(FULMARUS GLACIALIS)? by Lila Maya Colston-Nepali

Home , Prince Leopold Island

CAN GENOMIC TOOLS AID CONSERVATION OF AN ARCTIC SEABIRD,

THE NORTHERN FULMAR (FULMARUS GLACIALIS)?

Lila Maya Colston-Nepali

A thesis submitted to the Department of Biology in conformity with the requirements for the

degree of Master of Science

Queen’s University

Kingston, Ontario, Canada

September 2019

ABSTRACT

Human activities and climate change threaten Arctic ecosystems. Population genetics and genomics may help conservationists appropriately manage threatened species both by (1) determining the population genetic structure of a species, so that management can be designed to maximize conservation of genetic variation, and (2) enabling assessment of impacts on breeding populations of mortality during the nonbreeding season in migratory species. The northern fulmar (Fulmarus glacialis) is a seabird that breeds in colonies throughout the North Atlantic and

Pacific Oceans. Though not currently considered at risk, several concerns, such as increased levels of toxins and ingested plastics, warrant investigation to aid conservation of northern fulmars. Up to 1% of the global population is killed annually through unintentional capture in commercial fishing activities, and northern fulmar survival appears to be negatively affected by climate change. As northern fulmars are migratory, the impact of these mortality sources on specific colonies is often unknown but may be important to inform management strategies. In this thesis, I use restriction site-associated DNA sequencing (RADseq) to provide 6614 genome- wide single nucleotide polymorphisms (SNPs) to investigate the genetic structure of the Atlantic northern fulmar, using 127 samples from six breeding colonies, one non-breeding location, and fishing activities in the Baffin Bay-Davis Strait region. I found weak genetic differentiation and suggest that Atlantic northern fulmar populations are genetically connected, experiencing high levels of gene flow. Determining the exact breeding origin of the bycatch birds was difficult due to the lack of differentiation between colonies. However, the birds appear to be from Arctic

Canadian colonies, suggesting that the impact of the fisheries is on local colonies, and may be contributing to the 3% annual decline that has been observed at these locations.

ii ACKNOWLEDGEMENTS

Firstly, I would like to acknowledge that much of the sampling for this project occurred on the traditional territories of the Inuit people. Without the help and support of Inuit communities this project would have never been possible. I want to acknowledge my position as a settler scientist and hope that any outcomes from this project are in accordance with the National Inuit Strategy on Research.

I would like to further acknowledge the environmental impact of my project. Extraction of DNA requires numerous single use plastic tubes and pipette tips. The project also required long-distance travel by many researchers to collect samples and to attend conferences, and the use of high-performance computing clusters to analyze genomic data. Our lab has been working towards reducing our environmental impact, by attempting to recycle the plastic waste that we can and moving from the use of hazardous chemicals such as phenol and chloroform for extraction to more environmentally friendly methods such as salt extraction. I believe we still have a long way to go to reduce the impact that science has on the environment, and I will continue to work towards this throughout my career.

I would like to thank my supervisor, Vicki Friesen, for three years of guidance, support and compassion through both my Honors and Masters projects. Thank you for the incredible opportunity you gave me, everything you taught me, and all the encouragement you provided along the way! Thank you to Virginia Walker and Robert Colautti for serving on my committee, providing me with great advice, and being so lovely to work with! Thank you so much to my collaborators Jennifer Provencher, Mark Mallory, Kyle Elliot, Marie-Josée Fortin, Tony Gaston,

Grant Gilchrist, and Greg Robertson for all your wisdom and the incredible amount of work you put in to help this project happen. I feel honored to have had the opportunity to work with all of

iii you. Jenn, you have quite the fanbase of women in STEM here at Queen’s, and I feel so lucky that I have been able to learn from you.

Thank you to Tim Birt for all your help with lab work and logistics, and patiently providing help when I needed it. Thank you to the lovely Emma Gillesse and Sasha Main for their help with lab work. Thank you to Ryan Franckowiak, for being such a wonderful coach and mentor to me through every step. Your contribution to this lab is immeasurable! Thank you so much to Zhengxin Sun for library preparation- you were so kind and patient, and such a friendly face! Thank you to the Centre for Applied Genomics for sequencing. Thank you to the Centre for

Applied Computing, and Gang Liu, Hartmut Schmider and Jeff Stafford for providing help with the computing cluster. Thank you to Anna Tigano, my first mentor who taught me so much about bioinformatics, how to think like a scientist, and to be resilient. Thank you for continuing to provide your support and advice to me from afar! Thank you to Evelyn Jensen, Peiwen Li and

Yihan Wu for taking the time to help me with bioinformatics concerns, I so appreciate your kindness. You three are very talented and I know you will all make some amazing contributions to science! Thank you to Dave Fifield for providing me with field experience, being a great mentor, and showing me the incredible world of a seabird colony! I will never forget my time spent on Gull Island. Thank you to Noel McDermott for advising me with your knowledge of

Inuktitut.

Thank you to my wonderful lab mates for being there for me through both the laughter and the tears, you made this experience for me! I know we will be friends for life! In particular, thank you to Brody Crosby, Ryan Franckowiak, Bronwyn Harkness, Emma Lachance Linklater,

Drew Sauve, Becky Taylor and Russell Turner with a very special thanks to Katie Birchard, for being my lab mate, my roommate and one of my very best friends. I couldn’t have done this

iv without you. I would also like to thank the ‘BioBabes’: a group of the most supportive, incredible women who have been a rock to me throughout my degree. You inspire me, and I can’t wait to see all the amazing things you will all achieve over the years! Thank you to all of my friends in the department, the Biology community is very special and will always have a place in my heart. Thank you so much to Joanne Surrette and Sue Burrows for being the backstage heroes- none of us could do this without all of your help and support!

This project would not have been possible without the hard work of so many people.

Thank you to Jaypootee Aliqatuqtuq, Jonathon Aliqatuqtuq, Harry Alookie, John Alookie,

Uluriak Amarualik, Captain Joey Angnatok, John Ross Angnatok, Stevie Aulaqiaq, Stephanie

Avery-Gomm, Lisa Bjørnsdatter Kn, Maria Dam, Rian Dickson, Ryan Franckowiak, Tony

Gaston, Catherine Geoffroy, Michael Gendron, D. Jeffrey, Jaloo Kooneeliusie, Peter

Kooneeliusie, Alison Kopalie, Amy-Lee Kouwenberg, Mark Mallory, Captain Lloud Normore,

Morely Normore, Marcel O’Brian, Benny Saimat, Lee Shepard, Jeanie Toomasie, the community of Qikitarjauq, the Nattivak Hunters’ and Trappers’ Organization, the Canadian

Wildlife Service and the Norwegian Institute for Nature Research for sample collection. Thank you to the National Wildlife Research Centre for providing tissues and to Guy Savard, Lana

Dolgova, Kelsey Ha, and Bronwyn Harkness for providing sampling assistance.

Thank you to Environment and Climate Change Canada, the Natural Sciences and

Engineering Research Council, the Canadian Graduate Scholarships Program, the School of

Graduate Studies, and the Queen’s Biology Department for funding.

Finally, thank you to all my wonderful friends and family who have supported me through every step. In particular, I’d like to thank my parents. Ailsa Colston, thank you for raising me to be a strong woman and for providing never-ending support and love, no matter

v what time of day or night I needed you, and regardless of which time zones we were in. Thank you for always encouraging me to have fun and pursue happiness above success. Ganga Nepali, thank you for leading by example and teaching me that hard work, compassion and a strong moral compass can change the world. You have always done all you could to provide me the opportunities you never had, and I will always be so grateful. This thesis is as much a result of your hard work as it is mine.

I’d like to dedicate this thesis to Maina Nepali, my हजुरआमा (paternal grandmother), who was denied her right to learn to read, but whose sacrifices and struggles allowed me to be here submitting a Masters thesis today.

vi TABLE OF CONTENTS

Abstract...... ii

Acknowledgements...... iii

Table of Contents...... vii

List of Tables...... ix

List of Figures...... x

List of Abbreviations...... xi

Chapter 1: Introduction...... 1 Protecting genetic variation...... 1 Genetic diversity indices...... 4 The use of genomic tools to address conservation concerns...... 5 The northern fulmar...... 7 Conservation concerns...... 9 Study aims and predictions...... 10 Significance...... 11

Chapter 2: Methods...... 14 Sample collection and DNA extraction...... 14 Library preparation and sequencing...... 14 Assembling and identifying loci...... 15 Variant calling and filtering...... 16 Population genetic structure analyses...... 17 Population genetic structure...... 18 Searching for SNPs under selection...... 19 Genetic diversity indices...... 19 Isolation by distance...... 20 Gene flow between colonies...... 20 Population assignment...... 21 Unknown individuals...... 21 Machine learning methods...... 22 Training and holdout loci...... 23 Genetic stock identification...... 24

Chapter 3: Results...... 27 Sample processing...... 27 Population genetic structure analyses...... 27 Population genetic structure...... 27 Searching for SNPs under selection...... 28

vii Genetic diversity indices...... 28 Isolation by distance...... 28 Gene flow between colonies...... 28 Population assignment...... 29 Unknown individuals...... 29 Machine learning methods...... 29 Training and holdout loci...... 30 Genetic stock identification...... 31

Chapter 4: Discussion...... 50 Population genetic structure...... 50 Population assignment...... 55 Conservation implications...... 57

Future directions...... 62

References...... 65

Appendix...... 78

viii LIST OF TABLES

Table 1. Sampling locations, type of location, current subspecies designation, number of samples, and numbers of individuals after data filtering in this study...... 26

Table 2. Pairwise FST for the six colonies (above the diagonal), and probability of difference from 0 (below the diagonal)…...... 33

Table 3. Number of private alleles per colony using the reduced breeding data set of 105 individuals...... 34

Table 4. Pairwise estimates of FST between the birds of unknown origin and the six breeding colonies...... 35

Table 5. Private allele counts for all locations included in the study using the complete dataset of 127 individuals (not corrected for sample size)...... 36

Table 6. Assignment accuracy for different combinations of sampling sites and SNP datasets for the training and holdout method...... 37

Table 7. Assignment accuracy for each combination of groups using genetic stock identification methods...... 38

Table 8. Assignment results for unknown individuals...... 39

Table 9. Estimated number of breeding pairs of northern fulmars at the six colonies included in this study...... 64

ix LIST OF FIGURES

Figure 1. A) Map of the North Atlantic Ocean displaying the breeding distribution and the year-round migratory distribution of the northern fulmar (adapted from Gravley et al. 2019). B) Map of the North Atlantic Ocean indicating the six breeding colonies of fulmars that were sampled for this project...... 13

Figure 2. The first two principal components of a principal components analysis for breeding birds using all 6614 SNPS...... 41

Figure 3. The frequency of private alleles within populations (based on sample size corrected calculations, with a maximum possible of eight individuals)...... 42

Figure 4. Genetic distance (Slatkin’s linearized FST ) vs. log-transformed marine distance between colonies...... 43

Figure 5. The first two principal components of a principal components analysis using all 6614 SNPs and all birds included in the study...... 44

Figure 6. Assignment accuracies using all SNPs...... 45

Figure 7. The first two principal components of a principal components analysis using the reduced SNP dataset (253 SNPs)...... 46

Figure 8. Assignment accuracies using reduced SNP datasets...... 47

Figure 9. Principal components analysis using the reduced SNP dataset (122 SNPs)...... 48

Figure 10. Z-score distribution for all colonies compared to the expected distribution...... 49

x LIST OF ABBREVIATIONS

AUs Adaptive units bp Base pairs

C Celsius ddRADseq Double digest restriction site-associated DNA sequencing

DNA Deoxyribonucleic acid

ECCC Environment and Climate Change Canada

ESUs Evolutionary significant units

FST Wright’s fixation index

GSI Genetic stock identification

He Expected heterozygosity

K Number of genetic clusters

Km Kilometers

MC1R Melanocortin-1 receptor

MCMC Markov Chain Monte Carlo mtDNA Mitochondrial DNA

MUs Management units p Statistical probability value

PCA Principal component analysis

PCoA Principal coordinates analysis

PCR Polymerase chain reaction r Correlation coefficient

RAD Restriction site-associated DNA

xi RADseq Restriction site-associated DNA sequencing

SNPs Single nucleotide polymorphisms

xii CHAPTER 1: INTRODUCTION

The Earth is experiencing its sixth mass extinction event, with huge declines of biodiversity that are predicted to continue at a severe rate (Bellard et al. 2012, Ceballos et al. 2017). Climate changes, increasing anthropogenic activities, and the interaction between these pose a major threat in particular to Arctic ecosystems (Hoegh-Guldberg and Bruno 2010, Bennett et al. 2015,

Wauchope et al. 2016, Macias-Fauria and Post 2018). Sea ice is retreating, and late summer ice- cover was at its lowest in over 100 years in 2012 (Parkinson and Comiso 2013). The decline in sea ice results in habitat loss and fragmentation, changes the availability of food and trophic structure for marine organisms, alters the vegetation, and allows range expansion of some species (Bennett et al. 2015, Pecl et al. 2017). The loss of sea ice also allows greater industrial fishing (Christiansen et al. 2014), and expanded shipping routes (Chircop 2007). Other direct human impacts in the Arctic, such as the construction of pipelines for oil extraction, are also likely to increase (Sherval 2015). Greater human activity in the Arctic introduces contaminants, creates disturbance, and increases the rate of fisheries bycatch of non-target organisms. Human activities are likely to continue increasing (McGowan 2018), and the complex species interactions that are affected by these changes are insufficiently understood (Pecl et al. 2017).

Informed, proactive management plans are therefore needed for Arctic species.

Protecting genetic variation

Protection of biodiversity should operate on three levels: ecosystems, species and genes

(McNeely et al. 2010). Genetic diversity is often overlooked (Coates et al. 2018) but is essential for maintaining the evolutionary potential of a species. Standing genetic variation is an indicator for a species’ ability to adapt to environmental change including anthropogenic stressors (Barrett

1 and Schluter 2008), and a lack of genetic variation, especially in quantitative traits, may increase the risk of extinction in wild populations (Frankham 2005). Smaller populations are likely to have less genetic variation due to genetic drift (Lacy 1987, Frankham 1996), are more likely to lose beneficial variation, and may have maladaptive variation reach fixation. Furthermore, inbreeding in smaller populations may lead to extinction through inbreeding depression

(Frankham and Ralls 1998). Therefore, protecting high levels of genetic diversity should be a priority for conservation practitioners.

A species’ genetic variation may be geographically structured, consisting of a number of populations that are genetically differentiated, or distinct. The amount of differentiation is largely predicted by the amount of genetic drift that occurs within populations (which increases divergence among populations) and the amount of gene flow between populations (which decreases divergence) (Allendorf et al. 2013). Genetic drift, in turn, depends on population size, as smaller populations experience faster rates of drift. Time since separation also affects differentiation between populations, as populations formed by recent expansion may have experienced less genetic drift. Furthermore, selection acting upon populations may increase or decrease genetic differentiation, depending on the type of selection (Allendorf et al. 2013). For example, balancing selection acting upon several populations will act to decrease differentiation between these populations, whereas divergent directional selection will maintain large allele frequency differences between populations.

If populations are genetically differentiated the loss of a population will reduce overall genetic variation within a species (Ehrlich 1988). Conserving distinct populations with differing selection pressures will permit evolutionary processes to continue to act within natural populations (Allendorf et al. 2013). Protection of distinct populations is therefore important for

2 the persistence of a species, and conservation practitioners may define genetically divergent populations as separate ‘units’ requiring specific management plans. (Note that delineation of conservation units might also incorporate ecological data and political boundaries; Crandall et al.

2000, Fraser and Bernatchez 2001, Green 2005, Allendorf et al. 2013.) Conservation resources may be limited and understanding the evolutionary relatedness of populations allows more efficient management, by allowing prioritization of the most distinct populations (Ryder 1986).

