Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

An analysis of ecological traits as reproductive barriers between the MacGillivray’s (Geothlypis tolmiei)… Porter, Alison Nicole 2015

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2015_september_porter_alison.pdf [ 1.16MB ]
Metadata
JSON: 24-1.0167279.json
JSON-LD: 24-1.0167279-ld.json
RDF/XML (Pretty): 24-1.0167279-rdf.xml
RDF/JSON: 24-1.0167279-rdf.json
Turtle: 24-1.0167279-turtle.txt
N-Triples: 24-1.0167279-rdf-ntriples.txt
Original Record: 24-1.0167279-source.json
Full Text
24-1.0167279-fulltext.txt
Citation
24-1.0167279.ris

Full Text

An analysis of ecological traits as reproductive barriers between the MacGillivray’s (Geothlypis tolmiei) and Mourning (G. philadelphia) warblers   by Alison Nicole Porter B.Sc., Queen’s University, 2012  A THESIS SUMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  MASTER OF SCIENCE in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Zoology)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) May 2015 ©Alison Nicole Porter, 2015     ii  Abstract Speciation in birds is often thought to be influenced by ecological diversification, which may form barriers to reproduction between species. Ecological selection can act as a premating barrier by reducing the chance of interactions between species, or as a postmating barrier if hybrids have ecological traits that are unfavourable. My objective in this study is to understand the role of ecological traits in maintaining isolation between MacGillivray’s (Geothlypis tolmiei) and Mourning warblers (G. philadelphia), which are two songbird species that form a narrow hybrid zone in northeastern British Columbia. I generate ecological niche models for each species to investigate whether the niche might have a role in explaining the location and width of the hybrid zone by comparing niche and range limits, and whether there is evidence for niche divergence. I show that the species have niches that are similar but have diverged in climatic variability and precipitation measures. These differences may partly explain why the niche models predicted that geographic regions within a species’ own range were the most suitable for that species. The contact zone was the only region where both models predicted high suitability, suggesting that hybridization occurs only where conditions are suitable for both species. I also present two analyses of how genotype is related to ecological characteristics in the hybrid zone. I use the niche models to predict the relative suitability of locations where birds occupied breeding territories in the hybrid zone and compared the relative scores among a gradient of hybrid genotypes. At those same hybrid zone territories I measure microhabitat characteristics of territories and conducted a Principal Component Analysis to examine whether there is a relationship between microhabitat of territories with genotype. I found no evidence that genotype was related with ecological traits in the hybrid zone. This suggests that when under the same ecological conditions the two species are ecologically equivalent, and that hybrids likely do not experience a disadvantage in terms of the ecological traits of their breeding territories. Overall, these results suggest that niche divergence likely had only a small role in the diversification between the MacGillivray`s and Mourning warblers.       iii  Preface The research question and project for this thesis were designed in collaboration between me and my supervisor Dr. Darren Irwin. I carried out the field work for collecting microhabitat data and blood samples from the birds sampled for microhabitat analyses. Some additional blood samples used in genetic analyses were collected in the field by Darren Irwin and Alan Brelsford. I carried out DNA extraction, while Christine Grossen prepared genotype by sequencing libraries, and conducted most of the quality filtering and SNP calling. Methods for the steps conducted by Christine Grossen were written in collaboration between me and Christine Grossen. I conducted all analyses and other writing of the thesis. Julie Lee-Yaw worked in collaboration with me on aspects of ecological niche modelling. Specifically, Julie Lee-Yaw assisted with providing advice in devising methodology for generating and testing ecological niche models, and also provided helpful feedback on results and interpretations. Darren Irwin advised with helpful feedback on results and interpretations. Animal Care approval was obtained for this project (Canadian Council on Animal Care /National Institutional Animal User Training Program certificate number 5819 - 13; UBC Animal Care Certificate application number A13-0047). Bird banding and scientific permits were issued for this work from Environment Canada (BC-12-0027, 09-AB/SK-SC008, 10746 K).      iv  Table of contents  Abstract ........................................................................................................................................... ii Preface............................................................................................................................................ iii Table of contents ............................................................................................................................ iv List of tables ................................................................................................................................... vi List of figures ............................................................................................................................... viii Acknowledgements ........................................................................................................................ xi Chapter 1: Introduction ................................................................................................................... 1   1.1 Speciation in birds................................................................................................................... 1   1.2 The role of ecology in avian speciation .................................................................................. 1   1.3 Using hybrid zones to study speciation .................................................................................. 5   1.4 Studying ecological divergence from multiple perspectives .................................................. 6   1.5 Research goals and questions.................................................................................................. 9 Chapter 2: Does ecological divergence explain the position and narrowness of the MacGillivray’s/Mourning warbler (Geothlypis tolmiei/philadelphia) hybrid zone? .......... 11   2. 1 Introduction .......................................................................................................................... 11  2.1.1 Ecological divergence as a reproductive barrier ........................................................ 11  2.1.2 The MacGillivray’s and Mourning warbler hybrid zone ........................................... 13  2.1.3 Ecological traits of the MacGillivray’s and Mourning warblers ................................ 14  2.1.4 Research questions and predictions ............................................................................ 15   2.2 Methods................................................................................................................................. 16  2.2.1 Ecological niche modelling ........................................................................................ 16    2.2.1.1 Defining the study extent ....................................................................................... 16    2.2.1.2 Preparing the training dataset ................................................................................ 17    2.2.1.3 Model generation and cross-validation ................................................................. 19    2.2.1.4 Testing for niche divergence ................................................................................. 20    2.2.2.1 Microhabitat sampling ........................................................................................... 22    2.2.2.2 Bird sampling ........................................................................................................ 23    2.2.2.3 Microhabitat features of areas without birds ......................................................... 24  2.2.3 Developing the hybrid index ...................................................................................... 24  2.2.4 Relationship between genotype and ecological traits in the hybrid zone ................... 25    v     2.2.4.1 Relationship between genotype and microhabitat in the hybrid zone ................... 25    2.2.4.2 Relationship between genotype and niche model predictions in the hybrid zone . 26   2.3 Results ................................................................................................................................... 27  2.3.1 Ecological niche models ............................................................................................. 27  2.3.2 Niche and range limits comparison ............................................................................ 29  2.3.3 Niche identity test ....................................................................................................... 29  2.3.4 Niche background tests............................................................................................... 30  2.3.5 Hybrid index ............................................................................................................... 32  2.3.7 Microhabitat structure of parental species adjacent to the hybrid zone ..................... 33  2.3.8 The relationship between microhabitat and hybrid index in the hybrid zone ............ 34  2.3.9 Relationship between predicted suitability and hybrid index ..................................... 34   2.4 Discussion ............................................................................................................................. 35 Chapter 3: Conclusions ................................................................................................................. 58   3.1 General conclusions .............................................................................................................. 58  3.1.1 Slight divergence along a subset of niche axes influences avian speciation .............. 58  3.1.2 Slight niche divergence of species that meet at an ecotone contributes to hybrid zone narrowness ........................................................................................................................... 59  3.1.3 Using GIS to study hybrid zones and speciation ........................................................ 59   3.2 Future directions ................................................................................................................... 60  3.2.1 Future work in the MacGillivray’s and Mourning warbler system ............................ 60  3.2.2 Future work on the role of ecology in avian speciation ............................................. 62 References ..................................................................................................................................... 63 Appendices .................................................................................................................................... 74   Appendix A:  Sample preparation and data collection ............................................................... 74   Appendix B: Raw data analysis .................................................................................................. 75       vi  List of tables Table 1. A summary of the tests for niche divergence between species used in this study. Niche identity and background tests compare the same test statistic, Schoener’s D which measures niche overlap between two groups, to different null expectations. The aim of these tests is to first determine if species niches are different using the identity test. There are two possible explanations if the niches are found to be different. The two species could seem different because they occur in different ranges with different available conditions, or they could have divergent niches. The background test addresses the question of differences in available environment. The McCormack background test compares average differences between species to average differences between available conditions in their range (i.e. environmental background) along different niche axes, which are generated by conducting a principal component analysis on all species presence data and 10 000 points sampled randomly from each species’ background.C= conservatism, D=divergence………………………………………………………………….….40  Table 2. The results from niche model cross-validation show that Mourning warbler models (MOWA 1-5) performed better than MacGillivray’s warblers (MGWA 1-5), but that all models performed adequately. Sensitivity (i.e. the fraction of correctly predicted true presences), specificity (i.e. the fraction of correctly predicted true absences), the true skill statistic (TSS; i.e. actual model accuracy compared to expected random accuracy), and the area-under the curve statistic (AUC; i.e. statistic with lower subjectivity) were generally greater than 0.70, which is considered to indicate acceptable model performance (Araújo et al. 2005). ……………………41  Table 3. The percent contribution of climatic and vegetation variables included in niche models. The two variables with the highest contribution for the MacGillivray’s (MGWA) models were temperature seasonality and annual precipitation. In contrast, the two variables with the highest contribution for the Mourning (MOWA) models were annual mean temperature and the mean temperature of the wettest quarter. EVI= enhanced vegetation index, a measure of green vegetation cover………………………………………………………………………………….42 Table 4. The multivariate McCormack et al. (2010) background test compared the observed differences in PC scores of the two species (|dO|) to the differences of points sampled randomly from the two backgrounds. The test showed that the warblers are diverged along PC1    vii  (temperature and precipitation), and PC2 (precipitation during warm periods and variation in precipitation). The test showed that the two species have conserved June vegetation cover (EVI) and temperature during the warmest periods. CI Null dist= confidence interval for the null distribution. Variable 1-5 indicates the variables with the five highest loadings for each PC axis. Temp=Temperature. Variables that have first or second percent contribution in either species’ niche models are indicated with a subscript 1 or 2……………………………………………..43 Table 5. The 14 variables with the highest loadings on the first axis (PC1) of the principal component analysis on microhabitat variables measured in MacGillivray’s and Mourning warbler breeding territories in the hybrid zone and populations adjacent to the hybrid zone. Dbh= diameter at breast height. ………………………………………………………………………..44       viii  List of figures Figure 1. .(A) Range maps for the MacGillivray’s warbler (solid line) and the Mourning warbler (dashed line; range maps from Ridgely et al. [2003] edited based on occurrence data collected for niche models) overlaid on a map of the ecoregions of North America (Commission for Environmental Cooperation 2009). The Mourning warbler range was edited to the ecoregions that occur where the two species’ ranges meet (i.e. the boreal plains, taiga plains, and temperate prairies). Both ranges were edited to include only regions within Canada. (B) A map showing the edited ranges used to delineate the study populations for ecological niche model analyses. The blue and red ranges show the edited ranges of the MacGillivray’s (MGWA) and Mourning warblers (MOWA), respectively. The points on the map indicate the locations where a MacGillivray’s (blue) or Mourning warbler (red). These locations were used as the presence data for the ecological niche models.………………………………………………............................45 Figure 2. Examples of partial plots for the two variables with the highest contribution in the MacGillivray’s warbler and the Mourning warbler niche models. (A) The predicted probability of a MacGillivray’s warbler (MGWA) presence given temperature seasonality (°Cx1000), the variable with the highest contribution in all 5 MacGillivray’s warbler niche models. (B) The predicted probability of a MacGillivray’s warbler (MGWA) presence given annual precipitation (mm), the variable with the second highest contribution in all 5 MacGillivray’s warbler niche models. (C) The predicted probability of a Mourning warbler (MOWA) presence given annual mean temperature (°Cx10), the variable with the highest contribution in all 5 Mourning warbler niche models. (D) The predicted probability of a Mourning warbler (MOWA) presence given the mean temperature of the wettest quarter (°Cx10), the variable with the second highest contribution in all 5 Mourning warbler niche models……………………………………….…..47 Figure 3. Comparisons of the mean suitability predictions from five ecological niche models to the range limits (grey outline) of (A) the MacGillivray’s warbler, and (B) the Mourning warbler, show that areas with high predicted suitability are generally restricted to the species’ own range……………………………………………………………………………………………48 Figure 4.Density grids showing the ecological niches of (A) the MacGillivray’s warbler, and (B) the Mourning warbler, on the same scale which represents the available ecological space in two    ix  dimensions. Dashed lines and solid lines indicate 50% and 100%, respectively, of the available ecological space in a species’ range. (C) Results from the identity test showing that the observed value of Schoener’s D ( red line) is significantly lower than the null distribution of Schoener’s D generated by randomizing species’ identity (D=0.054, p=0.0198)………………………...…….49 Figure 5. Comparison of the observed Schoener’s D value between the MacGillivray’s and Mourning warblers with a null distribution of Schoener’s D generated by measuring niche differences between (A) the Mourning warbler’s niche with the MacGillivray’s warbler background (p= 0.0198), and (B) the MacGillivray’s warbler’s niche with the Mourning warbler background (p=0.6139). The results from A and B suggest that the two species’ niches may be more similar to each other than the ecological values of their ranges…………………………...50 Figure 6. Principal component analysis of SNPs generated using allopatric birds shows that PC1 separates MacGillivray’s (MGWA) and Mourning warblers (MOWA) sampled either in allopatric regions far from the hybrid zone or in allopatric regions adjacent to the hybrid zone (HZ). It also shows that there are many birds with intermediate genotypes in the hybrid zone………………………………………………………………………………………………52 Figure 7. The geographic cline fit to PC1 SNP scores of birds sampled for microhabitat analyses across a west-east transect through the hybrid zone. The cline was centred at a longitude of -15.51km, had a slope of 2.283e-06, and a width of 92.08km which is similar to the width of the cline described by Irwin et al. (2009)…………………………………………………………....53 Figure 8. (A) Principal component analysis of microhabitat in populations of MacGillivray’s (MGWA) and Mourning warblers (MOWA) adjacent to the hybrid zone (HZ). The first PC axis separates the two populations well. The first axis also shows some difference between bird territories and locations where birds were absent within a geographic area (i.e. between MGWA and western absences, and between MOWA and eastern absences). (B) In the hybrid zone there was no relationship detected between bird genotype (standardized hybrid index score) and habitat PC1 score (R2= -0.0277, p=0.947)………………………………………………………54 Figure 9. In the hybrid zone locations without birds (i.e. absences) were discriminated well from bird territories along the third microhabitat axis. Three birds that were outliers are not shown.    x  MGWA= MacGillivray’s warbler; MOWA=Mourning warbler. Hybrid= individuals with hybrid index scores along a continuum of genotypes between the parental species…………………….56  Figure 10. In the hybrid zone there is no significant relationship between standardized suitability difference scores and standardized hybrid index scores (R2= 0.006, p= 0.649)………………....57         xi  Acknowledgements I would first like to thank my supervisor Darren Irwin for all of his help and advice throughout my Master’s work. I am very appreciative of your support, particularly during the more challenging times. Thanks also Dolph Schluter and Eric Taylor for sitting on my committee and for helpful comments and advice.  