INVESTIGATING ECOLOGICAL NICHE DIFFERENTIATION AMONG

INVESTIGATING ECOLOGICAL NICHE DIFFERENTIATION AMONG WILD
CANIDS EXPERIENCING HYBRIDIZATION IN EASTERN NORTH AMERICA
A Thesis Submitted to the Committee on Graduate Studies
in Partial Fulfillment of the Requirements for the Degree of Master of Science
in the Faculty of Arts and Science
Trent University
Peterborough, Ontario, Canada
© Copyright by Josée-Anne Otis 2016
Environmental and Life Sciences M.Sc. Graduate Program
May 2016
Abstract
Investigating ecological niche differentiation among wild canids experiencing
hybridization in eastern North America
Josée-Anne Otis
Currently there are large areas of the North American landscape that are occupied by
Canis spp. hybrids of several varieties, leading to the logical question as to the genetic
structure and ecological function of Canis populations across the continent, and to what
extent hybrids reflect contemporary landscapes. This study illustrated patterns of niche
differentiation between parental canid species and their hybrids using individual high
quality genetic profile and species distribution models to support the intermediate
phenotype hypothesis. In general, hybrids demonstrated an intermediate habitat
suitability compared to its parental species, across most environmental variables used. A
similar trend was observed in the niche metric analysis, where we found that hybrids
exhibit intermediate niche breadth, with eastern coyotes and eastern wolves exhibiting the
broader and narrower niche, respectively. Our results demonstrate that the intermediate
phenotype hypothesis is supported even at a large scale and when involving highly
mobile large mammal species.
Keywords: hybridization, niche differentiation, intermediate phenotype , microsatellite
genotype, ecological niche modelling, canid, Canis lycaon, Canis latrans
ii
Acknowledgements
Je dédicace cette thèse à ma grande sœur Marie-Lison Otis, qui à mon âge n'avait pas
accouché d'une thèse, mais déjà de mes trois nièces adorées. Elle m'a traîné sur ses talons
dans les bois durant toute notre jeunesse et elle a très certainement contribué à mon
amour de la nature. Elle est un modèle à suivre sur plusieurs plans de ma vie, d’une très
grande écoute et a été ma fan #1 tout au long de mes études. Lison, j’espère être une aussi
bonne maman que toi lorsque la chance me sera donnée d’expérimenter la maternité. Je
souhaite aussi souligner le support de tous les membres de ma famille, qui ont toujours
cru en moi et qui ont su me ramener le sourire aux lèvres lorsque la pression académique
me montait à la tête. Merci !
The second person that I want to thank is my supervisor, Dr. Dennis Murray, who
gave me the chance to be involved in research through my own project and also within
his amazing Research Lab. I am glad that Dennis accepted the challenge of supervising
another “Frenchglish” writer (I guess my predecessors did a good job), I learned so much
during these few years. I also want to thank each of my Committtee members; Dr. Daniel
Thornton, Dr. Jeff Bowman and Dr. James Conolly, who have all contributed to the
accomplishment of this thesis by providing me helpful comments and knowledge within
their own areas of expertise. I special thanks to Dan who reviewed and commented all my
drafts before they reached the “big boss”.
I also want to say a big thank you to Dr. Linda Rutledge who helped me understand
the eastern wolf “story” and who has provided important recommendations and
comments on the genetic part of this project. I am grateful to Derek Walker and Nguyen
Thi Xuan for their incredible lab work, to Tracy Armstrong for her precious GIS help, to
iii
Erin Stewart Eves who provided me with all tools necessary to improve my English, to
the entire Murray’s Lab for their feedback and support, and especially to Michael Peers
for his detailed explanations to my multiple questions when I first joined the Lab.
From the bottom of my heart, for their invaluable help by reviewing some of my drafts
and their extended knowledge and experience in research and science, and especially for
their Friendship and amazing company through many outdoor adventures during my time
in Peterborough, I want to say thanks to Morgan Wehtje, Catarina Ferreira, Debbie
Jenkins, Philip Bertrand and Laura Maria Martinez. I also want to give a huge hug and
say thank you to the best officemate a grad student can dream of, to my dear P’tite ours,
Sarah Poole. You are precious friends that I hope to keep on my “side” forever and I hope
we will share new adventures soon! Finally, for all her kisses and endless positive energy,
I’m grateful to my dog, Fika, who has also slept at my feet during my long hours at the
computer.
Special thanks to all volunteer trappers from Quebec, especially Pierre-Yves Collin
who has coordinated most of the collect, Joseph Bopp at the NYSM, Vittorio Villacis at
NAFA, Andy MacDuff at the NYSDEC, and Kelly Leavesley at Manitoba CWS for
providing me with samples. Funding was provided by the Natural Sciences and
Engineering Research Council of Canada (NSERC), the Canada Research Chairs
program, the Fonds de Recherche Nature et Technologies du Québec (FQRNT), the
Foundation Desjardins scholarships, the Ontario Federation of Anglers and Hunters
(OFAH), and the David & Joyce Woods Graduate Scholarship program.
“Science is not only a disciple of reason but, also, one of romance and passion.”
~ Stephen Hawking (2010)
iv
Table of Contents
Abstract ............................................................................................................................... ii
Acknowledgements ............................................................................................................ iii
Table of Contents .................................................................................................................v
List of Figures .................................................................................................................... vi
List of Tables ................................................................................................................... viii
Chapter 1 General Introduction ...........................................................................................1
The rise of hybridization in a changing environment ......................................................2
The wolf-coyote system in eastern North America ..........................................................4
Taxonomic challenge and genetic assignment of wild canids .........................................6
The complex case of hybridization between eastern wolf and western coyote ...............7
Justification and relevance of the thesis ...........................................................................9
Chapter 2 Ecological niche differentiation across a wolf-coyote hybrid zone in
eastern North America. ......................................................................................................12
Abstract .........................................................................................................................13
Introduction ...................................................................................................................14
Methods .........................................................................................................................18
Results ...........................................................................................................................24
Discussion .....................................................................................................................28
Chapter 3 General Conclusions .........................................................................................39
Goals of the thesis revisited ...........................................................................................40
Future research directions ..............................................................................................42
Conclusion ......................................................................................................................44
Literature cited ...................................................................................................................45
Supporting information 1 ...................................................................................................55
Supporting information 2 ...................................................................................................56
Supporting information 3 ...................................................................................................64
v
List of Figures
Figure A. Conceptual model of hybridization dynamics of wild canids in eastern North
America. Eastern wolves, grey wolves, and western coyotes produce hybrids known as
Great Lakes boreal wolves and eastern coyotes. Direct hybridization is not observed
between grey wolves and western coyotes, but eastern wolves can interbreed with both
species and produce fertile offspring; through backcrossing, hybrids contribute to the
observed hybrid swarm........................................................................................................5
Figure 1. Geographic distribution of 648 wild canid samples from eastern North America
used in this study. Canids were classified in different groups using a Q-value threshold of
≥ 0.8, for a total of 110 eastern wolves, 487 eastern coyotes, and 51 hybrids; animals with
grey wolf ancestry are
excluded…………………………………………………………...…….……………….34
Figure 2. Habitat suitability models for wild canids in eastern North America, including
eastern wolves, eastern coyotes, and hybrids. Green and red represent areas with
predicted low and high suitability, respectively..…………………………………..........35
Figure 3. Habitat suitability response curves for individual environmental variables, when
occurring alone in models for eastern wolves, eastern coyotes and hybrid
canids...…...................................................................................................... ….........…..36
vi
Figure 4. Wild canid responses to individual land cover categories in univariate MaxEnt
models………………………………………………….…...………………..….........….37
Figure 5. Niche breadth of wild canids in relation to the background environment.