Several types of conservation units are used by conservation geneticists, potentially distinguished by different types of markers. Management units (MUs) are demographically independent units that experience restricted dispersal (Moritz 1994). MUs may be genetically diverged populations, determined via differentiation in neutral variation (Pallsbøll et al. 2007,

Funk et al. 2012), which indicates limited contemporary gene flow and/or occurrence of genetic drift. Differences in adaptive variation indicate differences in selection acting upon a population, and so adaptive variation may bias estimates of gene flow, and should not be included in the delineation of MUs (Liukart et al. 2003). Identifying MUs allows managers to monitor direct effects of stressors on specific populations and determine short-term local conservation strategies, such as fishing regulations (Allendorf et al. 2013). Alternatively, if populations are differentiated in both neutral and adaptive variation, they may be designated as evolutionary significant units (ESUs). ESUs contribute significantly to the evolutionary and adaptive potential of species (Funk et al. 2012).

Funk et al. (2012) suggested defining units for conservation in a hierarchical manner, and

Barbosa et al. (2018) demonstrated the utility of this method on Cabrera voles (Microtus cabrerae). Firstly, ESUs should be assessed using all loci, as ESUs reflect the evolutionary history of a species and will have been formed with both neutral and adaptive evolutionary

3 processes. The fixation index, FST, is a metric that measures deviations of genotype frequencies from Hardy-Weinberg proportions. FST outlier analyses search for single nucleotide polymorphisms (SNPs) that have an FST value that is significantly different from an expected range of neutral SNPs. SNPs that demonstrate significantly larger values may indicate divergent directional selection, and those with significantly smaller values may indicate balancing selection. Therefore, FST outlier analyses can be used to separate putatively neutral SNPs from putatively adaptive SNPs. (Note that detecting adaptive loci is difficult and has numerous caveats, see Shafer et al. 2015.) Putatively neutral SNPs can be used to define MUs within each

ESU, which can be prioritized based on genetic distinctiveness (see Genetic diversity indices, below). Finally, performing a clustering analysis with the FST outliers may identify putatively adaptive units (AUs) for conservation. The resulting conservation units can be presented to conservation practitioners to inform management decisions.

Genetic diversity indices

To help prioritize population units for conservation, the genetic distinctiveness of the unit (MU or ESU) should be understood. Genetic diversity can be indexed in a number of ways: Barbosa et al. (2018) employed private alleles to indicate uniqueness. Private alleles are only present in one population, or unit, and may exist due to genetic drift during long-term isolation, or selection

(Allendorf et al. 2013). When populations experience drift, genetic variation is lost, which is reflected in reduced heterozygosity of a population. Average expected heterozygosity, He, provides a metric of genetic diversity (Allendorf et al. 2013), and is not affected by sample size when many loci are used. However, heterozygosity does not provide a measure of rare alleles, which may be important for the evolutionary potential of a species (Greenbaum et al. 2014).

Allelic richness, the number of alleles at each locus, corrected for sample size, is a useful metric

4 alongside heterozygosity to assess the genetic diversity of a population (Allendorf et al. 2013).

Several other indices for diversity exist, but the combination of private alleles, expected heterozygosity, and allelic richness provides a reliable indicator of the genetic distinctiveness and diversity of a population or conservation unit.

The use of genomic tools to address conservation concerns

The shift from genetic to genomic approaches provides the ability to quantify both neutral and adaptive genetic variation, estimate diversity, and observe fine-scale population structure more accurately (Oubourg et al. 2010, Shafer et al. 2015, Supple and Shapiro 2018). These advances deliver greater power for the determination of effective conservation strategies. High throughput, reduced representation methods such as restriction-site associated DNA sequencing (RADseq) are increasingly used to address conservation concerns. RADseq subsamples the genome and provides thousands of SNPs for analysis (Davey and Blaxter 2011). First, restriction enzymes are used to digest the DNA into small fragments. DNA fragments are sequenced on a flow cell, where fluorescent primers allow the identification of base pairs in the DNA. Each base pair is analyzed, and the final product is millions of reads of DNA sequences, which can be processed with bioinformatics software. In the northern flicker (Colaptes spp.), RADseq detected low levels of differentiation, which previous analyses with allozymes and mitochondrial DNA

(mtDNA) did not (Aguillon et al. 2018). Similarly, distinct stock populations were found in the

Galapagos shark (Caharhinus galapagensis) with the use of neutral SNPs, but not mtDNA

(Pazmiño et al. 2017). Genomic data indicated that current stock designation in yellowfin tuna

(Thunnus albacares) should be revised to suit biological units better (Mullins et al. 2018). Many more studies have demonstrated the utility of genomic methods to understanding evolutionary

5 processes and addressing conservation concerns (e.g. Salmonidae, Lecaudey et al. 2018;

Aphelocoma jays, Zarza et al. 2016; Atlantic mackerel (Scomber scombrus L.), Rodríguez-

Ezpeleta et al. 2016; harbor porpoises (Phocoena phocoena), Lah et al. 2016; knob-scaled lizards (Xenosaurus spp.), Nieto-Montes de Oca et al. 2017; yellow warblers (Setophaga petechia), Bay et al. 2018; and the yellow-banded bumblebee (Bombus terricola), Kent et al.

2018).

Understanding sources of mortality is also necessary for effective conservation. In migratory species, this may be more difficult, as mortality often occurs away from the breeding or wintering grounds. In fact, in three species of raptor (osprey, Pandion haliaetus; Western marsh harrier, Circus aeruginosus; and Montagu’s harrier, C. pygargus), the mortality rate was six times higher during migratory periods than at stationary periods (breeding and wintering)

(Klassen et al. 2013). Mortality during migratory periods might be caused by anthropogenic activities (oil spills, hunting, fishing etc.). Identifying the breeding origin of affected individuals is needed to determine the overall impact of these activities on genetic variation within the species. Population specific genetic markers allow identification of breeding populations (such as in the Wilson’s warbler, Cardellina pusilla, Ruegg et al. 2014) and may help inform management strategies. The use of SNPs can provide more accurate assignment than microsatellites (Jeffery et al. 2018), and the application of many SNPs may allow assignment of individuals to populations, even when only weak genetic structure is detected (Benestan et al.

2015, Tigano et al. 2017, DeSaix et al. 2019).

6 The northern fulmar

Canada has an international responsibility to protect the Arctic species that inhabit or breed in

Canadian territory, including several species of seabirds. Seabirds represent a vital component of

Arctic marine ecosystems, as they are abundant, have broad ranges and are mostly apex predators. As migratory organisms, seabirds are present in numerous ecosystems on a global scale. Seabirds also have long life history strategies (they are slow to mature and have low reproductive rates), which makes them sensitive to changes in the environment, and they may even act as indicators for environmental change such as the presence of chemical contaminants, plastic pollution, habitat reduction, and climate change (Furness and Camphuysen 1997, Mallory et al. 2006, Piatt et al. 2007, Gaston et al. 2009, Jenouvrier et al. 2017, Rajpar et al. 2018).

Protection of seabirds is therefore important for the ecosystems in which they are widespread.

Northern fulmars (Fulmarus glacialis) are migratory seabirds that breed in coastal colonies throughout the North Atlantic and North Pacific Oceans. Northern fulmars are members of the Procellaridae family, the petrels. They are currently classified into three subspecies: F. g. glacialis which breeds in the Arctic ocean, F. g. auduboni which breeds in Europe, Iceland,

Greenland, and northeastern Canada, and F. g. rodgersii, which breeds in the Pacific Ocean

(Jouanin and Mougin 1979, Mallory et al. 2012). However, Kerr and Dove (2018) propose to elevate the Pacific subspecies into a separate species (F. rodgersii), and to dissolve the subspecies within the Atlantic Ocean to a single species (F. glacialis). Kerr and Dove (2013) estimated that the Atlantic and Pacific populations diverged 1.97 million years ago, and mitochondrial markers consistently separate the two populations.

The plumage of northern fulmars is variable, with light morphs, dark morphs and many intermediates. Plumage variation is greater in Pacific populations (Hatch and Nettleship 1998,

7 Mallory et al. 2012). Despite concordance between geography and frequency of morphs (at higher latitudes Atlantic birds display darker morphology and Pacific birds display lighter morphology, with the opposite occurring at lower latitudes), Kerr and Dove (2013) did not find an association between colour morphs and haplotypes at the melanocortin-1 receptor (MC1R) gene, which affects colour in other species (Theron et al. 2001), in either Atlantic or Pacific populations.

Northern fulmars generally breed in a small number of highly populated colonies

(Mallory et al. 2012). In the Atlantic, they have a broad distribution, from the high Arctic to the boreal North Atlantic. Colonies in Arctic Canada and northern Europe are believed to have existed for thousands of years (Gaston et al. 2006, Burg et al. 2003, Mallory et al. 2012).

However, Atlantic northern fulmars experienced major southward range expansions since the

18th century (Fisher 1952a). The breeding range expanded southward to boreal Europe in the eastern North Atlantic over the last 200 years (Stenhouse and Montevecchi 1999), and in the northwest Atlantic (Greenland, Newfoundland and Labrador) in the 1970’s (Mallory et al. 2012).

Northern fulmars do not have strong natal philopatry in Atlantic colonies, i.e. birds do not necessarily return to their natal colonies to breed, but adults demonstrate strong fidelity to the breeding colony that they select (Mallory et al. 2012).

Northern fulmars demonstrate long life history strategies: they live an average of 32 years

(with birds recorded as old as 50), do not begin breeding until they are 8-12 years old (dependent on sex), and only lay one egg per year (Fisher 1952a, Hatch 1987, Dunnet 1992). Due to their high abundance (approximately 6.1 million individuals in the Atlantic), wide distribution, and position in upper trophic levels (Mallory et al. 2012), northern fulmars are vital components of marine ecosystems.

8 Previous genetic work on northern fulmars found weak genetic structure in the Atlantic.

Burg et al. (2003) compared variation in mtDNA among populations from Iceland and the

United Kingdom. Weak population structure was observed, which Burg et al. (2003) attributed to the ability of fulmars to travel large distances and to breed successfully away from the natal colony. Further work by Burg et al. (2014) with the addition of historical samples and the nuclear aldolase gene supported the suggestion for high connectivity between colonies, and weak population structure. Finally, despite evidence to suggest strong genetic differentiation between

Atlantic and Pacific northern fulmars, and correlation of MC1R haplotypes with geography, Kerr and Dove (2013) also observed only weak differentiation amongst Atlantic colonies of northern fulmars.

Conservation concerns

Northern fulmars are not currently considered a species at risk of extinction, but several concerns warrant investigation for management. They have elevated levels of contaminants such as toxic trace elements (Buckman et al. 2004) and persistent organic pollutants in their livers (Mallory et al. 2006). Furthermore, globally northern fulmars have been found to ingest significant amounts of plastic, with increasing amounts in the stomachs of birds in Arctic Canada in recent years

(Van Franeker 1985, Provencher et al. 2009, Van Franeker et al. 2011, Avery-Gomm et al. 2012,

Trevail et al. 2015). Up to 1% of the global population is killed annually in gillnet and longline fishing in the North Atlantic, and northern fulmars are the major component of seabird bycatch in eastern Canada (Dunn and Steel 2001, Mallory 2006, Mallory et al. 2012, Hedd et al. 2015).

Finally, northern fulmar survival appears to be negatively correlated with climate variation

(winter North Atlantic oscillation, Grosbois and Thompson 2005; extreme weather, Mallory et

9 al. 2009). Arctic Canadian northern fulmars have experienced slow levels of decline since the

1970’s (Gaston et al. 2012). As mortality often occurs during the non-breeding season and/or at a distance from the breeding sites, the populations that individuals originate from are unknown.

Therefore, the colony-specific impact of different sources of mortality is unknown. A major fishery exists in the Baffin Bay-Davis Strait area: the Greenland Halibut fishery. Large numbers of northern fulmars have been reported to be caught as bycatch in the 0A and 0B regions

(subareas defined by the Northwest Atlantic Fisheries Organization, see Figure 1). Determining the colony of origin of these northern fulmars is a priority to conservation practitioners working in the area to assess the impact of the fisheries. Population viability analyses predict severe colony decline if bycatch birds originate from a limited number of colonies, but minimal overall impact if the birds originate from a wide range of colonies (Anderson et al. 2018). Due to their proximity to the fishery, three colonies may be most vulnerable: Qaqulluit, Akpait and Exeter

Sound. These colonies could experience up to an 11.7% decline over the next 66 years, depending on the extent of the bycatch. Qaqulluit and Akpait are included as sampling locations in this study (Figure 1).

Study aims and predictions

Genetic studies on northern fulmars have used few genetic markers and/or focused on European colonies (Burg et al. 2003, Kerr and Dove 2013, Burg et al. 2014). The use of high-throughput sequencing and a greater distribution of colonies may provide a more accurate investigation (see

The use of genomic tools to address conservation concerns, above). My goal was to use genomic tools to (1) investigate genetic differentiation of north Atlantic and Arctic Canadian colonies of northern fulmars, quantifying patterns of both neutral and adaptive genetic variation, potentially

10 identifying suggestions for conservation units. An additional objective was to (2) identify colony-specific genetic markers for population assignment and determine the respective breeding colonies of birds killed during the non-breeding season.

Due to significant levels of dispersal, I predicted that I would find little differentiation among colonies of northern fulmars, even using a greater distribution of colonies and the use of genomic markers in my study compared to previous work. The amount of divergence that populations experience due to genetic drift is affected by time since divergence. Recent major range expansion in northern fulmars may mean that older colonies, such as those in Arctic

Canada, are more differentiated than newer colonies, such as those in the boreal northeast

Atlantic. However, genetic drift is also affected by population size, occurring at a higher rate in smaller populations. Northern fulmars generally breed in large colonies, though newer colonies in the northwest Atlantic are significantly smaller (Mallory et al. 2012). Despite these contrasting effects on differentiation, I predicted that levels of genetic differentiation may vary amongst colonies, with greater differentiation between more northern colonies, due to greater time since divergence.

Because northern fulmars occupy a broad range in the Atlantic, different colonies may be subject to very different selective pressures. Kerr and Dove (2013) found an association between variation in the MC1R gene and geographical location. Therefore, I predicted that I would detect one or more putatively adaptive SNPs using FST outlier analysis.

Despite detecting low differentiation between colonies, Burg et al. (2003) found evidence for isolation by distance between European colonies. Therefore, I hypothesized that I may detect differentiation amongst colonies that are large distances apart. Further, each colony in Burg et al.’s study contained unique haplotypes (2003). Therefore, despite possibly low differentiation,

11 my goal was to identify panels of SNPs that will allow population assignment of birds killed in bycatch (as demonstrated by Benestan et al. 2015, Tigano et al. 2017 and DeSaix et al. 2019).

Significance

My work will contribute to a collaborative project aiming to assess the potential of seabirds to adapt to anthropogenic change. Environment and Climate Change Canada (ECCC) needs to understand the vulnerability of different seabird populations to determine policies that will best support seabird species in the future. My assessment of population differentiation in the northern fulmar will inform policies to delineate conservation units for management. Furthermore, ECCC has a vested interest in determining the colony of origin for birds killed during the non-breeding season (J. Provencher, ECCC, pers. comm.). My markers will allow ECCC to quantify the impact that human activities have on the population status of the northern fulmar.

Figure 1. A) Map of the North Atlantic Ocean displaying the breeding distribution and the year- round migratory distribution of the northern fulmar (adapted from Gravley et al. 2019).

B) Map of the North Atlantic Ocean indicating the six breeding colonies of fulmars that were sampled for this project (CV: Cape Vera, PLI: Prince Leopold Island, QAQ: Qaqulluit, AKP:

Akpait, FI: Faeroe Islands, BI: Bjørnøya). The Labrador Sea (a location at which non-breeding birds were sampled) and the 0A and 0B fisheries management subareas (the regions where the bycatch samples were caught) are also indicated.