I am very appreciative of all the support, advice and help that was shared with me from my collaborator on the niche model component of my thesis, Julie Lee-Yaw. She went above any of my expectations in a collaboration and has taught me a lot. Thanks also to Christine Grossen, my collaborator on the genomic component of my thesis. I am very appreciative for the support, advice and help from Irwin lab members Kira Delmore, Haley Kenyon, Dave Toews, Miguel Alcaide, Michelle Chen, and Silu Wang. I have been very lucky to have worked in a lab with such knowledgeable and supportive people. I also had a hard-working field assistant, Kate Broadley, who I would like to thank for helping me out even though we started work at a very early sunrise and cooked dinner under storms almost every day. Kim Gilbert and Amanda Moreira thanks for your advice and help, it was great to have such great office neighbours. I’m very grateful to my friends and family who have been supportive over the course of my Master’s work. I am very lucky to have had support from all of you, and it has helped me very much over the course of this work.  I received funding support to complete this work from the National Sciences and Engineering Research Council of Canada, Werner and Hildegard Hesse Fund for Ornithology, and the Beaty Biodiversity Research Centre.    1  Chapter 1: Introduction 1.1 Speciation in birds A fundamental problem in biology is explaining the causes of biodiversity. The source of diversity in birds has been of particular interest for biologists presumably because this group exhibits an exceptional number of adaptive radiations. For instance, the radiation of neotropical migratory birds, species which breed in temperate regions and migrate to wintering grounds in the tropics,  is thought to have been largely influenced by the glacial cycles of the Pleistocene (Hewitt 2004, Weir and Schluter 2004). Evidence for the role of the glaciations comes from the construction of molecular clocks that date the time of divergence of sister species to the Pleistocene epoch (Avise et al. 1998, Weir and Schluter 2004). Glacial cycles are thought to have facilitated speciation, the evolution of reproductive isolation, in birds by fragmenting the distributions of ancestral populations into allopatry. Allopatric speciation is considered to be the prevailing geographic mode of speciation in birds (Edwards et al. 2005). Allopatry could have been reached in one of two ways. Vicariance may have caused allopatry whereby glacial ice sheets fragmented an ancestral population (Edwards et al. 2005). Otherwise allopatry may have been reached through dispersal, where groups of an ancestral population followed the northward retreat of glaciers along different routes that lead them to different geographic regions (Edwards et al. 2005, Irwin et al. 2001). Once allopatry was reached by populations of the ancestral species genetic, morphological and behavioural differences may have accumulated through ecological and sexual selection or genetic drift. The populations may have adapted to new environments and evolved different ecological niches, which may have contributed to species divergence. Thus, there is a prevailing idea that divergent ecological selection had a large influence on avian speciation (Edwards et al. 2005) 1.2 The role of ecology in avian speciation Reproductive isolation may arise through divergent selection caused by taxa being exposed to different environmental conditions (Schluter 2001). Divergent ecological selection can have a role in the evolution of reproductive isolation by causing premating or postmating barriers.    2  Premating barriers may result from ecological divergence if adaptations to different ecological space reduce interspecific interactions (Coyne and Orr 2004).  A reduction in interspecific interactions via ecological divergence may be in the form of sister species adapted to conditions in spatially or temporally distinct ranges, or at a smaller scale being adapted to occupying different microhabitats within overlapping ranges. For example, the Baltimore (Icterus galbula) and Bullock’s orioles (I. bullockii) breed in distinct ranges possibly as a result of adaptation to different temperatures as the two species have been shown to exhibit different temperature tolerance (Swenson 2006). The adaptations to different temperatures in the orioles appear to have an important role in limiting sympatry and thus promoting isolation between the two species of orioles (Swenson 2006).  When ranges do overlap, premating barriers formed by divergent ecological selection may still promote reproductive isolation by reducing interactions between species. Breeding habitats of birds may select for divergent adaptations in traits such as those involved with foraging or nesting that are specific to the vegetation characteristics of a particular breeding microhabitat. These divergent adaptations to habitat may evolve before secondary contact occurs and be maintained afterwards, or divergence may evolve following secondary contact (Johannesson 2001). A classic example of adaptations to different foraging habitat is a former wood-warbler genus, Dendroica, in the northeastern spruce-fir forests of North America, where closely related species use different parts of the tree to forage (MacArthur 1958). Habitat partitioning between closely related bird species was also shown between the Bridled (Baeolophus wollweberi) and Juniper titmouse (B. ridwayi). The ranges of the titmice overlap, however they were found to have breeding habitat with different extents of vegetation cover (Weathers and Greene 1998). Thus, the chance of the species coming into direct contact with one another in sympatry is reduced. Habitat partitioning may have also produced further divergence as the two species of titmice were found to have different thermal tolerances that were correlated with the extent of vegetation cover in their habitat (Weathers and Greene 1998). Physiological and habitat or climatic correlations have been found to explain the broad distributions and local habitat partitioning between many other closely related species of birds (e.g. Cooper 2002; Root 1988; Hayworth and Weathers 1984).    3  Ecological selection may also lead to premating barriers in sympatry if mating occurs assortatively by microhabitat selection (Coyne and Orr 2004; Rice 1984). Selection for assortative mating by microhabitat may have evolved either before or after secondary contact occurs. Before pairing, female birds have been observed to visit the territories of multiple males and select the male occupying the highest quality habitat (Bensch and Hasselquist 1992). If this is a common behaviour across species and species-specific territory characteristics are preferred by females, it is possible that mate choice occurs assortatively based on the characteristics in a male’s territory (Adamík and Bureš 2007). Assortative mate choice based on the microhabitat in breeding territories may be important so that the appropriate species and abundance of vegetation is available to build nests, supply resources to raise a nest successfully, and for protection from predators (Sætre and Sæther 2010). Given that habitat quality is thought to be one of the most important predictors of reproductive success in birds (Johnson 2007), if habitat quality is species specific we would expect the selective pressures for assortative mating based on territory traits to be strong.   Adaptations to climate or other ecological conditions have the potential to not only form premating barriers, but also postmating barriers. Accordingly, ecological selection may act against phenotypes of hybrid offspring, which might be intermediate to or extreme compared to parentals, and promote isolation between species (Rieseberg et al. 1999; Coyne and Orr 2004). Intermediate or extreme traits for breeding habitat selection and for physiological or phenological adaptations may be unfavourable in birds and could cause reduced fitness of hybrids. If hybrids obtain breeding territories with characteristics that differ from the parental species, the hybrids may be at a disadvantage for acquiring food for survival and for raising offspring, as well as other habitat-dependant requirements including nest material and location and predator protection (Sætre and Sæther 2010). Furthermore, if mating occurs non-randomly based on microhabitat structure hybrids occupying an intermediate habitat may be at a disadvantage at acquiring a mate (Sætre and Sæther 2010).  An additional role that ecological divergence could play in promoting reproduction isolation between species does not involve adverse consequences for hybrids. Rather, hybrids of ecologically divergent species may instead be favoured in a sympatric zone, particularly at ecotones where environmental conditions are intermediate to those in allopatric areas of the    4  parental species ranges (e.g. Good et al. 2000; Moore and Koenig 1986; Delmore et al. 2013). In these hybrid zones bounded hybrid superiority occurs, where intermediate ecological traits may be best suited for intermediate environments, and therefore hybrids will have a higher fitness than either parental species in the hybrid zone (Moore 1977). Because hybrids are only favoured in a narrow region, ecological selection can be said to help maintain species’ borders by restricting the spread of hybrid genotypes beyond the contact zone.  Hybrid genotypes may not only be favoured by the environment if they are intermediate but also if they have phenotypes that are extreme (i.e. transgressive) to parental species. There is considerable evidence that in a number of cases hybrids exhibit extreme phenotypes compared to parental species (Rieseberg et al. 1999). In some instances these extreme phenotypes have been shown to be best suited to the environmental conditions in a hybrid zone and facilitate ecological adaptation (Rieseberg et al. 1999). Ecological selection may therefore favour hybridization in these cases and restrict the ranges of both the parentals and the hybrids forming isolating barriers. Divergence is not always required for ecological traits to promote speciation, as evidence suggests that the absence of ecological niche divergence may be a common contributor to the process of evolving reproductive isolation. Speciation without niche divergence can occur if a vicariance event results in a zone that is ecologically unsuitable and restricts gene flow between the divided ancestral species (Kozak and Wiens 2006). The newly divided populations would then evolve independently but not necessarily along ecologically divergent axes (Kozak and Wiens 2006). Thus, ecological barriers between two ecologically similar species are important for initiating allopatry and speciation in general given that allopatric speciation is the most frequent form of speciation (Wiens 2004; Wiens and Graham 2005; Wiens et al. 2010).  There is mounting evidence supporting the importance of niche conservatism in allopatric speciation.  Niche conservatism seems to be widespread including species thought to have diverged during the Pleistocene glaciations, challenging the prevailing idea that ecological divergence is largely responsible for the radiation of North American bird species (Hawkins et al. 2006; Lovette and Hochachka 2006). Many studies testing for patterns of niche divergence versus conservatism have shown that closely related species often have conserved ecological    5  niches (e.g. Peterson et al. 1999; Peterson and Nyári 2008; Kozak and Wiens 2006; Martínez-Meyer and Peterson 2006; Cadena et al. 2012) or that niches remain largely conserved with small differences along some niche axes (e.g. Peterson and Holt 2003; McCormack et al. 2010; Wellenreuther et al. 2012). Given that niche conservatism may be important to initiate allopatry, closely related species might frequently exhibit divergence only on a subset of niche axes, depending on the time since divergent selection on the niche began and the strength of selection (Wiens 2004). The argument that niche divergence is not ubiquitous in speciation was strengthened by a review by Peterson (2011) which showed that the majority of studies on niche evolution detected little evidence for niche divergence in cases where divergence had likely occurred only on short to moderate timescales (i.e. short timescale: up to 10 000 years; moderate timescale: 10 000-100 000 years; Peterson [2011]).  1.3 Using hybrid zones to study speciation To better understand the role of ecological diversification in promoting the evolution of reproductive barriers in birds, regions where species come into contact and interbreed can be studied. Hybrid zones are useful for studying speciation because they offer the opportunity to detect factors that maintain species differences (Barton and Hewitt 1989). During speciation taxa accumulate differences, some of which promote reproductive isolation by acting as a post- or pre-mating barrier. Occasionally these barriers are not complete and species ranges overlap with hybridization. In these instances we can study the species under the same environment as well as the outcomes of new genetic combinations (Barton and Hewitt 1989). One outcome of hybridization is the formation of a stable hybrid zone, where the stability is caused by some form of reproductive isolation acting to reduce gene exchange between the parental species (Barton and Hewitt 1985).  In some cases referred to as tension zones, reproductive isolation may be caused by endogenous selection against hybrids (Barton and Hewitt 1989). In other cases exogenous selection may be acting to maintain isolation between taxa by favouring certain genotypes at different sites along an environmental gradient (Moore and Price 1993). This second hybrid zone model is known as the geographic-selection gradient model (Moore and Price 1993). Alternatively, in a bounded-superiority model hybrids have higher fitness than parentals in the hybrid zone (Moore 1977).     6  Determining the reproductive barriers (i.e. those selecting against hybrids, or those selecting against parental species) can provide insight on traits that might have been important during speciation, such as those that might have promoted ecological diversification in birds. Hybridization is relatively common between closely related species of birds that diverged during the Pleistocene (Weir and Schluter 2004). With the advent of new technology and methods in geographic information systems (GIS) used over the past 15 years there have been some steps towards using hybrid zones to understand the role of ecological diversification (e.g. Cicero 2004; Swenson 2006; McCormack et al. 2010). Much of this technology however goes underused in studies of hybrid zones of birds to understand the role of ecological diversification in speciation.  1.4 Studying ecological divergence from multiple perspectives Advances in spatial data analysis represent new opportunities to explore the importance of ecology, and in particular niche divergence, for speciation. The understanding of a species’ ecology and whether it has influenced speciation has the opportunity to expand with the application of ecological niche models, which use large environmental and species datasets that are configured for use in GIS (Swenson 2008). The GIS-based niche models estimate the realized niche of a species or a focal population given the ecological conditions measured at locations where the focal species has been detected (Swenson 2008; Peterson 2001). Niche models use the environmental data at the locality points and contrast them with the environmental data at locations where the species is known to be absent to generate models that explain the probability of a species occurring at a location with a given set of environmental data. There are different modelling algorithms that can be used to generate these rules which include generalized linear models, generalized additive models, and machine-learning algorithms like MaxEnt (Phillips et al. 2006). The probabilities generated by the niche models are referred to as the “suitability” of a set of conditions for the species. Locations with high predicted suitability are those with conditions that are better suited to a species’ niche.  Ecological niche models have made advances in studying the role of ecology in speciation. Niche models can be generated for closely related species and compared to test for niche conservatism or divergence (Warren et al. 2008; McCormack et al. 2010). Analyses of niche models that investigate ecological divergence and conservatism are advantageous because they    7  provide an alternative to experiments that test niche limits or differences between species (e.g. common garden experiments or transplant experiments) which might not be feasible with certain species. Another advantage of these methods is that they utilize point locality data and environmental data from large geographic databases which have become increasingly more broadly available. Databases of presence points with geographic coordinates can sometimes include thousands of locality data for a given species, and map layers of ecological variables provide easy access to the ecological conditions at the locations the species was observed. Thus, there is existing data available that can be used in niche models to test for niche divergence or conservatism. This makes testing for niche conservatism or divergence in species of birds with broad geographic ranges possible.   The use of ecological niche models has been used in a number of studies investigating the role of ecology in the diversification of many types of taxa, particularly birds. Many of these studies analyse the differences in the suitability predicted by niche models between closely related species that occupy allopatric ranges. A key development for these analyses that was first proposed by Warren et al. (2008) and later modified by Broennimann et al. (2012) included analyses that not only quantified and tested differences between the niche models, but also tested whether those differences could be explained solely by the background ecological differences that may exist between the two species’ ranges.  Overall, studies that have conducted these two types of comparisons for niche divergence or conservatism on avian taxa found that the species being compared often experienced different ecological conditions, but have conserved many aspects of the niche (e.g. Zink et al. 2014; McCormack et al. 2010; Anciães and Peterson 2009; Peterson 2011). This pattern is also common in taxa other than birds (e.g. Wellenreuther et al. 2012; Pyron and Burbrink 2009). These results challenge the idea that ecological divergence is important in speciation, particularly for birds. However, challenges remain with interpreting these tests for niche divergence and conservatism for allopatric taxa.  Given that one of the main problems in distinguishing between niche conservatism and divergence using ecological niche models for species that are mainly allopatric is that observed differences may result from different environments available in their background, analysing the niches of species when they come into contact may help better understand how the niche might influence isolation. In sympatry both species experience the same suite of available    8  environmental conditions, thus any differences in niches could not be attributed to different background environments. Moreover, if hybrids are formed we can explore whether genotype is associated with niche model predictions about the relative suitability of sites for either species. If an association does exist, this could mean that the environment is imposing divergent selection on different genotypes, which might explain the width and position of the hybrid zone (Barton and Hewitt 1985). If no relationship is detected, the hybrid zone may be located at an ecotone because of low population density which tends to attract a tension zone, or the hybrid zone may be maintained by another source of selection (Barton and Hewitt 1985). Chatfield et al. (2004) analysed the correlation between hybrid index and a suitability score standardized to incorporate the outputs of suitability predictions from the niche models of two salamander species, and found no association. De La Torre et al. (2014) conducted an analysis to detect alleles under selection that are associated with environmental variables in a hybrid zone between two spruce species. Their analysis revealed a small number of alleles were detected to be under divergent selection and associated with environmental variables that were shown to be important in niche models (De La Torre et al. 2014). A limited number of other studies have detected evidence that suggests selection by the environment influences hybrid zone dynamics (e.g. Tarroso et al. 2014; Culumber et al. 2012). These are some of the apparently few examples that investigate the association between ecological niche model predictions and a hybrid index in sympatry, despite there being many studies published that used niche models to test for niche divergence in species that come into contact and hybridize (e.g. Wellenreuther et al. 2012; Cicero 2004; Martínez-Freiría et al. 2008; Swenson 2006).  