosystematics of Carex section Sci rpinae ( Cyperaceae ) The biosystematics of Carex section Scirpinae ( Cyperaceae ). University of New Hampshire. D UNLOP , D.A. , and G.E. C ROW . 1999. The taxonomy of Carex section Scirpinae (Cyperaceae). Rhodora 101: 163 Ð 199. E LLSTRAND , N.C. , and D.R. E LAM . 1993. Population Genetic Consequences of Small Population Size # : Implications for Plant Conservation. Annual Review of Ecology and Systematics 24: 217 Ð 242.

PAGE 79

66 E XCOFFIER , L. , and H.E.L. L ISCHER . 2010. Arlequin suite ver 3.5: A new series of programs to perform population genetics analyses under Linux and Windows. Molecular Ecology Resources 10: 564 Ð 567. F AGERIA , N.K. , H.R. G HEYI , and A. M OREIRA . 2011. Nutrient bioavailability in salt affected soils. Journal of Plant Nutrition 34: 945 Ð 962. F ANT , J.B. , K. H AVENS , J.M. K ELLER , A. R ADOSAVLJEVIC , and E.D. Y ATES . 2014. The influence of contemporary and historic landscape features on the genetic structure of the sand dune endemic, Cirsium pitcheri (Asteraceae). Heredity 112: 519 Ð 530. F ELTZ , C.J. , and G.E. M ILLER . 1996. An asymptotic test for the equality of coefficients of variation from k populations. Statistics in Medicine 15: 647 Ð 658. F ISCHER , J. , and D.B. L INDENMAYER . 2007. Landscape modification and habitat fragmentation: a synthesis. Global Ecology and Biog eography 16: 265 Ð 280. F OLL , M. , and O. G AGGIOTTI . 2008. A genome scan method to identify selected loci appropriate for both dominant and codominant markers: A Bayesian perspective. Genetics 180: 977 Ð 993. F RANKHAM , R. 2005. Genetics and extinction. Biologic al Conservation 126: 131 Ð 140. F RANKHAM , R. 2010. Where are we in conservation genetics and where do we need to go # ? 661 Ð 663. F RANKHAM , R. , C.J.A. B RADSHAW , and B.W. B ROOK . 2014. Genetics in conservation management: Revised recommendations for the 50/500 ru les, Red List criteria and population viability analyses. Biological Conservation 170: 56 Ð 63. G ARRISON , E. 2012. Vcflib: A C++ library for parsing and manipulating VCF files. Available at: https://github.com/vcflib/vcflib. G ARRISON , E. , and G. M ARTH . 2012. Haplotype based variant detection from short read sequencing. 1 Ð 9. Available at: http://arxiv.org/abs/1207.3907. G ILPIN , M.E. , and M.E. S OULƒ . 1986. Minimum viable populations: Processes of species extinction. In Conservation Biology: The Science of Scarc ity and Diversity, 19 Ð 34. G IVEN , D.R. , and J.H. S OPER . 1981. The arctic apine element of the vascular flora at Lake Superior. National Museums of Canada, Ottawa. G îMEZ F ERNçNDEZ , A. , I. A LCOCER , and S. M ATESANZ . 2016. Does higher connectivity lead to highe r genetic diversity? Effects of habitat fragmentation on genetic variation and population structure in a gypsophile. Conservation Genetics 17: 631 Ð 641. G ORHAM , E. 1991. Northern peatlands: role in the carbon cycle and probable responses to climatic warming . Ecological Applications 1: 182 Ð 195.

PAGE 80

67 G UINDON , S. , and O. G ASCUEL . 2003. A simple, fast and accurate method to estimate large phylogenies by maximum likelihood. Systematic Biology 52: 696 Ð 704. H AMILTON , J.A. , and C.G. E CKERT . 2007. Population genetic consequences of geographic disjunction: A prairie plant isolated on Great Lakes alvars. Molecular Ecology 16: 1649 Ð 1660. H AMRICK , A.J.L. , and M.J.W. G ODT . 1996. Effects of life history traits on genetic diversity in plant species . Philosophical Transactions of the Royal Society of London B Biological Sciences 351: 1291 Ð 1298. H EWITT , G.M. 2000. The genetic legacy of the Quarternary ice ages. Nature 405: 907 Ð 913. H UFFT , R.A. , M.E. D E P RENGER L EVIN , R.A. L EVY , and M.B. I SLAM . 2018. Us ing herbarium specimens to select indicator species for climate change monitoring. Biodiversity and Conservation 27: 1487 Ð 1501. H USON , D.H. , and C. S CORNAVACCA . 2012. Dendroscope 3: An interactive tool for rooted phylogenetic trees and networks. Systematic Biology 61: 1061 Ð 1067. J OMBART , T. , and I. A HMED . 2011. adegenet 1.3 1: new tools for the analysis of genome wide SNP data. Bioinformatics 27: 3070 Ð 3071. K RAB , E.J., J. R OENNEFARTH , M. B ECHER , G. B LUME W ERRY , F. K EUPER , J. K LAMINDER , J. K REYLING , ET AL . 2 018. Winter warming effects on tundra shrub performance are species specific and dependent on spring conditions. Journal of Ecology 106: 599 Ð 612. K RUCKEBERG , A.R. 1951. Intraspecific Variability in the Response of Certain Native Plant Species to Serpentine Soil. American Journal of Botany 38: 408 Ð 419. K RUCKEBERG , A.R. , and D. R ABINOWITZ . 1985. Biological Aspects of Endemism in Higher Plants Publis. Annual Review of Ecology and Systematics 16: 447 Ð 479. D E L A T ORRE , A.R. , D.R. R OBERTS , and S.N. A ITKEN . 2014. Genome wide admixture and ecological niche modelling reveal the maintenance of species boundaries despite long history of interspecific gene flow. Molecular Ecolo gy 23: 2046 Ð 2059. L EE , K.N. 1999. Appraising Adaptive Management. Conservation Ecology 3: 1 Ð 18. L I , H. , and R. D URBIN . 2009. Fast and accurate short read alignment with Burrows Wheeler transform. Bioinformatics 25: 1754 Ð 1760. L ISCHER , H.E.L. , and L. E XCOFFIER . 2012. PGDSpider: An automated data conversion tool for connecting population genetics and genomics programs. Bioinformatics 28: 298 Ð 299. L OWRY , D.B. 2012. Ecotypes and the controversy over stages in the formation of new species. Biological Journ al of the Linnaean Society 106: 241 Ð 257.