13 CHAPTER 2: METHODS

Sample collection and DNA extraction

Blood or muscle samples were collected from 148 northern fulmars, including 116 breeding adults or chicks from six colonies in the North Atlantic, 13 non-breeding birds in the Labrador

Sea, and 19 birds caught as accidental bycatch in the 0A and 0B fisheries management subareas

(Figure 1, Table 1). Samples are archived at -80° C at Queen’s University, Kingston. Tissue samples were digested with proteinase K (0.57mg/mL) (Macherey-Nagel, Düren, Germany) in

500 μL lysis buffer and then treated with 2.5 μL RNase A/T1 (0.014mg/mL RNase A and 35.7

U/mL RNase T1) (Thermo Fisher Scientific, Canada) for 30 mins at 37° C (Maniatis et al. 1982).

Next, DNA was extracted using either a standard phenol/chloroform protocol (Friesen et al.

1997), or NaCl (Aljanabi and Martinez 1997). Extracted DNA was further purified using ethanol precipitation (Sambrook et al. 1989). Purified DNA was subjected to electrophoresis through 2% agarose and assayed with a Nanodrop Microvolume Spectrophotometer (Thermo Fisher

Scientific, Wilmington, DE, USA) to check quality. Finally, DNA was quantified using a Qubit

3.0 flurometer and a Qubit dsDNA broad sensitivity assay kit (Invitrogen, California, USA).

DNA concentrations were standardized to approximately 25 ng/µl (range: 20-30 g/µl).

Quality of the DNA was highly variable. Due to the non-standard storage of some samples before extraction (e.g. bycatch samples were stored in coolers on ships for extended periods of time and some samples were many years old), some samples had lower DNA quality.

Library preparation and sequencing

The software ddRADseqtools (Mora-Márquez et al. 2017) was used to design the restriction enzyme protocol for this study, allowing in silico testing of different combinations of enzymes

14 and fragment sizes. Enzymes and fragment sizes were chosen to maximize depth of coverage without the loss of many sequencing reads. Restriction site-associated DNA (RAD) libraries were prepared using combinatorial barcodes and indexes, using a custom protocol designed by

Zhengxin Sun based on the protocol by Peterson et al. (2012). DNA was double digested using the enzymes MluCl and SbfI (at concentrations of 0.5µl and 0.25µl respectively) for three hours, incubated in a thermocycler at 37° C. After digestion, DNA was cleaned using SPRI beads.

Then, adapters were ligated to the DNA. Forty-eight barcodes and four indexes were used. For each unique combination per individual, 1µl of annealed adapters were used. Then samples were ready to be pooled and cleaned once again with SPRI beads. The library was then size selected for fragments between 125-400 bp (excluding barcodes and adaptors). Finally, the pool of samples was amplified using PCR for 14 cycles. Paired-end sequencing (2 x 125bp) was then completed on an Illumina Hi-Seq 2500 at the Centre for Applied Genomics in Toronto.

Assembling and identifying loci

Raw data from the Illumina sequencer was processed using the program FastQC, which allows quality control of sequencing output (Andrews 2010). To remove PCR duplicates, the programming language Python 3.5.4 (Python Software Foundation, https://www.python.org/) was used to run a deduplication function designed for the library preparation used

(github.com/Eljensen/ParseDBR_ddRAD). Next, data were processed using the software

STACKS v 2.2 (Catchen et al. 2013). The STACKS algorithm process_radtags was used to demultiplex the samples based on the unique combination of barcodes and indexes. Reads with uncalled bases and low-quality scores were removed. Next, reads were aligned to the northern fulmar reference genome (PUBMED 25504712, GenBank:JJRN00000000.1), using the

15 Burrows-Wheeler aligner, using the BWA-MEM algorithm (Li and Durbin 2010), which performs high-quality alignment (Keel and Snelling 2018). The software SAMtools (Li et al. 2009) was then used to convert alignment outputs into bam file formats, assess the alignment success, and remove non-mapping reads and secondary alignments.

Variant calling and filtering

Aligned reads were assembled using the STACKS pipeline ref_map.pl, which uses the alignment positions to assemble loci and call SNPs from each read. Next, the STACKS algorithm populations and the software VCFtools were used to filter the SNPs (Danecek et al. 2011). In library preparation and bioinformatic processing, artefacts may be introduced (such as PCR artefacts, locus dropout, alignment errors etc.), possibly leading to biases in downstream analyses

(O’Leary et al. 2018). Thorough filtering can reduce this effect and completing filtering in a hierarchical manner rather than applying multiple filters at once allows each parameter to be assessed and adjusted as necessary. The following filtering parameters were chosen to avoid the loss of too many loci and individuals, while remaining stringent enough to avoid potential SNP miscalls.

The STACKS algorithm populations was used to filter SNPs for a maximum heterozygosity of 0.75 (to avoid SNP miscalls that may have caused inflated heterozygosity) and to create a

VCF file for further filtering. To avoid linkage disequilibrium, only the first SNP per locus was included in the output file. In VCFtools the function depth was used to assess depth of coverage of each individual. Individuals that were identified with very low depth (with an average read depth of three or less) were removed from the dataset using the function remove. Next, the function site- mean-depth was used to assess the average coverage across all individuals per locus. Sites with a

16 mean depth below 10 were removed using the function min_mean, as sites with a low depth cannot be called with high certainty: these sites may be inconsistent from poor quality DNA and may include false homozygotes (in the absence of the alternative allele due to lack of coverage).

Using max_mean, sites with a mean depth more than two standard deviations above the mean across all loci were removed, as sites with unusually high depth may have been introduced by the sequence similarity thresholds in bioinformatics processing. Unusually high depth might indicate that loci have been oversplit, meaning that alleles from one locus have been split into different contiguous sequence alignments (contigs) (O’Leary et al. 2018). The function max-missing was used to remove any site with more than 20% missing data and the function maf was used to include sites with a minimum minor allele frequency of 1% (essentially one individual in 127).

(A separate dataset using a minimum minor allele frequency of 5% was also created. After finding very similar results for pairwise FST and principal components analysis using both datasets (see Appendix Table 1 and Figure 1), the dataset using the cutoff of 1% was chosen, as it contained more SNPs, which may be useful in population assignment). Next, the function missing-ind was used to identify the number of missing loci per individual. To avoid high levels of missing data, but retain enough individuals per population, a missing data cutoff of 30% of total loci was applied, and individuals above the threshold were removed from the dataset, using the function remove.

Finally, sites were assessed for deviations from linkage equilibrium in VCFtools using the function hap-r2, which outputs a file reporting r2 statistics, and for deviations from Hardy-

Weinberg expected proportions using the function hardy. Sites that deviated from Hardy-

Weinberg proportions in four or more of six populations were excluded from the dataset. Data

17 files were converted to the appropriate input formats (STRUCTURE, GENEPOP, BAYESCAN) for each subsequent analysis using the program PGDSpider v. 2.1.1.5 (Lischer and Excoffier 2011).

Population genetic structure analyses

For analyses of population genetic structure and gene flow, only individuals from known breeding-colonies were used. a) Population genetic structure

Population genetic structure was assessed using a number of analyses. First, the Bayesian clustering program STRUCTURE v. 2.3.4 (Pritchard et al. 2000) was used to determine the most likely number of genetic groups within the dataset, and to assign individuals to those groups.

STRUCTURE was run using several different models: a) the admixture model, with no prior population information, b) the admixture model with sampling site as prior information, and c) the no admixture model with sampling site as prior information, to thoroughly investigate genetic structure. The program was run using 1,000,000 Markov chain Monte Carlo (MCMC) iterations, following a burn-in of 100,000 iterations. Each value of K (the assumed number of genetic groups in the dataset) from 1 to 8 was run five times. STRUCTURE can be sensitive to unbalanced sample sizes (Puechmaille 2016), so three new files (the ‘resample’ files) were created choosing eight random individuals (as the lowest population size was nine) from each population using the

R package GENEPOPEDIT (Stanley et al. 2016). Each of the three resample files was run using both a) the admixture model, with no prior population information and b) the admixture model with sampling site as prior information. Therefore, a total of nine STRUCTURE analyses were completed. To infer the most likely number of genetic groups three methods were used: the

Evanno method (Evanno et al. 2005), as implemented in STRUCTURE HARVESTER (Earl and

18 vonHoldt 2012); Bayes’ Rule as outlined in the STRUCTURE manual (Pritchard et al. 2000); and the parsimony indicator implemented in KFinder (Wang 2019).

A principal components analysis (PCA) and a principal coordinates analysis (PCoA) were both implemented on the complete breeding dataset using the R packages adegenet (Jombart and Ahmed 2011) and ade4 (Dray and Dufour 2007). A PCA uses the presence/absence of alleles in molecular data to create the principal components, whereas a PCoA uses a matrix of similarities between the individuals to create the principal coordinates. Results were almost identical between the analyses, and the PCA was chosen for use in further tests.

Next, estimates of FST for pairwise comparisons of populations (Weir and Cockerham 1984,

Excoffier et al. 1992, Michalakis and Excoffier 1996) were calculated in the software GenoDive

(Meirmans 2004).

Finally, a K-means clustering analysis (Meirmans 2012) was performed using the software

GenoDive (Meirmans and Van Tienderen 2004). K-means clustering analyses attempt to estimate the most likely number of genetic groups, by minimizing within-group diversity and maximizing between-group diversity.

b) Searching for SNPs under selection

The program BayeScan (Foll and Gaggiotti 2008) was used to identify SNPs that are potentially under selection. The program was run using prior odds of both 100 and 1000, using 100,000 iterations following a burn-in of 50,000 iterations. Analyses were run with a) all populations compared to each other, b) the eastern populations compared with the western populations, c) higher latitude populations compared to lower latitude populations, and d) Bjornøya (the

19 population with the highest pairwise FST values across all populations) compared to all other populations.

c) Genetic diversity indices

Three genetic diversity indices were calculated for each population: private alleles, expected heterozygosity, and allelic richness. The number of private alleles, and their frequency amongst individuals, was calculated using the R package poppr v. 2.8.1 (Kamvar et al. 2015), allelic richness was calculated in the R package hierfstat (Goudet 2005), and expected heterozygosity

(Nei 1987) was calculated in GenoDive (Meirmans and Van Tienderen 2004). To control for sample size when calculating private alleles, the R package genepopedit (Stanley et al. 2016) was used to create ten files with eight individuals randomly selected from each population. Each of these files was assessed for number of private alleles, and the mean of these results taken.

d) Isolation by distance

To test for an effect of isolation by distance, i.e. an increase in genetic distance between colonies with increasing geographic distance, a Mantel test was implemented in the R package ade4 (Dray and Dufour 2007), using the function mantel.rtest. Pairwise comparisons of FST calculated previously (see Population structure, above) were converted into Slatkin’s linearized FST ((1-

FST)/ FST). Fulmars do not fly over large expanses of land and would be more likely to fly south than north around Greenland to traverse the Atlantic (Mark Mallory, pers. comm.). Marine distances (the shortest distance that does not cross land) are therefore more appropriate than

Euclidean distance, as the distance fulmars would travel between colonies. Marine distances

20 were calculated using Google Maps (Google Maps 2019), and log-transformed. Results were visualized using the R package ggplot2 (Wickham 2016).

e) Gene flow between colonies

The program BayesAss v. 3 (Wilson and Rannala 2003), using Bayesian analysis, molecular assignment and MCMC sampling, was implemented to estimate contemporary migration between populations. The program was run using 10,000,000 MCMC repetitions following a burn-in of 1,000,000 and sampling every 1000 iterations. The program was tested using a range of mixing parameters and different seed values. To ensure convergence, trace files were created for each run. The program was run using an input file containing all populations, as well as several input files containing smaller numbers of populations for testing purposes.

Population assignment

Population assignment was complicated, due to the low differentiation between colonies (see

RESULTS). Further, populations with sample sizes below 30 may not have high assignment accuracies (Benestan et al. 2015) and only one of my populations (Faeroe Islands) had more than

30 samples. However, some previous work had success despite low differentiation (Benestan et al. 2015, Tigano et al. 2017, DeSaix et al. 2019). Several analyses, such as genetic stock identification (GSI) (Anderson et al. 2008, Anderson et al. 2010), have identified reliable population markers, and newer techniques such as machine-learning using random forest algorithms are being developed for population assignment (Sylvester et al. 2017). Jeffery et al.

(2018) compared population assignment methods using genome-wide SNPs in Atlantic salmon

(Salmo salar), and observed greatest assignment accuracy using a panel of unlinked SNPs with

21 high FST. The following methods were therefore designed based on previous research and used in combination in attempt to attain the best assignment success possible from the dataset.

a) Unknown individuals

A PCA was run using all individuals, to investigate where unknown individuals cluster compared to the breeding birds using two components. Next pairwise values of FST were estimated between the birds from the Labrador Sea and the bycatch birds versus each other and each breeding colony. Finally, numbers of private alleles were calculated again for each population of birds, and each group of unknown birds, to determine if birds from the Labrador Sea or the bycatch birds contained alleles that were not shared with any of the known breeding locations.

b) Machine learning methods

The R package assignPOP uses supervised machine-learning methods (employing a ‘training’ dataset) both i) to determine the assignment accuracy of the baseline dataset using different proportions of training loci and training individuals, and ii) to assign unknown individuals to populations (Chen et al. 2018). The package can use either K-fold or Monte-Carlo cross validation to perform resampling using the baseline data to assess assignment accuracy. Different machine-learning models can be used: support vector machine, linear discriminant analysis, naïve Bayes, decision tree or random forest. Each model was tested, and the model random forest resulted in slightly higher assignment accuracy than other models. Monte-Carlo cross validation and random forest were chosen in all further assignment tests performed in assignPOP.

Accuracy was estimated for assignment to i) all six colonies, ii) all colonies except Akpait and

22 Bjørnøya, which had fewer than 15 individuals each, iii) Faeroe Islands vs. Bjørnøya vs. Arctic

Canadian colonies (chosen based on pairwise FST estimates), and iv) just the Canadian colonies.

Based on evidence that unlinked SNPs with the highest FST produce the most accurate population assignments (Jeffery et al. 2018), the assignPOP analyses were repeated using custom

SNP panels. The function genepop_toploc in the R package genepopedit (Stanley et al. 2016) selects the panel of unlinked loci with the highest pairwise FST between populations. Pairwise tests were run between each possible pair of colonies (15 tests total). Loci from each pairwise test with FST above 0.1 were selected, and duplicates were removed. Assignment accuracy was assessed using assignPOP. These tests were completed on i) all six colonies (using 253 SNPs), ii) all colonies except Akpait and Bjørnøya (using 139 SNPs), iii) Faeroe Islands vs. Bjørnøya vs. Arctic

Canadian colonies (using 122 SNPs), and iv) just the Canadian colonies (using 154 SNPs).

Principal components analyses were also employed for each of the test groupings, to visualize differentiation between groups using the reduced datasets, and to estimate where the unknown birds might cluster.

c) Training and holdout loci

DeSaix et al. (2019) employed the training and holdout technique outlined by Anderson (2010) using RADseq data on populations with low differentiation. With their framework, individuals from each population are split into ‘training’ sets (used to create the SNP panel for assignment), and ‘holdout’ sets (used to test the assignment panels) to prevent high-grading bias. SNP panels are created based on loci with the highest FST.

To follow this framework, samples from Akpait and Bjørnøya were not included, as the small sample sizes for these colonies could hinder the assignment process. To start, samples were

23 separated into two groups: the Faeroe Islands, and Arctic Canadian colonies. Twenty individuals from each group were used in the training dataset, and all others were used in the holdout dataset. The function genepop_toploc in the R package genepopedit (Stanley et al. 2016) was used to select the loci with the highest FST between the groups. Four SNP panels were created: i) the top

500 loci, ii) the top 1000 loci, iii) the top 2000 loci, and iv) all loci. The R package assignPOP

(Chen et al. 2018) was used to assign the individuals in the holdout dataset back to their respective group, and accuracy was assessed for each SNP panel.