Using the new statistical tests for niche divergence and conservatism together with analyses of relationships between niche characteristics and genotypes in contact zones is a strong approach to understand the role of the niche as a reproductive barrier. An approach that examines the role of the niche in maintaining isolation between species at these two scales could reveal whether the niches of allopatric taxa are divergent or conserved, and how this might explain range limits. The analysis at the hybrid zone scale could add to the results of how the niche may be a reproductive barrier from the large-scaled approach in a number of ways. For example, if two species were found to have divergent niches, analyses in the hybrid zone could show whether these differences are indeed maintained in the contact zone and might act as a reproductive barrier. If these differences were not maintained in the hybrid zone then the interpretation of the role of    9  niche divergence as an isolating barrier would weaken. Another strength of including analyses in the hybrid zone is that including genetic information could provide information about hybrid genotypes. Integrating genetic and ecological data could reveal whether hybrids might be at a disadvantage compared to parentals based on ecological traits, and could therefore suggest if these traits have a role as a reproductive barrier. However, few studies integrate analyses of niche differences in both allopatric and sympatric areas of species’ ranges, and there are scarce examples in the literature that examine relationships between niche characteristics and the genetic history of individuals in the hybrid zone. The link between tests for niche divergence or niche conservatism and how niche characteristics might act in sympatry to maintain isolation between species is therefore unknown.  1.5 Research goals and questions Given that there are only a small number of studies that have addressed the role of niche divergence or conservatism in the radiation of North American bird species, and fewer that use hybrid zones to better understand this, we lack empirical support for the idea that many of these species diversified in the Pleistocene through adaptations to different environmental conditions. Many of North American species come into contact with their closest relative and hybridize (Rohwer and Irwin 2011), and thus provide the opportunity to investigate broad-scale patterns of niche divergence or conservatism in allopatry and smaller-scale roles of the environment in determining species limits in sympatry and under the same available conditions.  The goal of this study was to detect if environmental factors have a role in maintaining isolation between two species of birds, the MacGillivray’s (Geothlypis tolmiei) and Mourning warblers (G. philadelphia), which come into contact and hybridize in northeastern British Columbia (Irwin et al. 2009). I built ecological niche models for both species to address several questions regarding the role of the niche as a reproductive barrier and a potential source of diversification. The niche models were compared using current techniques to detect if populations of the species that are likely involved in dynamics at coinciding range edges show divergence along niche axes of climatic and vegetation cover variables, and if their niches might reduce range overlap. I also undertook a smaller-scale approach to understanding the potential role of selection acting on breeding habitat use in maintaining isolation at the hybrid zone. For these analyses I genotyped birds across a transect    10  through the hybrid zone and measured microhabitat traits of their breeding territories to determine if genotype is associated with microhabitat structure, or with suitability predicted by the niche models.       11  Chapter 2: Does ecological divergence explain the position and narrowness of the MacGillivray’s/Mourning warbler (Geothlypis tolmiei/philadelphia) hybrid zone? 2. 1 Introduction 2.1.1 Ecological divergence as a reproductive barrier Divergent selection on ecological traits can produce reproductive barriers in a number of ways. Pre-mating reproductive barriers can result from divergent selection if they reduce interactions between taxa because they cannot occupy the same spatial or temporal environmental space (Coyne and Orr 2004). For example, species may not have the opportunity to interact because they are adapted to the environmental conditions in different geographic regions, resulting in no range overlap (Swenson 2006). Alternatively, sexual selection may favour assortative mating based on ecological traits (Coyne and Orr 2004; Rice 1984). Selection for assortative mating based on microhabitat selection in birds could occur based on differences in the types of vegetation that allow each species to build and raise a nest successfully (Sætre and Sæther 2010). Selection for divergent phenotypic or behavioural traits between species may also cause post-mating barriers. Hybrids may have ecological traits that are intermediate or different from either parental species, which may not be favoured by the environment (Schluter 2001). Alternatively, intermediate or different ecological traits of hybrids may be favoured over parental traits in a zone of secondary contact (Moore 1977). In the case of hybrids having superior fitness compared to parental species, a barrier is formed by ecological selection because hybrids are restricted to the zone of environmental conditions that favours hybrid genotypes (Moore 1977). Sister taxa of birds that diverged during the Pleistocene glaciations are thought to have evolved reproductive isolation in part by adapting to different ecological conditions. A primary explanation for this is that these sister taxa often occupy ranges with different environmental conditions. For example, a pattern common in North America for sister taxa originating in the Pleistocene is that one species occupies western North America, while the other species occupies the eastern side, with the range edges meeting along the Rocky Mountains (Rohwer and Irwin    12  2011). However, an association between species ranges and environmental conditions may not accurately reflect niche divergence (Warren et al. 2008; McCormack et al. 2010). This is because a species may not be found at all locations within its overall range, and there is likely variation in the environmental conditions within the range. Thus, a species’ ecological niche may only encompass a portion of the conditions available in the broader geographic region that it occupies.  To understand the role of the niche in promoting divergence between species, we can study the characteristics that explain where species occur within their ranges at various scales. At a broad geographic scale (e.g. across a large portion of a species range) ecological niche models can be used to predict the probability of a species being present at a given location either within its range or within the range of a close congener (Peterson 2001). Model predictions for different areas can be compared for both species to determine whether their predicted niches suggest niche divergence or niche conservatism (Warren et al. 2008). Niche divergence can only be declared when the difference between species niches is greater than the difference between the average environmental conditions of their ranges (Warren et al. 2008). If two species have predicted niches that are more similar to each other than the average environmental conditions in their ranges are, this would suggest that their niches have been conserved during speciation.  A second approach for understanding the role of the niche in maintaining isolation between species is to examine the associations between environmental characteristics and species occurrences when two species come into contact. Studying this association in zones of secondary contact eliminates the potential confounding role of differences in available environment when comparing niches as both species experience the same suite of available environmental conditions (Bronson et al. 2003). Furthermore, if hybrids are formed the characteristics of territories occupied by hybrids may reveal whether hybrid genotypes are likely to be at a disadvantage compared to parental species. Thus, areas where reproductive barriers break down can be informative for finding sources of reproductive isolation (Hewitt 1988; Swenson 2008). In this study I sought to determine whether two closely-related bird species, the MacGillivray’s (Geothlypis tolmiei) and Mourning warblers (G. philadelphia), have divergent ecological niches, and whether this may promote isolation by studying microhabitat structure and relative breeding territory suitability in a hybrid zone.    13  2.1.2 The MacGillivray’s and Mourning warbler hybrid zone The MacGillivray’s warbler and the Mourning warbler are one of many west-east pairs of related bird taxa in North America (Rohwer and Irwin 2011). The MacGillivray’s warbler breeds on the western side of the Rocky Mountains from northern British Columbia to the southwestern United States (Pitocchelli 1995, 2011). The Mourning warbler’s breeding range extends from eastern Canada to just east of the Rocky Mountains in British Columbia (Pitocchelli 1995, 2011). The northernmost part of the Mourning warbler range reaches northeastern British Columbia, while the range extends as far south as Michigan (Pitocchelli 1995, 2011). The two species can be distinguished by the presence (MacGillivray’s warbler) or absence (Mourning warbler) of a white eye-arc, wing-tail length measures, and by two genetic markers (Irwin et al. 2009). The MacGillivray’s warbler and the Mourning warbler come into contact and hybridize in an area approximately 78 km east of the Pine Pass, which is a mountain pass through the crest of the northern Rocky Mountains’ Hart Ranges, in northeastern British Columbia (Fig. 1a; Irwin et al. 2009). An analysis of variation in genetic markers and morphometric traits of birds along a west-east transect passing through the hybrid zone showed a narrow transition from MacGillivray’s warbler-like traits to Mourning warbler-like traits (Irwin et al. 2009). According to hybrid zone theory, the narrow width suggests that some form of selection is maintaining isolation between the species (Hewitt 1988); however the isolating mechanisms remain unknown. One common premating barrier in birds is song, which is used by males as a signal to attract mates (Irwin 2000). Divergent song between species that come into contact in a hybrid zone may therefore reduce interbreeding and contribute to the maintenance of a narrow hybrid zone. Kenyon et al. (2011) analyzed song variation across the MacGillivray’s and Mourning warbler hybrid zone and in allopatric populations to determine whether divergent song may be promoting isolation in the hybrid zone. Song, however, was convergent between the species in the hybrid zone and not strongly correlated with genetic background, revealing no support for song as a pre-mating barrier in this system.  An alternative explanation for the maintenance of the narrow MacGillivray’s and Mourning warbler hybrid zone could involve divergent ecological adaptations of the parental species. The MacGillivray’s and Mourning warblers are one of many avian sister species that diverged during    14  the Pleistocene (Weir and Schluter 2004). Pleistocene ice sheets are thought to have fragmented ancestral bird species into western and eastern refugia, leading to the divergence of the isolated populations through adaptation to differential environmental conditions of the separate refugia (Weir and Schluter 2004). If separation of ancestral populations into western and eastern refugia promoted ecological divergence of west-east sister taxa, the current distributions of the MacGillivray’s and Mourning warblers may limit secondary contact to a small overlap at a narrow ecotone. Ecological divergence may also promote isolation between the two species within the hybrid zone if the species occupy breeding territories with different vegetation. Selection for species-specific breeding habitat may be important for assortative mating, or for raising a successful nest. Thus, the two species may be isolated by divergence in geographically broad-scaled ecological traits that define their ranges, but also fine-scaled ecological traits that explain the specific conditions where they breed within their ranges. 2.1.3 Ecological traits of the MacGillivray’s and Mourning warblers Divergent ecological adaptations can produce stable hybrid zones when the distributions of parental species meet at an ecotone (Moore 1977; Barton and Hewitt 1985). The MacGillivray’s and Mourning warbler hybrid zone is located at an ecotone between the montane cordillera ecozone to the west and the boreal plains ecozone in the east. Allopatric populations of the MacGillivray’s warblers west of the hybrid zone breed in a cool and moist climate due to the warm Pacific air rising over the Rocky Mountains and foothills (Demarchi 2011). Towards the east where allopatric Mourning warbler populations breed the landscape flattens and the climate becomes drier as Pacific air descends from the Rocky Mountains (Demarchi 2011). Therefore, it appears that the two species occupy ranges with distinct ecological characteristics.  Surveys of microhabitat structure in allopatric breeding territories of MacGillivray’s and Mourning warblers provide suggestions that the species occupy similar types of habitat; however there are notable differences. The breeding habitat of both species tends to be in new growth or riparian areas with a dense understory of shrubs, grasses and few trees (Pitocchelli 2011, 2013). Both species have been described as habitat specialists (Finch 1989; Sodhi and Paszkowski 1995). The species makeup of the vegetation in the territories varies between species, with MacGillivray’s warbler typically occupying habitat with more coniferous plants compared to the    15  Mourning warbler (Pitocchelli 2011, 2013). This may reflect the difference in climate between the two ranges as the MacGillivray’s warbler is found in more moist and higher elevation areas, which are regions where the tree community is largely made up of conifers. Although the microhabitat of the MacGillivray’s warbler is well studied (Pitocchelli 2013) and there have been a few studies that measured Mourning warbler habitat, these studies were done in allopatry. Less is known about the microhabitat of breeding territories in the hybrid zone and specifically whether the associations between microhabitat and the two species are maintained in sympatry. At broader scales it is unknown whether the associations of each species’ range with different climatic zones accurately represents niche divergence or whether the species occupy similar climate conditions within these broader zones. Without proper analysis we may fail to detect the environmental characteristics that explain where the species occur, and how adaptations to the environment might play a role as a reproductive barrier.  2.1.4 Research questions and predictions To detect if divergent ecological niches have a role in explaining the location and maintenance of the narrow MacGillivray’s and Mourning warbler hybrid zone, I developed ecological niche models to test for divergence on a subset of niche axes of climatic and vegetation cover variables, and measured microhabitat structure along a transect through the hybrid zone. I compared the two species’ niche models using newly developed methods to test whether they exhibited niche divergence or not along the subset of niche axes. Evidence for niche divergence would be detected if the niches of the two species were not identical, and if the differences between the two species were greater than the differences between the conditions available in their respective ranges (Warren et al. 2008). Additionally, evidence for climatic and vegetation cover variables acting as a barrier between the two species would be provided by ecological niche models if they predict a decline in suitability outside the species’ known distribution (Swenson 2008). I also asked whether niche characteristics detected in allopatry were maintained in sympatry, as this could provide evidence for niche divergence promoting isolation. My final objective was to determine whether microhabitat components of territories occupied during the breeding season differed between parental species both in allopatry and sympatry, and if microhabitat use was related to genotype in the hybrid zone. If microhabitat preferences have a    16  role in promoting isolation in the MacGillivray’s and Mourning warbler hybrid zone, I predicted that the two species segregate into distinct microhabitats within the hybrid zone, with the species maintaining differences that have been observed in allopatry. For instance, I would predict that MacGillivray’s warblers occupy territories with a higher percentage of coniferous plants than Mourning warblers in sympatry as this has been noted in allopatry (Pitocchelli 2005, 2011). The absence of evidence for ecological divergence from both ecological niche modelling and microhabitat structure analyses could suggest that some form of selection other than ecological selection has a more important role in maintaining the narrowness and position of the MacGillivray’s and Mourning warbler hybrid zone. In this study I refer to the two parental forms and their hybrids but note that there is a full continuum of hybrid genotypes as my quantitative analyses take into account. 2.2 Methods 2.2.1 Ecological niche modelling 2.2.1.1 Defining the study extent To examine whether ecological factors restrict the distributions of the MacGillivray’s and Mourning warblers to a narrow zone of overlap, I first developed ecological niche models. To construct the models I used MaxEnt, which is a machine learning algorithm that predicts the suitability of environmental conditions for a focal species (Phillips et al. 2006). MaxEnt models are trained using environmental data from locations where a species is present on the landscape, with the goal of distinguishing these conditions either from background conditions or those found where the species is known to be absent. The resulting models can be used to predict the suitability of a study area given the environmental conditions. Ecological niche models can be constructed to model the entire range of a species, or for populations of interest within a smaller portion of the species’ total range (i.e. the study extent). I restricted the populations of interest to those breeding in the part of the MacGillivray’s warbler range within Canada, and the part of the Mourning warbler range within the ecoregions that meet the eastern border of the MacGillivray’s warbler range (Fig. 1a). These ecoregions are: the taiga    17  plains, boreal plains and temperate prairies. I chose to delineate the study populations using these methods for the Mourning warbler to focus on modelling the populations that reach the western edge of the range where it meets the MacGillivray’s warbler’s range. This also eliminates the bias from far-eastern Mourning warbler populations where ecological factors are different yet unlikely to have a role in determining the range’s western limits. Similarly, I restricted the study populations to Canada to avoid bias from populations of both species that are not likely to have a role in barriers between the two species as they do not come into contact in the United States. I downloaded range shapefiles for both species from NatureServe.com (Ridgely et al. 2003) and edited them to the study extent using ArcGIS 10.1 (ESRI Inc. 2012, Redlands, CA, USA; Fig. 1b). I first edited the eastern and western edges of the MacGillivray’s and Mourning warblers ranges, respectively, to reflect point count and breeding evidence data collected from the British Columbia Breeding Bird Atlas, the Atlas of the Breeding Birds of Alberta, the Manitoba Breeding Bird Atlas, the North American Breeding Bird Survey (BBS), and Irwin et al. (2009).  