PAGE 81

68 L UDWIG , B. , S. C HADDE , and L. P ROUT . 1996. Establishment Record for Horseshoe Bay: Research Natural Area within Hiawatha National Forest Mackinac County, Michigan. M ARTIN , B. 2018. raxml_ascbias. Available at: https ://github.com/btmartin721/raxml_ascbias/. M C L AUGHLIN , W.T. 1932. Atlantic Coastal Plain Plants in the Sand Barrens of Northwestern Wisconsin Author. Ecological Monographs 2: 335 Ð 383. M ODESTO , I.S. , C. M IGUEL , F. P INA M ARTINS , M. G LUSHKOVA , M. V ELOSO , O.S. P AULO , and D. B ATISTA . 2014. Identifying signatures of natural selection in cork oak (Quercus suber L.) genes through SNP analysis. Tree Genetics and Genomes 10: 1645 Ð 1660. M OORE , L. , W.K. L AUENROTH , D.M. B ELL , and D.R. S CHLAEPFER . 2015. Soil, water, and t emperature explain canopy phenology and greeness on the Shortgrass Steppe. Great Plains Research 25: 121 Ð 138. O LSON , M.S. , N. L EVSEN , R.Y. S OOLANAYAKANAHALLY , R.D. G UY , W.R. S CHROEDER , S.R. K ELLER , and P. T IFFIN . 2013. The adaptive potential of Populus bal samifera L. to phenology requirements in a warmer global climate. Molecular Ecology 22: 1214 Ð 1230. O MERNIK , J.M. 1995. Ecoregions # : A Framework for Managing Ecosystems. The George Wright Forum 12: 35 Ð 50. P ARCHMAN , T.L. , Z. G OMPERT , J. M UDGE , F.D. S CHILKEY , C.W. B ENKMAN , and C.A. B UERKLE . 2012. Genome wide association genetics of an adaptive trait in lodgepole pine. Molecular Ecology 21: 2991 Ð 3005. P EMBROOK , J.W. 2014. Systematics of Carex section Scirpinae Tuck. (Cyperaceae), with insight to the origin of edaphic endemics. University of Colorado Denver. P ETERSON , B.K. , J.N. W EBER , E.H. K AY , H.S. F ISHER , and H.E. H OEKSTRA . 2012. Double digest RADseq: An inexpensive method for de novo SNP discovery and genotyping in model and non model species. PLoS ONE 7: 1 Ð 11. P URITZ , J.B. , C.M. H OLLENBECK , and J.R. G OLD . 2014. dDocent # : a RADseq, variant calling pipeline designed for population genomics of non model organisms. PeerJ 2: 1 Ð 14. R C ORE T EAM . 2014. R: A language and environment for statistical computing. Availab le at: http://www.r project.org/. R AJ , A. , M. S TEPHENS , and J.K. P RITCHARD . 2014. FastSTRUCTURE: Variational inference of population structure in large SNP data sets. Genetics 197: 573 Ð 589. R AJAKARUNA , N. 2018. Lessons on Evolution from the Study of Edaphi c Specialization. Botanical Review 84: 39 Ð 78. R AJAKARUNA , N. 2004. The Edaphic Factor in the Origin of Plant Species. International Geology Review 46: 471 Ð 478.

PAGE 82

69 R ELLSTAB , C. , F. G UGERLI , A.J. E CKERT , A.M. H ANCOCK , and R. H OLDEREGGER . 2015. A practical guide to environmental association analysis in landscape genomics. Molecular Ecology 24: 4348 Ð 4370. R ESCHKE , C. , R. R EID , J. J ONES , T. F EENEY , H. P OTTER , A.W. G ROUP , T.N. C ONSERVANCY , and G.L. P ROGRAM . 1999. Conserving Great Lakes Alvars. The Nature Conservancey Great Lakes Program, Chicago, IL. R EZNICEK , A.A. 1994. The disjunct coastal plain flora in the great lakes region. Biological Conservation 68: 203 Ð 215. R ICHTER B OIX , A. , M. Q UINTELA , G. S EGELBACHER , and A. L AURILA . 2011. Genetic analysis of di fferentiation among breeding ponds reveals a candidate gene for local adaptation in Rana arvalis. Molecular Ecology 20: 1582 Ð 1600. R ODA , F., L. A MBROSE , G.M. W ALTER , H.L. L IU , A. S CHAUL , A. L OWE , P.B. P ELSER , ET AL . 2013. Genomic evidence for the parallel evolution of coastal forms in the Senecio lautus complex. Molecular Ecology 22: 2941 Ð 2952. R OSENBERG , N.A. 2004. DISTRUCT # : a program for the graphical display of population structure. 137 Ð 138. S ANDOVAL C ASTILLO , J. , N.A. R OBINSON , A.M. H ART , L.W.S. S TRAIN , and L.B. B EHEREGARAY . 2018. Seascape genomics reveals adaptive divergence in a connected and commercially important mollusc , the greenlip abalone (Haliotis laevigata), along a longitudinal environmental gradient. Molecular Ecology 27: 1603 Ð 1620. S CHAEFE R , C. , and D. L ARSON . 1997. Vegetation, Environmental Characteristics and Ideas on the Maintenance of Alvars on the Bruce Peninsula, Canada. Journal of Vegetation Science 8: 797 Ð 810. S HACKLEFORD , R. 2003. Conservation assessment for Bulrush Sedge (Carex sc irpoidea Michx.): Subspecies scirpoidea and convoluta (KŸkenth.) Dunlop. Milwaukee, Wisconsin. S TAMATAKIS , A. 2014. RAxML version 8 # : a tool for phylogenetic analysis and post analysis of large phylogenies. 30: 1312 Ð 1313. S TARK , K.E. , J.T. L UNDHOLM , and D. W. L ARSON . 2004. Arrested development of soil on alvars of ontario, Canada: Implications for conservation and restoration. Natural Areas Journal 24: 95 Ð 100. S TEBBINS , G.L. , and J. M AJOR . 1965. Endemism and Speciation in the California Flora. Ecological mon ographs 35: 1 Ð 35. S UBBARAO , G. V. , O. I TO , W.L. B ERRY , and R.M. W HEELER . 2003. Sodium A Functional Plant Nutrient. Critical Reviews in Plant Sciences 22: 391 Ð 416.