Next, the above process was repeated using each of the four colonies with sufficient sample sizes but using 13 individuals for the training dataset: Faeroe Islands, Prince Leopold

Island, Cape Vera and Qaqulluit.

The process was then repeated for only the three Canadian colonies with sufficient sample sizes: Prince Leopold Island, Cape Vera and Qaqulluit. However, this time estimates of pairwise FST using each possible colony comparison were used, instead of across all colonies (as was implemented above, and by DeSaix et al. 2019). SNP datasets were created for i) all SNPs above an FST estimate of 0.1, ii) all SNPs above an FST estimate of 0.5, and iii) all SNPs.

Assignments were completed on the unknown birds using the SNP datasets and colony comparisons that produced the highest assignment accuracy.

d) Genetic stock identification

Genetic stock identification (GSI), determining the stock composition of mixed stocks, is commonly used in fisheries research and management. Researchers of Pacific northern fulmars have had success applying GSI to population assignment of Pacific birds caught in fishing bycatch (D. Baetscher, pers. comm.). Previous GSI methodology could overestimate the

24 accuracy of results, but Anderson et al. (2008) designed a method to determine the accuracy of genetic stock identification using leave-one-out cross validation to provide less biased accuracy estimates. The R package rubias implements genetic stock identification by genotyping the reference samples and using a conditional likelihood model to infer the origin of the unknown samples, but accounts for unequal sample sizes amongst populations (available at https://github.com/eriqande/rubias).

The package rubias was used both to assess the genetic reference data for assignment accuracy and to estimate the origin of the unknown individuals. The program can also test whether the unknown individuals do not come from any of the reference populations, by providing a z-score, a statistic based on the log-likelihood of the unknown individual belonging to the reference groups. The z-scores for the unknown individuals can be compared to a simulated normal distribution. If the fit is poor, the samples may have originated from populations not included in the reference data.

GSI was performed to ascertain assignment accuracy to different groupings of populations: i) all six known breeding colonies, ii) the Faeroe Islands compared to the Arctic

Canadian colonies, iii) the four Arctic Canadian colonies, iv) the southern Arctic Canadian colonies (Akpait and Qaqulluit) compared to the northern colonies (Cape Vera and Prince

Leopold Island), and v) the northern colonies compared to Qaqulluit (pairing the Northern colonies together, as these colonies had a pairwise FST of 0.00). All datasets included all 6614

SNPs. The population divisions were chosen based on results from earlier assignment tests, and to investigate which comparisons produce the highest accuracies.

25 Table 1. Sampling locations, type of location, current subspecies designation, number of samples, and numbers of individuals after data filtering. See Figure 1 for locations.

Sampling Breeding/Non Current Total number of Total number of location -breeding subspecies samples individuals after location designation data filtering

Cape Breeding F. g. glacialis 19 19 Vera Prince Breeding F. g. glacialis 22 18 Leopold Island Qaqulluit Breeding F. g. glacialis 19 17 Akpait Breeding F. g. glacialis 13 11 Faeroe Breeding F. g. auduboni 31 Islands Bjørnøya Breeding F. g. glacialis 9 9 Labrador Non-breeding Unknown 13 10 Sea Bycatch Non-breeding Unknown 19 12

26 CHAPTER 3: RESULTS

Sample processing

PCR duplication rate ranged from 12.3% to 25.4% for the three indexes. After deduplication, approximately 346.1 million reads were retained. After demultiplexing and quality filtering in

STACKS, approximately 309.5 million total reads remained, with a range of 12,049-16,081,801 reads per individual, and ~15,000 loci. During filtering with VCFTOOLS, 11 individuals were removed due to low depth of coverage, and seven were removed due to missing data above 30%.

The final dataset consisted of 127 individuals (105 breeding birds and 22 birds of unknown breeding location) and 6614 loci (Table 1).

Population genetic structure analyses a) Population genetic structure

All analyses indicated little population genetic structure. Results from the program STRUCTURE produced inconsistent results for the best value of K across the three methods for each of the nine runs. However, the parsimony indicator suggested that the best estimate of K was one for six of nine runs. When a K value greater than one was suggested (not shown), all individuals displayed

80% posterior probability of assignment to the first group, and a small percentage to the other group(s). Based on these results, the most likely number of genetic groups is one. The principal components analysis (Figure 2) demonstrated little differentiation between groups, which were essentially indistinguishable. Estimates of pairwise FST between colonies were low, and many comparisons were not significantly greater than 0 (Table 2) (such as between Akpait and Cape

Vera). However, some pairwise estimates were significantly greater than 0 (such as between the

Faeroe Islands and Cape Vera), indicating that some genetic structuring is present. Bjørnøya

27 displayed significant pairwise FST values for comparisons with all other five colonies. Finally, K- means clustering suggested the presence of only one genetic cluster in the data.

b) Searching for SNPs under selection

None of the BayeScan runs identified any SNPs as potentially under selection.

c) Genetic diversity indices

Allelic richness and expected heterozygosity were very similar among populations. All six colonies contained a large number (> 220) of private alleles (Table 3). However, the Faeroe

Islands contained the largest number of private alleles, even after correcting for sample size.

Most private alleles were only present in one individual per population (81%), and none were present in more than 50% of individuals per population (Figure 3).

d) Isolation by distance

The Mantel test did not detect a significant correlation between genetic and geographic (marine) distance between colonies (r = 0.144, p = 0.08). However, pairwise FST did tend to increase with increasing distance between colonies (Figure 4).

e) Gene flow between colonies

All runs of BayesAss failed to converge and so could not be used to estimate contemporary migration between populations.

28 Population assignment a) Unknown individuals

The principal components analysis for all birds failed to differentiate the groups of individuals from the Labrador Sea, those caught in the bycatch and the six breeding populations (Figure 5).

Both groups of birds of unknown origin displayed high within-group variation, with individuals from neither group clustering closely together.

Pairwise FST estimates between birds sampled from the Labrador Sea or bycatch versus the breeding colonies were variable (Table 4), but generally higher than the FST values between the breeding colonies (Table 2). The birds from the Labrador Sea appear to be most differentiated from those from Bjørnøya, and least differentiated from samples from the Faeroe

Islands (Table 4). The bycatch birds also appear to be most differentiated from those from

Bjørnøya but least differentiated from those from Qaqulluit. The birds from the Labrador Sea and the bycatch also demonstrated high pairwise FST estimates relative to other comparisons.

The inclusion of the birds from unknown breeding location in t ests for private alleles resulted in a large reduction of private alleles from the breeding colonies (Table 5), suggesting shared variation between the breeding colonies and the unknown birds. The birds from the

Labrador Sea and those from the bycatch had several unique alleles at low frequency.

b) Machine learning methods

Using all SNPs and all colonies, assignment accuracy was very low using machine learning

(~30%, Figure 6a). With Akpait and Bjørnøya removed, assignment accuracy was slightly higher but still low (~40%, Figure 6b). Accuracy tests for the Faeroe Islands vs. Bjørnøya vs. Arctic

29 Canadian colonies, and between just the Canadian colonies were similarly low (<30%, Figure 6c and d).

The reduced SNP datasets provided slightly more clarity than the complete dataset. For all colonies, the PCA differentiated the six colonies slightly more than the original one (Figure

7). The unknown birds seem to cluster farthest from Bjørnøya. Surprisingly, the assignment accuracy seemed to be even lower than using the full dataset (Figure 8a). With Akpait and

Bjørnøya removed, PCA did not provide a clear result. The assignment accuracy for this dataset also appeared worse than using the full dataset (Figure 8b). For Faeroe Islands vs. Bjørnøya vs.

Arctic Canadian colonies, the PCA quite distinctly separated the colonies, with Bjørnøya particularly distinct (Figure 9). The unknown birds had almost no overlap with Bjørnøya, with the Labrador Sea birds clustering closer to the Faeroe Islands, and the bycatch birds clustering closer to the Arctic Canadian birds. The assignment accuracy was higher than using the whole

SNP dataset (~50%, Figure 8c). For just the Canadian colonies, the PCA did not provide accuracy for population assignment. Assignment accuracy was somewhat higher than the whole

SNP dataset but still low (Figure 8d).

None of the assignment accuracies for the reference data was over 50%, and therefore no assignment tests were run on unknown individuals, as results would have been unreliable.

c) Training and holdout loci

For the analysis based on the Faeroe Islands vs. Arctic Canada, the use of the top 1000 SNPs provided the greatest assignment accuracy (overall accuracy 84%) (Table 6). Assignment tests on the unknown birds suggested that 50% of birds from the Labrador Sea originated from the

Arctic colonies, and the other 50% originated from the Faeroe Islands. Meanwhile, 17% of birds

30 caught in the bycatch were assigned to the Faeroe Islands, and 83% were assigned to Arctic

Canada. Assignment accuracy tests indicated under-estimation of assignment to the Faeroe

Islands (91% accuracy of assignment to Arctic Canadian colonies, 55% accuracy of assignment to the Faeroe Islands). In general, the result would suggest a greater proportion of birds found in the Labrador Sea originated from the Faeroe Islands, and a greater proportion of birds caught in the bycatch originated from Arctic Canada.

Overall assignment accuracy using the four colonies with sufficient sample sizes (Faeroe

Islands, Prince Leopold Island, Cape Vera and Qaqulluit) was very low (ranging from 14.3-

24%). Using the three Canadian colonies with sufficient sample sizes (Prince Leopold Island,

Cape Vera and Qaqulluit) produced reasonable overall assignment accuracy (ranging from 67-

75%) but assignment to individual colonies was very low for some colonies. Therefore, assignment tests were not run on the unknown samples using these reference datasets.

d) Genetic stock identification

The accuracy using genetic stock identification to assign birds to all six colonies, or the four

Arctic Canadian colonies were unsatisfactory (43% and 36.8% respectively) (Table 7). The distribution of z-scores for all colonies did not fit a normal distribution (Figure 10) and suggested that approximately one-third of the unknown individuals originate from populations not included in our reference dataset.

Assignment accuracy using the Faeroe Islands and Arctic Canada was very high (98%) with a slight under-estimation of individuals from the Faeroe Islands (96% assignment accuracy to Faeroe Islands and 100% assignment accuracy to Arctic Canada). Assignment of unknown individuals suggested that 70% of individuals from the Labrador Sea originated from the Faeroe

31 Islands, and 30% from Arctic Canadian colonies. All bycatch birds were assigned to Arctic

Canada (Table 8).

For the Arctic regions (north and south), the assignment accuracy was 70%. However, there was a large skew in accuracies, with 97% accuracy of assignment to southern colonies, but only 33% accuracy to northern colonies. Results then likely represent large under-estimation of northern colonies. Of the individuals from the Labrador Sea that were identified as belonging to

Arctic Canadian colonies, all were identified as originating from northern colonies. Of the bycatch individuals, 42% were identified as belonging to northern colonies, and 58% from southern colonies.

Finally, for Prince Leopold Island and Cape Vera compared to Qaqulluit, the overall assignment accuracy was 66% (73% accuracy of assignment to Prince Leopold and Cape Vera, and 60% to Qaqulluit). Of the individuals from the Labrador Sea that were identified as belonging to Arctic Canadian colonies, two were identified as originating from Qaqulluit, and one from Prince Leopold and Cape Vera. Nine bycatch individuals were identified as originating from Qaqulluit, one from Prince Leopold and Cape Vera, and two individuals demonstrated similar likelihood of assignment to either group.

32 Table 2. Pairwise FST for the six colonies (above the diagonal), and probability of difference from 0 (below the diagonal).

Cape Prince Qaqulluit Akpait Faeroe Bjørnøya Vera Leopold Islands Island

Cape Vera - 0.000 0.004 0.000 0.005 0.003

Prince 0.443 - 0.000 0.000 0.005 0.008 Leopold Island Qaqulluit 0.002 0.457 - 0.002 0.008 0.010 Akpait 0.531 0.945 0.202 - 0.003 0.009 Faeroe 0.001 0.001 0.001 0.094 - 0.009 Islands Bjørnøya 0.001 0.003 0.005 0.001 0.013 -

Significantly different from 0 at α = 0.05 is shown in bold.

33 Table 3. Number of private alleles per colony using the reduced breeding data set of 105 individuals.

Colony Number of private Average value corrected for sample size of alleles 8 (min-max)

Cape Vera 94 260 (204-300) Prince Leopold 53 230 (183-297) Island Qaqulluit 135 304 (222-359) Akpait 18 224 (187-261) Faeroe Islands 604 421 (344-505) Bjørnøya 31 333 (274-394)

34 Table 4. Pairwise estimates of FST between the birds of unknown origin and the six breeding colonies.

Cape Prince Qaqulluit Akpait Faeroe Bjørnøya Labrador Bycatch Vera Leopold Islands Sea Island Labrador 0.012 0.009 0.013 0.011 0.0035 0.018 - 0.015 Sea Bycatch 0.006 0.0034 0.000 0.004 0.006 0.011 0.015 -

Significantly different from 0 at α = 0.05 is shown in bold.

35 Table 5. Private allele counts for all locations included in the study using the complete dataset of

127 individuals (not corrected for sample size).

Colony Private alleles

Cape Vera 47 Prince Leopold Island 12 Qaqulluit 50 Akpait 6 Faeroe Islands 236 Bjørnøya 15 Labrador Sea 6 Bycatch 22

36 Table 6. Assignment accuracy for different combinations of sampling sites and SNP datasets for the training and holdout method.

Sample site Dataset Overall accuracy Accuracy per group combination (1) Faeroe Islands (2) 500 SNPs 77% Faeroe Islands: 45% Arctic Canada Arctic Canada: 84% 1000 SNPs 84% Faeroe Islands: 55% Arctic Canada: 91% 2000 SNPs 79% Faeroe Islands: 55% Arctic Canada: 84% All SNPs 81% Faeroe Islands: 0% Arctic Canada: 92% (1) Faeroe Islands (2) 500 SNPs 38% Faeroe Islands: 32% Prince Leopold Island Prince Leopold Island: (3) Cape Vera (4) 17% Qaqulluit Cape Vera: 71% Qaqulluit: 40% 1000 SNPs 30% Faeroe Islands: 21% Prince Leopold Island: 17% Cape Vera: 43% Qaqulluit: 60% 2000 SNPs 24% Faeroe Islands: 5% Prince Leopold Island: 17% Cape Vera: 57% Qaqulluit: 60% All SNPs 14% Faeroe Islands: 5% Prince Leopold Island: 33% Cape Vera: 100% Qaqulluit: 20% (1) Prince Leopold SNPs with FST above 75% Prince Leopold Island: 0% Island (2) Cape Vera 0.1 Cape Vera: 83% (3) Qaqulluit Qaqulluit: 100% SNPs with FST above 67% Prince Leopold Island: 0% 0.05 Cape Vera: 50% Qaqulluit: 100% All SNPs 70% Prince Leopold Island: 40% Cape Vera: 33% Qaqulluit: 100%

37 Table 7. Assignment accuracy for each combination of groups using genetic stock identification methods.

Sample site Overall accuracy Accuracy per group combination All breeding colonies 43% Cape Vera: 33% Prince Leopold Island: 7% Qaqulluit: 33% Akpait: 0% Faeroe Islands: 96% Bjørnøya: 0% Faeroe Islands vs. 99% Faeroe Islands: 96% Arctic Canadian Arctic Canada: 100% colonies

Arctic Canadian 37% Cape Vera: 61% colonies Prince Leopold Island: 33% Qaqulluit: 33% Akpait: 0% Arctic regions 70% North: 97% South: 33% Prince Leopold and 66% Prince Leopold and Cape Vera: 73% Cape Vera vs. Qaqulluit: 60% Qaqulluit

38 Table 8. Assignment results for unknown individuals based on: Assignment Test 1) test between

Faeroe Islands and Arctic Canada using training and holdout loci; Assignment Test 2) test

between Faeroe Islands and Arctic Canada using genetic stock identification; Assignment Test 3)

test between Arctic regions using genetic stock identification; and Assignment Test 4) test

between Northern Arctic colonies and Qaqulluit using genetic stock identification. (Individuals

identified as originating from the Faeroe Islands in Test 2 were not applicable for Test 3 and 4).