Histograms represent the expected niche breadth if an animal selects habitats at random
from the background, and the triangle indicates observed niche breadth of each of the
three canid groups. Note that the X-axis is presented at a different scale for each
panel.………......................................................................................................................38
vii
List of Tables
Table 1. Estimates of the relative contribution from environmental variables to habitat
suitability models for eastern wolves, eastern coyotes and hybrids..................................33
Table 2. Comparative niche overlap and test of background similarity between eastern
wolves, eastern coyotes and hybrids. I statistics are presented, and the outcome of the test
depends on the canid group used as observed vs. background……….........................….33
viii
Chapter 1
General Introduction
“No one definition has satisfied all naturalists; yet every naturalist knows vaguely what
he means when he speaks of a species”
~ Charles Darwin (1872)
1
General Introduction
The rise of hybridization in a changing environment
Ongoing and widespread anthropogenic modification of natural habitats has led to
clear trends in species decline and extirpation (Olden, 2006). More subtly, these changes
also have contributed to variability in species overlap and co-existence, such that species
formerly occupying distinct ranges that were separated by dispersal or reproductive
barriers now often are found in close proximity and sympatry (Olden, 2006, Crispo et al.,
2011). The broader outcome of these subtle changes has been an increase in hybridization
(i.e. reproduction between two genetically distinct groups (Harrison, 1993), hereafter
called “parental species”), often resulting in viable offspring (Barton & Hewitt, 1985;
Rhymer & Simberloff, 1996). It follows that hybridization can lead to increased habitat
and ecosystem homogeneity and a redistribution of species associations and positions
within the ecosystem (McKinney & Lockwood, 1999; Olden, 2006). Such changes have
had a profound influence on a variety of plant and animal systems (e.g., Ryan et al., 2009;
Neira et al., 2005), which has prompted a major conservation crisis surrounding the
question of hybridization, its species- and ecosystem-level effects, and whether future
changes in species composition and distribution can be forecasted based on current
understanding of hybridization dynamics (Allendorf et al., 2001; Seehausen, 2004;
Seehausen et al., 2007).
The conservation implications of hybridization are particularly challenging to address
and predict because the process of genetic introgression (i.e. “the incorporation of alien
genes into a new, reproductively integrated population system”, Rieseberg & Wendel,
1993) is highly variable and case-dependant. It is important to distinguish that while
2
hybridization can occur naturally across a range of plant and animal species (Hubbs,
1955; Grant & Grant, 1992; Ellstrand et al., 1996), it is the clear increase in hybridization
owing to anthropogenic change that is the major conservation concern (Allendorf et al.,
2001). Introgressive hybridization is characterized by haplotype mixing of nuclear and
mitochondrial genes between species (van Dongen et al., 2013), which is the first major
threat to the genetic integrity of a population (Mallet, 2005). However, there are also
potential advantages to genetic introgression via increased biodiversity through the
establishment of novel genotypes (Anderson & Stebbins, 1954; Arnold, 1992).
In theory, hybrids may closely reflect local adaptation (Anderson, 1948), such that
parental species may sometimes be the product of evolutionary lags and therefore lack
contemporary adaptation to changing environments (Lewontin & Birch, 1966; Baskett &
Gomulkiewicz, 2011). Instead, hybrids may be strongly selected in response to new or
recently-vacated niches (Anderson & Stebbins, 1954). Accordingly, phenotypic
characteristics of hybrids, which are often intermediate to those of parental species
(Rieseberg et al., 1999), may be especially linked to environments that are subject to
strong anthropogenic effects; these environments are often subject to habitat loss and
fragmentation or climate change. Yet, it is important to recognize that hybrids can be
quite variable in both their genetic composition and levels of genetic introgression, and
that further variability can arise through epigenetics or other sources of phenotypic
plasticity (Bullini, 1994; Landry et al., 2007). Therefore, revealing the ecological function
and significance of hybrids in contemporary landscapes represents an especially vexing
challenge requiring a thorough understanding of genetics, evolution, and ecology.
3
The wolf-coyote system in eastern North America
Initial attempts to differentiate between North American canids were based on
phenotypic observations of individuals (e.g., color pattern, size, or skull characteristics),
which led to the recognition of a variety of sub-species and ecotypes (Kolenosky &
Stanfield, 1975; Nowak, 1995). The recognition of large variability in wolf phenotypes
led to the identification of 26 races, including several that were thought to have resulted
from hybridization with coyotes (Young & Goldman, 1944). This hypothesis has been
supported by recent advances in genetic methods and techniques allowing phenotypic
variation to be investigated at a much finer scale than what was available previously (e.g.,
Grewal et al., 2004; Wilson et al., 2009; Rutledge et al., 2010a). Hybridization is possible
among several of the currently-recognized species of Canis spp. in North America; these
species include eastern wolves (Canis lycaon) which currently occur in southern Ontario
and Quebec, grey wolves (C. lupus), which are present across much of boreal Canada and
Alaska, red wolves (C. rufus) which occurred historically across the southeastern United
States but now are restricted in a recovery area in eastern North Carolina, and western
coyotes (C. latrans) which historically occupied the central plains of North America.
However, even if they come into contact, each canid species does not necessarily readily
hybridize because of their different evolutionary history (Wilson et al., 2000; 2009),
disparate mate selection patterns (Lehman et al., 1991), or any combination of these
factors. Yet, in more contemporary landscapes the eastern wolf has become the vector for
hybridization between coyote, grey wolf, and red wolf (Wheeldon & White, 2009; Wilson
et al., 2009), leading to hybrid zones across much of the North American continent.
Indeed, currently red wolf-coyote hybrids occur outside the North Carolina recovery area,
eastern wolf-coyote hybrids (eastern coyotes) are found across eastern North America,
4
and eastern wolf-grey wolf hybrids are present in the southern boreal forest and Great
Lakes area (Wheeldon & White, 2009). Within each of these populations there is
additional variability due to differential levels of backcrossing with parental species
(Wilson et al., 2009; Figure A). This means that currently there are large areas of the
North American landscape that are occupied by Canis spp. hybrids of several varieties,
leading to questions about the genetic structure and ecological function of Canis
populations across the continent, and about whether hybrids reflect contemporary
landscapes and the natural selection processes acting therein.
Figure A. Conceptual model of hybridization dynamics of wild canids in eastern North
America, based on the 3-species hypothesis. Eastern wolves, grey wolves, and western
coyotes produce hybrids known as Great Lakes boreal wolves and eastern coyotes. Direct
hybridization is not observed between grey wolves and western coyotes, but eastern
wolves can interbreed with both species and produce fertile offspring; through
backcrossing, hybrids contribute to the observed hybrid swarm.