PAGE 83

70 W ESTERGAARD , K.B. , S. F IOR , L.P. B RUEDERLE , H.K. S TEN¯IEN , N. Z EMP , and A. W IDMER . 2018. Population genomic evidence for plant glacial survival in Scandinavia. Molecular Ecology 1 Ð 45. W HITE , S.N. , N.S. B OYD , and R.C. V AN A CKER . 2015. Temperature Thresholds and Growing Degree Day Models for Red Sorrel (Rumex acetosella) Ramet Sprouting, Emergence, and Flowering in Wild Blueberry. Weed Science 63: 254 Ð 263. W ILLIS , S.C. , C.M. H OLLENBECK , J.B. P URITZ , J.R. G OLD , and D.S. P ORTNOY . 2017. Haplotyping RAD loci: an efficient method to filter paralogs and account for physical linkage. Molecular Ec ology Resources 17: 955 Ð 965. W ORCH , S., K. R AJESH , V.T. H ARSHAVARDHAN , C. P IETSCH , V. K ORZUN , L. K UNTZE , A. B …RNER , ET AL . 2011. Haplotyping, linkage mapping and expression analysis of barley genes regulated by terminal drought stress influencing seed qual ity. BMC Plant Biology 11: 1 Ð 14. Z HOU , Y. , L. Z HANG , J. L IU , G. W U , and O. S AVOLAINEN . 2014. Climatic adaptation and ecological divergence between two closely related pine species in Southeast China. Molecular Ecology 23: 3504 Ð 3522.

PAGE 84

71 Appendi x : A. ! Descriptions of soil and leaf tissue sampling sites for Carex scirpoidea Michx . subsp. convoluta (KŸk) Dunlop . Habitat descriptions and associated species observations made on site by Leo Bruederle and Nick Bard. B. ! Weights for all soil samples taken from Ca rex scirpoidea Michx . subsp. convoluta ( KŸk . ) Dunlop (Cyperaceae) population sites before and after drying out and removal of rocks, if applicable. Numbers following the site name are the transect number, and the number of the sampling point along that tra nsect. C. ! Soil chemical and physical parameters measured across the ranges of Carex scirpoidea Michx . subsp. convoluta (Kuk . ) Dunlop and subsp. scirpoidea (Cyperaceae) . D. ! Average daily soil temperature for all days determined to be in the growing season (> 0¼C) for all populations of Carex scirpoidea Michx . subsp. convoluta (KŸk.) Dunlop and subsp. scirpoidea (Cyperaceae) . Due to missing data, extrapolations were generated for subsp. convoluta using quadratic GLMs, which comprise the first 5 plots. E. ! Average number of alleles, and associated standard deviation across 7384 measured loci for 11 populations of Carex scirpoidea Michx . subsp. convoluta (KŸk . ) Dunlop and subsp. scirpoidea (Cyperaceae) . F. ! Maximum likelihood unrooted phylogeny of all population s of Carex scirpoidea subsp. convoluta (KŸk . ) Dunlop and subsp. scirpoidea ( Cyperaceae) .

PAGE 85

72 Appendix A : Descriptions of s oil and leaf tissue sampling sites for Carex scirpoidea Michx . subsp. convoluta (KŸk . ) Dunlop (Cyperaceae) . Habitat descriptions and associated species observations made on site by Leo Bruederle and Nick Bard. Site State/Province Country Latitude & Longitude Elevation (in m) Habitat Description Associated Species Cabot Head , Ontario, CA N 45.244, 81.304 183.79 Open meadow with limestone pavement with abundant dense clumps of C. scirpoidea subsp. convoluta. Surrounded by Pinus sp., little cedar. Grazed by pigs possibly. Potentilla fruticosa, Aster sp., Solidago sp., Poa sp., Rosa sp., Pinus sp. Horseshoe Bay , Michigan, USA 45.987, 84.745 199.34 Cedar glade swamps, marl substrate, mostly open canopy, lots of dead cedars, grey soil even at top layer. Schizacryium scoparium, Picea glauca, Thuja occidentalis, Juniperus horizontalis, Arctostaphylos uva ursi, Andro meda sp., Gentiana sp., Parnassea palustris, Potentilla fruticosa Manitoulin Island, Ontario , CAN 45.596, 82.121 181.66 Right on lake shoreline. Swale between white spruce, cedar spruce glade and limestone pavements. Open canopy. Like others, an intermediate zone. Viburnum triloba, Schizacryium scoparium, Juniperus horizontalis, Potentilla fruticosa, Fragaria virginiana, T huja occidentalis, Thalictrum sp., Arctostaphylos uva ursi.

PAGE 86

73 Maxton Plains, Michigan, USA 46.075, 83.6617 174.65 Limestone pavement with scattered grasses, exposed rocks, early colonizing mosses. Juniperus horizontalis, Rhus radacans, Prunus sp ., Bryophyta sp ., Potentilla fruticosa, Schizacryium scoparium Thompson's Harbor, Michigan, USA 45.351, 83.580 178.43 Open swale with cedar. Intermediate zone, between cedar glade and lakeshore, between cedar glade and lake shore. Near trail. Common C. sci rpoidea subsp. convoluta . Lots of Great lakes endemics. Juniperus horizontalis, Potentilla fruticosa, Arctostaphylos uva ursi, Carex flayva, Shephardia canadensis, Gentiana sp. , Thuja occidentalis, Schizacryium scoparium, Aster sp. , Picea glauca or P. mariana , Rudbeckia sp.

PAGE 87

74 Appendix B: Weights for all soil samples taken from Carex scirpoidea Michx . subsp. convoluta ( KŸk . ) Dunlop (Cyperaceae) population sites before and after drying out and removal of rocks, if applicable. Numbers following the site name are the transect number, and the number of the sampling point along that transect. Abbreviations for sites are as follows: MP = Maxton Plains, MI, USA; TH = Thompson's Harbor, MI, USA; HB = Horseshoe Bay, MI, USA; CH = Cabot Head, ON, CA N ; MI = Manito ulin Island, ON, CAN. Site Before (g) After (g) Rock wt. (g) Water wt (g) H 2 O % MP 1 3 146.4 108.92 0 37.48 26 MP 2 8 79.8 47.27 0 32.53 41 MP 3 8 88.59 51.6 0 36.99 42 TH 1 5 142.69 104.46 12.35 38.23 29 TH 2 7 175.69 129.27 14.48 46.42 29 TH 3 5 142.6 103.85 5.25 38.75 28 HB 1 1 261.53 151.41 0 110.12 42 HB 2 2 201.52 107.23 0 94.29 47 HB 3 7 288.28 176 0 112.28 39 CH 1 4 212.61 146.67 0 65.94 31 CH 2 5 185.75 125.61 0 60.14 32 CH 3 8 177.22 118.6 0 58.62 33 MI 1 5 185.91 123.03 0 62.88 34 MI 2 2 204.39 164.37 0 40.02 20 MI 3 4 225.9 169.29 0 56.61 25