The edited Mourning warbler range shapefile was then cropped to a shapefile consisting of the ecoregions of interest, and the ranges of both species were cropped to the Canadian border. All data used to calibrate and test the ecological niche models were restricted to this study extent. 2.2.1.2 Preparing the training dataset I trained the models using environmental data at locations where the MacGillivray’s and Mourning warblers were observed during point counts for the British Columbia Breeding Bird Atlas, the Manitoba Breeding Bird Atlas, and the BBS (Fig. 1b.). These surveys conduct point counts at designated locations along routes, which in the case of the provincial atlases are visited only once throughout the atlas’ duration (BC atlas 2008-2012, Manitoba Atlas 2010-2013). The point count locations surveyed by the BBS are intended to be visited annually, however the duration and years that certain locations are visited varies based on participation of volunteer surveyors. I used a BBS point count location in the models only once, regardless of how many times the focal species was recorded at that location. To restrict the period from which the BBS point counts were collected to when technology for accurately determining the geographic coordinates of a point count was more available, I used data collected by the BBS from 2000 through to 2013. I calibrated the models using absences from these point count datasets. I    18  defined absences as the point count locations where the species being modelled was never recorded. A total of 5823 locations for MacGillivray’s warblers were used to generate models, with 34543 locations where the MacGillivray’s warbler was absent across the species’ study area. I used 1227 locations where the Mourning warbler was present and 28870 locations where the Mourning warbler was absent across its study area to generate the Mourning warbler niche model dataset. The environmental data were downloaded as map layers and edited using ArcGIS for use in ecological nice models. I retrieved map layers of 19 bioclimatic variables from WorldClim.org (Hijmans et al. 2005) at a 1km resolution and projected them to a Canada Albers Equal Area Conic projection. The climatic data are averages calculated for 1950 through 2000. I downloaded the NASA MODIS Terra (available at https://lpdaac.usgs.gov/data_access) monthly vegetation index product at 1km resolution using the ModisDownload function in R (http://r-gis.net/?q=ModisDownload) for May and June between 2000 and 2013. The monthly vegetation index is enhanced vegetation index (EVI), a measure of green vegetation determined from satellite images. EVI for the month of May and June may be relevant to vegetative components of the niches because it is when most birds have arrived to the breeding grounds and started breeding activities across the study extent. Using the Modis Reprojection Tool (MRT; Dwyer and Schmidt 2006) I then arranged the EVI tiles into a mosaic and projected the resulting layer to the same projection, extent, and resolution as the climatic data with a bilinear resampling method. I averaged the resulting 14 rasters producing one raster layer of average May EVI values, and another for mean June EVI.  Given that bird locality data and environmental data were collected during different time periods, I assumed that the climate had not changed between the times of data collection. The bioclimatic and EVI variables may be associated with the species’ breeding phenology, physiological limits, and the resources required for survival. I did not restrict the climatic variables to those measuring conditions during the breeding season because conditions during the non-breeding season in the breeding range are known to influence the phenology of migrating birds during migration and the breeding season (Forchhammer et al. 2002; Hüppop and Hüppop 2003; Marra et al. 2005; Hurlbert and Liang 2012). To reduce model overfitting which is a problem when MaxEnt models are complex and have correlated predictors (e.g. Elith et al. 2006), I identified highly correlated variables (r>0.7) and    19  removed 11 variables. Highly correlated variables were identified by uniformly sampling at 10% of the raster cells across the study area (i.e. both species’ ranges). I kept 10 variables: mean June EVI, annual mean temperature, mean diurnal range, temperature seasonality, temperature annual range, mean temperature of wettest quarter, mean temperature of warmest quarter, annual precipitation, precipitation seasonality, and precipitation of warmest quarter. I extracted the environmental data from the raster layers at the coordinate position of all presence and absence localities for both species to use as model input. 2.2.1.3 Model generation and cross-validation With the reduced training datasets I randomly assigned the presences and absences into five groups to build the models using a cross-validation approach. For each species I built five different niche models using all training data but withholding data from one group. Each of the five models was therefore built with a different set of training data withheld. The models were generated with environmental variables as the predictors and presence/absence as the response. To further reduce model overfitting, I set MaxEnt parameters to use only hinge features.  Hinge features model variable relationships that are more complex than those modeled by linear, quadratic, or product relationships but do not increase model complexity and therefore avoid overfitting the model to the data (Phillips and Dudík 2008). The withheld data was used as a testing dataset for that particular model by measuring the ability of the model to predict a true presence or absence with validation statistics. For each step the sensitivity (i.e. fraction of correctly predicted true presences), specificity (i.e. fraction of correctly predicted true absences), the true skill statistic (TSS; i.e. actual model accuracy compared to expected random accuracy), and the area-under the curve statistic (AUC; i.e. statistic with lower subjectivity) were calculated to measure the performance of the model (Allouche et al. 2006). I chose to use multiple validation statistics because they each have different strengths and weaknesses (e.g. Lobo et al. 2008; Allouche et al. 2006). All five models for both species were projected across the study area to generate a map of predicted suitability. I took the mean suitability across the five model’s predictions for all cells in the study area as the final predicted suitability for either the MacGillivray’s or Mourning    20  warbler. Maps showing predicted suitability across the study area were used to compare areas of high predicted suitability to the range limits of the two species.  2.2.1.4 Testing for niche divergence The use of spatial data to indirectly evaluate the extent of niche divergence between species requires some care if species are found in different geographic regions because model predictions of species presence may be autocorrelated with the conditions within the species’ range (Warren et al. 2014). Several recent tests account for spatial autocorrelation in comparison of the environmental space occupied by sister species have been proposed in the literature. I compared inferences from these approaches to evaluate the extent of niche divergence between the MacGillivray’s and Mourning warblers. These approaches test whether two species meet two criteria for niche divergence. Niche divergence between two groups can be confirmed if the characteristics of the niches are different, and that those differences greater than the average differences between the environments of the ranges in which the two species occur (i.e. their environmental backgrounds; Warren et al. 2008; McCormack et al. 2010). Alternatively, there would be evidence for niche conservatism if the predicted niches are found to have no significant differences, or if the differences between the two species’ background is greater than the differences between the species’ niches. Warren et al. (2008) developed a method to test for niche identity between two taxa where observed niche differences are compared to a null expectation (i.e. the niche identity test; Table 1). The null distribution is made by building niche models after dividing the pool of occurrence points for both focal taxa into two datasets and randomizing taxon identity, predicting suitability across the study area for each dataset, calculating Schoener’s  D, and repeating these steps 100 times. Schoener’s D compares two niche models and ranges from 0, which indicates no identity, to 1, which indicates that two niches are identical. The Schoener’s D calculated from the empirical niche models for the focal taxa is compared to the null distribution, and if the observed D lies beyond the 95% percentile niches are inferred to be significantly similar. If the observed D is less than the 5% percentile niches are more different than what is expected if species identity was random.    21  A test for comparing the differences in background conditions to the differences of species occurrences was also developed by Warren et al. (2008, 2010). Their background test (Table 1) builds two null distributions of Schoener’s D calculated from comparing the predicted suitability of niche models built using occurrence data from one focal taxa and background data from that focal taxa’s background, to niche models built using random background points from the other focal taxa’s study area as the occurrences and another set of random points to use as the background points, and repeating this process 100 times. Two null distributions are generated as the test is conducted for both species. The observed D is compared to the two null distributions, with the null expectation that the two species differ because they occur in backgrounds with different environmental conditions. Thus, under the null hypothesis the two species are inferred to be as different ecologically as the background. If the observed D falls outside the central 95% of either null distribution the null expectation is rejected, and divergence is supported if the observed value is less than the lower confidence limit of the null distribution.  I performed both the identity and background tests as described by Warren et al. (2008), but with modifications presented by Broennimann et al. (2012) to avoid certain biases in the original tests.  For instance, Warren et al. (2008) tests may underestimate the degree of overlap between two species niches if the species occur in ranges where environmental conditions are not equally available. Thus, Broennimann et al. (2012) modified the tests by controlling for the density of available environments and the density at which certain environmental conditions are occupied (Table 1). Broennimann et al. (2012) analysed the performance of four niche modelling approaches, as well as seven multivariate approaches for their modified test. Based on the results of this analysis I used the PCA-env multivariate approach, which had the highest performance overall. The PCA-env approach calibrates a Principal Component Analysis (PCA) on the environmental data from all background and occurrence points, and calculates D by comparing the scores from the first and second multivariate axis instead of the suitability scores generated by niche models as when using the Warren et al. (2008) approach. For the PCA-env method I used all 21 environmental variables since PCA can function with collinear variables. I used a second multivariate approach to test for niche divergence between the MacGillivray’s and Mourning warblers. The second multivariate test was proposed by McCormack et al. (2010) and compares the difference between the mean PC scores of the occurrence points from the two    22  focal groups to a distribution of differences between the mean PC scores of random points drawn from one focal group’s background and random points from the other group’s background (Table 1). To conduct this test, I trained a PCA on all MacGillivray’s and Mourning warbler occurrence points and 10 000 points selected at random from each of the two species’ allopatric backgrounds. Again I used all 21 environmental variables for this PCA. The null distribution of mean PCA differences between MacGillivray’s and Mourning warbler backgrounds was made by calculating the difference between the two means of 999 bootstrapped samples from the background points for the scores on the first four PC axes. I also calculated the difference between mean scores for both species’ occurrence points on the first four PC axes.  The observed values from the second multivariate approach were compared to the central 95% of the null distributions, which were calculated as the range between the 2.5 and 97.5 percentiles of the distribution. For the first 4 axes I compared the occurrence difference value to the distribution generated from the bootstrapped samples to determine if niche divergence or conservatism could be detected.  If the observed difference were within the central 95% of the null distribution the differences between the two species we would conclude that occurrences can be explained by the differences between their backgrounds. Alternatively, conservatism would be supported if the occurrences are less different than the background (i.e. occurrences have a difference of less than 2.5 percentile), while divergence would be supported if the observed difference is greater than the null distribution’s 97.5 percentile. For the first four PC axes I also looked for the variables with the highest importance in each the MacGillivray’s and the Mourning warblers’ niche models based on how heavily each variable correlated with each PC axis. If an axis has a variable that is highly important for a niche model, it is likely to be more biologically relevant in explaining the differences between the niches of the two species. 2.2.2 Microhabitat structure of breeding territories 2.2.2.1 Microhabitat sampling  To detect whether the taxa are segregated by microhabitat in sympatry I measured the microhabitat structure of breeding territories along a west-east transect during the 2013 breeding season. The transect extended from allopatric MacGillivray’s warbler populations in Mackenzie,    23  BC, eastwards through the MacGillivray’s and Mourning warbler hybrid zone centre near Chetwynd, BC, to adjacent allopatric Mourning warbler populations in western Alberta. A total of 38 bird territories were sampled within the hybrid zone area as defined by Irwin et al. (2009), and 10 were sampled in each MacGillivray’s warbler and Mourning warbler areas of allopatry adjacent to the hybrid zone.  To define the breeding territories defended by males, territories were mapped for 30 minutes using a handheld GPS by marking the locations where a male was observed singing. For most birds, two points that were at least 10m apart were randomly selected from each set of mapped territory points as the centres of 5m radius vegetation plots. For the first five birds sampled three plots were measured, and one bird’s territory was large enough to only fit one vegetation plot. Within the plots I estimated the percent ground cover of each plant species. Ground cover plants were defined as those under 0.5m tall. All shrubs (i.e. plants with woody stems 0.5-2m tall) were counted and identified to species or genus, as were all trees (i.e. plants with woody stems above 2m tall) within the plot. Every tree was also classified by diameter at breast height (dbh) into 10cm categories. The data from the two or three plots were averaged for each individual territory. This habitat sampling protocol was adapted from the BBIRD field protocol for avian habitat sampling (Martin et al. 1997).  2.2.2.2 Bird sampling I captured all birds that had their territories sampled for microhabitat structure using a recording of a local MacGillivray’s warbler to attract them toward a mistnet setup inside their territory. Each bird was banded with a Canadian Wildlife Services aluminum band, and birds that were located in an area with a high density of other banded MacGillivray’s and Mourning warblers were given a unique combination of 1-3 colour bands for visual identification. A 20μl sample of blood was taken from the brachial vein of captured birds and stored in 500μl of lysis buffer for genetic analysis and the birds were released.      24  2.2.2.3 Microhabitat features of areas without birds A similar protocol was used to measure the vegetation at locations where MacGillivray’s and Mourning warblers were absent as described under section 2.2.2.1 Microhabitat sampling. These sites were sampled to compare locations with birds present to those where birds were absent. This comparison was done to determine if birds select only a subset of available microhabitat conditions. A total of 23 locations across the east-west transect (3 in the allopatric region of the MacGillivray’s warbler range, 10 in the hybrid zone, and 10 in the allopatric part of the Mourning warbler range) were selected by stopping every 10km along most routes used to survey for birds. At each 10km mark the centre of a 5m radius vegetation plot was marked at an assigned distance away from the road by rotating through 10, 20, and 30m. To confirm that the marked vegetation plot was not occupied by a MacGillivray’s or Mourning warbler a recording of a local MacGillivray’s warbler was played for a minimum of 2 minutes, and another 2 minutes were spent checking for any indication of a bird defending a territory. If a bird was detected at one of these locations or if the location was inaccessible (e.g. steep drop), I moved 1km down the road and checked that the plot was not occupied by a MacGillivray’s or Mourning warbler. This protocol was conducted until the absence of the focal birds was confirmed. The same vegetation variables were measured in the absence plots as in occupied territories.  2.2.3 Developing the hybrid index We used a genotype-by-sequencing method to determine the genetic background of birds sampled for microhabitat analyses. We also included some birds previously sampled by Irwin et al. (2009) and from the Chicago Field Museum that breed in allopatric regions far from the hybrid zone to aid in delineating the genotypes of both parental species. DNA was extracted from blood and tissue samples using a standard phenol-chloroform extraction technique. A genotype-by-sequencing (GBS) library was prepared using a modified version of the protocol by Elshire et al. (2011), which is detailed in Appendix A and B. This generated a dataset of 3845 SNPs which I used for constructing a hybrid index. A hybrid index score was developed using the SNP dataset to analyse the ecological differences among a gradient of genotypes in the hybrid zone. I also used the hybrid index to assign birds    25  used in the hybrid zone analyses as MacGillivray’s warbler, Mourning warbler, or hybrid for qualitative description purposes. I developed the hybrid index score by conducting a PCA on the SNP dataset and using the values on the first axis as the index scores because they discriminated between the two parental species well (see section 2.3.5). The PCA was generated using birds sampled from allopatric regions of the two species’ ranges far from the hybrid zone and that were sampled by other members of the Irwin lab in previous years, and applied to the birds sampled for microhabitat analyses. The SNP dataset had missing data because in genotype by sequencing individuals are not all sequenced at the same loci. Therefore I used the snpgdsPCA function from the SNPRelate package (Zheng et al. 2012) in R v.2.15.2 (R Core Team 2012), which is a PCA function created to use with missing SNP data from genotype by sequencing. To verify that the first PC axis accounted for the genetic structure between the two species we used STRUCTURE 2.5.4 (Pritchard et al. 2000) to calculate the assignment probabilities of all sampled birds to two groups. I compared the assignment probabilities to the PC1 scores using Pearson’s correlation coefficient.  