5
Taxonomic challenge and genetic assignment of wild canids
The origin and taxonomy of North American canid are still highly debated, however,
the different opinions follow two general hypotheses; 2-species vs. 3-species hypothesis
(see Kyle et al., 2006 for a full review). The reasons behind this debate are based on the
fact that different molecular markers have been used by different research teams, and the
names given to the different canid groups have not been consistent across the literature,
leading to misunderstandings that have helped to feed the debate. Moreover, there are
also some genetic challenges encountered when studying introgressive hybridization,
such as the difficulty to distinguish incomplete lineage sorting (i.e. the persistence of
ancestral alleles in both populations) of introgression events (Pamilo & Nej, 1988;
Hudson & Turelli, 2003). These debates highlight the inevitable need to use advance
genetic techniques to assign eastern North American wild canid individuals to a specific
reference group. Different kinds of genetic markers are currently used for canid
assignment: 1) bi-parentally inherited nuclear autosomal microsatellites markers, 2)
maternally inherited marker (mtDNA), 3) Y-chromosome haplotypes, and most recently
4) restriction site associated DNA (RAD). The introgression of eastern wolf mtDNA and
Y-chromosomes into Great Lakes boreal wolves (Rutledge et al. 2010a) and eastern
coyotes (Way et al., 2010), as well as the potential incomplete lineage sorting of mtDNA
between eastern wolves and coyotes (Wheeldon &White, 2009), all for caution in the use
of these type of markers to base species’or group assignment when working with wild
canids. These discrepancies are one of the reasons why we decided to use 12
microsatellites markers in our analysis, given that RAD-sequencing was not sufficiently
affordable given the scope and extent of this study.
6
The most commonly used and recognized assignment technique for wild canids is
based on a Bayesian analytical framework using microsatellite markers (e.g., Veradi et
al., 2006; Benson et al., 2012; Wheeldon et al., 2013). This approach is implemented in
the software program Structure (Pritchard et al., 2000; Falush et al., 2003) that outputs a
proportional ancestry value for each individual based on a certain threshold value (Qvalue). In this study we used a threshold of Q≥0.8 to determine parental groups, which is
the same threshold used by COSEWIC for the recent assessment of the eastern wolf
(COSEWIC 2015). Therefore, we considered this threshold as relevant to our study
system. However, we concede the arbitrary nature of this threshold to investigate hybrid
status (Vänä & Primmer, 2006), however, this was the reason why we complemented our
analysis of population structure (using program Structure) with a principal components
analysis. The individual assignment using Structure is known to vary to a certain degree
depending of the threshold used, but Q≥0.8 is conservative and assignments are expected
to have ~10% variability based on a previous investigation done by Benson et al. (2012).
The complex case of hybridization between eastern wolf and western coyote
As implied above, historically coyotes and eastern wolves were divided spatially and
presumably occupied different niches. Before European settlement, the coyote was
distributed in the central part of North America, in the prairies and deserts. In contrast, the
eastern wolf was found in the eastern forests dominated by deciduous trees, and fed
mostly on white-tailed deer (Odocoileus virginianus) (Forbes & Theberge 1996; Wilson
et al., 2010; Kyle et al., 2006). During colonization and settlement, wolves were
effectively “extirpated” from most of the United States and southeastern Canada by
culling, hunting, and habitat destruction (Fritts et al., 2003). The western coyote
7
responded to the newly-vacated wolf niche by expanding its distribution to include
eastern North America (Lehman et al., 1991; Kays et al., 2010; Bozarth et al., 2011).
Coincident loss and fragmentation of much of the eastern forest and replacement with
agricultural habitat also contributed to coyote expansion. At the beginning of the 19th
century, a population of eastern wolves was relegated to the current boundaries of
Algonquin Provincial Park, Ontario. There are two recent hypothesis concerning coyotes’
colonization route, according to Kays et al. (2010), coyotes dispersed into Ontario
through both northern and southern corridors around the Great Lakes, although the
alternate hypothesis is that coyotes arrived in Canada through Lake Huron and Lake Erie
(Wheeldon et al., 2010). Regardless of their dispersal pathway, around mid-1900s
coyotes first came into contact with a small and fragmented eastern wolf population in
eastern Canada with which it hybridized and gave rise to eastern coyotes, otherwise
known as ‘Tweed wolf’ or ‘coywolf’ (Kays et al., 2010; Way et al., 2010).
Way (2007) found an increasing longitudinal size gradient among coyotes from
western to eastern North America, suggesting two possible processes promoting this
phenotypic response: i) a result of selection imposed by increasing reliance on whitetailed deer as prey and therefore the need for larger body size to successfully capture this
novel food source (Crête et al., 2001), or ii) hybridization with eastern wolves. Further
evidence from the analysis of genotypes and morphology (cranial and mandibular) of
coyotes in eastern United States suggests that coyote expansion has been largely driven
by hybridization with eastern wolves. Phenotypically, in eastern North America hybrids
(i.e. eastern coyote) are intermediate in size compared to parental species (Kays et al.,
2008; Way et al., 2010; Benson et al., 2012) and appear to respond favourably to
8
anthropogenic modification of natural forested landscapes into broken forests, agricultural
fields, and the rural-suburban interface (Gompper, 2002; Kays et al., 2008; Grubbs &
Krausman, 2009).
Benson et al. (2012) addressed these questions in greater detail by analyzing the
relation between genotype and morphology of the different Canis types present in
Algonquin Provincial Park and beside. They identified eastern wolves as being heavier
and having longer bodies compared to eastern coyotes and eastern coyote-eastern wolf
hybrids. Likewise, this study also revealed a qualitative pattern of body mass change
across a larger spatial scale ranging from that of the Great Lakes boreal wolf to the
eastern wolf to the eastern coyote, with their related hybrids having intermediate body
mass relative to their parental species. This gradient may very well imply that body mass
is a direct reflection of local adaptation by canids to contemporary landscape (Larivière &
Crête, 1993). However, the Benson et al. (2012) analysis was conducted at a spatial
extent that was too small to capture evolutionary processes and landscape-level niche
differentiation. Indeed, to date investigations into the hybridization dynamics of North
American Canis have not included assessment of other aspects of their niche occupancy,
including habitat characteristics and the climate features that influence habitat.
Justification and relevance of the thesis
An important feature of wolf-coyote hybridization in North America is that it is driven
proximally by anthropogenic habitat change following European settlement (Wilson et
al., 2009; Stronen et al., 2012). Therefore, this phenomenon is not considered as a natural
process. On one hand, because some eastern wolf or red wolf populations are very small
and occupy restricted spaces, it may be that hybridization with coyotes functions as a
9
logical evolutionary step (Kyle et al., 2006). Alternatively, hybrids may never truly
replace wolves in their ecological niche (Richens & Hugie, 1974; Crête et al., 2001; Kays
et al., 2008), so genetic introgression from coyotes may actually be deleterious to
conservation of the wolf genome and its corresponding function as top predator.
Moreover, eastern coyote range expansion is ongoing and there is limited understanding
of its potential consequences for ecosystems. For instance, re-establishment of a top
Canis predator can greatly affect the ecosystem, potentially reverberating to the function
and dynamics of resident coyotes (Hebblewhite & Smith, 2010). However, because the
ecological role of the eastern wolf is so poorly known at the scale of its distribution in
eastern North America, it is difficult to predict whether displacement from eastern
coyotes or eastern wolf – eastern coyote hybrids is a relevant consideration. Yet, because
eastern wolves are currently listed as threatened by the Committee on the Status of
Endangered Wildlife in Canada (COSEWIC, 2015), the potential threat posed by eastern
coyote range expansion, the hybridization with and the displacement of remaining eastern
wolves is a significant conservation concern.