PAGE 88

Appendix C: All s oil chemical and physical parameters measured across the ranges of Carex scirpoidea Michx . subsp. convoluta ( Kuk . ) Dunlop and subsp. scirpoidea (Cyperaceae) . Asterisks (*) imply concentrations of exchangeable cations. The following sites were sampled: Horseshoe Bay , MI , USA ; Maxton Plains , MI , USA ; Thompson's Harbor , MI , USA ; Cabot Head , ON , CA N ; Manitoulin Island , ON , CA N ; Solvagtind , G R L; Kjelvatn , NO R ; Pembina , MN , USA ; Kjerringa , NO R ; High Creek Fen , CO , USA ; Eagle Summit , AK , USA; Twelve Mile Summit , AK , USA; Traill , G R L; Mestervig , G R L. ID Solid matter % Loss on ignition % pH Al mg/kg As mg/kg B mg/kg Ba mg/kg Be mg/kg Ca mg/kg Cd mg/kg Co mg/kg Cr mg/kg Cu mg/kg Carex scirpoidea subsp. convoluta Thompson's Harbor 97.3 19.29 7.36 5472 3.6 13.7 25 0.3 113647 0.5 1.6 5472 3.6 Thompson's Harbor 96.5 22.51 7.43 5024 2.2 13.9 26.6 0.3 117583 0.5 1.4 5024 2.2 Thompson's Harbor 96.4 23.27 7.43 5247 2.9 13 27.2 0.3 110451 0.6 1.5 5247 2.9 Horseshoe Bay 93.4 16.2 7.65 1166 2.4 12.4 43.3 0.3 362647 0.2 0.9 1166 2.4 Horseshoe Bay 95.1 18.45 7.49 1452 2.8 13.2 41.8 0.3 345156 0.4 1.1 1452 2.8 Horseshoe Bay 89.8 15.48 7.89 1285 3.5 10.8 45.6 0.3 385123 0.3 0.8 1285 3.5 Maxton Plains 96.8 18.23 7.44 16062 2.8 24.6 71 0.4 34366 1.5 4.8 16062 2.8 Maxton Plains 92.4 43.14 7.33 15695 5.8 38.4 91.9 0.5 48969 2.3 4 15695 5.8 Maxton Plains 90.1 51.34 7.25 15752 8.7 39.5 98.6 0.5 35557 2.8 5.1 15752 8.7 Manitoulin Island 97.7 14.22 7.47 6779 2.5 9 43.7 0.3 24010 0.9 3.2 6779 2.5 Manitoulin Island 98 13.25 7.42 6667 1.3 9.1 37.9 0.3 25869 0.9 2.9 6667 1.3 Manitoulin Island 98.1 12.88 7.38 7272 2.6 8.8 48.3 0.3 25297 0.8 2.9 7272 2.6 Cabot Head 95.5 23.58 7.53 25155 16.4 71.5 108.1 1 97238 2 9.5 25155 16.4 Cabot Head 95.8 25.66 7.64 16809 14.6 52.9 89.9 0.6 122812 1.7 7 16809 14.6 Cabot Head 95 27.35 7.59 24401 16 63.6 117.1 1 96239 2.3 10 24401 16 75

PAGE 89

ID Solid matter % Loss on ignition % pH Al mg/kg As mg/kg B mg/kg Ba mg/kg Be mg/kg Ca mg/kg Cd mg/kg Co mg/kg Cr mg/kg Cu mg/kg Carex scirpoidea subsp. scirpoidea Twelve Mile Summit 86.3 80.18 5.86 7554 47.7 3 218.5 0.3 18376 4.4 29.9 11.9 107.9 Twelve Mile Summit 86.5 79.81 5.55 10154 23.1 4.9 182.6 0.3 18658 1.9 8.4 16.2 103.8 Twelve Mile Summit 86 80.25 5.92 10432 7.2 5.4 178.3 0.3 23194 0.9 4.6 14.4 101.2 Eagle Summit 84.7 83.02 5.88 8337 8.2 5.9 221.6 0.3 26133 1.8 4.6 10.4 52.3 Eagle Summit 85.1 79.92 6.32 9196 8.3 7.7 210.6 0.3 28669 1.6 4.2 11.8 57.6 Eagle Summit 83.9 82.68 6.33 8564 18.2 7.1 293.1 0.3 30040 2.4 12.8 9.1 46 High Creek Fen 88.6 68.65 7.51 8406 19.6 42.6 462.4 0.8 66410 3.1 4.2 11.7 8.9 High Creek Fen 92.8 35.92 7.59 9797 7.6 29 467.7 0.3 73106 1.7 2.5 8.1 4.9 High Creek Fen 92.4 35.65 7.66 2757 10.1 27.6 942.2 0.3 213495 2.6 2.2 1.9 3 Pembina 93.2 29.43 7.68 10916 2.1 34.2 190.3 0.3 125418 0.9 3.2 13.1 11 Pembina 93.9 26.12 7.62 11643 2.3 26.7 193.8 0.3 128043 0.9 3.4 13.9 9.4 Pembina 93.1 28.34 7.6 12180 2.4 36.2 176.4 0.3 82306 1 3.5 14.7 12.7 Mestervig 97.9 7.74 6.89 20494 1.8 13.7 252.1 0.9 4809 1.5 8.1 33.3 14 Mestervig 98.7 3.98 7.13 23389 1.9 18.9 255.1 1 3340 1.6 9.4 36.2 15.6 Mestervig 99.1 2.8 7.05 17098 1.6 12.8 180.7 0.7 2265 1.2 6.8 27 11.4 Traill ¿ 99.3 2.77 6.96 9813 1.6 4.3 57.3 0.4 3125 0.7 4.1 18.2 8.3 Traill ¿ 98.6 5.12 6.9 13573 1.7 5.8 89.6 0.6 4483 0.9 5.3 20.1 11.4 Traill ¿ 99 3.77 7.09 11529 1.7 4.8 67.5 0.5 3950 0.8 4.7 17.1 8.7 Solvagtind 95.4 15.07 6.37 16636 6.6 2.2 85.1 0.5 8267 1.5 10.9 27.8 38.7 Solvagtind 92.8 29.37 6.28 21895 3.1 2.8 118.3 0.3 11932 2 11.8 80.8 43.3 Solvagtind 95.1 16.14 6.42 24589 4.4 1.7 143.7 0.3 10922 2.2 13 91.4 58.8 Kjelvatn 97.5 5.51 6.14 23265 2.6 0.6 73.7 0.5 4244 2.7 15.9 37 18.9 Kjelvatn 97.3 7.78 6.18 18215 5.7 0.6 55.3 0.5 4499 2.4 14.5 29.7 21 Kjelvatn 97.7 6.74 5.88 17118 1.1 0.7 62 0.5 3411 2.2 11.8 27.5 19.5 Kjerringa 97.9 4.9 7.24 37238 0.7 1.4 57 1.1 20384 2.7 10.6 50.5 13.2 Kjerringa 97.4 8.54 6.35 34910 0.7 1.1 36.2 1 16037 2.9 11.1 44.3 11.7 Kjerringa 97.2 9.88 7.38 30731 1.6 2.8 33.4 0.9 42818 2.4 9.2 35.1 10.2 76