Classification as MacGillivray’s or Mourning warbler was determined by assigning birds that fell within 2 standard deviations of the mean score on the first PC axis of birds from allopatric populations of MacGillivray’s or Mourning warblers located far from the hybrid zone (i.e. the same birds used to train the PCA). Birds with scores between the parental ranges were classified as hybrids. I standardized the hybrid index scores by longitude by taking the residuals from a sigmoidal cline fitted to hybrid index scores across the longitudinal transect through the hybrid zone. Standardizing the hybrid index by longitude was conducted to avoid confounding effects of geography on analyses in the hybrid zone because ecological conditions varied with longitude. 2.2.4 Relationship between genotype and ecological traits in the hybrid zone 2.2.4.1 Relationship between genotype and microhabitat in the hybrid zone To detect whether differences exist between the vegetative components measured in MacGillivray’s warbler, Mourning warbler, and hybrid breeding territories I conducted a PCA on the 104 measured microhabitat variables. These microhabitat variables were the percentage of ground cover, shrubs or trees for each plant species found within a vegetation plot, and the total    26  number of plant species, as well as total number of shrubs, and trees. For the microhabitat analyses I compared PC scores of the populations of parental species adjacent to the hybrid zone to identify the axes along which the parental species differed, which I identified by looking for distinct clusters of one species. Along the axis where parental species differed I analysed the relationship between standardized hybrid index and PC scores of birds sampled in the hybrid zone using linear regression. My objective with this analysis was to determine if microhabitat differences observed between the species’ populations outside of the hybrid zone were maintained in the hybrid zone. If those differences were maintained I predicted that those microhabitat traits would predict genotype of birds in the hybrid zone.  2.2.4.2 Relationship between genotype and niche model predictions in the hybrid zone In addition to testing for niche divergence in populations that are likely involved in the dynamics along the borders of the MacGillivray’s and Mourning warbler ranges, I used predictions from the ecological niche models described above to investigate whether niche differences exist in the hybrid zone. Investigating niche differences in the hybrid zone is valuable because the background environment is the same for both species. Thus, any differences detected in predicted suitability between hybrid index scores could not be attributed to differences in available environments. Additionally, analysis in the hybrid zone could reveal whether the environmental space occupied by hybrids has a role in maintaining isolation. For instance, if genotype is related to niche suitability I would expect hybrids to occupy territories that are different compared to territories occupied by parental species. I analysed the relationship between the suitability predicted by the ecological niche models and standardized hybrid index by developing a method to standardize and combine the predictions made by both the MacGillivray’s warbler models and the Mourning warbler models at all locations where the 58 birds were sampled during the 2013 breeding season, and two birds sampled by Irwin lab members previously. At each sampled location I extracted data for the same 10 environmental variable rasters used to train the models and applied both species niche models to generate two suitability measures. The suitability score predicted by one species niche model was added to the distribution of predicted suitability scores of all allopatric presence localities for that species, and the percentile of the suitability score was used. For each bird I    27  subtracted the percentile taken from the distribution of predicted suitability for the Mourning warbler from the percentile taken from the distribution of predicted suitability for the MacGillivray’s warbler to obtain a combined and standardized suitability score, which I will refer to as the standardized suitability difference score (SSDS). The SSDS could range between 1 and -1, where a score of 1 meaning a site has the highest suitability score for a MacGillivray’s warbler and the lowest suitability score for a Mourning warbler, and vice versa for a score of -1. A SSDS of 0 would indicate that a location is equally suitable for both species. This means that a score of 0 could mean that location is equally good or poor for both species, so the predicted suitability from both species models at these locations should be examined. If SSDS and genotype are related, I predicted that hybrids would occupy territories that have a mean SSDS of 0. The mean SSDS of 0 could indicate that hybrids occupy territories that are poorly suited for both parental species, which could suggest the hybrids are at a disadvantage. Alternatively, if hybrids have a mean score of 0 they might occupy territories that are equally good for both parental species and not experience a disadvantage. I constructed a linear model with standard difference score as the explanatory variable and standard hybrid index score as the response to understand the relationship between these variables. If a relationship exists between these variables, a territory with a positive standardized suitability score would be predicted to be occupied by a bird with a MacGillivray’s warbler-like hybrid index.  I also tested whether SSDS differs at locations where birds were detected compared to where birds were absent by conducting a nested ANOVA, with presence/absence nested in geographic locations (i.e. allopatric west, allopatric east, and the hybrid zone). 2.3 Results 2.3.1 Ecological niche models The ecological niche models generated for the MacGillivray’s and Mourning warbler performed adequately during model validation on the testing datasets. In ecological niche modelling validation statistics above 0.8 are considered to demonstrate good performance of models, while statistics below 0.7 are generally considered to be poor (Araújo et al. 2005). All five ecological    28  niche models for the MacGillivray’s warbler had high sensitivity and TSS scores during cross-validation although the models did not perform as well predicting true absences as specificity was low (Table 2). Cross-validation AUC scores for these five models ranged between 0.585 and 0.756, with three models having AUC scores above 0.75 and thus performed adequately at predicting presences and absences in the testing dataset. Relatively low cross-validation statistics for the McGillivray’s warbler models may be in part explained by the fact that the species occurs in a geographic range that has lower climatic variability. The performance of the five Mourning warbler ecological niche models during cross-validation was high as all test statistics for all models were above 0.72 (Table 2). Similar to the MacGillivray’s warbler models specificity consistently was the lowest statistic, however AUC scores were all above 0.84 suggesting the models are robust.  The variables with the highest contribution for distinguishing between bird presence and absence differed considerably between the MacGillivray’s and Mourning warbler niche models (Table 3). The two most important variables in all MacGillivray’s warbler models were temperature seasonality and annual precipitation. Models predict that a MacGillivray’s warbler presence is highly probable at locations with temperature seasonality between approximately 6°C and 9.5°C (Fig. 2a). The probability of a MacGillivray’s warbler presence also increases with annual precipitation (Fig. 2b). In contrast, annual mean temperature and mean temperature of the wettest quarter are the variables with the highest contribution for all five Mourning warbler models (Table 2). The models predict a high probability of bird presence where the annual mean temperature is between -1.5°C and 1.2°C (Fig. 2c), where the mean temperature of the wettest quarter is between 13°C and 17°C (Fig. 2d). This suggests that Mourning warblers are absent in locations that are cold and have large amount of snowfall. The variables that were most important in the models of each species tend to show little importance for the models of the other species. While predicting the presence of the MacGillivray’s warbler was based on climatic variability, models for the Mourning warbler depended little on variability. Additionally the relationships between probability of presence and several environmental variables such as annual mean temperature, temperature seasonality were opposite between models of the two species.    29  2.3.2 Niche and range limits comparison The average suitability predicted across the study area by the five MacGillivray’s warbler ecological niche models indicated that niche and range limits coincide in that predicted suitability was high inside the boundaries of the species’ known range, and outside the boundaries suitability was low (Fig. 3a). The same pattern of high suitability within the actual range was observed with the average predicted suitability of the models for the Mourning warblers (Fig. 3b). However, unlike its sister species, the Mourning warbler models predicted high suitability in the south-west part of the study area, which is outside its own range. Given that there is a region of very low (<0.1) suitability between this region of high suitability and the regions of high suitability within the actual range, these results also suggest that the Mourning warbler is limited to its range by niche requirements. Thus, the models suggest that both species may be limited to a narrow range of overlap by their ecological niches.  2.3.3 Niche identity test To determine if two species have divergent niches the first criterion to examine is whether the niches are significantly different from each other. The niche identity test applies a PCA to all observed presences and random points from the allopatric ranges of both species and compares the observed niche identity statistic D, which measures the difference between PC scores of the observed presences, to a null distribution of D values calculated by randomizing species identity of the occurrence points (Broennimann et al. 2012). Tests for niche identity between the MacGillivray’s and Mourning warbler show that the two species have non-identical niches. The first two axes of the PCA-env test explained 76.24% of the variation, with PC1 being strongly correlated with precipitation measures and temperature, and PC2 with measures of vegetation cover and mean temperature of the warmest quarter. The multivariate PCA-env method showed that the MacGillivray’s warbler had a wide range of available environmental space (Fig. 4a) compared to the Mourning warbler (Fig. 4b), but that they overlapped. Despite the overlap between the available environments, the two species were separated along the two axes. The test of niche identity confirmed that the two species occupy different environmental space, as niche identity was significantly lower than expected by chance (D=0.054, p=0.0198; Fig. 4c).     30  2.3.4 Niche background tests The niche background tests address the second criterion for determining if two niches are divergent. Niche background tests examine whether differences observed between two niches in the identity test are explained by species occupying ranges with different ecological conditions (Warren et al. 2008). To declare divergence, species niches must be more different than the ranges’ backgrounds. There is some evidence that the MacGillivray’s and Mourning warblers have niches that are slightly more different than the differences between their background environments. The MacGillivray’s and Mourning warblers are fairly ecologically similar but do differ in certain aspects of the niche. However, the results varied somewhat with the test conducted.  The background test using the multivariate PCA-env method presented by Broennimann (2012) compares the observed niche identity D to two different null distributions. Each null distribution is generated by calculating D between the PC scores of one species occurrences to the PC scores of random locations in the other species’ background. The PCA used in this test is generated using all occurrence and random background points. This test examines whether D differs significantly when one species occurrences are exchanged with random points. If D is does not differ, differences between species niches can be explained by the fact that the two species occupy ranges that are ecologically different. The result from the PCA-env test was consistent with conservatism when testing Mourning warbler in the MacGillivray’s warbler background (p= 0.0198; Fig. 5a), while the result for the test in the other direction was inconclusive (p=0.6139; Fig. 5b). This test therefore does not support niche divergence, and instead suggests that the two species niches show a pattern of conservatism.  Contradictory to the PCA-env background test, I found support for niche divergence from the McCormack et al. (2010) multivariate background test, although the species may have conserved the use of some aspects of the niche. For this test a PCA is generated on all occurrence and 10000 points randomly sampled from each species’ backgrounds. This test differs from the background test of Broennimann et al. (2012) in that the differences in mean PC scores between the two species’ occurrences is compared to null distributions generated from the difference    31  between mean  PC scores of resampled background points, and that each PC axis can be analysed separately.  The first four PC axes explained a combined 89.98 % of the variation of the environmental data, and divergence was supported on 2 of the 4 axes (Table 4). Along the first axis, which explained 53.61% of the variation in environmental conditions, niche divergence was detected. The difference between the means of the two species’ scores was slightly, yet significantly, greater than what was expected based on differences between random background samples (p<0.05; Table 4). These results suggest that the MacGillivray’s warbler niche includes lower temperature and low variation in temperature compared to the Mourning warbler’s niche. The first axis is probably of high biological significance because the two most important variables in the MacGillivray’s warbler niche model loaded heavily on this axis. The top two variables in the Mourning warbler niche models loaded moderately on the first axis. Thus, I detected divergence in variables predicted to play an important a role in governing the occupancy of sites within at least one of the species’ ranges.   On the second axis, which explained 17.92% of the variation in environmental conditions, niche conservatism was detected as the difference between the presences was significantly smaller than that of the background (p<0.05; Table 4). Values on this axis were interpreted as measures of temperature during the warmest periods of the year and vegetation cover.  Annual mean temperature was the variable with the fifth highest loading on the second axis, and was the variable with the highest contribution in Mourning warbler niche models.  A strong signal of divergence was detected on the third axis, which explained 13.54% of the variation as the difference between the average scores for the parental species presence data was significantly greater than that of the background data (p<0.05; Table 4). This suggests that the niches of the two species are defined by different amounts of precipitation. Specifically, the MacGillivray’s warbler occupies space that had significantly higher precipitation than the Mourning warbler. The fourth axis, which explained 3.865% of the variation, suggested that the two species niches show conservatism in the amount of vegetative cover (p<0.05; Table 4). The conflicting results from the two types of background tests (i.e. described by Broennimann et al. [2012], and by McCormack et al. [2010]) can be reconciled by examining the density grid    32  PCA plots produced by the PCA-env method for the MacGillivray’s and Mourning warblers (Fig. 4a, Fig. 4b, respectively). The PCA-env density grid plots show that the pattern of conservatism is much greater on the second axis than the signature of divergence along the first axis.  Thus, when both axes are considered as in the Broennimann et al. (2012) tests a pattern of conservatism is detected. I can therefore conclude that the two species show divergence for certain aspects of the niche loading on the first and third axis, and conservatism for those along the second and fourth axes.  2.3.5 Hybrid index I developed a hybrid index using a PCA on 3845 SNPs for use in analyses investigating the relationship between genotype and ecological traits of breeding habitats in the MacGillivray’s and Mourning warbler hybrid zone. The first axis of the PCA explained 4.3% of the SNP variation and discriminated between the allopatric groups well (Fig. 6). The scores on the first axis were highly correlated with the assignment probabilities from STRUCTURE (r = 0.979, p<2.2e-16). Genetic differentiation is often low between neighbouring populations (e.g. Delmore et al. 2015, Kenyon 2013), accordingly, the SNP PC1 scores reflect the differentiation between the two species well despite the low percent of variation explained by the first axis. For descriptive purposes I categorized individuals in the hybrid zone as pure MacGillivray’s or Mourning warblers as those with a PC score within two standard deviations of the mean PC1 score of birds sampled in allopatric regions far from the hybrid zone. The mean PC1 score for allopatric MacGillivray’s warblers was -0.111± 0.012 standard deviation, while the allopatric Mourning warblers had a mean PC1 score of 0.129±0.029 standard deviation. Individuals in the hybrid zone with scores between -0.135 and -0.088 were classified as MacGillivray’s warblers, and individuals with scores ranging from 0.070 to 0.187 were classified as Mourning warblers. Birds with scores between the allopatric ranges were classified as hybrids. After applying these classification criteria to the 38 birds in the hybrid zone there were 23 hybrids, 8 Mourning warblers, and 5 MacGillivray’s warblers.  I standardized the hybrid index scores by geography by fitting a geographic cline to the hybrid index scores and using the residuals as the standardized hybrid index. The cline fit to the hybrid    33  index scores was centered at a longitude of -1551km, had a slope of 2.283e-06, and a width of 92.08km (Fig. 7). Residuals of the cline, which I used as the standardized hybrid index scores for analyses in the hybrid zone ranged between -0.07230 and 0.19240, with birds with the lowest scores assigned to MacGillivray’s warblers and birds with the highest assigned to Mourning warblers. 2.3.7 Microhabitat structure of parental species adjacent to the hybrid zone The first 13 axes of the PCA on microhabitat variables trained on birds found in allopatric populations adjacent to the hybrid zone explained a total 75% of the variation in the microhabitat data and each axis explained at least 3% of the variation alone. The first axis, which explained 10.57% of the variation in microhabitat, best separated the microhabitat features of territories occupied by MacGillivray’s warblers in allopatry adjacent to hybrid zone from those occupied by Mourning warblers adjacent to the hybrid zone (Fig. 8a.). Along all other 12 axes both allopatric populations of the two species and locations without birds were very similar. The comparison of the species’ territories along the first axis reveals that the two species differ in the species of ground cover plants and shrubs, and the number of older trees (Table 5). Mourning warblers in allopatric populations adjacent to the hybrid zone tend to occupy territories with a higher percentage of prickly rose (Rosa acicularis), dead trees, Hornemann’s willowherb (Epilobium hornemannii) and western sarsaparilla (Aralia nudicaulis), but a lower percentage of elderberry (Sambucus cerulea), ferns, stinging nettle (Urtica dioica), devil’s club (Oplopanax horridus), horsetails (Equisetum spp.), and thimbleberry (Rubus parviflorus) compared to MacGillivray’s warblers. Mourning warblers also occupied territories with more trees with a dbh between 20cm and 30cm, whereas the MacGillivray’s warbler’s territories mainly had trees with dbh under 20cm. Both species occupied territories with dense understoreys largely made up of woody shrubs, cow parsnip and fireweed (Epilobium augustifolium). The most common woody shrubs found in the territories were mountain alder (Alnus tenuifolia), highbush cranberry (Viburnum edule), prickly rose, thimbleberry, and willows (Salix spp.). Twinberry (Lonicera involucrata) was the most abundant shrub species in the territories. Trees growing in the territories were typically young and were most often willows or mountain alder.    