While not ignoring or discrediting the functional importance of the ongoing taxonomic
debate regarding the status of North American Canis (for full taxonomic review see
Chambers et al., 2012), the present study focused on revealing patterns of niche overlap
and niche differentiation between eastern wolves, eastern coyotes, and their hybrids.
Specifically, this study uses geospatial and genetic data to investigate the extent of
differentiation between the three Canis groups in terms of their landscape and habitat
features. This system is ideal for studying niche shifts and hybridization in mammals, and
is a logical step toward developing an effective conservation strategy for eastern wolves.
10
Therefore, our main objectives are to: 1) describe niche ecology and niche metrics of the
eastern wolf, eastern coyote and their hybrids, 2) assess whether hybridization has led to
differential niche ecology of hybrids, and 3) assess whether hybrids can potentially
displace eastern wolves. These efforts may further serve as a basis for: i) evaluating
potential areas for eastern wolf (or red wolf) reintroduction in North America, and ii)
assessing potential for habitat restoration to promote viability in extant eastern wolf
populations that are facing ingress from eastern coyotes and hybrids.
11
Chapter 2
Ecological niche differentiation across a wolf-coyote hybrid zone
in eastern North America.
“The hope of the future lies not in curbing the influence of human occupancy – it is
already too late for that – but in creating a better understanding of the extent of that
influence and a new ethic for its governance.”
Aldo Leopold (1986)
12
Abstract
Anthropogenic activity in the form of habitat alteration or disturbance has had a
profound effect on the distribution, abundance, and interactions of most species occurring
in natural landscapes, including hybridization. Hybridization can alter a variety of species
attributes, including niche dynamics, which fundamentally determine the role of a given
species or hybrid in the ecosystem. Notwithstanding the potentially important role of
hybridization on ecological communities, there is a gap in our understanding of whether
hybrids retain the basic niche characteristics of their parental types, especially for large
carnivores that have high behavioural plasticity in their niche characteristics. We sought
to test whether eastern wolf (Canis lycaon) – coyote (C. latrans) hybrids exhibited
intermediate habitat niche characteristics compared to their progenitors. Species
distribution models revealed clear patterns of niche differentiation, with hybrids tending
to be found in areas having intermediate environmental niche attributes (e.g., maximum
temperature, road density and human density) and niche breadth, compared to parental
groups. Niche overlap between hybrids and either parental groups was greater than
between the two parental groups, further suggesting an intermediate environmental niche
of hybrids. Our results indicate that even among widely-ranging and highly plastic wild
mammals, hybrids tend to exhibit intermediate niche characteristics. Because the
intermediate phenotype hypothesis was supported using wild canids, which exhibit a
particularly high degree of behavioural plasticity, our results elevate the hypothesis as
being observable even using coarse data at a large spatial scale.
13
Introduction
Current patterns of environmental change are responsible for dramatically altering the
distribution and abundance of organisms, while also varying the structure and function of
ecosystems (Crispo et al., 2011). Human-caused changes in landscape and habitat
attributes are occurring at a pace that far exceeds natural rates of environmental change,
leading to a variety of consequences including increased contact between closely related
species that were formerly allopatric and reproductively isolated (Stockwell et al.,2003;
Olden, 2006). Hybridization is one possible outcome of novel species contact, and
hybrids are widely reported for plant and animal species that previously were not known
to interbreed (Huxel, 1999; Seehausen et al., 2007). Hybrids often exhibit changes in
morphology or behaviour compared to parental species, and these changes ultimately
determine the niche space occupied by the hybrid individual (Rieseberg et al., 2003;
Ellington & Murray, 2015). It follows that hybridization tends to cause especially
complex changes in niche characteristics when hybrid offspring are fertile and there is
further genetic introgression between hybrids and parental groups (i.e. backcrossing),
potentially leading to a wide range of genotypes and phenotypes in a given population.
Without selection against hybrids or hybrid speciation, an hybrid swarm can arise, which
is a widespread introgression that can lead to the complete homogenisation of the parental
species genomes (Allendorf et al., 2001), which is a direct lost of biodiversity.
Increasingly, it is understood that while hybrids can have lower fitness when occurring in
the same environmental space as parental groups, in human-altered landscapes the
availability of new or vacant niches may favour hybrid survival and productivity (Barton,
2001; Seehausen, 2004). However, the expected selective advantage of hybrids in novel
14
environments (Rieseberg et al., 2003; Seehaussen, 2004) has been markedly difficult to
demonstrate empirically, in large part because environments favoring hybrids can be
especially complex and rapidly-changing, and hybrid success is influenced by both
adaptive and non-adaptive changes (Arnold, 1997; Stone, 2000). Under some conditions
hybrids can even out-perform parental species, leading to a range of possibilities in the
distribution and abundance of hybrids and parentals across the landscape (see Rhymer &
Simberloff, 1996; Ellstrand & Schierenbeck, 2000).
Although competition between newly sympatric and hybridizing species has been
fairly well investigated (e.g., Fitzpatrick & Shaffer, 2007; van Dongen et al., 2013), the
potential for hybridization to shift community dynamics has received considerably less
attention (but see Neira et al., 2005; Ryan et al., 2009). For example, one way that
hybridization can affect species interactions is through changes in niche ecology, but in
general the prevalence and extent of shifts in space, habitat or diet among hybrids and
parental groups is not well known. It follows that understanding how hybridization
impacts individual niche characteristics should be prioritized to properly predict
persistence of different genotypes/phenotypes across niche space (Howard et al., 1993;
Seehausen et al., 2007).
In this paper, we combined genetic and spatial analysis to investigate the ecological
niche of wild canids in eastern North America. Eastern wolves (Canis lycaon) hybridize
with coyotes (Canis latrans sp.), leading to fertile offspring and a range of Canis
genotypes/phenotypes across the eastern continental landscape (Lehman et al., 1991;
Wilson et al., 2009; Kays et al., 2010). Historically, wolf-coyote hybridization was
facilitated by human-caused decline in wolf numbers in eastern North America, combined
15
with alteration of natural forests to more open habitats; both changes favoured
colonization by coyotes and subsequent interbreeding with eastern wolves (Gompper,
2002; Murray & Waits, 2007). Contrary to coyotes, eastern wolves prefer forested habitat
and are more sensitive to altered landscapes (Benson et al., 2012). Consequently, the
distribution of coyote-like canids (hereafter ‘eastern coyotes’) expanded across most of
eastern North America while the distribution of eastern wolves became largely restricted
to central Canada (Kyle et al., 2006). Way (2007) found an increasing longitudinal size
gradient among coyotes from western to eastern North America, implying a cline in the
degree of hybridization and genetic introgression; evidence from genetic and
morphological analyses supports that coyote expansion was largely driven by
hybridization with wolves (Kays et al., 2010). To this day, eastern wolves, eastern
coyotes, and their relatives continue to interbreed, leading to an increasing diversity of
canid groups across the landscape of eastern North America.