Sampling Individual Result 1 Result 2 Result 3 Result 4 location Labrador Sea K15-426-33 Faeroe Islands Faeroe Islands Not applicable Not applicable K15-426-35 Faeroe Islands Faeroe Islands Not applicable Not applicable K15-426-38 Faeroe Islands Faeroe Islands Not applicable Not applicable K15-426-41 Arctic Canada Faeroe Islands Not applicable Not applicable K15-426-42 Arctic Canada Faeroe Islands Not applicable Not applicable K15-426-44 Faeroe Islands Faeroe Islands Not applicable Not applicable K15-426-47 Faeroe Islands Arctic Canada North Qaqulluit K15-426-48 Arctic Canada Arctic Canada North Prince Leopold and Cape Vera K15-426-49 Arctic Canada Arctic Canada North Qaqulluit K15-426-51 Arctic Canada Faeroe Islands Not applicable Not applicable Bycatch B18-2 Faeroe Islands Arctic Canada South Qaqulluit B18-3 Arctic Canada Arctic Canada South Qaqulluit B18-6 Faeroe Islands Arctic Canada North Prince Leopold and Cape Vera B18-7 Arctic Canada Arctic Canada North Qaqulluit B18-9 Arctic Canada Arctic Canada South Qaqulluit B18-11 Arctic Canada Arctic Canada North Qaqulluit B18-12 Arctic Canada Arctic Canada South Qaqulluit B18-13 Arctic Canada Arctic Canada South Qaqulluit

39 B18-15 Arctic Canada Arctic Canada North Prince Leopold and Cape Vera: 57% Qaqulluit: 43% B18-17 Arctic Canada Arctic Canada North Prince Leopold and Cape Vera: 51% Qaqulluit: 49% B18-18 Arctic Canada Arctic Canada North Qaqulluit B18-19 Arctic Canada Arctic Canada North Qaqulluit

Figure 2. The first two principal components of a principal components analysis for breeding birds using all 6614 SNPs.

Figure 3. The frequency of private alleles within populations (based on sample size corrected calculations, with a maximum possible of eight individuals).

Figure 4. Genetic distance (Slatkin’s linearized FST ) vs. log-transformed marine distance between colonies.

Figure 5. The first two principal components of a principal components analysis using all 6614

SNPs and all birds included in the study: six breeding colonies (in grey), birds found in the

Labrador Sea (pink), and birds caught in the bycatch (blue).

Figure 6. Assignment accuracies using all SNPs and A) all colonies, B) all colonies except

Bjørnøya and Akpait, C) the Faeroe Islands, Bjørnøya, and Arctic Canada, and D) the Arctic

Canadian colonies. Colour of box indicates the proportion of loci used in training set.

Figure 7. The first two principal components of a principal components analysis using the reduced SNP dataset (253 SNPs) and A) all breeding colonies, and B) breeding colonies and unknown birds.

Figure 8. Assignment accuracies using reduced SNP datasets and A) all colonies, B) all colonies except Bjørnøya and Akpait, C) the Faeroe Islands, Bjørnøya, and Arctic Canada, and D) the

Arctic Canadian colonies.

Figure 9. Principal components analysis using the reduced SNP dataset (122 SNPs) and A) the

Faeroe Islands, Arctic Canada and Bjørnøya, and B) the addition of the unknown birds.

48 0.4

0.3

d 0.2

x E

0.1

0.0

-4 0 4 Z score

Figure 10. Z-score distribution for all colonies (blue) compared to the expected distribution

(black).

49 CHAPTER 4: DISCUSSION

As predicted, I found evidence for weak genetic structure in Atlantic northern fulmars. My results from the use of high-throughput genomic sequencing support previous work (Burg et al.

2003, Kerr and Dove 2013, Burg et al. 2014), providing strong evidence for a genetically mostly homogenous Atlantic population. Assignment tests strongly suggested that birds from the

Labrador Sea represent northern fulmars from across the Atlantic range, whereas the birds caught in bycatch all originate from Arctic Canada.

Population genetic structure

Population genetic structure may be weak due to similarities in selective forces acting upon different populations, limited genetic drift within populations, historical association, and/or contemporary gene flow (Hedrick 2013). The wintering range of northern fulmars is large and shared among birds from many colonies (Mallory et al. 2008). Therefore, birds from across the range probably experience similar selection pressures during the non-breeding season. However, due to the large breeding range of northern fulmars, colonies are likely to experience different selective forces. For example, the Arctic Canadian colonies included in my study occur in cold water currents, such as the Baffin Island Current, whereas the European colonies in my study are influenced by warm water currents such as the North Atlantic Current and Norwegian Current

(Friesen 2015). Further, I did not find evidence for SNPs under balancing selection using the program BayeScan (although the proportion of the genome that I sequenced was small, likely less than 1% (Valencia et al. 2018)). Thus, selection is unlikely to explain the genetic similarity of colonies. Alternatively, because northern fulmars breed in large colonies (Table 9) and have long generation times (~19 years, Jones et al. 2008), and colonies were only established since the

50 recession of the Pleistocene glaciers, the effects of genetic drift may be minimal (Allendorf et al.

2013). Further, northern fulmars may have undergone significant demographic changes recently enough that migration-drift equilibrium has not been reached. For example, northern fulmars have experienced large range expansions since the Pleistocene, particularly in the last ~250 years. Finally, the weak genetic structure may reflect historical associations. Lombal et al.

(unpubl. data) suggested that historical processes have a larger influence on the genetic structure of seabird populations than contemporary processes. In a review of population differentiation in seabirds, Friesen et al. (2007) calculated that the divergence time between colonies of northern fulmars is less than four times the effective population size times generation time, indicating that northern fulmars may not have experienced complete lineage sorting, and are probably not at migration-drift equilibrium. Many natural populations may not be at migration-drift equilibrium due to insufficient time since separation (McCauley 1993). If gene flow is limited, and a population is at equilibrium, genetic and geographic distance should be positively correlated

(Hutchison and Templeton 1999). If the association is not significant, then either populations are not at equilibrium, or gene flow is not limited. There may be a positive association between genetic and geographic distance on a smaller scale, but if the association is not significant across the region then equilibrium has not been reached yet. Indeed, I found a positive but not significant correlation between genetic and geographic distance, whereas Burg et al. (2003) found significant isolation by distance using only colonies in Iceland, the United Kingdom and the Faeroe Islands. Therefore, historical association likely has a lingering influence on the population structure of Atlantic northern fulmars, which have yet to reach migration-drift equilibrium. Additionally, there may be on-going gene flow (see below).

51 Contemporary gene flow is also likely contributing to weak genetic structure in Atlantic northern fulmars. Unfortunately, I was unable to estimate the amount of gene flow using available software. The program BayesAss to measure migration was likely inappropriate for my dataset. The program assumes that migration rates are low and requires strong differentiation

(FST > 0.1) to produce accurate estimates (Meirmans 2013). I had low differentiation amongst colonies, and likely high migration rates. When data do not fit the assumptions of the program, convergence problems are likely, and indeed I was unable to get the runs to converge. However, the low frequency of private alleles occurring in multiple individuals per population suggests strong gene flow between colonies. Private alleles arise largely through mutation, and their frequency is indicative of time since population divergence. A high frequency of alleles found in only one population is negatively associated with gene flow (Slatkin 1985). Despite the large total number of private alleles, the majority were present in only one individual per population

(>1000) (Figure 3). Further, indirect evidence for gene flow between colonies of northern fulmars exists in the literature: at Eynhallow in Scotland, less than 5% of banded chicks returned to breed (Dunnet and Anderson 1979). Based on survival estimates of northern fulmars, a minimum of 89% of banded birds that survived to breed are likely to have emigrated elsewhere for breeding. The natural history of northern fulmars also suggests a high incidence of breeding away from the natal colony. Northern fulmars are pelagic seabirds capable of travelling large distances, with evidence for foraging over 1000 km from their colony (Hatch et al. 2010).

Juveniles are thought to spend a minimum of three years away from land before starting to prospect for breeding sites (Macdonald 1977). Juveniles were found to remain closer to the natal colony after several years, but the potential for inter-colony mixing seems high. Finally, the large range expansion of northern fulmars in the last 200 years (Stenhouse and Montevecchi 1999)

52 further supports the ability of northern fulmars to breed successfully away from the natal colony.

Studies on numerous other organisms suggest that substantial levels of contemporary gene flow correlate with the absence of genetic structure (such as in the freshwater fish, the empire gudgeon (Hypseleotris compressa), McGlashan and Hughes 2000; the arctic fox (Alopex lagopus), Carmichael et al. 2007; the oceanic bottlenose dolphin (Tursiops truncatus), Quérouil et al. 2007; the Canada lynx (Lunx canadensis), Row et al. 2012; and the prairie katydid

(Neoconcephalus biovacatus), Ney and Schul 2017). Several of these studies also found no significant effect of isolation by distance (McGlashan and Hughes 2000, Quérouil et al. 2007,

Ney and Schul 2017). A lack of isolation by distance and low estimates of FST for pairwise comparisons may indicate little limitation on long distance dispersal, and the occurrence of gene flow over large distances. However, the lack of isolation by distance does not necessarily indicate panmixia among all colonies, but perhaps indirect gene flow due to mixing between many combinations of colonies.

As predicted, some genetic structure is present amongst the northern fulmar colonies in

my study, with several significant estimates of FST for pairwise comparisons of colonies. Burg et al. (2003) suggested that gene flow between northern fulmar colonies is high, but that colonies are not completely panmictic given significant isolation by distance and presence of unique mitochondrial haplotypes at each colony. Hutchison and Templeton (1999) outline scenarios considering the relationship between genetic and geographic distance, and what these might indicate about the evolutionary forces acting on populations. When there is no correlation

between genetic and geographic distance and pairwise estimates of FST are low, gene flow may have a stronger effect than genetic drift, resulting in a panmictic population. Alternatively, genetic drift might have a stronger effect if populations are fragmented. Natural populations,

53 however, usually experience both gene flow and drift, with the relationship between genetic and geographic distance becoming increasingly significant before reaching equilibrium. Considering my results for isolation by distance (Figure 4), northern fulmar populations are likely not panmictic nor fragmented but have not yet reached equilibrium.

Arctic populations of seabirds often have lower genetic differentiation than species from lower latitudes (Friesen et al. 2007). Historical demographic factors are likely to have contributed to this, such as association in a small number of glacial refugia during the Last

Glacial Maximum resulting in homogenous populations. Several other arctic seabird species have weak or no population genetic structure, such as the thick-billed murre (Uria lomvia,

Tigano et al. 2017), common murre (Uria aalge, Morris-Pocock et al. 2008), and dovekie (Alle alle, Wojczulanis-Jakubas et al. 2014). Wojczulanis-Jakubas et al. (2014) suggested three factors that may have contributed to a lack of structure in dovekies. Firstly, range expansion and changes to population size have occurred recently, and the dovekie populations may not be at migration-drift equilibrium. Secondly, the evolutionary history of the dovekie, such as shared glacial refugia, may have reduced differentiation. Finally, high rates of contemporary dispersal between colonies may have led to the lack of genetic structure observed.

Despite being highly mobile, many seabird species do have strong genetic structure.

Friesen (2015) and Friesen et al. (2007) reviewed population differentiation and speciation in seabirds, considering what factors may result in genetic structure. Here I discuss some of these factors and how they may have impacted the northern fulmars. Physical isolation can create differentiation, as many seabirds do not fly over large areas of ice or land. The colonies in my study are large distances apart, but accessible by flying over water. Northern fulmars do not appear to be limited by large distances, so physical isolation is unlikely to be an important factor.

54 However, many other factors can prevent gene flow without physical barriers. Natal philopatry represents a potential restriction to gene flow in seabirds (Friesen 2015). Northern fulmars are not known to display strong natal philopatry (above), perhaps explaining the lack of structure in my dataset. The non-breeding and foraging distributions may impact genetic differentiation between populations. Populations that overlap during the non-breeding season or have a similar foraging distribution may be more likely to exchange genes than those that do not. Atlantic northern fulmars are essentially panmictic during the wintering period (Mallory et al. 2008) and are opportunistic feeders (Mallory et al. 2010). Therefore, there is large potential for overlap between birds from a wide range of colonies. The pattern of gene flow may also impact genetic differentiation. If gene flow occurs according to a stepping-stone model (where dispersal is primarily between neighboring populations), we would expect to see isolation by distance.

However, when gene flow is more random amongst colonies, then just one migrant per generation may be enough to homogenize colonies genetically (Wright 1969, Friesen et al.

2007). Finally, Friesen (2015) suggested that changes to the environment of a species could increase gene flow, as some breeding locations may be more suitable and receive more immigrants. The Arctic has been experiencing greater human activity and disturbance in recent years, which may impact the selection of breeding colonies by young northern fulmars in future generations.

Population assignment

Population assignment of Atlantic northern fulmars was challenging due to the weak genetic differentiation between colonies, and small sample sizes for most colonies. Several of the methods used for assignment did not produce satisfactory accuracy, and so assignment tests on the unknown individuals were not run using these methods. PCA, private alleles, and genetic

55 stock identification indicated that both groups of unknown birds have high within-group variation, and likely comprise northern fulmars from a range of colonies, as well as colonies that

were not included in this study. Pairwise FST estimates suggest that neither group shares much variation with Bjørnøya, but that bycatch birds may share greater variation with birds from

Qaqulluit, and birds from the Labrador Sea with the Faeroe Islands. Interestingly, the two groups of unknown birds share less variation with each other than with many of the colonies in the study. If the two groups were made of a similar mixture of birds, the groups would have likely

demonstrated a lower estimate of pairwise FST.

The machine-learning algorithm implemented in assignPOP did not produce sufficient assignment accuracy, despite machine-learning techniques proving reliable in Atlantic salmon

(Salmo salar) and Alaskan Chinook salmon (Oncorhynchus tshawystscha) (Sylvester et al.

2017). Genetic stock identification generally produced the highest assignment accuracy and indicated that most individuals from the Labrador Sea originated from Europe, and that all the birds caught in the bycatch originated from Arctic Canada. Determining the specific colony of origin within Arctic Canada proved more challenging. Regionally, birds from the Labrador Sea originating from Arctic Canada are likely from northern colonies. Bycatch birds were somewhat split in origin for each region, but as my assignments likely underestimate the number of birds from northern colonies the majority likely originated from this area. Results for assignment to either Prince Leopold Island and Cape Vera, or to Qaqulluit contradict the regional results, as most individuals appear to be from Qaqulluit, a southern colony. Due to the uncertainty in these results, and the likelihood that several birds originated from unsampled colonies, these results should be interpreted with caution. However, in general, the results strongly suggest that bycatch birds caught in the 0A and 0B fisheries management areas originated from Arctic Canada, and

56 the birds from the Labrador Sea represent a more mixed group, with many individuals originating from Europe. Both groups probably represented birds from multiple source colonies.