34  The absence locations sampled for microhabitat in allopatric regions of the MacGillivray’s and Mourning warbler ranges adjacent to the hybrid zone had scores along the first principal component axis similar to the territories occupied by birds, but tended to have scores lower than birds in the same region (Fig. 8a.). The absence locations in the Mourning warbler range therefore had a higher percentage of species that were common in bird territories, and the absence locations in the MacGillivray’s range were Mourning warbler-like.  2.3.8 The relationship between microhabitat and hybrid index in the hybrid zone The differences in microhabitat PC1 scores that I detected between the MacGillivray’s and Mourning warblers in populations adjacent to the hybrid zone were not maintained in the hybrid zone. I detected no relationship between microhabitat PC1 scores and standardized hybrid index in the hybrid zone (R2= -0.0277, p=0.947; Fig. 8b). Locations occupied by birds did differ from locations without birds along the third microhabitat axis, which explained 7.4% of the variation in habitat (Fig. 9). Locations without birds had understoreys with lower densities of shrubs and groundcover, and tended to have larger trees. Lodgepole pine was common in the locations without birds, as was paper birch. 2.3.9 Relationship between predicted suitability and hybrid index Although I detected evidence that a few aspects of the niche differs between MacGillivray’s and Mourning warblers in allopatry with the background tests, in sympatry where the same background conditions are available the two species and their hybrids occupy breeding territories that do not differ in standardized suitability predicted by niche models. No relationship between SSDS and the hybrid index scores standardized for geography of birds in the hybrid zone was found (R2= 0.006, p= 0.649; Fig. 10). All birds sampled in the hybrid zone occupied territories with raw suitability scores greater than the lowest suitability scores measured in allopatry for both species’ models (minimum suitability score of allopatric MacGillivray’s warblers= 0.000876, minimum suitability score of allopatric Mourning warblers=0.00363). This indicates that locations in the hybrid zone where birds were breeding were suitable for both species. The SSDS did vary significantly between geographic regions (F3, 76= 56.291, p < 2e-16); however suitability did not differ within a geographic region between locations where birds were present    35  compared to where birds were absent (F3, 76= 0.963, p=0.415). Western allopatric territories were more suitable for MacGillivray’s warblers (mean SSDS= 0.486± 0.177) and the eastern allopatric territories had environmental conditions more suitable to Mourning warblers (mean SSDS= -0.673± 0.081). The standardized suitability scores of the territories in the hybrid zone were intermediate to those of the allopatric regions (mean SSDS= 0.013± 0.270). Consistent with the patterns of suitability scores at occupied territories, random locations without MacGillivray’s warblers in the western allopatric region adjacent to the hybrid zone were more suitable for MacGillivray’s warblers (mean SSDS= 0.491± 0.106), eastern absences were suitable for Mourning warblers (mean SSDS= -0.518± 0.174), and standard suitability at absence locations in the hybrid zone were intermediate to the two adjacent allopatric locations (mean SSDS= 0.071± 0.252).  2.4 Discussion The goal of this study was to understand the role of ecological traits in maintaining a narrow hybrid zone between MacGillivray’s and Mourning warblers. I addressed this question by comparing broad scale patterns of niche characteristics of the two species using ecological niche models, as well as microhabitat traits of breeding territories in the hybrid zone. The comparison of niche characteristics between populations of the two species at range borders revealed how niche divergence along a subset of niche axes may limit the ranges of the two species to a narrow zone of overlap, thus reducing secondary contact. Analyses of ecological traits of breeding territories in the hybrid zone helped explain whether ecological selection may promote isolation between the two species when they do come into contact. Overall my comparison of the ecological niches showed that the MacGillivray’s and Mourning warblers are ecologically similar, but may have diverged in some aspects of the niche. To conduct this comparison I generated ecological niche models for both species and used them to test for divergence of the niche as a whole, as well as in specific features of the niche. I detected evidence that the MacGillivray’s and Mourning warblers niches have been conserved overall when comparing the complete ecological niche. However, this test and the multivariate test that considers whether divergence may have occurred in separate parts of the niche provided conflicting results regarding divergence as the multivariate test showed evidence for divergence    36  in a couple niche characteristics. The results from the multivariate test suggest that niche of the MacGillivray’s warbler includes higher variation in temperature and more precipitation than that of the Mourning warbler. Alternatively, the niches of the two species overlap in vegetation cover and warm temperature. It is possible that the niches of the two species may have diverged slightly to adapt to different climates, but that ecological selection was generally very similar during speciation resulting in two ecologically similar species. The conflict between the tests for divergence that investigated the niche was a whole (i.e. the PCA-env test) and the test that compared specific aspects of the niche separately can be reconciled when comparing the magnitude of similarity to that of divergence observed between the two species. The PCA-env test essentially tested the for niche divergence using a score made up of a combination of the first two niche axes. The signature of conservatism along the second niche axis was very strong, and this is may have made the signature of divergence along the first axis undetectable in the PCA-env test. The third niche axis, which is not considered in the PCA-env test but explains a similar amount of variance as the second axis, has a strong signature of divergence but goes unnoticed by the PCA-env test. Conclusions made from inspecting each axis may better reflect true differences and similarities in niches. Similar arguments about tests that combine the aspects of the niche into one measure may overlook difference have been made when conflicting tests of divergence are observed between these different types of background similarity tests (e.g. McCormack et al. 2010). Although the differences between the niches are slight, they may indicate important divergent adaptations.  The divergent aspects of the two species niches may facilitate reproductive isolation by limiting the ranges to only a narrow zone of overlap because the limits of the range match predicted niche limits (i.e. areas where models predict high suitability). If range limits were not governed by ecological conditions, we would expect niche models to predict highly suitable conditions beyond a species known range (Glor and Warren 2010). Thus, opportunities for the MacGillivray’s and Mourning warblers to come into contact may be reduced to its current narrow width (Irwin et al. 2009) by their niches. This result suggests that the location of the hybrid zone may be explained to some extent by the slight differences detected between the two species’ niches.    37  The coincidence of niche and range limits, as well as the correlations between divergent aspects of the niche with longitude suggests that the location and possibly the width of the hybrid zone are determined by an environmental gradient. This provides support for the geographical selection-gradient model of hybrid zones, in which the width and location of a hybrid zone is determined by ecological variables (Moore and Price 1993). Stable hybrid zones described by this model are maintained by selection that varies with the environment. However, I detected no relationship between standard hybrid index score with standard suitability score in the hybrid zone. It is possible that divergent selection along the environmental gradient begins beyond the hybrid zone. Within the hybrid zone the two species may be locally adapted to the same selective pressures but unable to persist in the conditions beyond the hybrid zone and into the other species’ niche. The environmental conditions in the hybrid zone may therefore include the overlapping aspects of both species’ niches.  Within the hybrid zone, hybrids do not appear to breed in territories that have different climatic conditions or amount of vegetation cover than parental species as there was no relationship between standard suitability score and standard hybrid index score. My study also did not find convincing evidence that standard hybrid index was predicted by certain microhabitat vegetation features. In territories of allopatric populations at the edges of the hybrid zone the two species do breed in territories that have different predicted suitability and microhabitat from each other and from territories in the hybrid zone. However, since these differences were not maintained in the hybrid zone after controlling for the general genetic and habitat gradient across the zone, they likely do not act as an isolating barrier in sympatry. These results propose a compelling argument against ecological divergence acting as an isolating mechanism because studies in the hybrid zone remove the possibly confounding effect of environmental background in studies of niche divergence between allopatric populations. In fact, these results suggests that hybrids are not at a disadvantage compared to parental species regarding the selection of breeding territory, which is known to be an important determinant of fitness in birds (Johnson 2007). This suggests that ecological selection in terms of breeding habitat use against hybrids may not have a large role in maintaining the narrow hybrid zone.   Collectively, my results propose that ecological selection has a role in explaining the location and possibly a small role in explaining the narrowness of the hybrid zone between the    38  MacGillivray’s and Mourning warblers. The two speices’ niches remain largely conserved, but divergent aspects of the niche may limit interspecific encounters to the narrow hybrid zone. This suggests that divergence on certain niche axes may have promoted speciation. Few studies have conducted extensive tests of niche divergence using ecological niche models in hybrid zones to understand speciation in birds. There is some evidence in other avian sister species systems that come into secondary contact which also suggests that the hybrid zone is maintained by divergent adaptations resulting from ecological speciation in allopatry. Swenson (2006) built niche models which suggested that temperature plays a key role in explaining the location of four avian hybrid zones, and attributed this to adaptation to different conditions in Pleistocene refugia. Similar to my study, it appears that ecological niches may explain the location of a hybrid zone to a larger extent than the width. The two species are ecologically similar as they have conserved a number of niche attributes, including amount of vegetation cover and temperature during warm periods. Also, when exposed to the same suite of environmental background conditions in the zone of secondary contact the microhabitats of two species are ecologically indistinguishable. It is therefore likely that selection on breeding habitat use is not strong enough to maintain isolation on its own. This is the claim made by McCormack et al. (2010), who investigated the role of niche divergence in Aphelocoma jays, a group of North American species thought to have diverged during the Pleistocene glaciations. The jays showed divergence along several niche axes, and conservatism on others. McCormack et al. (2010) concluded that this did not provide sufficient evidence to argue that hybrid zones formed by the species were primarily maintained by exogenous selection. Ecological selection acting on breeding habitat and climate use may be contributing to the maintenance of isolation between MacGillivray’s and Mourning warblers, but there are likely other sources of isolation.  One source of ecological selection that may contribute to the isolation between MacGillivray’s and Mourning warblers is habitat and climate during the non-breeding season. The phenology and fitness of migratory birds is known to be strongly influenced by climatic and habitat conditions in the wintering grounds and along the migratory route (Marra et al. 1998; Norris et al. 2004; Gordo 2007; Bauer et al. 2008). Niches axes explaining the conditions in the non-breeding season may be divergent between the two species because they have different wintering    39  ranges and migratory routes (Pitocchelli 1995, 2011). The habitat and climate of the wintering grounds may cause carry-over effects into the breeding season, such as in the American redstart (Setophaga ruticilla), where individuals in higher quality wintering territories left for spring migration earlier and had higher reproductive success than those in poor-quality winter territories (Marra et al. 1998; Norris et al. 2004). If MacGillivray’s and Mourning warbler hybrids overwinter in territories that differ from parental species they may experience reduced fitness, thus ecological selection on wintering habitat could cause post-mating isolation.  Another possible source of post-mating isolation could be divergent migratory behaviour. MacGillivray’s and Mourning warblers exhibit a migratory divide, as they take migratory routes along the western and eastern coasts of North America, respectively (Rohwer and Irwin 2011). Hybrids may take an intermediate and potentially inferior route, which could lead to selection against hybrids maintaining the narrow contact zone (Rohwer and Irwin 2011). Hybrids have been shown to take migratory routes intermediate to parental species and be under strong selection in Swainson’s thrush (Delmore and Irwin 2014). This is first example of direct evidence for hybrids using intermediate and potentially suboptimal migratory strategies and provides a strong argument for divergent migratory behaviour as an isolating mechanism in North American songbirds. Selection associated with divergent migratory behaviour, perhaps in concert with the slight divergence in niche characteristics, may therefore best explain the maintenance of the narrow MacGillivray’s and Mourning warbler hybrid zone.    40  Table 1. A summary of the tests for niche divergence between species used in this study. Niche identity and background tests compare the same test statistic, Schoener’s D which measures niche overlap between two groups, to different null expectations. The aim of these tests is to first determine if species niches are different using the identity test. There are two possible explanations if the niches are found to be different. The two species could seem different because they occur in different ranges with different available conditions, or they could have divergent niches. The background test addresses the question of differences in available environment. The McCormack background test compares average differences between species to average differences between available conditions in their range (i.e. environmental background) along different niche axes, which are generated by conducting a principal component analysis on all species presence data and 10 000 points sampled randomly from each species’ background. C= conservatism, D=divergence. Name of test Null hypothesis Test statistic Generation of null distribution (s) Possible inferences Sources Identity test Niche differences are independent of species identity Schoener’s D:  Measures niche overlap between two species.   1: complete niche overlap  0: no niche overlap Calculate D for 100 resampled sets of locality data with random species assignment. C: Observed D>97.5% quantile D: Observed D<2.5% quantile Warren et al. (2008) Broennimann et al. (2012) Background test Niche differences are independent of background differences Calculate D between locality data for species A and random background data for species B.   Repeat to generate a 2nd null distribution exchanging species A and B. C: Observed D>97.5% quantile of both null distributions. D: Observed D<2.5% quantile of both null distributions. No inferences: if the two results do not agree.  Warren et al. (2008) Broennimann et al. (2012) McCormack test Niche differences of species along niche axes equal differences of background points dO:  Difference between two species’ average scores along a PC axis Calculate the difference between random points from both species’ backgrounds for each PC axis Varies with PC axis.  C: dO>97.5% quantile D: dO<2.5% quantile  McCormack et al. (2010)     41  Table 2. The results from niche model cross-validation show that Mourning warbler models (MOWA 1-5) performed better than MacGillivray’s warblers (MGWA 1-5), but that all models performed adequately. Sensitivity (i.e. the fraction of correctly predicted true presences), specificity (i.e. the fraction of correctly predicted true absences), the true skill statistic (TSS; i.e. actual model accuracy compared to expected random accuracy), and the area-under the curve statistic (AUC; i.e. statistic with lower subjectivity) were usually greater than 0.70, which is considered to indicate acceptable model performance (Araújo et al. 2005).  Model Sensitivity Specificity TSS AUC MGWA 1 0.8274678 0.5618758 0.8274678 0.7563903 MGWA 2 0.8015464    0.5852635 0.8015464 0.5852635 MGWA 3 0.7742489    0.6106528 0.7742489 0.6106528 MGWA 4 0.8668385    0.5257672 0.8668385 0.7539833 MGWA 5 0.8248927    0.5607179 0.8248927 0.7535916 MOWA 1 0.8408163    0.7743332 0.8408163 0.869978 MOWA 2 0.8373984    0.7279183 0.8373984 0.855225 MOWA 3 0.7673469 0.7748528 0.7673469 0.8497024 MOWA 4 0.8170732    0.7816072 0.8170732 0.8681548 MOWA 5 0.7673469    0.7729477 0.7673469 0.8422404        42  Table 3. The percent contribution of climatic and vegetation variables included in niche models. The two variables with the highest contribution for the MacGillivray’s (MGWA) models were temperature seasonality and annual precipitation. In contrast, the two variables with the highest contribution for the Mourning (MOWA) models were annual mean temperature and the mean temperature of the wettest quarter. EVI= enhanced vegetation index, a measure of green vegetation cover.  Percent contribution Variable name MGWA models MOWA models Annual Mean Temperature 11 15.9 15 17.5 11.3 33.5 33.4 31.4 33.6 33.6 Mean Diurnal Range  2.1 1.9 1.8 2.1 2 5.8 5.1 4.5 4.8 5.3 Temperature Seasonality  25.7 24.9 29.2 24.2 26.8 1.4 1.7 2.2 1.9 2.5 Temperature Annual Range  0.1 0.1 0 0 0.1 5.5 4.2 4.8 6.4 6 Mean Temperature of Wettest Quarter 14 14.6 10.8 12.6 12.1 32.7 31.9 31.3 29.8 31.3 Mean Temperature of Warmest Quarter 9.3 6.5 6.5 5.2 8.9   0.3 2.2 1.6 2.1 0.8 Annual Precipitation 20.8 20.4 22.2 21.5 22.2 4.9 5.2 4.5 5.3 4.6 Precipitation Seasonality  7.1 6.2 5.8 7.5 7.3 2.1 2.2 3.2 1.8 2.7 Precipitation of Warmest Quarter 3.5 3.2 2.6 2.9 2.3 8.4 8.1 9.2 8 7.1 Mean June EVI 6.3 6.3   6.1 6.5 6.8 5.5 5.8 7.2 6.4 6.1       43  Table 4. The multivariate McCormack et al. (2010) background test compared the observed differences in PC scores of the two species (|dO|) to the differences of points sampled randomly from the two backgrounds. The test showed that the warblers are diverged along PC1 (temperature and precipitation), and PC2 (precipitation during warm periods and variation in precipitation). The test showed that the two species have conserved June vegetation cover (EVI) and temperature during the warmest periods. CI Null dist= central 95% of the null distribution. Variable 1-5 indicates the variables with the five highest loadings for each PC axis. Temp=Temperature. Variables that have first or second percent contribution in either species’ niche models are indicated with a superscript 1 or 2.   