Using spatial distribution of canids in eastern North America, we assessed whether
patterns of occurrence and habitat use conformed to the intermediate phenotype
hypothesis (Anderson, 1948), which relates the broader patterns of occupancy among
hybrid individuals compared to those of parental groups (see also Heaney & Timm, 1985;
Tauleigne-Gomes & Lefèbvre, 2008). Specifically, we predicted that: 1) hybrid canids
(i.e. individuals with a high degree of genetic admixture compared to eastern wolf or
eastern coyote parental groups) occupy an intermediate habitat niche compared to
parental groups; 2) hybrid niche overlap with both parental groups whereas niches of
parental groups exhibit limited overlap; and 3) hybrids have a broader niche compared to
parental groups. Because our study also allowed us to investigate eastern wolf and eastern
16
coyote niche breadth, we predicted that: 4) eastern wolves are more specialized in their
use of resources and thus have a narrower niche breadth than other canids; and 5) eastern
wolves occur in more natural landscapes than eastern coyotes. Our study is among the
first to rigorously assess patterns of landscape-level niche differentiation across a full
hybrid zone, while also serving as a conservative test of the intermediate phenotype
hypothesis within an hybrid swarm.
17
Methods
Study area and sampling effort
We described the potential realized niche of eastern wolves, eastern coyotes, and their
hybrids by delineating a study area that bounded known contemporary locations of wild
canids having eastern wolf ancestry (see Kyle et al. 2006; Wilson et al. 2009; Bozarth et
al. 2011). This area spanned eastern Manitoba, central Ontario, Quebec and Maine,
through New England, New York and Pennsylvania, and around the Great Lakes,
including northern portions of Minnesota, Wisconsin, and Michigan. The northern
boundary of the study area was defined by the southern distribution of grey wolves,
where eastern wolves are currently excluded (Wheeldon & Patterson, 2012). The southern
boundary was defined by the potential spread of eastern wolf genes known to be present
in eastern coyotes. Notably, the contemporary primary source of eastern wolf genes is
through hybridization between eastern wolves in Ontario and Quebec and nearby coyotes.
Genetic data source and preparation
Georeferenced canid samples were obtained from a variety of sources including
published and unpublished data for a total of 1193 samples, see Supporting Information 1
for the data sources list. An additional 237 samples were obtained from museums,
governments, fur auction houses, and individual trappers spanning New York,
Pennsylvania, Manitoba, Ontario and Quebec. The raw dataset comprised 1430
contemporary (1995-2014) samples collected across 3 Canadian provinces and 10 States,
with location accuracy being <1000 m for most (85%) samples (J-A. Otis, unpubl.).
DNA Extraction & Genetic Analysis
18
DNA was extracted from tissue samples by magnetic bead using a MagneSil Blood
Genomic Max Yield System (Promega) on a Janus 96-well automated workstation
(Perkin-Elmer). DNA was quantified by Quant-It Picogreen Assay (Invitrogen) and
subsequently standardized to 2.5 ng/uL to ensure accurate, consistent amounts of template
DNA were used for polymerase chain reaction (PCR). Samples were genotyped at 12
autosomal microsatellite loci (cxx225, cxx2, cxx123, cxx377, cxx250, cxx204, cxx172,
cxx109, cxx253, cxx442, cxx410, cxx147) according to methods described in Rutledge et
al. (2010) and Wheeldon et al. (2010). Previous studies on this wolf-coyote complex have
demonstrated the efficiency of this set of microsatellites for investigations concerning
hybridization among canids in eastern North America (see Rutledge et al., 2012a;2015).
Data screening
We excluded samples with more than four missing alleles from the analysis, and the
dataset was screened with the R package Allelematch 2.03 (Galpern et al., 2012) to
remove potential duplicate individuals. Secondly, within the Algonquin Provincial Park
area we used only samples from Rutledge et al. (2010), deposited in the Dryad depository
(http://datadryad.org/resource/doi:10.5061/dryad.q9d6s/1), leading to the exclusion of
known relatives (8% of samples, J-A. Otis, unpubl.). An additional 677 samples were
excluded because of high grey wolf (C. lupus) ancestry, providing a sample size of 648
canids in our final dataset.
Genotyped individuals were assigned to Canis type with the F-model for correlated
allele frequencies in the Bayesian clustering program Structure v.2.3.4 (Pritchard et al.,
2000; Falush et al., 2003) to classify the remaining individuals (i.e. final dataset) into one
of three categories: i) eastern wolf, ii) eastern coyote or iii) hybrid (i.e. admixed) (see
19
Supplementary Information 2 for parameter details). Individuals were classified into
groups using Q ≥ 0.8 (proportional ancestry value) for parental groups. It is understood
that animals considered here as from eastern wolf and eastern coyote parental groups
already have been introgressed to a small degree, but they have been identified as
genetically distinct groups (e.g., Rutledge et al., 2010a; Way et al., 2010; Rutledge et al.,
2012; Wheeldon & Patterson, 2012). Accordingly, we ran a principal components
analysis (PCA) in the R package adegenet (Jombart, 2008; Fig. S2.5, Appendix S2) to
confirm the results obtained with Structure (Fig. 1).
Environmental variables
We used several measures of habitat and human disturbance to examine patterns of
space use by canid groups, including: 1) land use class (cropland, deciduous forest,
evergreen forest, flooded areas, grassland, scrubland, sparse vegetation, urban/bare areas),
based on reclassified GlobCover data (ESA 2010); 2) forest percentage cover (DeFries et
al., 2000); 3) human density (WCS 2005); and 4) road density (km/grid; ESRI 2014). We
also used representative climate variables from the WorldClim database (Hijmans et al.,
2005), including: 5) annual mean temperature; 6) minimum temperature of the coldest
month; and 7) maximum temperature of the warmest month. In addition, we included
environmental data that could have relevance to canid land use patterns and niche
ecology, including: 8) average winter (October-March) snow depth; and 9) snow cover,
using the North America Regional Reanalysis dataset (Mesinger et al., 2006). These nine
layers were resampled at 5 km2 grid cell size to account for variable levels of accuracy.
We addressed collinearity between variables by running pairwise Pearson correlations
and removing variables that were strongly correlated (r >0.85, Milanovich et al., 2010).
20
When pairs of variables were correlated, we retained only the most biologically
meaningful variable. The reduced set of seven variables (Table 1) was used in our focal
analysis.
Ecological niche modeling
We used MaxEnt (Philips et al., 2006) to develop ecological niche models for each
canid group. MaxEnt is a maximum entropy approach that models habitat suitability
patterns using presence observations. MaxEnt generates pseudo-absences to contrast with
observations to develop a habitat suitability map as a function of the environmental
variables. A primary assumption with this approach is that sampling was conducted at
random, which may not be met by our data given the large variety of data sources and
opportunistic sampling protocols underlying acquisition of canid tissue samples
(Newbold, 2010). To the fullest extent possible, spatial bias related to differences in
sampling efforts was controlled by randomly subsampling occurrence data separately for
each canid group, to retain a single sample per 10 km2 grid cell (Kremen et al., 2008;
Boria et al., 2014). Following the procedures of Elith et al. (2010), we also developed a
bias grid to down-weight the influence of spatial aggregation of the presence records.
Modeling details and implementation of the EMNeval procedure are fully described in
Supporting Information 3.
Environmental niche similarity
Niche overlap between eastern wolf, eastern coyote and hybrid was calculated using
logistic probabilities of occupancy in each grid cell according to the I statistic,
implemented in ENMTools (Warren et al., 2008; 2010). Here, we used the logistic output
of MaxEnt (average of the 10 suitability score maps) and compared suitability value of
21
each grid cell between different models. The I statistic ranges from 0 to 1, where models
having similar suitability values for each grid cell receive higher scoring. Note that
ordination (see Broennimann et al., 2012) also could have served to measure niche
similarity, but different calibration of our three models, different weighting for individual
environmental variables, and variable sampling intensity, limited the suitability of this
method for our purposes.