My results also demonstrate the utility of genomic data for population assignment of seabirds and potentially other highly mobile organisms. Population assignment of seabirds relies heavily on the genetic structure of the species in question (Edwards et al. 2001). Strongly differentiated seabird populations can often be resolved using only a few genetic markers or mtDNA (such as in the Black-footed albatrosses (Phoebastria nigripes), Walsh and Edwards

2005). However, many seabird colonies have a large amount of genetic variation and lack many fixed differences (Edwards et al. 2001). Gómez-Díaz and González-Solís (2007) assessed the use of genetic, morphometric and biogeochemical markers for population assignment of Calonectris shearwaters. Using mtDNA, birds were correctly assigned to species and subspecies designations. However, population-level assignment was unsuccessful using genetic markers, and required the use of biogeochemical markers instead. Assignment, therefore, can be difficult and Edwards et al. (2001) suggested the need for multiple polymorphic loci. My study indicates that the use of many genomic markers may allow population assignment even in weakly differentiated seabird populations. Fishing activities and impacts with offshore energy installations are global threats to pelagic seabirds (Melvin and Parrish 2001) and genomic approaches to population assignment may prove useful in future studies.

Conservation implications and taxonomy

The Faeroe Islands are the only colony in this study that would be classified as F. g. auduboni, the subspecies that Kerr and Dove (2018) suggested should be dissolved, instead classifying all

Atlantic northern fulmars as F. glacialis. My results demonstrate little differentiation between northern fulmars at the Faeroe Islands and other colonies, supporting the results of Kerr and

57 Dove (2018). However, the Faeroe Islands had the largest number of private alleles. This is particularly interesting considering that the Faeroe Islands is also the youngest colony included

in my study, only established in the 17thth Century (Fisher 1952a), so I expected this colony to be the least genetically diverse. The high allelic diversity suggests that this colony may have been founded from a colony that is not included in my samples, or from multiple colonies. Expansion in Europe is thought to have originated from Iceland (Fisher 1952b). As Icelandic samples were not included in this study, I cannot determine if the Faeroe Islands were colonized from Iceland.

Burg et al. (2014) also found higher haplotype diversity in the Faeroe Islands, and evidence for population growth. Genetic diversity and population growth were attributed to birds immigrating from large colonies and contributing to the colony diversity, providing further evidence for gene flow amongst northern fulmar colonies. The addition of other colonies from the subspecies F. g. auduboni is necessary to determine if there is a pattern of higher diversity at these colonies, and to assess more thoroughly the validity of F. g. auduboni as a subspecies. My research indicates that the Faeroe Islands may be an important colony that contributes to overall species diversity, but I do not have evidence that this colony is a separate subspecies.

My results also provide no evidence of multiple ESUs for northern fulmars in the

Atlantic. The six populations included in my study appear to be experiencing gene flow and have weak genetic differentiation. However, as outlined above, the populations are not completely panmictic. Determining whether the populations represent multiple MUs is more difficult.

According to Funk et al. (2012), MUs can be identified using neutral markers, which reflect neutral processes such as genetic drift and gene flow. Using the criterion by Funk et al. (2012), I would not define separate MUs amongst the colonies. However, MUs are classically defined as demographically independent units (Moritz 1994). Determining whether the colonies are

58 demographically independent (meaning that population dynamics are more strongly impacted by local birth or death rates, rather than migration) is difficult. My results suggest that gene flow does occur between colonies. However, even with high rates of migration, small but statistically

significant estimates of FST suggest that populations are connected, but that they are not panmictic

(Waples and Gaggiotti 2006). Further, northern fulmars are known to display high site fidelity, meaning that a northern fulmar that migrates to a different breeding colony will likely continue to return to that breeding colony for the rest of its life. Finally, northern fulmar colonies have been experiencing different population dynamics. Colonies on the east coast of Baffin Island have been decreasing at a rate of approximately 3% per year since the 1970s (Mallory et al. in prep). Meanwhile, Atlantic Canadian colonies (in Newfoundland and Labrador and northern

Québec) have been increasing since the 1970s (Stenhouse and Montevecchi 1999). Considering the differences in population status, and my assignment results indicating that fishing bycatch impacts colonies in the surrounding region, populations seem to be affected by local factors regardless of the influence of gene flow. In this case, separate colonies may be considered as separate MUs.

If northern fulmar colonies are demographically independent, Atlantic northern fulmars may be a meta-population. A meta-population is a group of demographically independent populations that could re-establish each other in the case of extirpation (Esler 2001). If the

Atlantic northern fulmars represent a meta-population, then declining colonies may avoid extirpation with migration from other colonies. Fangel et al. (2017) suggested that this may occur in northern fulmar colonies, based on the observation of improved attendance at a colony in Norway following a year of decline. Northern fulmars prefer cooler water temperatures

(annual mean temperature < 5° C) (Salmonsen 1965, Mallory 2008). As global temperatures

59 continue to change over the next few decades, the locations that northern fulmars choose to breed or winter at may be impacted, and colony sizes might shift as a result.

My results strongly suggest that the birds caught in fishing activities in 0A and 0B are from Arctic Canadian colonies. Bycatch in this area is mainly northern fulmars, with large amounts of seabird mortality occurring near Qaqulluit (Hedd et al. 2015). My results seem to reflect this, with all bycatch birds being identified as Canadian. Many bycatch birds were identified as coming from Qaqulluit. Therefore, bycatch in this area appears to be most heavily impacting surrounding colonies. Anderson et al. (2018) outline the projected impact of bycatch on the northern fulmar depending on the breadth of impact with three spatial scenarios: colonies within 200 km of locations where bycatch was recorded in the Baffin Bay-Davis Strait region

(including Qaqulluit and Akpait); colonies within 500 km; or all colonies in the region (from

Arctic Canada to Western Greenland, and including Prince Leopold Island and Cape Vera). My results indicate that bycatch may be affecting colonies in a wider range than 200 km, with suggestion that some bycatch birds originate from the more northerly Arctic colonies included in my study. Determining if the colonies from Greenland are also affected should be a next step to assess appropriately the impact of the bycatch. Depending on the extent of colonies affected by the bycatch, the colonies in this region could experience a decline between 0.3-4.6% over the next three generations. The estimated decline is much lower than the rate of decline that was observed in colonies on eastern Baffin Island (Mallory et al. in prep). Three colonies on the east coast of Baffin Island were surveyed aerially in August 2018, including Qaqulluit, to compare to previous colony estimates from the 1970s. The colonies were estimated to have an average annual decline of 3%. This discrepancy may be due to a severe underestimation of bycatch rates:

Anderson et al. (2018) suggested that the rate of bycatch is difficult to determine accurately, and

60 that bycatch in this area could be up to three times higher than officially recorded. Projected population declines are likely underestimations as a result. Furthermore, fishing activities are likely to increase, increasing the rate of bycatch further. If bycatch rate doubles, projected declines for the colonies in the Baffin Bay-Davis Strait region could be between 4.6-9.9% over the next three generations of northern fulmars (Anderson et al. 2018). Finally, other impacts on northern fulmars (such as colony disturbance, climate change and chemical or plastic contamination) may be contributing to the decline recorded at the eastern Baffin Bay colonies.

The colony name Qaqulluit is derived from ‘Qaqulluk’ (ᖃᖁᓪᓗᒃ), the Inuktitut word for northern fulmar (Qikiqtani Inuit Association 2018), and the colony was previously thought to be the largest northern fulmar colony in the world (Mallory 2016). Qaqulluit, then, may be an important source of immigrants for other colonies. Based on the rate of decline, and my population assignment results, this important colony is likely being impacted by fishing activities in the region, and possibly other anthropogenic activities.

Fisheries occur throughout the North Atlantic, and northern fulmars are a large component of seabird bycatch in many of these (Løkkeborg 2011, Kuhn and van Franeker 2012,

Fangel et al. 2017). The colonies in these areas are likely also the most impacted by these fishing activities. Northern fulmars in northern Norway have been experiencing severe decline and are now considered endangered on a national level (Kålås et al. 2015). The decline may be associated with several small-vessel fisheries in the area (Fangel et al. 2015, Fangel et al. 2017).

Therefore, bycatch is likely impacting northern fulmars on a local scale globally. Løkkeborg

(2011) provides an overview of methods to reduce seabird bycatch. Demersal longline fisheries are a major contributor to northern fulmar mortality, and the use of paired streamer lines (also known as ‘bird scaring lines’, with streamers to discourage birds from the area) can greatly

61 reduce the rate of bycatch. My results suggest that we need to invest more resources into reducing seabird bycatch. As outlined earlier, northern fulmars are sensitive to changes in their environment (Mallory 2006), and are susceptible to pollutants, plastics and climate variation

(Mallory et al. 2006, Provencher et al. 2009, Grosbois and Thompson 2005) as well as bycatch.

For these reasons, continuing to monitor the reduction in colony sizes is important.

From a conservation standpoint, my results are positive: the Atlantic northern fulmars are not fragmented, meaning they are less affected by genetic drift and can maintain overall high genetic diversity. Despite this, many colonies are experiencing decline, and bycatch is likely a large factor. Northern fulmar populations may be able to replenish each other, perhaps acting as a meta-population. My recommendation would be to treat different areas of northern fulmar breeding habitat (e.g. eastern Baffin Island, northern Baffin Island and surrounding colonies,

Bjørnøya, and boreal European colonies) as MUs and to monitor these populations and consider local impacts. The absence of separate ESUs highlights the necessity of collaborative conservation efforts for the future status of northern fulmars. Northern fulmars appear to be impacted by local threats, but they may also be affected by human activities during the non- breeding season. Atlantic northern fulmars occupy a wide range, and their migratory behaviour exposes them to threats in many different jurisdictions. Therefore, conservation efforts and government policies concerning northern fulmars, their habitat, and the environmental challenges that they face need to be coordinated across the range.

Future directions

My study has provided insight into the genetic structure of the Atlantic northern fulmar, and the sources of the birds caught in the bycatch in the 0A 0B fisheries management regions. This work would be improved with the addition of more colonies. The addition of colonies from Greenland

62 would help assess how wide the impact of the fisheries in 0A and 0B is. Additional European colonies and Icelandic samples might provide insight into the greater genetic diversity found in the Faeroe Islands. More colonies and greater numbers of samples at existing colonies might improve the assignment accuracy in population assignment. Including bycatch samples from other locations would also help determine if, as suggested here, impacts of bycatch are localized.

More generally, greater investment into research in the Arctic will be necessary to fully understand the implications that climate change and anthropogenic activities will have on Arctic species, as well as how to best protect them.

63 Table 9. Estimated number of breeding pairs of northern fulmars at the six colonies included in this study.

Colony Estimated number of breeding pairs

Cape Vera 9,000 (Gaston et al. 2006)

Prince Leopold Island 16,000 (Gaston et al. 2006)

Qaqulluit 25,000 (Mallory et al. in prep)

Akpait 15,000 (Akpait National Wildlife Area Management Plan 2018)

Faeroe Islands 600,000 (rough estimate, inclusive of all colonies on the Islands) (S. Hammer, pers. comm.)

Bjørnøya 8,000 (NINA Report 2013)

64 REFERENCES

Aguillon, S. M., Campagna, L., Harrison, R. G., and Lovette, I. J. 2018. A flicker of hope: genomic data distinguish northern flicker taxa despite low levels of divergence. Auk, 135:748- 766.

Aljanabi, S. M., and Martinez, I. 1997. Universal and rapid salt-extraction of high-quality genomic DNA for PCR-based techniques. Nucleic Acids Res., 25:4692-4693.

Allendorf, F. W., Luikart. G., and Aitken, S. N. 2013. Conservation and the genetics of populations. Hoboken: John Wiley & Sons.

Anderson, C. M., Iverson, S. A., Black, A., Mallory, M. L., Hedd, A., Merkel, F., and Provencher, J. F. 2018. Modelling demographic impacts of a growing Arctic fishery on a seabird population in Canada and Greenland. Mar. Environ. Res., 142:80-90.

Anderson, E. C. 2010. Assessing the power of informative subsets of loci for population assignment: standard methods are upwardly biased. Mol. Ecol. Resour., 10:701-710.

Anderson, E. C., Waples, R., and Kalinowski, S. 2008. An improved method for predicted the accuracy of genetic stock identification. C. J. Fish. Aquat. Sci., 65:1475-1486.

Andrews, S. 2010. FastQC: a quality control tool for high throughput sequence data. Available at: http://bioinformatics.babraham.ac.uk/projects/fastqc

Avery-Gomm, S., O’Hara, P. D., Kleine, L., Bowes, V., Wilson, L. K., and Barry, K. L. 2012. Northern fulmars as biological monitors of trends of plastic pollution in the eastern North Pacific. Mar. Poll. Bull., 64:1776-1781.

Barbosa, S., Mestre, F., White, T. A., Paupério, J., Alves, P. C., and Searle, J. B. 2018. Integrative approaches to guide conservation decisions: using genomics to define conservation units and functional corridors. Mol. Ecol., 27:3452-3465.

Barrett, R. D. H., and Schluter, D. 2008. Adaptation from standing variation. Trends Ecol. Evol., 23:38-44.

Bay, R. A., Harrigan, R. J., Underwood, V. L., Gibbs, H. L., Smith, T. B., and Ruegg, K. 2018. Genomic signals of selection predict climate-driven population declines in a migratory bird. Science, 359:83-86.

Bellard, C., Bertelsmeier, C., Leadley, P., Thuiller, W., and Courchamp, F. 2012. Impacts of climate change on the future of biodiversity. Ecol. Lett., 15:365-377.

Benestan, L., Gosselin, T., Perrier, C., Sainte-Marie, B., Rochette, R., and Bernatchez, L. 2015. RAD genotyping reveals fine-scale genetic structuring and provides powerful population

65 assignment in a widely distributed marine species, the American lobster (Homarus americanus). Mol. Ecol., 24:3299-3315.

Bennett, J. R., Shaw, J. D., Terauds, A., Smol, J. P., Aerts, R., Bergstrom, D. M., Blais, J. M., Cheung, W. W., Chown, S. L., Lea, M. A., et al. 2015. Polar lessons learned: long-term management based on shared threats in Arctic and Antarctic environments. Front. Ecol. Environ., 13: 316-324.

Brown, R. G. B. 1968. Fulmar distribution: a Canadian perspective. Ibis, 112:44-51.

Buckman, A. H., Norstrom, R. J., Hobson, K. A., Kanovsky, N. J., Duffe, J., and Fisk, A. T. 2004. Organochlorine contaminants in seven species of Arctic seabirds from northern Baffin Bay. Environ. Pollut., 128:327-338.

Burg, T. M., Bird, H., Lait, L., and Brooke, M. de L. 2014. Colonization pathways of the northeast Atlantic by northern fulmars: a test of James Fishers ‘out of Iceland’ hypothesis using museum collections. J. Avian Biol., 45:209-218.

Burg, T. M., Lomax, J., Almond, R., Brooke, M. de L., and Amos, W. 2003. Unravelling dispersal patterns in an expanding population of a highly mobile seabird, the northern fulmar (Fulmarus glacialis). Proc. R. Soc. Lond. B., 270:979-984.

Carmichael, L. E., Krizan, J., Nagy, J. A., Fuglei, F., Dummond, M., Johnson, D., Veitch, A., Berteaux, D., and Strobeck, C. 2007. Historical and ecological determinants of genetic structure in arctic canids. Mol. Ecol., 16:3466-34683.

Catchen, J., Hohenlohe, P., Bassham, S., Amores, A., and Cresko, W. 2013. Stacks: an analysis tool set for population genomics. Mol. Ecol., 22:3124-3140.

Ceballos, G., Ehrlich, P. R., and Rodolfo, D. 2017. Biological annihilation via the ongoing sixth mass extinction signaled by vertebrate population losses and declines. Proc. Natl. Acad. Sci. U.S.A., 114:E6089-E6096.

Chen, K-Y., Marschall, E. A., Sovic, M. G., Fries, A. C., Gibbs, H. L., and Ludsin, S. A. 2018. assignPOP: an R package for population assignment using genetic, non-genetic, or integrated data in a machine-learning framework. Methods Ecol. Evol., 9:439-446.

Chircop, A. 2007. Climate change and the prospects of increased navigation in the Canadian Arctic. WMU J. Marit. Aff., 6:193-205.