PAGE 90

ID Fe mg/kg Ga mg/kg K mg/kg Li mg/kg Mg mg/kg Mn mg/k Mo mg/kg Na mg/kg Ni mg/kg P mg/kg Pb mg/kg S mg/kg Sc mg/kg Carex scirpoidea subsp. convoluta Thompson's Harbor 5066 2.2 1340 8.7 33656 103 0.8 226 6.8 274 12.1 787.7 1.2 Thompson's Harbor 4772 2.2 1263 8.3 31546 115 0.7 179 5.7 302 14.2 732.3 1.2 Thompson's Harbor 4966 2.2 1241 8.6 32016 109 0.8 172 5.7 300 16.4 714.5 1.1 Horseshoe Bay 942 2.2 301 7.2 1664 111 0.9 62 2.9 142 9.4 6027 0.2 Horseshoe Bay 1105 2.2 338 6.5 1587 117 0.9 60 4.1 162 11.9 5698 0.3 Horseshoe Bay 1009 2.2 283 7.4 1616 120 0.8 59 3.2 152 11.6 5927 0.2 Maxton Plains 15913 5.2 3750 13.2 18887 346 0.7 313 12.1 1049 32.6 1074 3 Maxton Plains 15503 5.4 4354 14.8 23882 454 0.8 256 12.6 1727 67.7 2144 2.9 Maxton Plains 16752 5.4 4275 13.5 14969 588 0.9 263 15.9 1800 80.7 2651 3 Manitoulin Island 8553 2.7 1447 8.5 11999 193 0.5 291 9.9 284 19 466.9 2 Manitoulin Island 7877 3.5 1643 9.4 12731 146 0.6 311 8.7 255 13.6 457.1 2 Manitoulin Island 8207 3.1 1874 9.1 12810 193 0.4 367 8.6 243 12.1 399.4 2.1 Cabot Head 27833 6.7 7312 23.3 55105 1519 0.8 320 17.6 646 20.5 903.1 5 Cabot Head 22416 4.4 5187 16 67185 1792 0.9 283 13.7 611 18.4 878.3 3.6 Cabot Head 27333 6.5 6258 20.9 53819 2091 0.8 283 19.3 750 27.2 1055 4.9 77

PAGE 91

ID Fe mg/kg Ga mg/kg K mg/kg Li mg/kg Mg mg/kg Mn mg/k Mo mg/kg Na mg/kg Ni mg/kg P mg/kg Pb mg/kg S mg/kg Sc mg/kg Carex scirpoidea subsp. scirpoidea Twelve Mile Summit 41762 3.5 936 2.6 1151 2146 1.2 106 32.9 1581 3.8 4066 3.1 Twelve Mile Summit 11271 2.2 1662 4.3 1771 2430 1.1 149 31.6 1754 5.6 3735 4.8 Twelve Mile Summit 7428 2.2 1488 4.6 1632 1208 0.4 152 32.5 1604 3.5 4274 5 Eagle Summit 8977 2.2 1299 2.8 1588 1877 0.9 139 23.3 1984 4.8 4704 3.1 Eagle Summit 7167 2.2 1664 3.2 1670 1776 0.6 126 20.2 2003 4.7 5276 4 Eagle Summit 11205 2.2 1494 3.1 1596 4822 0.9 114 22.1 1847 6 4274 2.8 High Creek Fen 25392 3.1 1782 7.9 10456 103 1.6 395 9 670 41.9 9560 1.1 High Creek Fen 21138 3.6 2196 8.6 7031 108 0.4 240 3.3 418 21 4198 1.2 High Creek Fen 39452 2.8 677 7.7 14506 205 0.4 309 1.5 459 18.1 3408 0.3 Pembina 9688 2.8 1717 11.3 14841 512 0.4 182 7.5 972 5.6 1387 2 Pembina 10219 3.2 1873 11.9 14622 420 0.4 194 7.8 883 5.8 1149 2.1 Pembina 10831 3.2 2030 11.5 17064 580 0.4 173 7.8 1070 6.2 1419 2.3 Mestervig 17583 6.7 6290 10.7 3423 357 0.4 212 19.5 500 9.1 460.9 4.7 Mestervig 20040 7.3 8468 10 3291 376 0.4 232 24.4 411 9.7 295.8 5 Mestervig 15418 5.8 5874 7.8 2312 278 0.4 179 16.7 303 8 174.8 3.7 Traill ¿ 9528 3.5 2296 6.9 2633 147 0.4 184 8.5 277 4.7 143.5 2.2 Traill ¿ 11887 4.4 2915 9.9 3532 218 0.4 237 11.8 311 6 232.2 3 Traill ¿ 10515 4 2571 8.1 3012 199 0.4 194 10.1 316 5.3 152.8 2.7 Solvagtind 18136 6.2 2865 15.3 4859 622 0.4 289 23.9 733 14.1 880.7 7.3 Solvagtind 24433 7.7 4759 25.3 12092 478 0.4 362 47.1 1114 9.3 1948 9 Solvagtind 27810 9.8 6809 28.7 14341 678 0.4 469 56.8 742 6.6 955.3 9.5 Kjelvatn 33448 9.8 3549 38.6 10979 582 0.4 226 27.6 580 7.6 443.4 6.4 Kjelvatn 29386 8.4 2448 34.3 7767 781 0.4 189 19.6 671 7.2 651.6 5.5 Kjelvatn 28142 7.9 2821 25.7 7207 547 0.4 178 18.5 543 8.4 506.3 4.9 Kjerringa 38752 11.2 4051 28.2 11190 353 0.4 2622 23.2 328 28.1 229 6.4 Kjerringa 41162 10.4 2411 26.3 9507 337 0.4 2168 21.1 372 83.8 380.8 5.9 Kjerringa 35630 8.4 2448 21.1 7925 308 0.4 2465 18.6 320 32 408.1 4.9 78