Because sample size and background environments differed between each canid group,
niche overlap values could not be compared directly (Peterson, 2011). Rather, we used a
null model approach which tests that niches of two populations are more similar than
expected based on chance alone, using a background similarity test (Warren et al., 2008).
This method assesses whether species are ecologically different than expected given
differences in local environmental backgrounds of the regions where they occur. Our
approach involved producing 100 random niche models for each species (within their
respective buffered minimum polygon convex) by using GME (Geospatial Modelling
Environment; Beyer, 2012) to generate random points to match the number of presence
data. Then, we ran habitat suitability models based on these random points using the
default settings in MaxEnt. To calculate niche similarity between the 3 canid groups, we
calculated pairwise niche overlap between the real niche model of a species (i.e. observed
group) and the 100 random models from the background of each other species (i.e.
background group). Because test results depend on which group is selected as observed
versus background, each pair of groups were compared both ways for a total of 6 pairwise
comparisons. This allowed us to compare the actual niche overlap between eastern
wolves and hybrids (or eastern coyotes) to that expected by chance alone (i.e. the overlap
22
expected if either hybrids or eastern wolves were using their environments randomly).
Statistical significance indicates that the ecological niches of the species differs from that
expected at random, so this is treated as a two-tailed test. If the niche similarity value fell
on the right or left of the random distribution (at the α = 0.05 level), we considered that
the group had higher or lower niche similarity than expected, respectively. If the value
fell within the random distribution, we inferred that niche differences could be related to
differences in the local environmental background only (Warren et al., 2008).
Environmental niche breadth
We used Levin’s inverse concentration metric to determine the environmental niche
breadth of each canid group, implemented in ENMTools (Warren et al., 2008). The
values of this metric indicate minimum to maximum niche breadth ranging from 0 to 1,
respectively. To compare niche breadth of the eastern wolf against other groups, we
accounted for differences in their geographic background by generating 100 random
models for each species within the buffered minimum convex polygon for the species and
calculating niche breadth using the 100 replicate models based on random presences in
their respective range. This formed a distribution of null expected niche breadth for each
canid, which was then compared to observed niche breadth from models derived from
presence records. Environmental niche breadth values that were further to the left of the
distribution indicated increased habitat selection (Peers et al., 2012).
23
Results
Ecological niche modeling
The habitat model for eastern wolves revealed considerable habitat suitability in
southeastern Ontario and southwestern Quebec, the southwestern region of the Great
Lakes (northern Wisconsin, northeastern Minnesota), and northern New England (Maine,
northern New Hampshire, northern New York) (Figure 2). Yet, for eastern wolves overall
only 6% of the study area had a habitat suitability value > 0.50. The AUC for the eastern
wolf model was high (0.873), implying that it strongly differentiated occurrence from
background locations. According to estimates of the relative contribution of
environmental variables to the eastern wolf model (Table 1), minimum temperature of the
coldest month was the variable contributing the most explanatory power (33.7%),
followed by percentage tree cover (32.9%) and road density (19.4%). Eastern wolf
distribution was bimodally associated with winter temperature, and negatively associated
with low forest cover and high road density (Figure 3). As expected, relative to eastern
coyotes and hybrids, eastern wolves were most likely to occur in evergreen forests, and
also had a high degree of habitat association with deciduous forests (Figure 4). The low
contribution of land use types to the eastern wolf model compared to other canid groups
is probably explained by the fact that few types are found within its distribution and these
are well represented by model variables (e.g., tree cover and road density are correlated to
degree of landscape modification). Likewise, the importance of minimum temperature
might be related to the restricted geographic distribution of eastern wolves (Figure 1)
compared to other canid groups.
24
The eastern coyote model revealed reasonably high habitat suitability across the entire
study area, with notable hotspots in the central Great Lakes’ region of Ontario, portions of
the New England states, and the Gaspé Peninsula in Quebec (Figure 2). The AUC value
for the eastern coyote model was reasonably high (0.706) but qualitatively lower than that
for eastern wolves, likely reflecting the more generalized habitat selection patterns of
eastern coyote. The strongest predictors of eastern coyote habitat suitability were land use
types (36.0%), maximum temperature of the warmest month (25.5%), and minimum
temperature of the coldest month (18.4%) (Table 1). Indeed, all land cover types except
evergreen forest had high (>0.50) suitability for eastern coyotes, and they had higher
suitability than eastern wolves or hybrids in 6 of 8 land cover categories (Figure 4).
For hybrid canids, suitable habitats included much of south central Ontario and the
region spanning West Virginia to New Brunswick (Figure 2). The AUC value for hybrids
(0.677) also indicated reasonable model fit, and the environmental suitability for hybrids
was strongly influenced by land use types (47.7%), snow depth (34.5%), and to a lesser
extent, road density (10.3%) (Table 1). The strong influence of snow depth, only on
hybrids, probably reflects sampling constraints rather than a truly biotic influence.
Qualitatively, hybrid response curves tended to be more similar to those of eastern
coyotes than eastern wolves (Figure 3), but as predicted by the intermediate phenotype
hypothesis, to a large extent their responses were intermediate to those of both parental
groups. Indeed, for 4 of the 8 land use categories, hybrid suitability was between that
observed for eastern wolves and eastern coyotes, with deciduous forests being especially
suitable for hybrids (Figure 4).
25
Pairwise niche overlap
Pairwise niche overlap between canids was largest between hybrids and eastern
coyotes, and lowest between eastern wolves and eastern coyotes (Table 2). Further, our
analysis revealed significant niche differentiation between canid pairs, but the
interpretation of this difference is variable according to the canid group used as the
observed group vs. background group. The observed eastern wolf niche model did not
differ from random when compared to the other canid groups, perhaps owing to the fact
that the available habitat of the two other species occurred in a much larger spatial area
(Figure 2), and therefore available environments included areas outside the training data
for eastern wolves (see Nakazato et al., 2010). However, when niche overlap for hybrids
was compared to the eastern wolf background, hybrids were less similar than expected;
when hybrids were compared to the eastern coyote background, they were more similar
than expected (Table 2). Eastern coyotes were not differentiated when compared to the
hybrid background, but eastern coyotes were less similar than expected to eastern wolves.
Therefore, we conclude that when compared to a background representing eastern
wolves, both eastern coyotes and hybrids exhibited significant niche differentiation,
whereas hybrids showed a lack of differentiation when compared to eastern coyotes.
Niche breadth
Our habitat suitability models revealed that the 3 wild canids differ substantially in
their estimated niche breadth. As predicted, niche breadth was narrowest for eastern
wolves (Levin’s concentration: 0.311, p < 0.01) and broader for eastern coyotes (0.954; p
= 0.02) and hybrids (0.845, p = 0.12). When compared to available habitat within their
environment, eastern wolves were the most highly selective canid, as their distribution
26
fell far outside the random distribution (Figure 5). Likewise, eastern coyote niche breadth
fell slightly outside the random distribution. In accordance with our predictions, niche
breadth for hybrids was largely intermediate to either parental group, but had similar
dimensions to those expected relative to the background environment.