Christiansen, J. S., Mecklenburg, C. W., and Karamushko, O. V. 2014. Arctic marine fishes and their fisheries in the light of global change. Global Change Biol., 20:352-359.

Coates, D. J., Byrne, M., and Moritz, C. 2018. Genetic diversity and conservation units: dealing with the species-population continuum in the age of genomics. Front. Ecol. Evol., 6:165.

66 Crandall, K. A., Bininda-Emonds, O. R., Mace, G. M., Wayne, R. K. 2000. Considering evolutionary processes in conservation biology. Trends Ecol. Evol., 15:290-295.

Danecek, P., Auton, A., Abecasis, G., Albers, C. A., Banks, E., DePristo, M. A., Handsaker, R., Lunter, G., Marth, G., Sherry, S. T., et al. 2011. The variant call format and VCFtools. Bioinformatics, 27:2156-2158.

Davey, J. W., and Blaxter, M. L. 2011. RADSeq: next-generation population genetics. Brief. Funct. Genomics, 9:416-423.

DeSaix, M. G., Bulluck, L. P., Eckert, A. J., Viverette, C. B., Boves, T. J., Reese, J. A., Tonra, C. M., and Dyer, R. J. 2019. Population assignment reveals low migratory connectivity in a weakly structured songbird. Mol. Ecol., 28:2122-2135.

Dray, S., and Dufour, A. B. 2007. The ade4 package: implementing the duality diagram for ecologists. J. Stat. Softw., 22:1-20.

Dunn, E., and Steel, C. 2001. The impact of longline fishing on seabirds in the north-east Atlantic: recommendations for reducing mortality. RSPB, Sandy. NOR Rapportserie Report No 5-2001.

Dunnet, G. M. 1992. A forty-three year study on the fulmars on Eynhallow, Orkney. Scot. Birds, 16:155-159.

Dunnet, G. M., J. C. Ollason, and A. Anderson. 1979. A 28-year study of breeding fulmars Fulmarus glacialis in Orkney. Ibis, 21:293-300.

Earl, D. A., and vonHoldt, B. M. 2012. STRUCTURE HARVESTER: a website and program for visualizing STRUCTURE output and implementing the Evanno method. Conserv. Genet. Resourc., 4:359-361.

Ehrlich, P. R. 1988. The loss of diversity: causes and consequences. Pages 21-27, Wilson E.O., ed. Biodiversity. National Academy Press, Washington, D.C.

Edwards, S. V., Silva, M. C., Burg, T. M., Friesen, V. L., and Warheit, K. I. 2001. Molecular genetic markers in the analysis of seabird bycatch populations. Pages 115-140 in Melvin, E. F., and Parrish, J. K., editors. Seabird bycatch: trends, roadblocks and solutions. University of Alaska Sea Grant, Anchorage, Alaska, USA.

Environment and Climate Change Canada. 2018. Akpait National Wildlife Area Management Plan (Proposed). Environment and Climate Change Canada, Canadian Wildlife Service Northern Region.

Esler, D. 2001. Applying metapopulation theory to conservation of migratory birds. Conserv. Biol., 14:366-372.

67 Evanno, G., Regnaut, S., and Goudet, J. 2005. Detecting the number of clusters of individuals using the software STRUCTURE: a simulation study. Mol. Ecol., 14:2611-2620.

Excoffier, L. Smouse, P. E., and Quattro, J. M. 1992. Analysis of molecular variance inferred from metric distances among DNA haplotypes- application to human mitochondrial-DNA restriction data. Genetics, 131:479-491.

Fangel, K., Aas, Ø., Vølstad, J. H., Bærum, K. M., Christensen-Dalsgaard, S., Nedreaas, K., Overdik, M., Wold, L. C., and Anker-Nilssen, T. 2015. Assessing incidental bycatch of seabirds in Norwegian coastal commercial fisheries: empirical and methodological lessons. Glob. Ecol. Conserv., 4:127-136.

Fangel, K., Bærum, K. M., Christensen-Dalsgaard, S., Aas, Ø., and Anker-Nilssen, T. 2017. Incidental bycatch of northern fulmars in the small-vessel demersal longline fishery for Greenland halibut in coastal Norway 2012-2014. ICES J. Mar. Sci., 74:332-342.

Fisher, J. 1952a. The fulmar. Collins.

Fisher, J. 1952b. A history of the fulmar, Fulmarus, and its population problems. Ibis, 94:334- 354.

Foll, M., and Gaggiotti, O. 2008. A genome-scan method to identify selected loci appropriate for both dominant and co-dominant markers: a Bayesian perspective. Genetics, 180:977-993.

Frankham, R. 1996. Relationship of genetic variation to population size in wildlife. Cons. Biol., 10:1500-1508.

Frankham, R. 2005. Genetics and extinction. Biol. Cons., 126:131-140.

Frankham, R., and Ralls, K. 1998. Inbreeding leads to extinction. Cons. Biol., 392:441-442.

Fraser, D. J., and Bernatchez, L. 2001. Adaptive evolutionary conservation: towards a unified concept for defining conservation units. Mol. Ecol., 10:2741-2752.

Friesen, V. L. 2015. Speciation in seabirds: why are there so many species...and why aren't there more? J. Ornithol., 156: S27-S39.

Friesen, V. L., Burg, T. M., and McCoy, K. D. 2007. Mechanisms of population differentiation in seabirds. Mol. Ecol. 16:1765-1785.

Friesen, V. L., Congdon, B. C., Walsh, H. E., and Birt, T. P. 1997. Intron variation in marbled murrelets detected using analyses of single-stranded conformational polymorphisms. Mol. Ecol., 6:1047-1058.

Funk, W. C., McKay, J. K., Hohenlohe, P. A., and Allendorf, F W. 2012. Harnessing genomics for delineating conservation units. Trends in Ecol. Evol., 27:489-496.

Furness, R. W., and Camphuysen, C. J. 1997. Seabirds as monitors of the marine environment. ICES J. Mar. Sci., 54:726-737.

Gaston, A. J., Bertram, A. W., Boyne, J. W., Chardine, G., Davoren, A. W., Diamond, A., Hedd, A., Montevecchi, W. A., Hipfner, J. M., Lemon, M. J. F., et al. 2009. Changes in Canadian seabird populations and ecology since 1970 in relation to changes in oceanography and food webs. Environ. Rev., 17:267-286.

Gaston, A. J., Mallory, M. L., and Gilchrist, H. G. 2012. Populations and trends of Canadian Arctic seabirds. Polar Biol., 35:1221-1232.

Gaston, A. J., Mallory, M. L., Gilchrist, H. G., and O’Donovan, K. 2006. Status, trends and attendance patterns of the Northern Fulmar, Fulmarus glacialis, in Nunavut, Canada. Arctic, 59:165-178.

Google Maps. 2019. International Cartographic Association. (https://www.google.ca/maps/@50.7990075,-150.9767058,3.87z)

Gómez-Díaz, E., and González-Solís, J. 2007. Geographic assignment of seabirds to their origin: combining morphologic, genetic, and biogeochemical analyses. Ecol. Appl., 17:1484-1498.

Goudet, J. 2005. HIERFSTAT, a package for R to compute and test hierarchical F-statistics. Mol. Ecol. Resour., 5:184-186.

Gravley, M. C., Sage, G. K., Ramey, A. M., Hatch, S. A., Gill, V. A., Rearick-Whitney, J. R. R., Petersen, A., and Talbot, S. L. 2019. Development and characterization of polymorphic microsatellite markers in northern fulmar, Fulmarus glacialis (Procellariiformes), and cross- species amplification in eight other seabirds. Genes Genom, 41:1015-1026.

Green, D. M. 2005. Designatable units for status assessment of endangered species. Cons. Biol., 19:1813-1820.

Greenbaum, G., Templeton, A. R., Zarmi, Y., and Bar-David, S. 2014. Allelic richness following population founding events- a stochastic modeling framework incorporating gene flow and genetic drift. PLoS One, 9:e115203.

Grosbois, V., and Thompson, P. M. 2005. North Atlantic climate variation influences survival in adult fulmars. Oikos, 109:273-290.

Hatch, S. A. 1987. Adult survival and productivity of northern fulmars in Alaska. Condor, 89:685-696.

Hatch, S. A., and Nettelship, D. N. 1998. Northern fulmar (Fulmarus glacialis). In: Poole A, Birds of North America online. Ithaca, NY: Cornell Lab of Ornithology.

69 Hatch, S. A., Gill, V. A., and Mulcahy, D. M. 2010. Individual and colony-specific wintering areas of Pacific Northern Fulmars (Fulmarus glacialis). Can. J. Fish. Aquat. Sci., 67:386-400.

Hedd, A., Regular, P. M., Wilhelm, S. I., Rail, J-F., Drolet, B., Folwer, M., Pekarik, C., and Robertson, G. J. 2015. Characterization of seabird bycatch in eastern Canadian waters 1998- 2011, assessed from onboard fisheries observer data. Aquat. Conserv. Mar. Fresw. Ecosyst., 26:530-548.

Hedrick, P. W. 2013. Adaptive introgression in animals: examples and comparison to new mutation and standing variation as sources of adaptive variation. Mol. Ecol., 22:4606-4618.

Hoegh-Guldberg, O., and Bruno, J. F. 2010. The impact of climate change on the world’s marine ecosystems. Science, 328:1523-1528.

Hutchison, D. W., and Templeton, A. R. 1999. Correlation of pairwise genetic and geographic distances measures: inferring the relative influences of gene flow and drift on the distribution of genetic variability. Evolution, 53:1898-1914.

Jeffery, N. W., Wringe, B. F., McBride, M. C., Hamilton, L. C., Stanley, R. R. E., Bernatchez, L., Kent, M., Clément, M., Gilbey, J., Sheehan, T. F., et al. 2018. Range-wide regional assignment of Atlantic salmon (Salmo salar) using genome-wide single-nucleotide polymorphisms. Fish. Res., 206:163-175.

Jenouvrier, S., Desprez, M., Fay, R., Barbraud, C., Weimerskirch, H., Delord, K., and Caswell, H. 2017. Climate change and functional traits affect population dynamics of a long-lived seabird. J. Anim. Ecol., 87:906-920.

Jombart, T., and Ahmed, I. adegenet 1.3-1: new tools for the analysis of genome-wide SNP data. Bioinformatics, 27:3070-3071.

Jones, O. R., Gaillard, J. M., Tuljapurkar, S., Alho, J.S., Armitage, K. B., Becker, P. H., Bize, P., Brommer, J., Charmantier, A., Charpentier, M., et al. 2008. Senescence rates are determined by ranking on the fast-slow life-history continuum. Ecol. Lett., 11:664-673.

Jouanin, C., and Mougin, J. L. 1979. Order Procellariiformes. In: Check-list of birds of the world, 48-121 (Mayr, E., and Cottrell. G. W.). Museum of Comparative Zoology, MA. Cited in: Mallory, M. L., Hatch, S. A., and Nettleship, D. N. 2012. Northern Fulmar (Fulmarus glacialis) version 2.0 in The Birds of North America (P.G. Rodewald, editor). Cornell Lab of Ornithology, Ithaca, New York, USA.

Kålås, J. A., Dale, S., Gjershaug, J. O., Husby, M., Luslevant, T., Strann, K. B., and Strøm, H. 2015. Fugler (Aves). Norsk rødliste for arter 2015. Artsdatabanken, Trondheim. Cited in: Fangel, K., Bærum, K. M., Christensen-Dalsgaard, S., Aas, Ø., and Anker-Nilssen, T. 2017. Incidental bycatch of northern fulmars in the small-vessel demersal longline fishery for Greenland halibut in coastal Norway 2012-2014. ICES J. Mar. Sci., 74:332-342.

70 Kamvar, Z. N., Brooks, J. C., and Grünwald, N. J. 2015. Novel R tools for analysis of genome- wide population genetic data with emphasis on clonality. Front. Genet., 6:208.

Keel, B. N., and Snelling, W. M. 2018. Comparison of Burrows-Wheeler transform-based mapping algorithms used in high-throughput whole-genome sequencing: application to illumina data for livestock genomes. Front. Genet., 9:35.

Kent, C. F., Dey, A., Patel, H., Tsvetkov, N., Tiwari, T., MacPhail, V. J., Gobeil, Y., Harpur, B. A., Gurtowski, J., Schatz, M. C., et al. 2018. Conservation genomics of the declining North American bumblebee Bombus terricola reveals inbreeding and selection on immune genes. Front. Genet., 9:316.

Kerr, C. R., and Dove, C. J. 2013. Delimiting shades of gray: phylogeography of the Northern Fulmar, Fulmarus glacialis. Ecol. Evol., 3:1915-1930.

Kerr, C. R., and Dove, C. J. Submitted April 2018. Split Northern Fulmar Fulmarus glacialis into two species.

Klassen, R. H. G., Hake, M., Strandberg, R., Koks, B. J., Trierweiler, C., Exo, K-M., Bairlein, F., and Alerstam, T. 2013. When and where does mortality occur in migratory birds? Direct evidence from long-term satellite tracking of raptors. J. Anim. Ecol., 83:176-184.

Kuhn, S., and van Franeker, J. 2012. Plastic ingestion by the northern fulmar (Fulmarus glacialis) in Iceland. Mar. Pollut. Bull., 64:1252-1254.

Lacy, R. C. 1987. Loss of genetic diversity from managed populations: interacting effects of drift, mutation, immigration, selection, and population subdivision. Cons. Biol., 1:143-158.

Lah, L., Trense, D., Benke, H., Berggren, P., Gunnlaugsson, Þ., Lockyer, C., Özturk, B., Pawliczka, I., Roos, A., Siebert, U., et al. 2016. Spatially explicit analysis of genome-wide SNPs detects subtle population structure in a mobile marine mammal, the harbor porpoise. PloS One, 11:e0162792.

Lecaudey, L. A., Schliewen, U. K., Osinov, A., G., Taylor, E. B., Bernatchez, L., and Weiss, S. J. 2018. Inferring phylogenetic structure, hybridization and divergence times with Salmonidae (Teleostei: Salmonidae) using RAD-sequencing. Mol. Phylogenetics Evol., 124:82-99.

Li, H., and Durbin, R. 2010. Fast and accurate long-read alignment with Burrows-Wheeler Transform. Bioinformatics, Epub. [PMID: 20080505].

Li, H., Handsaker, B., Wysoker, A., Fennell, T., Ruan, J., Homer, N., Marth, G., Abecasis, G., Durbin, R., and 1000 Genome Project Data processing subgroup. 2009. The sequence alignment/map (SAM) format and SAMtools. Bioinformatics, 25:2078-2079.

Lischer, H. E. L., and Excoffier, L. 2012. PGDSpider: an automated data conversion tool for connecting population genetics and genomics programs. Bioinformatics, 28:298-299.

Liukart, G., England, P. R, Tallmon, D. Jordan, S., and Taberlet, P. 2003. The power and promise of population genomics: from genotyping to genome typing. Nature. Rev. Genet., 4: 981-994.

Løkkeborg, S. 2011. Best practices to mitigate seabird bycatch in longline, trawl and gillnet fisheries- efficiency and practical applicability. Mar. Ecol. Prog. Ser., 435:285-303.

Macdonald, M. A. 1977. An analysis of the recoveries of British-ringed fulmars. Bird Study, 24:208-214.

Macias-Fauria, M., and Post, E. 2018. Effects of sea ice on Arctic biota: an emerging crisis discipline. Biol. Lett., 14:20170702.

Mallory, M. L. 2006. The Northern Fulmar (Fulmarus glacialis) in Arctic Canada: Ecology, threats, and what it tells us about marine environmental conditions. Environ. Rev. 14:187-216.

Mallory, M. L. 2016. Recommended Revisions to: Key Marine Habitat Sites for Migratory Birds in Nunavut and the Northwest Territories. Wolfville, NS: CWS OP 109 (Draft).