PC1 PC2 PC3 PC4 |dO| 4.548 0.502 2.513 0.213 95% CI Null Dist. 4.293, 4.435 2.199, 2.289 0.825, 0.907 0.228, 0.278 Niche divergence conservatism divergence conservatism % Variance explained 53.61 17.91 13.54 3.86 Variable 1 Temperature Annual Range Mean Temperature of Warmest Quarter Precipitation of Warmest Quarter Mean June EVI Variable 2 Precipitation of Coldest Quarter Mean May EVI Isothermality Mean May EVI Variable 3 Min Temperature of Coldest Month Mean June EVI Precipitation Seasonality Mean Temperature of Warmest Quarter Variable 4 Temp Seasonality1 Max Temperature of Warmest Month Precipitation of Wettest Month Precipitation of Warmest Quarter Variable 5 Annual Precipitation2 Annual Mean Temperature1 Mean Temperature of Wettest Quarter2 Precipitation Seasonality Interpretation Temp, Precipitation Temp warm periods, EVI Precipitation EVI     44  Table 5. The 14 variables with the highest loadings on the first axis (PC1) of the principal component analysis on microhabitat variables measured in MacGillivray’s and Mourning warbler breeding territories in the hybrid zone and populations adjacent to the hybrid zone. Dbh= diameter at breast height.  Variable name PC1 Loading % Ground cover- western sarsaparilla -0.239 % Trees- dead -0.214 % Shrubs- prickly rose -0.214 % Ground cover- Hornemanns’ willowherb -0.212 Number of tree species- dbh >20cm -0.210 % Shrubs western sarsaparilla -0.202 % Ground cover- prickly rose -0.200 % Ground cover- fern 0.163 % Ground cover-stinging nettle 0.153 % Shrubs- elderberry 0.151 % Trees- alder 0.137 % Ground cover-horsetail 0.124 % Trees- thimbleberry 0.121 % Shrubs- devil’s club 0.121        45   Figure 1.(A) Range maps for the MacGillivray’s warbler (solid line) and the Mourning warbler    46  (dashed line; range maps from Ridgely et al. [2003] edited based on occurrence data collected for niche models) overlaid on a map of the ecoregions of North America (Commission for Environmental Cooperation 2009). The Mourning warbler range was edited to the ecoregions that occur where the two species’ ranges meet (i.e. the boreal plains, taiga plains, and temperate prairies). Both ranges were edited to include only regions within Canada. (B) A map showing the edited ranges used to delineate the study populations for ecological niche model analyses. The blue and red ranges show the edited ranges of the MacGillivray’s (MGWA) and Mourning warblers (MOWA), respectively. The points on the map indicate the locations where a MacGillivray’s (blue) or Mourning warbler (red). These locations were used as the presence data for the ecological niche models.               47  Figure 2. Examples of partial plots for the two variables with the highest contribution in the MacGillivray’s warbler and the Mourning warbler niche models. (A) The predicted probability of a MacGillivray’s warbler (MGWA) presence given temperature seasonality (°Cx1000), the variable with the highest contribution in all five MacGillivray’s warbler niche models. (B) The predicted probability of a MacGillivray’s warbler (MGWA) presence given annual precipitation (mm), the variable with the second highest contribution in all five MacGillivray’s warbler niche models. (C) The predicted probability of a Mourning warbler (MOWA) presence given annual mean temperature (°Cx10), the variable with the highest contribution in all five Mourning warbler niche models. (D) The predicted probability of a Mourning warbler (MOWA) presence given the mean temperature of the wettest quarter (°Cx10), the variable with the second highest contribution in all five Mourning warbler niche models.    48  Figure 3. Comparisons of the mean suitability predictions from five ecological niche models to the range limits (grey outline) of (A) the MacGillivray’s warbler, and (B) the Mourning warbler, show that areas with high predicted suitability are generally restricted to the species’ own range.    49  (C)  Figure 4.Density grids showing the ecological niches of (A) the MacGillivray’s warbler, and (B) the Mourning warbler, on the same scale which represents the available ecological space in two dimensions. Dashed lines and solid lines indicate 50% and 100%, respectively, of the available ecological space in a species’ range. (C) Results from the identity test showing that the observed value of Schoener’s D ( red line) is significantly lower than the null distribution of Schoener’s D generated by randomizing species’ identity (D=0.054, p=0.0198).    50  (A)  (B)  Figure 5. Comparison of the observed Schoener’s D value between the MacGillivray’s and Mourning warblers with a null distribution of Schoener’s D generated by measuring niche differences between (A) the Mourning warbler’s niche with the MacGillivray’s warbler background (p= 0.0198), and (B) the MacGillivray’s warbler’s niche with the Mourning warbler    51  background (p=0.6139). The results from A and B suggest that the two species’ niches may be more similar to each other than the ecological values of their ranges      52  Figure 6. Principal component analysis of SNPs generated using allopatric birds shows that PC1 separates MacGillivray’s (MGWA) and Mourning warblers (MOWA) sampled either in allopatric regions far from the hybrid zone or in allopatric regions adjacent to the hybrid zone (HZ). It also shows that there are many birds with intermediate genotypes in the hybrid zone.       53   Figure 7. The geographic cline fit to PC1 SNP scores of birds sampled for microhabitat analyses across a west-east transect through the hybrid zone. The cline was centred at a longitude of -1551km, had a slope of 2.283e-06, and a width of 92.08km which is similar to the width of the cline described by Irwin et al. (2009).       54  (A)   (B)  0 5 10-0.050.000.050.100.150.20Habitat PC1Standardized hybrid index score   55  Figure 8. (A) Principal component analysis of microhabitat in populations of MacGillivray’s (MGWA) and Mourning warblers (MOWA) adjacent to the hybrid zone (HZ). The first PC axis separates the two populations well. The first axis also shows some difference between bird territories and locations where birds were absent within a geographic area (i.e. between MGWA and western absences, and between MOWA and eastern absences). (B) In the hybrid zone there was no relationship detected between bird genotype (standardized hybrid index score) and habitat PC1 score (R2= -0.0277, p=0.947).      56   Figure 9. In the hybrid zone locations without birds (i.e. absences) were discriminated well from bird territories in the hybrid zone along the third microhabitat axis. Three birds that were outliers are not shown. MGWA= MacGillivray’s warbler; MOWA=Mourning warbler; Hybrid= individuals with hybrid index scores along a continuum of genotypes between the parental species. Species and hybrid classification based on hybrid index scores.      57   Figure 10. In the hybrid zone there is no significant relationship between standardized suitability difference scores and standardized hybrid index scores (R2= 0.006, p= 0.649).      58  Chapter 3: Conclusions 3.1 General conclusions 3.1.1 Slight divergence along a subset of niche axes influences avian speciation The results from the ecological niche models developed for the MacGillivray’s and Mourning warblers, as well as quantification of microhabitat in their stable and narrow hybrid zone showed that both divergence and conservatism in different aspects of the niche likely played a role in speciation. The two species are overall quite ecologically similar in both allopatry and sympatry, but occupy environmental space in allopatry that differs in a number of variables such as annual temperature range and climate during cold periods. This suggests that the two species may have evolved differently in response to divergent environmental surroundings. However, those environmental surroundings may have also shared similarities and contributed to a much larger signature of conservatism in vegetation cover during the breeding season and climate during warm periods. The similarities between the species’ niches may allow them to come into contact and provide the opportunity to form hybrids within the narrow hybrid zone as this was the only region where both niche models predicted high suitability in a continuous expansion of their ranges (i.e. not separated by unsuitable regions). The patterns of niche evolution between the two species therefore seem to partly explain the distributions of the two species as well as the location and narrowness of their hybrid zone.  Divergence along only a small subset of niche axes in the evolution of MacGillivray’s and Mourning warblers suggested by my study may be characteristic of a group of avian sister species pairs in North America. Many pairs of avian sister species that diverged during the Pleistocene have a similar west-east distribution pattern as the MacGillivray’s and Mourning warblers and come into contact and hybridize in the same area of northeastern British Columbia (e.g. the Audubon and Myrtle warblers [Brelsford and Irwin 2009], the Pacific and Winter wren [Toews and Irwin 2008], the Townsend’s and Black-throated green warblers [Toews et al. 2011], Red-breasted and Yellow-bellied Sapsuckers [Seneviratne et al. 2012]; Weir and Schluter 2004). Given this pattern in distributions and the existence of a suture zone, which is a geographic    59  clustering of hybrid zones (Remington 1968), these pairs of sister species may have similar patterns of niche evolution as those suggested by my study. Thus, divergence in only a subset of niche characteristics may be a universal pattern in the evolution of North American birds. 3.1.2 Slight niche divergence of species that meet at an ecotone contributes to hybrid zone narrowness Despite the detection of divergence in some niche characteristics between MacGillivray’s and Mourning warblers in allopatry, I found that when the two species come into contact and hybridize they are ecologically very similar. Under the geographic-selection gradient model, a gradient in environmental conditions such as an ecotone selects for certain genotypes to maintain a narrow hybrid zone (Moore and Price 1993). However, in the MacGillivray’s and Mourning warbler hybrid zone I detected no relationship between environmental suitability or microhabitat and genotype after accounting for longitude. This lends little support to the geographic-selection gradient model and suggests that exogenous selection based on breeding season habitat and climate does not have a large role in promoting isolation in the hybrid zone. However, the allopatric populations directly adjacent to the hybrid zone do have different habitat suitability characteristics. This adds support to my hypothesis that the two species come into contact in a narrow region suitable for both species, but that neither species can expand the hybrid zone due to niche divergence. It is however possible those environmental variables not captured in the niche models pose selective pressures on different genotypes, and that other factors such as hybridization itself restricts the species from expanding their ranges into greater overlap. 3.1.3 Using GIS to study hybrid zones and speciation The GIS techniques provide the opportunity to answer questions about niche divergence between species and how it might be involved in promoting isolation in a hybrid zone; however they have rarely been applied to these questions. The results from my research on the MacGillivray’s and Mourning warblers as well as their hybrid zone show the utility of using GIS-based ecological niche models to understand the ecological differences of species that are largely allopatric but have a small zone of overlap. My study showed the importance of using multiple tests that address niche divergence or conservatism at different scales: one that tests for divergence in the    60  overall niche, and another that tests for differences along a number of niche axes. This approach can reveal smaller signals of divergence or conservatism in certain aspects of the niche that may be overlooked in a test that looks at the overall niche, and therefore provides a more detailed and accurate understanding of niche comparisons between species. Additionally, by integrating the predictions of niche models from two species with genetic data I was able to address questions regarding the role of the niche in promoting isolation between the two species in sympatry. 3.2 Future directions 3.2.1 Future work in the MacGillivray’s and Mourning warbler system The ecological niche models make predictions about the suitability of a given suite of environmental variables based on correlations with species presence. It is therefore important to determine if axes where divergence was detected between the niche models of MacGillivray’s and Mourning warblers are in fact important selective pressures. If the species have diverged in a trait through ecological speciation we may be able to test predictions regarding physiological limits. Temperature was found by Swenson (2006) to be an important predictor of niche differences between Bullock’s orioles and Baltimore orioles, and is thought to play a role in the hybrid zone’s structure and location. The difference in temperature between the two orioles was consistent with heat stress experiments conducted by Rising (1969). Further support for divergence along axes of temperature and precipitation variables such as those detected in my study between the MacGillivray’s and Mourning warblers could be demonstrated by physiology-based studies. My results from the predictions of ecological suitability made by niche models and tests of divergence suggest that in allopatry the niches of the two species have some differences. However I found very little evidence of divergent use of the environment in the hybrid zone, suggesting that they are very similar ecologically in sympatry. In addition, hybrids appear to occupy environmental space that is indistinguishable from the parental species in the hybrid zone. This result could be explained by local adaptation selecting for the same ecological traits in sympatry. Determining if the two species and their hybrids are under the same ecological selective pressures to be similar in sympatry could be achieved using niche models and niche    61  divergence tests. To conduct this analysis niche models could be generated for allopatric populations of the two species at specific longitudinal distance intervals from the hybrid zone’s edges. The models of populations with corresponding distances from the hybrid zone for both species could then be compared using niche divergence tests. Evidence for local adaptation in the hybrid zone would be detected using this method if niche divergence tests showed that populations further away from the hybrid zone are more different than those close to the hybrid zone. This analysis on local adaptation was beyond the scope of this study because there was not sufficient data available for Mourning warbler occurrences in Alberta or Saskatchewan.  Ecological selection from non-breeding season habitat may be acting on MacGillivray’s and Mourning warblers instead of breeding. My study focussed on the niche axes that explain breeding habitat; however research shows that non-breeding habitat can have a strong influence on the phenology and fitness of migratory species (Marra et al. 1998; Norris et al. 2004). Carry-over effects from the non-breeding season could influence the fitness of hybrids and cause post-mating isolation if they overwinter or take stop-overs during migration in unfavourable habitat. Ecological niche models for the non-breeding season could provide insight into whether divergence in niche axes of the non-breeding season might influence isolation between the species.  Given that MacGillivray’s and Mourning warblers are ecologically similar in sympatry, but with some aspects of the niche divergence in allopatry, it is unlikely that ecological selection alone maintains isolation between the two species. Song was found to be convergent in the hybrid zone by Kenyon et al. (2011) and therefore it is unlikely to act as a premating barrier. However, the two species are known to take divergent migratory routes between their breeding and wintering grounds (Rohwer and Irwin 2011), and migratory behaviour is known to be genetically-based (Helbig 1991). Thus, hybrids may take an intermediate and unfavourable migratory route. Hybrids have been shown to take intermediate migratory routes or use migratory strategies that differ from parental subspecies of Swainson’s thrush that meet at a migratory divide (Delmore and Irwin 2014). These alternative strategies and routes used by hybrids have been shown to be associated with genomic areas under selection, suggesting divergent migratory behaviour of hybrids may indeed be an important isolating mechanism (Delmore et al. 2015). Future work on    62  determining the reproductive barriers between the MacGillivray’s and Mourning warblers should look at the role of divergent migratory behaviour as a post-mating isolating mechanism. 3.2.2 Future work on the role of ecology in avian speciation Hybridization in birds is common, and many pairs of sister species of North American birds have similar west-east distribution patterns and come into contact and form stable hybrid zones in northeastern British Columbia, similar to the MacGillivray’s and Mourning warblers (Toews and Irwin 2008; Brelsford and Irwin 2009; Irwin et al. 2009; Toews et al. 2011; Seneviratne et al. 2012). The role of ecology in maintaining these other hybrid zones has not yet been studied. Investigating ecology’s role in these other systems could reveal whether the patterns observed in the MacGillivray’s and Mourning warbler system are general of pairs of sister species with similar current distribution and hybridization patterns, as well as sister species that diverged under comparable conditions during the Pleistocene (Weir and Schluter 2004). This could provide empirical evidence for the idea that these species had similar patterns of evolution to adapt to environmental conditions. Given that I found the MacGillivray’s and Mourning warblers are ecologically similar, it may also indicate that sources of reproductive barriers such as sexual selection, migratory behaviour, or intrinsic incompatibilities play a larger role in the evolution of reproductive isolation than habitat or climate in these other North American sister species.      63  References Adamík, P., and S. Bureš. 2007. Experimental evidence for species-specific habitat preferences in two flycatcher species in their hybrid zone. Naturwissenschaften 94:859-863. Allouche, O., Tsoar, A., and R. Kadmon. 2006. Assessing the accuracy of species distribution models: prevalence, kappa and the true skill statistic (TSS). Journal of applied ecology 43:1223-1232. Anciães, M., and A. T. Peterson. 2009. Ecological niches and their evolution among Neotropical manakins (Aves: Pipridae). Journal of Avian Biology 40:591-604.  Araújo, M. B., Pearson, R. G., Thuiller, W., and M. Erhard. 2005. Validation of species–climate impact models under climate change. Global Change Biology 11:1504-1513. Avise, J. C., Walker, D., and G. C. Johns. 1998. Speciation durations and Pleistocene effects on vertebrate phylogeography. Proceedings of the Royal Society of London B: Biological Sciences 265:1707-1712. Barton, N. H., and G. M. Hewitt. 1985. Analysis of hybrid zones. Annual Review of Ecology and Systematics 16:113-148. Barton, N. H., and G. M. Hewitt. 1989. Adaptation, speciation and hybrid zones. Nature 341: 497-503. Bauer, S., Van Dinther, M., Høgda, K.-A., Klaassen, M., and J. Madsen. 2008. The consequences of climate-driven stop-over sites changes on migration schedules and fitness of Arctic geese. Journal of Animal Ecology 77:654-660.  Bensch, S., and D. Hasselquist. 1992. Evidence for active female choice in a polygynous warbler. Animal Behaviour 44:301-311. Brelsford, A., and D. E. Irwin. 2009. Incipient speciation despite little assortative mating: the Yellow-rumped warbler hybrid zone. Evolution 63:3050-3060.    64  Broennimann, O., Fitzpatrick, M. C., Pearman, P. B., Petitpierre, B., Pellissier, L., Yoccoz, N. G., Thuiller, W., Fortin, M.-J., Randin, C., Zimmerman, N. E., Graham, C. H., and A. Guisan. 2012. Measuring ecological niche overlap from occurrence and spatial environmental data. Global Ecology and Biogeography 21:481-497. Bronson, C. L., Grubb, T. C., and M. J. Braun. 2003. A test of the endogenous and exogenous selection hypotheses for the maintenance of a narrow avian hybrid zone. Evolution 57:630-637. Cadena, C. D., Kozak, K. H., Gómez, J. P., Parra, J. L., McCain, C. M., Bowie, R. C. K., Carnaval, A. C., Moritz, C., Rahbek, C., Roberts, T. E., Sanders, N. J., Schneider, C. J., VanDerWal, J., Zamudio, K. R., and C. H. Graham. 