27
Discussion
Our main objective was to evaluate patterns of niche differentiation among groups of
North American wild canids that have experienced extensive hybridization and genetic
introgression during the last four centuries. We predicted that hybrids would exhibit
intermediate behavioural phenotypes relative to eastern wolf and eastern coyote groups,
and our results confirmed that hybrids had environmental, land cover, and anthropogenic
niche characteristics that were in between parental groups. Likewise, niche overlap was
greatest for hybrids relative to parental groups. Collectively, these results bolster the
perception that niche characteristics of hybrid organisms is intermediate to those of
parental groups (e.g., Anderson, 1948; Barton 2001; Choler et al., 2004), and suggest that
genetic admixture and/or introgression can help to shape niche dynamics and large scale
animal-environment interactions. That the intermediate phenotype hypothesis was
supported at a large spatial scale and using observations of North American wild canids,
which are known for their strong behavioural plasticity and high degree of introgression
(Atwood et al., 2004; Sacks et al., 2008), emphasizes the generality and relevance of the
hypothesis in explaining patterns of hybrid spatial structuring and niche characteristics
within a hybrid swarm.
While originally conceived to explain differential patterns of habitat selection among
parental species and their F1 hybrids, the predictions emerging from the intermediate
phenotype hypothesis (Anderson, 1948; Barton, 2001) are upheld even in a complex
system involving wild canids ranging across expansive landscapes with high habitat
heterogeneity. First, the extensive hybridization and backcrossing observed in wild canids
of eastern North America offers a particularly vexing taxonomic problem. At its origin,
28
the loss of geographic barriers to reproduction between formerly-distinct coyotes and
eastern wolves arose through human-caused changes to native habitat combined with
intensive human exploitation; these changes led to increased spatial overlap,
interbreeding, and hybridization between canid species (Kyle et al., 2006; Murray &
Waits, 2007). Currently, wild canids having various levels of genetic introgression range
across a variety of available habitat and environmental conditions in eastern North
America, illustrating their high capacity for spatial overlap, dispersal and colonization,
and behavioural plasticity. It follows that the genetic complexity and spatial heterogeneity
in the wild canid system offers an especially challenging scenario for testing questions of
hybridization and niche differentiation.
The intermediate phenotype hypothesis predicts that hybrids will thrive in transition
zones and ecotones whereas parental groups will be better suited to core habitats (Barton,
2001; Choler et al., 2004; Tauleigne-Gomes & Lefèbvre, 2008). The intermediate
phenotype hypothesis can be extended to contemporary conditions by positing that
hybrids will be more apt to colonize and occupy human-altered landscapes. It follows that
hybrids may occur especially in environments where ecosystem structure and function
and attendant patterns of biodiversity are recently or dramatically altered (Fitzpatrick &
Shaffer, 2007). Although it is not well understood why hybrids are especially suited to
altered environments, it seems that lessened competition with parental groups in such
areas is an important factor (Seehausen, 2004). Regardless, while F1 hybrids may
naturally exhibit considerable distinction from parental species and therefore should have
a tendency toward use of intermediate habitats, it is understood that further backcrossing
and introgression among fertile hybrids should progressively diminish the expected
29
distinction between hybrids and parental groups. In our wild canid system, there is
evidence that hybridization between eastern wolves and coyotes began >300,000 years
(Rutledge et al., 2010b), with a strong acceleration of interbreeding in recent centuries
(Way et al., 2010). Therefore, in theory the distribution of canid genotypes and
phenotypes in the broader population should follow a gradual cline (Wilson et al., 2009;
Rutledge et al., 2010a). In this system, hybridization has affected body size (Way, 2007),
habitat and diet preferences (see Kays et al., 2008, 2010; Benson et al., 2012), as well as
fine-scale space use patterns (Ellington & Murray, 2015). Although backcrossing and
introgression should reduce the likelihood of successfully differentiating between canid
groups in terms of their niche breadth and niche overlap, our results provide convincing
evidence that even within a hybrid swarm, highly-admixed individuals can exhibit an
intermediate ecological phenotype compared to parental groups.
Previous studies examining phenotypic differences among wild canids (e.g., Kays et
al., 2008; Benson et al., 2013) were conducted across a small spatial scale or do not
encompass the full suite of phenotypes found along the wolf-coyote gradient in eastern
North America. However, because hybrid phenotype selection is dependent on available
habitats and the presence of competition (Grant & Grant, 2002; Abbott et al., 2013), it is
important that niche differentiation be investigated across an appropriate scale involving
the complete hybrid zone and extent of potential phenotypic responses in the population.
When comparing habitat suitability of parental groups and hybrids within the full
spectrum of the eastern wolf-coyote complex, we found that hybrids have intermediate
habitat preferences; elsewhere, intermediate phenotypes of hybrids were associated with
differences in morphology or vocalization (e.g., van Dongen et al., 2013; Page et al.,
30
2001), but rarely in terms of habitat differences (but see Heaney & Timm, 1985; Choler et
al., 2004). In particular, hybrids in our system show intermediate tolerance to humancaused landcover changes, whereas eastern wolf-like animals occur primarily in forested
habitat. The strong contribution of maximum and minimum temperature within the
eastern coyote model probably reflects the factors restricting their broader distribution,
although we note that our analysis excluded animals found mostly in north-central North
America due to their stronger grey wolf ancestry. Regardless, our results clearly illustrate
the intermediate niche dimensions of highly-admixed canids and their ability to
successfully colonize human-altered landscapes, compared to parental groups.
Our niche breadth analysis reinforced that hybrids exhibit intermediate niche
characteristics, which is in line with our predictions and adds support to the contention
that niche dimensions widen among highly admixed animals (Kays et al., 2010; Thornton
& Murray, 2014). Notwithstanding this, results from the niche overlap analysis were
more challenging to interpret but generally showed pairwise differences consistent with
our predictions, especially for hybrids and eastern coyotes.
Currently, eastern wolves are recognized as a threatened species in Canada
(COSEWIC 2015), whereas they are listed as special concern in Ontario (COSSARO
2004 [Committee on the Status of Species at Risk in Ontario], available online). Given
the current, highly-restricted distribution of eastern wolves in Canada (Fig. 1) and the
high habitat suitability for non-wolf canids across contemporary landscapes in eastern
North America demonstrated by our models, our analysis highlights potential
conservation challenges for eastern wolves. For example, continued introgression could
lead to increased homogeneity in the canid population and further genetic dilution of
31
parental groups, as is reported in passerine songbirds (Parkes, 1951; Gill, 2004). It
follows that such loss of diversity will alter niche breadth dynamics among extant wild
canids, which may have broader community- or ecosystem-level consequences.
Accordingly, it is critical that the genetic integrity of eastern wolves be monitored
through time, and if maintenance of the eastern wolf genotype is a priority, management
actions should be considered (see Gese et al., 2015, Mank et al., 2004). However, we note
that thus far active management has had low success in establishing self-sustaining wolf
populations when faced with hybrid swarms involving non-wolf canids (Murray et al.,
2015), making the likely success of future programs open to debate. An interesting area of
future research would be to divide hybrids in two different categories - “early” and
“contemporary” generations - to investigate if there is any kind of hybrid selection in the
long term and if these two different hybrids groups differ ecologically (see Ryan et al.,
2009).
In conclusion, our study extends understanding of the patterns of spatial distribution
and niche dynamics in a closely-related and hybridizing species complex. Of particular
importance, our results demonstrate that the intermediate phenotype hypothesis is
supported even at a large scale and when involving highly mobile large mammal species.