PAGE 92

ID Se mg/kg Sr mg/kg Ti mg/kg V mg/kg Y mg/kg Zn mg/kg C % N % C/N Vol. w t . g/l Al* mg/kg B* ug/kg Ba * mg/kg Carex scirpoidea subsp. convoluta Thompson's Harbor 2.2 54.3 113 11 3.4 29.3 11.08 0.453 24.49 971 2.2 257 4 Thompson's Harbor 2.2 54.5 112 10 3.3 27.6 12.86 0.485 26.52 851 3.1 381 5.3 Thompson's Harbor 2.2 53.7 124 9.9 3.3 34.4 12.74 0.493 25.83 858 2.8 216 5.5 Horseshoe Bay 2.2 0.1 31 1.8 0.6 20 15.73 0.488 32.25 640 1.8 996 4.3 Horseshoe Bay 2.2 829.4 36 2.3 1 23.8 16.26 0.537 30.27 629 0.7 1404 4.6 Horseshoe Bay 2.2 0.1 31 1.8 0.6 22 15.84 0.436 36.37 632 0.2 748 4.5 Maxton Plains 2.2 31.6 350 28.3 5.9 60.3 9.005 0.772 11.67 930 6 619 6.7 Maxton Plains 2.2 29.5 215 26.1 5.7 117.8 21.32 1.72 12.39 593 5.9 690 11.1 Maxton Plains 2.2 25 172 28.9 6.4 145.8 26.46 2.194 12.06 574 4.7 550 12.3 Manitoulin Island 2.2 32.9 236 18.2 5 62.5 6.884 0.315 21.88 959 3 366 8.6 Manitoulin Island 2.2 32 221 18.2 4.1 45.5 6.467 0.272 23.74 940 2.8 298 5.1 Manitoulin Island 2.2 33 202 19.2 4.1 47.3 5.844 0.242 24.19 922 2.7 326 8 Cabot Head 2.2 34.2 137 36.7 17 54.1 13.38 0.709 18.85 813 5.2 415 8.7 Cabot Head 2.2 36.4 111 26 13.3 40.1 15.06 0.631 23.88 768 5.1 537 10.1 Cabot Head 2.2 33.1 145 35.3 16 67.8 14.75 0.807 18.27 742 7.2 606 11.5 79

PAGE 93

ID Se mg/kg Sr mg/kg Ti mg/kg V mg/kg Y mg/kg Zn mg/kg C % N % C/N Vol . w t g/l Al * mg/kg B * ug/kg Ba * mg/kg Carex scirpoidea subsp. scirpoidea Twelve Mile Summit 2.2 93.1 149 16.2 26.2 98.6 44.18 2.85 15.5 177 6.1 2790 77.5 Twelve Mile Summit 2.2 97 282 18.5 23 104.2 43.58 2.61 16.7 197 5.5 2394 49 Twelve Mile Summit 2.2 119.1 261 13.7 18.1 57.7 43.9 2.83 15.5 185 2.3 1923 56.5 Eagle Summit 2.2 156.7 172 12.4 26 58.4 45.54 2.83 16.1 207 4.7 1909 83.3 Eagle Summit 2.2 165.8 189 11.7 22.6 85.6 44.23 2.86 15.4 193 2.4 1688 70.1 Eagle Summit 3.1 176.2 149 12.5 21.1 90.2 45.27 2.77 16.4 202 5.5 1737 98.6 High Creek Fen 4.9 199.6 131 24.3 7.9 59.6 37.48 2.72 13.8 199 4.4 4491 194.9 High Creek Fen 2.2 156.7 138 10.9 4.5 25.8 18.32 1.38 13.3 416 1.7 1824 170.2 High Creek Fen 2.2 405.3 46 4.3 1 22.8 22.12 1.31 16.9 382 0.9 1614 194.4 Pembina 2.2 98.5 256 21.2 4.8 51.4 14.97 1.08 13.8 695 2 2105 37.5 Pembina 2.2 104.3 270 22.7 5.2 41 13.62 0.88 15.4 738 1.4 1757 34 Pembina 2.2 81.8 293 23.7 5.4 65.2 14.07 1.12 12.6 721 1.8 2095 39.5 Mestervig 2.2 13.3 386 31.2 10.4 45.9 3.24 0.26 12.6 985 6.4 576 118.3 Mestervig 2.2 11.7 328 34.7 10.3 48.6 1.42 0.12 11.5 1176 2 399 106.9 Mestervig 2.2 9.2 262 26.6 7.2 38.4 0.93 0.08 11.2 1308 3.6 527 102.3 Traill ¿ 2.2 14 367 20.9 8.1 22.5 1.1 0.1 10.9 1189 3.9 417 19.6 Traill ¿ 2.2 19.2 398 22.3 9.6 29.9 2.1 0.17 12.3 988 3.5 487 33.6 Traill ¿ 2.2 17.3 379 20.6 10.4 28.2 1.58 0.12 13.2 1084 4 468 24.1 Solvagtind 2.2 33.6 564 33.1 58.5 41.7 7.19 0.63 11.4 547 6.4 1803 16 Solvagtind 2.2 26.5 1014 49.2 49 61.7 13.01 1.12 11.6 510 4.1 1898 21.3 Solvagtind 2.2 28.8 717 57.8 63.7 58 7.75 0.58 13.3 670 3.8 1420 19.3 Kjelvatn 2.2 21.3 610 46.9 17.1 85.9 1.97 0.19 10.2 928 4.1 1188 11.2 Kjelvatn 2.2 22.8 714 41.6 18 66.4 3.29 0.3 11.1 910 3 1240 6.5 Kjelvatn 2.2 17.3 799 38.4 13.2 62.7 2.88 0.26 11.2 948 3.1 1339 8.2 Kjerringa 2.2 341.3 630 24.9 18.2 115.1 2.1 0.13 16.5 1016 4.9 1241 11.3 Kjerringa 2.2 293.2 636 24.5 15.7 209.2 3.89 0.26 14.8 822 4.8 1640 4.8 Kjerringa 2.2 434.1 904 16.4 16.4 110.4 5.21 0.28 18.4 818 2.7 2.2 434.1 80