2012. Latitude, elevational climatic zonation and speciation in New World vertebrates. Proceedings of the Royal Society of London. Series B: Biological Sciences 279:194-201.  Chatfield, M. W. H., Kozak, K. H., Fitzpatrick, B. M., and P. K. Tucker. 2010. Patterns of differential introgression in a salamander hybrid zone: inferences from genetic data and ecological niche modelling. Molecular Ecology 19:4265-4282. Cicero, C.  2004. Barriers to sympatry between avian sibling species (Paridae: Baeolophus) in local secondary contact. Evolution 58:1573-1587. Commission for Environmental Cooperation. 2009. Ecological Regions of North America. Commission for Environmental Cooperation, Montreal, Canada. Cooper, S. J. 2002. Seasonal metabolic acclimatization in mountain chickadees and juniper titmice. Physiological and Biochemical Zoology 75:386-395. Coyne, J.A., and H.A. Orr, H.A. 2004. Speciation. Sinauer Associates, Sunderland, MA.  Culumber, Z. W., Shepard, D. B., Coleman, S. W., Rosenthal, G. G., and M. Tobler. 2012. Physiological adaptation along environmental gradients and replicated hybrid zone structure in swordtails (Teleostei: Xiphophorus). Journal of evolutionary biology 25:1800-1814.    65  Danecek, P., Auton, A., Abecasis, G., Albers, C. A., Banks, E., DePristo, M. A., Handsaker, R., Lunter, G., Marth, G., Sherry, S. T., McVean, G., and R. Durbin. 2011. The variant call format VCFtools. Bioinformatics 27:2156-2158.  De La Torre, A. R., Roberts, D. R., and S. N. Aitken. 2014. Genome‐wide admixture and ecological niche modelling reveal the maintenance of species boundaries despite long history of interspecific gene flow. Molecular ecology 23:2046-2059. Delmore, K. E., Brenneman, R.A., Lei, R., Bailey, C. A., Brelsford, A., Louis, E. E., and S. E. Johnson. 2013. Clinal variation in a brown lemur (Eulemur spp.) hybrid zone: Combining morphological, genetic and climatic data to examine stability. Journal of evolutionary biology 26:1677-1690. Delmore, K. E., and D. E Irwin. 2014. Hybrid songbirds employ intermediate routes in a migratory divide. Ecology Letters 17:1211-1218. Delmore, K. E., Hübner, S., Kane, N. C., Schuster, R., Andrew, R. L., Câmara, F., Guigo, R., and D. E. Irwin. 2015. Genomic analysis of a migratory divide reveals candidate genes for migration and implicates selective sweeps in generating islands of differentiation. Molecular Ecology 24:1873-1888. Demarchi, D.A. 2011. 2011. Ecosystem Information Section Third Ed. Ministry of Environment, Vancouver, British Columbia.  DePristo, M. A., Banks, E., Poplin, R., Garimella, K. V., Maguire, J. R., Hartl, C., Philippakis, A. A., del Angelo, G., Rivas, M. A., Hanna, M., McKenna, A., Fennell, T. J., Kernytsky, A. M., Sivachenko, A. Y., Cibulskis, K., Gabriel, S. B., Altshuler, D., and M. J. Daly. 2011. A Framework for variation discovery and genotyping using next-generation Dna sequencing data. Nature Genetics 43:491-498. Dwyer, M. J., and G. Schmidt. 2006. The MODIS reprojection tool. In: Earth science satellite remote sensing (Qu, J. J., Gao, W., Kaftos, M., Murphy, R. E., and V. V Salmonson, eds), pp. 162-177. Springer, Berlin Heidelberg.     66  Edwards, S. V., Kingan, S. B., Calkins, J. D., Balakrishnan, C. N., Jennings, W. B., Swanson, W. J., and M. D. Sorenson. 2005. Speciation in birds: genes, geography, and sexual selection. Proceedings of the National Academy of Sciences 102:6550-6557. Elith J., Graham C. H., Anderson R.P., Dudík M., Ferrier S., Guisan, A., Hijmans, R. J., Huettmann, F., Leathwick, J. R., Lehmann, A., Li, J., Lohmann, L. G., Loiselle, B. A., Manion, G., Moritz, C., Nakamura, M., Nakazawa, Y., McC. Overton, J., Peterson, A. T., Phillips, S. J., Richardson, K., Scachetti-Pereira, R., Schapire, R. E., Soberón, J., Williams, S., Wisz, M. S., and N. E. Zimmermann. 2006. Novel methods improve prediction of species’ distributions from occurrence data. Ecography 29:129–51 Elshire, R. J., Glaubitz, J. C., Sun, Q., Poland, J. A., Kawamoto, K., Buckler, E. S., and S. E. Mitchell. 2011. A robust, simple genotype-by-sequencing (GBS) approach for high diversity species. PloS ONE 6:e19379. Finch, D. M. 1989. Habitat use and habitat overlap of riparian bird sin three elevational zones. Ecology 70:866-880. Forchhammer, M. C., Port, E., and N. C. Stenseth. 2002. North Atlantic Oscillation timing of long- and short-distance migration. Journal of Animal Ecology 71:1002-1014. Glor, R. E., and D. Warren. 2010. Testing ecological explanations for biogeographic boundaries. Evolution 65:673-683. Good, T. P., Ellis, J. C., Annett, C. A., and R. Pierotti. 2000. Bounded hybrid superiority in an avian hybrid zone: effects of mate, diet, and habitat choice. Evolution 54:1774-1783. Gordo, O. 2007. Why are bird migration dates shifting? A review of weather and climate effects on avian migratory phenology. Climate Research 35:37-38. Hawkins, B. A., Diniz‐Filho, J. A. F., Jaramillo, C. A., and S. A. Soeller. 2006. Post‐Eocene climate change, niche conservatism, and the latitudinal diversity gradient of New World birds. Journal of Biogeography 33:770-780.    67  Hayworth, A. M., and W. W. Weathers. 1984. Temperature regulation and climatic adaptation in black-billed and yellow-billed magpies. Condor 86:19-26. Helbig, A. J. 1991. Inheritance of migratory direction in a bird species: a cross-breeding experiment with SE-and SW-migrating blackcaps (Sylvia atricapilla). Behavioral Ecology and Sociobiology 28:9-12. Hewitt, G. M. 1988. Hybrid zones-natural laboratories for evolutionary studies. Trends in Ecology & Evolution 3:158-167. Hewitt, G. M. 2004. Genetic consequences of climatic oscillations in the Quaternary. Philosophical Transactions of the Royal Society of London B: Biological Sciences 359:183-195. Hijmans, R. J., Cameron, S. E., Parra, J. L., Jones, P. G., and A. Jarvis. 2005. Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology 25:1965-1978. Hüppop, O., and K. Hüppop. 2003. North Atlantic Oscillation and timing of spring migration in birds. Proceedings of the Royal Society of London Series B- Biological Sciences 270:233-240.  Hurlbert, A. H., and Z. Liang. 2012. Spatiotemporal variation in avian migration phenology: citizen science reveals effects of climate change. PloS ONE 0.1371/journal.pone.0031662. Irwin, D. E. 2000. Song variation in an avian ring species. Evolution 54:998-1010.  Irwin, D. E., Bensch, S., and T. D. Price. 2001. Speciation in a ring. Nature 409:333-337. Irwin, D. E., Brelsford, A., Toews, D. P. L., MacDonald, C., and M. Phinney. 2009. Extensive hybridization in a contact zone between MacGillivray’s warblers Oporornis tolmiei and Mourning warblers O. philadelphia detected using molecular and morphological analyses. Journal of Avian Biology 40:539-552.  Johannesson, K. 2001. Parallel speciation: a key to sympatric divergence. Trends in Ecology & Evolution 16:148-153. Johnson, M. D. 2007. Measuring habitat quality: a review. The Condor109:489-504.    68  Kenyon, H. L., Toews, D. P. L., and D. E. Irwin. 2011. Can song discriminate between MacGillivray’s and Mourning warblers in a narrow hybrid zone? Condor 113:655-663. Kenyon, H. L. 2013. Song in a hybrid zone between Townsend’s (Setophaga townsendi) and black-throated green (S. virens) warblers. MSc Thesis, University of British Columbia, Vancouver, BC.  Kozak, K. H., and J. Wiens. 2006. Does niche conservatism promote speciation? A case study in North American salamanders. Evolution 60:2604-2621. Langmead, B., and S. Salzberg. 2012. Fast gapped-read alignment with Bowtie 2. Nature Methods 9:357-359. Lindgreen, S. 2012. AdapterRemoval: easy cleaning of next-generation sequencing reads. BMC Research Notes 5:337. Lobo, J. M., Jiménez-Valverde, A., and R. Real. 2008. AUC: a misleading measure of the performance of predictive distribution models. Global Ecology and Biogeography 17:145-151. Lovette, I. J., and W. M. Hochachka. 2006. Simultaneous effects of phylogenetic niche conservatism and competition on avian community structure. Ecology 87:S14-S28. MacArthur, R. H. 1958. Population ecology of some warblers of northeastern coniferous forests. Ecology 39:599-619. Marra, P. P., Hobson, K. A., and R. T. Holmes. 1998. Linking winter and summer events in a migratory bird by using stable-carbon isotopes. Science 282-1884-1886.  Marra, P. P., Francs, C. M., Mulvihill, R. S., and F. R. Moore. 2005. The influence of climate on the timing and rate of spring bird migration. Oecologia 142:307-315.  Martin, T.E., Paine, C. R., Conway, C. J., Hochachka, W. M., Allen, P., and W. Jenkins. 1997. BBIRD field protocol. Unpublished report, Montana Cooperative Wildlife Research Unit, University of Montana, Missoula.     69  Martínez-Meyer, E., and A. T. Peterson. 2006. Conservatism of ecological niche characteristics in North American plant species over the Pleistocene-to-Recent transition. Journal of Biogeography 33:1779-1789. McCormack, J. E., Zellmer, A. J., and L. L. Knowles. 2010. Does niche divergence accompany allopatric divergence in Aphelocoma jays as predicted under ecological speciation?: insights from tests with niche models. Evolution 64:1231-1244. Moore, W. S. 1977. An evaluation of narrow hybrid zones in vertebrates. Quarterly Review of Biology 52:263-277. Moore, W. S., and W. D. Koenig. 1986. Comparative reproductive success of yellow-shafted, red-shafted, and hybrid flickers across a hybrid zone. The Auk 103:42-51. Moore, W. S. and J. T. Price. 1993. Nature of selection in the northern flicker hybrid zone and its implications for speciation theory. In: Hybrid Zones and the Evolutionary Process (R.G. Harrison, ed.), pp. 196–225. Oxford University Press, New York.  Norris, D. R., Marra, P. P., Kyser, T. K., Sherry, T. W., and L. M. Ratcliffe. 2004. Tropical winter habitat limits reproductive success on the temperate breeding grounds in a migratory bird. Proceedings of the Royal Society of London Series B- Biological Sciences 271:59-64. Peterson, A. T., Soberón, J., and V. Sánchez-Cordero. 1999. Conservatism of ecological niches in evolutionary time. Science 285:1265-1267. Peterson, A. T. 2001. Predicting species' geographic distributions based on ecological niche modeling. The Condor 103:599-605. Peterson, A.T., and R.D. Holt. 2003. Niche differentiation in Mexican birds: using point occurrences to detect ecological innovation. Ecology Letters 6:774-782. Peterson, A. T., and Á. S. Nyári. 2008. Ecological niche conservatism and Pleistocene refugia in the thrush-like mourner, Schiffornis sp., in the neotropics. Evolution 62:173-183.    70  Peterson, A. T. 2011. Ecological niche conservatism: a time-structured review of evidence. Journal of Biogeoggraphy 38:817-827. Phillips, S. J., Anderson, R. P., and R. E. Schapire. 2006. Maximum entropy modeling of species geographic distributions. Ecological Modelling 190:231–259. Phillips, S. J., and M. Dudík. 2008. Modeling of species distributions with Maxent: new extensions and a comprehensive evaluation. Ecography 31: 161-175. Pitocchelli, J. 2011. Mourning Warbler (Geothlypis philadelphia), The Birds of North America Online (A. Poole, Ed.). Ithaca: Cornell Lab of Ornithology; Retrieved from the Birds of North America Online: http://bna.birds.cornell.edu/bna/species/072. Pitocchelli, J. 2013. MacGillivray's Warbler (Geothlypis tolmiei), The Birds of North America Online (A. Poole, Ed.). Ithaca: Cornell Lab of Ornithology; Retrieved from the Birds of North America Online: http://bna.birds.cornell.edu/bna/species/159. Pritchard, J. K., Stephens, M., and P. Donnelly. 2000. Inference of population structure using multilocus genotype data. Genetics 155:945-959. Pyron, R. A., and F. T. Burbrink. 2009. Lineage diversification in a widespread spcies: roles for niche divergence and conservatism in the common kingsnake, Lampropeltis getula. Molecular Ecology 18:3443-3457. R Core Team. 2012. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, http://www.R-project.org/. Remington, C. L. 1968. Suture-zones of hybrid interaction between recently joined biotas. In: Evolutionary biology, pp. 321-428. Springer US. Rice, W. R. 1984. Disruptive selection on habitat preference and the evolution of reproductive isolation: a simulation study. Evolution 38:1251-1260.    71  Ridgely, R. S., Allnutt, T. F., Brooks, T., McNicol, D. K., Mehlman, D. W., Young, B. E., and J. R. Zook. 2003. Digital Distribution Maps of the Birds of the Western Hemisphere, version 1.0. NatureServe, Arlington, Virginia, USA.  Rieseberg, L. H, Archer, M., and R. K. Wayne. 1999. Transgressive segregation, adaptation and speciation. Heredity 83:363-372. Rising, J. D. 1969. A comparison of metabolism and evaporative water loss of Baltimore and Bullock orioles. Comparative Biochemistry and Physiology 31:915-925. Rohwer, S., and D. E Irwin. 2011. Molt, orientation, and avian speciation. The Auk 128:419-425. Root, T. 1988. Energy constraints on avian distributions and abundances. Ecology 69:330-339. Sætre, G. P., and S. A. Saether. 2010. Ecology and genetics of speciation in Ficedula flycatchers. Molecular Ecology 19:1091-1106. Schluter, D. 2001. Ecology and the origin of species. Trends in Ecology & Evolution 16:372-380. Seneviratne, S. S., Toews, D. P. L., Brelsford, A., and D. E. Irwin. 2012. Concordance of genetic and phenotypic characters across a sapsucker hybrid zone. Journal of Avian Biology 43:119-130. Sodhi, N. S., and C. A. Paszkowski. 1995. Habitat use and foraging behavior of four Parulid warblers in a second-growth forest. Journal of field ornithology 66:277-288.  Swenson, N. G. 2006. Gis‐based niche models reveal unifying climatic mechanisms that maintain the location of avian hybrid zones in a North American suture zone. Journal of Evolutionary Biology 19:717-725. Swenson, N. G. 2008. The past and future influence of geographic information systems on hybrid zone, phylogeographic and speciation research. Journal of Evolutionary Biology 21:421-434.    72  Tarroso, P., Pereira, R. J., Martínez‐Freiría, F., Godinho, R., and J. C. Brito. 2014. Hybridization at an ecotone: ecological and genetic barriers between three Iberian vipers. Molecular ecology 23:1108-1123. Toews, D. P. L., and D. E. Irwin. 2008. Cryptic speciation in a Holarctic passerine revealed by genetic and bioacoustics analyses. Molecular Ecology 17:2691-2705. Toews, D. P. L., Brelsford, A., and D. E. Irwin. 2011. Hybridization between Townsend’s Dendroica townsendii and black-throated green warblers D. virens in an avian suture zone. Journal of Avian Biology 42:434-446. Warren, D. L., Glor, R. E., and M. Turelli. 2008. Environmental niche equivalency versus conservatism: quantitative approaches to niche evolution. Evolution 62:2868-2883. Warren, D. L., Glor, R. E., and M. Turelli. 2010. ENMTools: a toolbox for comparative studies of environmental niche models. Ecography 33:607-611. Warren, D. L., Cardillo, M., Rosauer, D. F., and D. I Bolnick. 2014. Mistaking geography for biology: inferring processes from species distributions. Trends in Ecology and Evolution 29:572-580. Warren, W. C., Clayton, D. F., Ellegren, H., Arnold, A. P., Hillier, L. W., et al. 2010. The genome of a songbird. Nature 464:757-762. Weathers, W. W., and E. Greene. 1998. Thermoregulatory responses of bridled and juniper titmice to high temperature. Condor 100:365-372. Weir, J. T., and D. Schluter. 2004. Ice sheets promote speciation in boreal birds. Proceedings of the Royal Society B: Biological Sciences 271:1881-1887.  Wellenreuther, M., Larson, K. W., and E. I. Svensson. 2012. Climatic niche divergence or conservatism? Environmental niches and range limits in ecologically similar damselflies. Ecology 93:1353-1366.    73  Wiens, J. J. 2004. Speciation and ecology revisited: phylogenetic niche conservatism and the origin of species. Evolution 58:193-197. Wiens, J. J., and C. H. Graham. 2005. Niche conservatism: integrating evolution, ecology, and conservation biology. Annual Review of Ecology, Evolution, and Systematics 36:519-539. Wiens, J. J., Ackerly, D. D., Allen, A. P., Anacker, B. L., Buckley, L. B., Cornell, H. V., Damschen, E. I., Davies, T. J., Grytnes, J.-A., Harrison, S. P., Hawkins, B. A., Holt, R. D., McCain, C. M., Stephens, P. R. 2010. Niche conservatism as an emerging principle in ecology and conservation biology. Ecology Letter 13:1310-1324.  Zheng, X., Levine, D., Shen, J., Gogarten, S. M., Laurie, C., and B. S. Weir. A High-performance Computing Toolset for Relatedness and Principal Component Analysis of SNP Data Bioinformatics 2012; doi: 10.1093/bioinformatics/bts606. Zink, R. M. 2014. Homage to Hutchinson, and the role of ecology in lineage divergence and speciation. Journal of Biogeography 41: 999-1006.      74  Appendices Appendix A:  Sample preparation and data collection We used a genotype-by-sequencing method to determine the genetic background of birds sampled for microhabitat analyses. We also included some birds previously sampled by Irwin et al. (2009) and from the Chicago Field Museum that breed in allopatric regions far from the hybrid zone to aid in delineating the genotypes of both parental species. DNA was extracted from blood and tissue samples using a standard phenol-chloroform extraction technique. A genotype-by-sequencing (GBS) library was prepared using a modified version of the protocol by Elshire et al. (2011). Please see Elshire et al. (2011) for adapter and primer sequences. For each sample, 110ng of DNA were added to 5.8µl of common adaptors (0.4 ng/µl), 5.8µl of barcoded adaptors (0.4ng/µl), 1µl of Pst1-HF (20,000 units/ml) and 2 µl of 10x NEB4 buffer. After the digest (2 hours at 37°C), 1.6 µl T4 ligase (400,000 U/mL), 5 µl of the provided 10x buffer and 23.4µl ultrapure water were added to each sample and incubated at 16°C over night. After inactivation of the enzyme at 65°C for 10 minutes, 15 µl of ligation mixture were cleaned using 23µl of AMPure XP beads (Beckman-Coulter) and washed twice on a magnetic plate with 200µl of 70% ethanol. 40µl of TE were used for resuspension of the ligated DNA. A PCR was run on each sample in a 25µl reaction volume containing 2µl of the cleaned ligated DNA, 5µl of 5x Phusion Buffer, 0.5µl of 10mM dNTPs, 0.125µl of primer GBS-PrimerA (100µM) and GBS-PrimerB (100µM), 18µl of ultrapure water and 0.25µl of PhusionTaq. PCR cycling conditions were as follows: 98°C for 30 seconds followed by 20 cycles of 98°C for 10 seconds, 65°C for 30 seconds and 72°C for 30 seconds and a final elongation at 72°C for 5 minutes. After visual inspection on a 2.5% agarose gel, 5.2µl of each amplification reaction were pooled and run on a 1.5% agarose gel. A band in the size range 300-400bp was cut out and the final library was extracted using a gel DNA recovery kit (Zymoclean) and eluted into ultrapure water. The paired-end sequencing was performed on an Illumina HiSeq 2000.      75  Appendix B: Raw data analysis Quality trimming was performed using Trimmomatic 0.32 (www.usadellab.org/cms/index.php?page = trimmomatic ) with the settings TRAILING:3 SLIDINGWINDOW:4:10 MINLEN:50. To avoid non-independent information AdapterRemoval (version 1.5.2, Lindgreen 2012) was run on all paired reads for the removal of adapters from read-throughs and to collapse overlapping reads into single reads. Unpaired reads from Trimmomatic were concatenated with paired and collapsed reads from AdapterRemoval. Barcode splitting was performed using Sabre (https://github.com/najoshi/sabre). For the input barcode file we used a concatenation of the barcodes and the enzyme cut site and ran Sabre allowing one mismatch. The barcodes, which identified each individual, were variable in length ranging from 4 to 8bp. Demultiplexed reads were aligned to a build of the Zebra finch genome (Warren et al. 2010) with repetitive sequences masked (TaeGut3.2.4.75.dna_rm). For the read alignment we used bowtie2 (version 2.2.2, Langmead et al. 2012) with the settings of “very-sensitive-local”. UnifiedGenotyper (GATK, version 3.1-1, DePristo et al. 2011) was used for genotype calling before the following hard filters were applied using VariantFiltration of GATK (QD < 2.0, MQ < 30.0, -12.5 > MQRank-Sum > 12.5, FS > 40.0, HaplotypeScore > 12.0, ReadPosRank-Sum < -8.000, QUAL < 30.0, AN < 20). After visual inspection of the distribution of heterozygosity, a threshold of 0.8 was applied to avoid SNPs with very high heterozygosity. Further quality filtering was performed using VCFTools (Danecek et al. 2011). A minimal genotype quality (GQ) of 20 was required, as was a genotyping rate of 0.75 and a minor allele frequency per SNP of 0.05.  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0167279/manifest

Comment

Related Items