Mallory, M. L., Akearok, J. A., Edwards, D. B., O'Donovan, K., and Gilbert, C. D. 2008. Autumn migration and wintering of Northern Fulmars (Fulmarus glacialis) from the Canadian High Arctic. Polar Biol., 31:745-750.

Mallory, M. L., Braune, B. M., and Forbes, M. R. L. 2006. Contaminant concentrations in breeding and non-breeding northern fulmars (Fulmarus glacialis L.) from the Canadian high Arctic. Chemosphere, 64:1541-1544.

Mallory, M. L., Gaston. A. J., Forbes, M. R., and Gilchrist, H. G. 2009. Influence of weather on reproductive success of northern fulmars in the Canadian high Arctic. Polar Biol., 32:529-538.

Mallory, M. L., Hatch, S. A., and Nettleship, D. N. 2012. Northern Fulmar (Fulmarus glacialis) version 2.0 in The Birds of North America (P.G. Rodewald, editor). Cornell Lab of Ornithology, Ithaca, New York, USA.

Mallory, M. L., Karnovsky, N. J., Gaston, A. J., Hobson, K. B., Provencher, J. F., Forbes, M. R., Jr. Hunt, G. L., Byers, T., and Dick, T. A. 2010. Temporal and spatial patterns in the diet of Northern Fulmars Fulmarus glacialis in the Canadian High Arctic. Aquat. Biol., 10:181-191.

Maniatis, T., Frich, E. F., and Sambrook, J. 1982. Molecular cloning: a laboratory manual. NY, USA: Cold Spring Harbor Laboratory.

McCauley, D. E. 1993. Genetic consequences of extinction and recolonization in fragmented habitats. Pages 217-233 in O. M. Kareiva, J. G. Kingsolver, and R. B. Huey, eds. Biotic interactions and global change. Sinauer, Sunderland, MA. Cited in: Hutchison, D. W., and Templeton, A. R. 1999. Correlation of pairwise genetic and geographic distances measures:

72 inferring the relative influences of gene flow and drift on the distribution of genetic variability. Evolution, 53:1898-1914.

McGlashan, D. J., and Hughes, J. M. 2000. Low levels of genetic differentiation among populations of the freshwater fish Hypseleotris compressa (Gobiidae: Eleotridinae): implications for its biology, population connectivity and history. Heredity, 86:222-233.

McGowan, A. 2018. Confronting the challenges of a warming arctic. Environment: Science and Policy for Sustainable Development, 60: 12-15.

McNeely, J. A., Miller, K. R., Reid, W. V., Mittermeier, R. A., and Werner, T. B. 1990. Conserving the world’s biological diversity. IUCN, World Resources Institute, Conservation International, WWF-US and the World Bank: Washington D.C.

Melvin, E. F., and Parrish, J. K. 2001. Seabird bycatch: trends, roadblocks and solutions. University of Alaska Sea Grant, Anchorage, Alaska, USA.

Meirmans, P. G. 2013. Nonconvergence in Bayesian estimation of migration rates. Mol. Ecol., 14:726-733.

Meirmans, P. G., and Van Tienderen, P. H. 2004. GENOTYPE and GENODIVE: two programs for the analysis of genetic diversity of asexual organisms. Mol. Ecol. Notes, 4: 792-794

Meirmans, P.G. 2012. AMOVA-based clustering of population genetic data. J. Hered., 103:744- 750.

Michalakis, Y., and Excoffier, L. 1996. A generic estimation of population subdivision using distances between alleles with special reference for microsatellite loci. Genetics, 142:1061-1064.

Mora-Márquez F, García-Olivares V, Emerson BC, López de Heredia U. 2017. ddRADseqTools: a software package for in silico simulation and testing of double-digest RADseq experiments. Mol. Ecol. Resour., 17:230-246.

Moritz, C. 1994. Defining “evolutionarily significant units” for conservation. Trends Ecol. Evol., 9:373-375.

Morris-Pocock, J. A., Taylor, S. A., Birt, T. P., Damus, M. Piatt, J. F., Warheit, K. I., and Friesen, V. L. 2008. Population genetic structure in Atlantic and Pacific Ocean common murres (Uria aalge): natural replicate tests of post-Pleistocene evolution. Mol. Ecol., 17:4859-4873.

Mullins, R. B., McKeown, N. J., Sauer, W. H. H., and Shaw, P. W. 2018. Genomic analysis reveals multiple mismatches between biological and management units in yellowfin tuna. ICES J. Mar. Sci., 75:2145-2152.

73 Nei, M. 1987. Molecular Evolutionary Genetics. Columbia University Press, New York.

Ney, G., Schul, J. 2017. Low genetic differentiation between populations of an endemic prairie katydid despite habitat loss and fragmentation. Conserv. Genet., 18:1389-1401.

Nieto-Montes de Oca, A., Barley, A., J., Meza-Lázaro, R. N., García-Vázquez, U. O., Zamora- Abrego, J. G., Thomson, R. C., and Leaché, A. D. 2017. Phylogenomics and species delimitation in the knob-scaled lizards of the genus xenosaurus (Squamata: Xenosauridae) using ddRADseq data reveal a substantial underestimation of diversity. Mol. Phylogenet. Evol., 106:241-253.

NINA Report 2018. The status and trends of seabirds breeding in Norway and Svalbard. Norwegian Institute for Nature Research.

O’Leary, S. J., Puritz, J. B., Willis, S. C., Hollenbeck, C. M., and Portnoy, D. S. 2018. These aren’t the SNPs you’re looking for: principles of effective SNP filtering for molecular ecologists. Mol. Ecol., 27:3193-3206.

Oubourg, N. J., Pertoldi, C., Loeschcke, V., Bijlsma, R., and Hedrick, P. W. 2010. Conservation genetics in transition to conservation genomics. Cell Press, 26: 177-187.

Pallsbøll, P. J., Bérubé, M., and Allendorf, F. W. 2007. Identification of management units using population genetic data. Trends. Ecol. Evol., 22:11-16.

Parkinson, C. L., and Comiso, J. C. 2013. On the 2012 record low Arctic sea ice cover: combined impact of preconditioning and an August storm. Geophys. Res. Lett., 40:1356-1361.

Pazmiño, D. A., Maes, G. E., Simpfendorfer, C. A., Salinas-de-León, P., and van Herwerden, L. 2017. Genome-wide SNPs reveal low effective population size within confined management units of the highly vagile Galapagos shark (Carcharhinus galapagenesis). Conserv. Genet., 18:1151-1163.

Pecl, G. T., Araújo, M. B., Bell, J. D., Blanchard, J., Bonebrake, T. C., Chen, I. C., Clark, T. D., Colwell, R. K., Danielsen, F., Evengård, B. et al. 2017. Biodiversity redistribution under climate change: impacts on ecosystems and human well-being. Science, 355 p. eaai9214.

Peterson, B., Weber, J., Kay, E., Fisher, H., and Hoekstra, H. 2012. Double digest RADseq: an inexpensive method for de novo SNP discovery and genotyping in model and non-model species. PLoS One, 7:e37135.

Piatt, J. F., Sydeman, W. J., and Wiese, F. 2007. Introduction: a modern role for seabirds as indicators. Mar. Ecol. Prog. Ser., 352:199-204.

Pritchard, J. K., Stephens, M., and Donnelly, P. 2000. Inference of population structure using multilocus genotype data. Genetics,155:945-959.

74 Provencher, J. F., Gaston, A. J., and Mallory, M. L. 2009. Evidence for increased ingestion of plastics by northern fulmars (Fulmarus glacialis) in the Canadian Arctic. Marine Poll. Bull., 1078- 1096.

Puechmaille, S. J. 2016. The program STRUCTURE does not reliably recover the correct population structure when sampling is uneven.: subsampling and new estimators alleviate the problem. Mol. Ecol. Resour., 16:608-627.

Quérouil, S., Silva, M. A., Freitas, L., Prieto, R., Magalhães, S., Dinis, A., Alves, F., Matos, J. A., Mendonça, D., Hammond, P. S., et al. 2007. High gene flow in oceanic bottlenose dolphins (Tursiops truncatus) in the North Atlantic. Conserv. Genet., 8:1405-1419.

Qikiqtani Inuit Association. 2018. Qikiqtaaluk Inuit Qaujimajatuqangit and Inuit Qaujimajangit Iliqqusingitigut for the Baffin Bay and Davis Strait Marine Environment. Prepared by Heidi Klein, Sanammanga Solutions Inc. for submission to the Nunavut Impact Review Board for the Baffin Bay and Davis Strait Strategic Environmental Assessment.

R Core Team. 2018. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing. Vienna, Austria. https://www.R-project.org/.

Rajpar, M. N., Ozdemir, I., Zakaria, M., Sheryar, S., and Rab, A. 2018. Seabirds as bioindicators of marine ecosystems, Seabirds, Heimo Mikkola, IntechOpen. doi: 10.5772/intechopen.75458.

Rodríguez-Ezpeleta, N., Bradbury, I. R., Medibil, I., Álvarez, P., Cotano, U., and Irigoien, X. 2016. Population structure of Atlantic mackerel inferred from RAD-seq-derived SNP markers: effects of sequence clustering parameters and hierarchical SNP selection. Mol. Ecol. Resour., 16:991-1001.

Row, J. R., Gomez, C., Koen, E. L., Bowman, J., Murray, D. L., and Wilson, P. J. 2012. Dispersal promotes high gene flow among Canada lynx populations across mainland North America. Conserv. Genet., 13:1259-1268.

Ruegg, K. C., Anderson, E. C., Paxton, K. L., Apkenas, V., Lao, S., Siegal, R. B., Desante, D. F., Moore, F., and Smith, T. B. 2014. Mapping migration in a songbird using high-resolution genetic markers. Mol. Ecol., 23:5726-5739.

Ryder, O. A. 1986. Species conservation and systematics: the dilemma of subspecies. Trends. Ecol. Evol,.1:9-10.

Salomonsen, F. 1965. The geographical variation of the fulmar (Fulmarus glacialis) and the zones of the marine environment in the North Atlantic. Auk, 82:327-355.

Sambrook, J. Fritsch, E. F., and Maniatis, T. 1989. Molecular Cloning: A Laboratory Manual. 2nd edition. Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press.

75 Shafer, A. B. A., Wolf, J. B. W., Alves, P. C., Bergström, L., Bruford, M. W., Brännström, Colling, G., Dalén, L., De Meester, L., Ekblom, R., et al. 2015. Genomics and the challenging translation into conservation practice. Trends. Ecol. Evol., 30:78-87.

Sherval, M. 2015. Canada’s oil sands: the mark of a new ‘oil age’ or a potential threat to Arctic security? Extr. Ind. Soc., 2:225-236.

Slatkin, M. 1985. Rare alleles as indicators of gene flow. Evolution, 39:53-65.

Stanley, R. R. E., Jefferey, N. W., Wringe, B. F., Dibacco, C., and Bradbury, I. R. 2017. GENEPOPEDIT: a simple and flexible tool for manipulating multilocus molecular data in R. Mol. Ecol., 17:12-18.

Stenhouse, I. J., and Montevecchi, W. A. 1999. Increasing and expanding populations of breeding northern fulmars in Atlantic Canada. Waterbirds, 22:382-391.

Supple, M. A., and Shapiro, B. 2018. Conservation of biodiversity in the genomics era. Genome Biol., 19:131.

Sylvester, E. V., Bentzen, P., Bradbury, I. R., Clément, M., Pearce, J., Horne, J., and Beiko, R. G. 2017. Application of random forest feature selection for fine-scale genetic population assignment. Evol. Appl., 11:153-165.

Theron, E., Hawkins, K., Bermingham, E., Ricklefs, R. E., and Mundy, N. I. 2001. The molecular basis of an avian plumage polymorphism in the wild: a melanocortin-1-receptor point mutation is perfectly associated with the melanic plumage morph of the bananquit, Coereba flaveola. Curr. Biol., 11:550-557.

Tigano, A., Shultz, A. J., Edwards, S. V., Roberstson G. J., and Friesen, V. L. 2017. Outlier analyses to test for local adaptation to breeding grounds in a migratory Arctic seabird. Ecol. Evol., 7:2370-2381.

Trevail, A. M., Gabrielsen, G. W., Kühn, S., and van Franeker, J. F. 2015. Elevated levels of ingested plastic in a high Arctic seabird, the northern fulmar (Fulmarus glacialis). Polar Biol., 38:975-981.

Tuck, L. M. 1971. The occurrence of Greenland and European birds in Newfoundland. Bird- Banding, 42:184-209.

Valencia, L. M., Martins, A., Ortiz, E. M., and Di Fiore, A. 2018. A RAD-sequencing approach to genome-wide marker discovery, genotyping, and phylogenetic inference in a diverse radiation of primates. PLoS One, 13:e0201254.

Van Franeker, J. A. 1985. Plastic ingestion in the North Atlantic Fulmar. Mar. Poll. Bull., 16:367-369.

76 Van Franeker, J. A., Blaize, C., Danielsen, J., Fairclough, K., Gollan, J., Guse, N., Hansen, P-L., Heubeck, M., Jensen, J-K., et al. 2011. Monitoring plastic ingestion by the northern fulmar Fulmarus glacialis in the North Sea. Environ. Pollut., 159:2609-2615.

Wang, J. 2019. A parsimonious estimator of the number of populations from STRUCTURE-like analysis. Mol. Ecol. Resourc., 19:970-981.

Walsh, H. E., and Edwards, S. V. 2005. Conservation genetics and Pacific fisheries bycatch: mitochondrial differentiation and population assignment in black-footed albatrosses (Phoebastria nigripes). Conserv. Genet., 6:289-295.

Waples, R. S., and Gaggiotti, O. 2006. What is a population? An empirical evaluation of some genetic methods for identifying the number of gene pools and their degree of connectivity. Mol. Ecol., 15:1419-1439.

Wauchope, H. S., Shaw, J. D., Varpe, Ø., Lappo, E. G., Boertmann, D., Lanctot, R. B., and Fuller, R. A. 2016. Rapid climate-driven loss of breeding habitat for Arctic migratory birds. Global Change Biol., 23:1085-1094.

Weir, B. S., and Cockerham, C. C. Estimating F-statistics for the analysis of population structure. Evolution, 38:1358-1370.

Wickham, H. 2016. ggplot2: Elegant graphics for data analysis. Springer-Verlag, New York.

Wilson, G. A., and Rannala, B. 2003. Bayesian inference of recent migration rates using multilocus genotypes. Genetics, 163: 1177-1191.

Wojczulanis-Jakubas, K., Kilikowska, A., Harding, A. M., Jakubas, D., Karnovsky, N. J., Steen, H., and Johnsen, A. 2014. Weak population genetic differentiation in the most numerous Arctic seabird, the little auk. Polar Biol., 37:621-630.

Wright, S. 1969. Evolution and the genetics of populations, Vol. 2: The theory of gene frequencies. University of Chicago Press, Chicago. Cited in: Friesen, V. L. 2015. Speciation in seabirds: why are there so many species….and why aren't there more? J. Ornithol., 156: S27- S39.

Zarza, E., Faircloth, B. C., Tsai, W. L. E., Bryson Jr., R. W., Klicka, J., and McCormack, J. E. 2016. Hidden histories of gene flow in highland birds revealed with genomic markers. Mol. Ecol., 25:5144-5157.

77 APPENDIX

Supplementary Table 1. Pairwise FST for the six colonies.

Cape Prince Qaqulluit Akpait Faeroe Vera Leopold Islands Island

Cape Vera - Prince Leopold Island 0.000 - Qaqulluit 0.004 0.000 - Akpait 0.000 0.000 0.002 - Faeroe Islands 0.005 0.005 0.008 0.003 - Bjørnøya 0.005 0.004 0.007 0.005 0.007

Supplementary Figure 1. The first two principal components of a principal components analysis for breeding birds using all 6614 SNPs.