PAGE 94

ID Be * ug/kg Ca * mg/kg Cd * ug/kg Co * ug/kg Cr * ug/kg Cu * ug/kg Fe * mg/kg K * mg/kg Li * ug/kg Mg * mg/kg Mn * mg/kg Mo * ug/kg Na * mg/kg Carex scirpoidea subsp. convoluta Thompson's Harbor 12 7458 42 14 20 277 1.74 56.4 140 312.3 4.14 193 6.2 Thompson's Harbor 12 9499 40 14 20 277 2.25 80.8 149 433.8 6.66 189 6.6 Thompson's Harbor 12 8748 36 14 60 277 2.15 77.3 163 404 5.19 202 6.5 Horseshoe Bay 12 16008 38 14 96 277 2.28 37.9 397 221.9 4.43 339 10.3 Horseshoe Bay 12 16367 59 14 54 277 1.87 40.7 305 236.9 4.29 359 7.1 Horseshoe Bay 12 15627 44 14 20 277 1.66 32.7 343 83.46 4.91 409 6 Maxton Plains 12 5594 69 14 20 277 3.73 127.7 158 1209 2.98 233 5.8 Maxton Plains 91 11190 146 64 87 277 4.05 254.3 187 2493 9.6 333 6.5 Maxton Plains 12 13473 97 14 20 277 3.43 338.6 232 2965 10.23 295 7.6 Manitoulin Island 12 3887 104 14 45 277 3.29 43.6 82 733 9.35 149 5.5 Manitoulin Island 12 3336 114 14 20 277 2.98 48.9 82 690.2 5.06 146 5.5 Manitoulin Island 12 3211 77 14 20 277 2.91 59.3 82 652.5 9.03 127 7.3 Cabot Head 12 6157 35 14 93 277 3.1 145.6 200 1345 9.56 180 8.2 Cabot Head 12 6004 41 14 20 277 3.46 117.7 129 1240 10.45 180 7.1 Cabot Head 12 7292 50 14 128 277 4.17 152.9 202 1499 11.08 199 8.6 81

PAGE 95

ID Be * ug/kg Ca * mg/kg Cd * ug/kg Co * ug/kg Cr * ug/kg Cu * ug/kg Fe * mg/kg K * mg/kg Li * ug/kg Mg * mg/kg Mn * mg/kg Mo * ug/kg Na * mg/kg Carex scirpoidea subsp. scirpoidea Twelve Mile Summit 12 13724 279 148 20 277 28.65 263.4 1132 342.6 45.49 19 39.7 Twelve Mile Summit 12 13005 397 88 20 277 6.84 388.9 925 372 121.4 19 40.8 Twelve Mile Summit 12 16442 126 14 20 277 3.69 282.6 1103 374.4 25.85 19 50.3 Eagle Summit 12 18495 385 14 20 277 7.01 319.3 1282 626.4 96.06 19 55 Eagle Summit 12 20619 198 14 20 277 3.6 347 1247 630 17.7 19 42.7 Eagle Summit 12 22259 250 14 20 277 9.14 339.5 1678 663.3 31.08 19 36.3 High Creek Fen 397 28377 113 128 20 277 26.79 292 2380 4848 10.5 312 301.5 High Creek Fen 12 21925 26 14 20 277 17.7 96.8 1611 2681 6.43 19 119.9 High Creek Fen 12 26431 7 42 20 277 17.96 128.7 1933 3137 11 19 146 Pembina 12 15339 7 14 50 277 1.16 151.7 255 1560 2.69 19 20 Pembina 12 14175 7 14 20 277 1.14 156 220 1723 3.78 19 18.5 Pembina 12 14882 7 14 20 277 1.13 186.2 220 1523 2.08 19 17.4 Mestervig 12 3030 13 14 20 277 4.98 47.9 239 95.61 1.89 19 22.9 Mestervig 12 1793 9 14 20 277 1.85 44 220 75.61 1.21 19 28.3 Mestervig 12 1223 6 14 20 277 2.76 31.2 147 55.32 0.88 19 13.4 Traill 12 1186 7 14 20 277 2.53 24.6 106 57.9 0.84 19 8.9 Traill 12 2226 7 14 20 277 2.42 29.8 216 103.3 0.74 19 13.2 Traill 12 1630 7 14 20 277 3.21 30.2 163 69.68 0.58 19 8.5 Solvagtind 12 3705 34 14 20 604 1.87 122.6 82 143.4 4.56 19 34.2 Solvagtind 12 7432 48 14 20 277 2.28 161.8 82 218.2 9.91 19 35.4 Solvagtind 12 6046 16 14 20 463 2.41 102.4 82 161.4 4.57 19 26.6 Kjelvatn 12 2010 11 14 20 277 4.02 41.6 82 64.33 1.54 19 17.2 Kjelvatn 12 2428 17 14 37 277 4.07 38.4 82 50.35 1.55 19 14.3 Kjelvatn 12 1553 16 14 20 277 3.78 40.7 82 56.13 1.46 19 13.2 Kjerringa 12 3839 15 14 20 277 8.72 61 90 51.41 2.82 19 41.6 Kjerringa 12 2617 28 14 20 277 6.09 49.7 82 90.51 3.28 19 33.8 Kjerringa 12 6025 31 14 20 277 11.55 31.2 119 76.2 6.07 19 18 82

PAGE 96

ID Ni * ug/kg P * mg/kg Pb * ug/kg S * mg/kg Se * ug/kg Si * mg/kg Sr * ug/kg Ti * ug/kg V * ug/kg Zn * ug/kg H * mg/kg I on ic exch ange capacity * D eg ree of base saturation * Carex scirpoidea subsp. convoluta Thompson's Harbor 20 9.87 349 15.2 239 15.7 7313 34 56 537 0 399.7 100 Thompson's Harbor 20 13.12 506 15.2 239 20.1 8219 104 52 719 0 512.3 100 Thompson's Harbor 27 11.5 487 16 239 17.6 8803 26 130 604 0 472.2 100 Horseshoe Bay 20 8.36 519 247.1 239 10.8 77678 26 52 1343 0 818.7 100 Horseshoe Bay 46 10.96 527 230.8 239 11.2 81378 26 52 1889 0 837.7 100 Horseshoe Bay 48 7.55 519 164.9 239 7.8 53263 26 52 1307 0 787.9 100 Maxton Plains 20 8.17 400 15.4 239 25.4 3374 266 52 616 0 382.3 100 Maxton Plains 50 21.51 731 29.6 239 24.9 5120 100 279 1854 0 770.6 100 Maxton Plains 20 27.05 758 33.2 239 24 5925 26 52 1729 0 925.6 100 Manitoulin Island 27 9.07 483 10.5 239 16.6 3436 163 52 3123 0 255.9 99.9 Manitoulin Island 20 8.32 323 13 239 18.2 2473 49 106 2026 0 224.9 99.9 Manitoulin Island 25 8.91 232 10.9 239 17.9 2371 64 52 2534 0 216.1 99.8 Cabot Head 20 1.67 232 18.7 239 25.8 5272 180 106 190 0 422.3 99.9 Cabot Head 20 1.5 217 12.3 239 27.3 4308 156 52 299 0 405.3 99.9 Cabot Head 20 1.12 192 17.3 239 30.9 5791 291 124 277 0 491.9 99.9 83

PAGE 97

ID Ni * ug/kg P * mg/kg Pb * ug/kg S * mg/kg Se * ug/kg Si * mg/kg Sr * ug/kg Ti * ug/kg V * ug/kg Zn * ug/kg H * mg/kg I on exch . C a p* D eg . of base sat * Carex scirpoidea subsp. scirpoidea Twelve Mile Summit 689 16.28 476 91 239 71.7 73123 26 52 8353 326.3 1046 69 Twelve Mile Summit 573 22.21 426 89.4 239 86.5 69466 26 52 11176 404.8 1096 63 Twelve Mile Summit 536 21.75 363 81.2 239 60.2 90516 26 52 5596 335.4 1194 72.1 Eagle Summit 328 19.52 356 121.1 239 44.7 119673 26 52 6141 316.9 1302 75.6 Eagle Summit 167 26.35 426 115.7 239 60.2 129965 26 52 5246 202.9 1293 84.4 Eagle Summit 147 22.33 412 103.2 239 50.9 143503 309 52 5933 233.1 1407 83.5 High Creek Fen 20 13.1 893 516.1 239 155.1 108295 26 52 2763 0 1836 100 High Creek Fen 20 5.66 311 194.9 239 108.8 71746 26 52 1106 0 1323 100 High Creek Fen 20 4.13 312 265.7 239 158 90274 26 52 574 0 1587 100 Pembina 20 11.41 243 17.2 239 94.8 19587 324 182 243 0 898.6 100 Pembina 20 10.72 227 15.2 239 75.5 18557 26 52 256 0 854 100 Pembina 20 13.86 253 14.6 239 117.3 22243 26 52 198 0 873.5 100 Mestervig 21 0.95 78 7 239 18.4 5333 451 52 449 13.1 174.3 92.5 Mestervig 30 0.33 69 6.2 239 11.9 3286 185 52 465 7.4 105.5 93 Mestervig 20 0.33 69 3.2 239 11.9 2286 166 52 475 7.9 74.8 89.5 Traill ¿ 20 0.81 66 4.2 239 19 2961 227 52 968 4.8 69.8 93.1 Traill ¿ 20 1.15 80 5.7 239 19.7 5400 198 52 654 17 137.7 87.8 Traill ¿ 20 0.93 70 4.1 239 22 3895 214 52 878 5.2 93.4 94.4 Solvagtind 273 5.53 432 29.8 239 45.9 5390 26 52 700 76 276.7 72.7 Solvagtind 132 8.8 402 48 239 40.3 9071 539 52 930 81.9 476 82.9 Solvagtind 231 3.87 235 22.6 239 23.5 7113 78 52 863 57.7 376 84.8 Kjelvatn 66 2.39 133 10.1 239 16.3 7480 235 52 873 53.8 160.7 66.8 Kjelvatn 57 2.45 121 12.7 239 8.8 9713 254 68 1028 54.2 180.6 70.3 Kjelvatn 54 2.77 127 11.3 239 9.3 5945 178 52 1728 59 142.2 58.9 Kjerringa 55 2.04 196 4.4 239 13.8 29120 166 52 445 5.6 204.8 97.2 Kjerringa 78 2.89 812 10.2 239 11.8 15495 166 75 753 44.5 184.9 76.1 Kjerringa 41 2.94 230 8 239 12.9 28438 41 52 670 0 308.7 99.9 8 4

PAGE 98

85 Appendix D: Average daily soil temperature for all days determined to be in the growing season ( > 0¼C) for all populations of Carex scirpoidea Michx . subsp. convoluta (KŸk.) Dunlop and subsp. scirpoidea (Cyperaceae) . Due to missing data, extrapolations were generated for subsp. convoluta using quadratic GLMs, which comprise the first 5 plots.

PAGE 99

86

PAGE 100

87

PAGE 101

88

PAGE 102

89

PAGE 103

90

PAGE 104

91

PAGE 105

92 Appendix E: Average number of alleles, and associated standard deviation across 7384 measured loci for 11 populations of Carex scirpoidea Michx . subsp. convoluta (KŸk) Dunlop and subsp. scirpoidea. Plant site abbreviations are as follows: HB = Horseshoe Bay, MI, USA; TH = Thompson's Harbor, MI, USA; MP = Maxton Plains, MI, USA; MI = Manitoulin Island, ON, CA N ; CH = Cabot Head, ON , CA N ; AM = Anvil Mountain, AK, USA; B = Butte, MT, USA, HCF = High Creek Fen, CO, USA; P = Pembina, MN, USA; E R = Escanaba River, MI, USA; BB = Brig Bay, NL , CA N . Taxon Population Average Number of Alleles Standard Deviation Carex scirpoidea subsp. convoluta TH 2.004 0.256 HB 2.004 0.257 MP 2.003 0.239 MI 2.004 0.249 CH 2.005 0.269 Carex scirpoidea subsp. scirpoidea AM 2.003 0.229 B 2.005 0.270 HCF 2.004 0.256 P 2.005 0.264 ER 2.005 0.268 BB 2.005 0.263

PAGE 106

93 Appendix F: Maximum likelihood unrooted phylogeny of all populations of Carex scirpoidea Michx. subsp. convoluta (KŸk . ) Dunlop ( blue, preceded by c) and subsp. scirpoidea ( red, preceded by s) (Cyperaceae) . Plant site abbreviations are as follows: HB = Horseshoe Bay , MI, USA; TH = Thompson's Harbor , MI, USA; MP = Maxton Plains , MI, USA; MI = Manitoulin Island ON, CA N ; CH = Cabot Head ON, CA N ; AM = Anvil Mountain A K, USA; B = Butte MT, USA, HCF = High Creek Fen CO, USA; P = Pembina MN, USA; ER = Escanaba River MI, USA; BB = Brig Bay NL, CA N . ASC_GTRGAMMA nucleotide substitution model was used, and plotted with RAxML (Stamatakis, 2014) . Bootstrap support indicated at nodes; 500 iterations.