This generality allows us to substantially improve our ability to evaluate, and even
predict, potential impacts of hybridization under scenarios of environmental change.
Ultimately, a more robust integration of genetic information in species niche modeling
research will contribute strongly to increased understanding of the patterns and processes
involved in hybridization and population dynamics.
32
Table 1. Estimates of the relative contribution from environmental variables to habitat
suitability models for eastern wolves, eastern coyotes and hybrids.
Variables
Maximum T°
Minimum T°
Snow depth
Percent tree cover
Land cover
Human density
Road density
Eastern
Eastern Hybrids
wolves
coyotes
Percent contribution
4.6
25.5
1.4
33.7
18.4
5.9
8.5
4.5
34.5
32.9
8.1
0.2
1.0
36.0
47.7
0
0.7
0
19.4
6.8
10.3
Table 2. Comparative niche overlap and tests of background similarity between eastern
wolves, eastern coyotes and hybrids. I statistics are presented, and the outcome of the test
depends on the canid group used as observed vs. background.
Canid groups
(observed)
eastern wolves
eastern coyotes
eastern wolves
hybrids
eastern coyotes
hybrids
Canid groups
(background)
eastern coyotes
eastern wolves
hybrids
eastern wolves
hybrids
eastern coyotes
33
I
0.717
0.717
0.745
0.745
0.977
0.977
More or less similar
than expected
NS
Less (p<0.01)
NS
Less (p<0.01)
NS
More (p<0.01)
Figure 1. Geographic distribution of 648 wild canid samples from eastern North America
used in this study. Canids were classified in different groups using a Q-value threshold of
≥ 0.8, for a total of 110 eastern wolves, 487 eastern coyotes, and 51 hybrids; animals with
grey wolf ancestry are excluded.
34
Figure 2. Habitat suitability models for wild canids in eastern North America, including
eastern wolves, eastern coyotes, and hybrids. Green and red represent areas with
predicted low and high suitability, respectively.
35
Figure 3. Habitat suitability response curves for individual environmental variables,
when occurring alone in models for eastern wolves, eastern coyotes and hybrid canids.
36
Figure 4. Wild canid responses to individual land cover categories in univariate MaxEnt
models.
37
*
*
Figure 5. Niche breadth of wild canids in relation to the background environment.
Histograms represent the expected niche breadth if an animal selects habitats at random
from the background, and the triangle indicates observed niche breadth of each of the
three canid groups. Note that the X-axis is presented at a different scale for each panel.
38
Chapter 3
General Conclusions
“What’s the use of having developed a science well enough to make predictions if, in the
end, all we’re willing to do is stand around and wait for them to come true?”
~ Frank Sherwood Rowland (1996)
39
General Conclusion
Goals of the thesis revisited
Worldwide, large predator populations have declined due largely to habitat loss and
human persecution (Ripple et al., 2014). Although these two threats will continue to play
an important role in the ability of large canids to persist in North America, hybridization,
and/or synergism between hybridization and other forms of anthropogenic change
(Rutledge et al., 2012b), may impose marked short-term impacts on carnivore populations
(Doherty et al., 2015). For example, although large tracts of suitable habitat exist for red
wolves, the success of reintroduction efforts in North Carolina has been threatened by
hybridization with coyotes (Gese et al., 2015; Murray et al., 2015). The potential threat
posed by eastern coyote range expansion, either due to hybridization with other canids,
and/or displacement of remaining eastern wolves, is a significant conservation concern.
Additionally, the divergence in abundance of the parental species during an introgressive
hybridization event is most likely to menace the genetic integrity of the rarer species than
the one of the abundant species (Rhymer & Sinberdof, 1996). This speaks to the pressing
need for a better understanding of the ecology of hybrids and their relationship with
parental species, especially in areas with heterogenous human impacts. For instance,
Detwiler et al. (2005) suggested that hybridization among African cercopithecine
monkeys is a major threat for small parental populations isolated by habitat
fragmentation, but the identification of species-specific habitat characteristics could
prevent their extinction.
In this thesis, the wolf-coyote hybrid complex in eastern North America served as a
case study to test whether hybrids show an intermediate phenotype, as evidenced by niche
40
characteristics that reflect traits common to both parental groups. The wolf-coyote system
was ideally suited to explore this question because the genetic basis for evaluating canid
hybrid dynamics has been extensively studied (e.g., Wilson et al., 2009; Rutledge et al.,
2010a; Wheeldon & Patterson, 2012). Moreover, the large spatial scale over which
hybridization has occurred in this system, and the high degree of behavioral plasticity
exhibited by wild canids (Atwood et al., 2004; Sacks et al., 2008), provides an especially
robust test of the intermediate phenotype hypothesis. However, the modelling exercise
applied in this study could easily be applied to a variety of other taxa and hybridization
scenarios, such as bobcat and lynx (Schwartz et al., 2004) or tiger salamanders
(Fitzpatrick & Shaffer, 2007). Indeed, given that the intermediate phenotype hypothesis is
supported in the wolf-coyote complex, it is reasonable to infer that it forms the basis of
niche differentiation in a large variety of other hybrid systems.
The canid niche models that I developed showed differences in the distribution of the
three groups with reasonable confidence, despite that they overlap geographical region
with a broadly similar landscape. Therefore, my findings support that there are
fundamental ecological differences between groups that promote niche differentiation,
and these are most likely related to their genotype divergence. However, as with all
studies of a similar nature, inference derived from this study is limited by the assumption
that sample locations represent source rather than sink populations (Pulliam, 2000), which
is less realistic among highly mobile species experiencing ongoing colonization and range
expansion, as is the case with eastern coyote. Furthermore, the models assume that
species distributions, and the environmental and ecological factors affecting their spatial
dynamics, are largely static. In the case of interbreeding animals with variable levels of
41
introgression, the models also assume that genotype-environment relationships also are
fixed and that these are consistent among individuals in the three groups we examined
(i.e., eastern wolves, eastern coyotes, hybrids). It is possible that some of these
relationships did undergo a significant degree of variation. Nevertheless, my results
indicate that over the long term eastern wolves may be challenged to persist in eastern
North America due to both their relatively low density and to the region’s high suitability
for the long term establishment of eastern coyotes and hybrids. According to my results,
outside of Algonquin Provincial Park, dispersing eastern wolves likely will need to
compete intensively against other canids for an increasingly limited and unsuitable
habitat; the latter canid groups have a distinct advantage in these areas. Thus, the future
status of this wild canid hybridization complex may be strongly determined by the extent
and degree of human disturbance across the landscape. Accordingly, the protection of
highly forested habitat with low human disturbance, as well as active management efforts
(e.g.,, the existing harvest ban buffer established around Algonquin Provincial Park to
restore social structure, see Rutledge et al., 2010b) will be paramount if the eastern wolf
genotype is to persist.
Future research directions
Hybridization and introgression are becoming increasingly prevalent across a variety
of formerly distinct taxa and across a broad range of ecosystems (Harrison, 1993; Crispo
et al., 2011). Proximally, outcomes of these phenomena are often difficult to predict with
confidence (Crispo et al., 2011), making it crucial to develop a more robust understanding
of the processes involved during these events in natural systems. With this thesis, I
showed that hybridized and highly–admixed forms of wild canids found across a
42
heterogeneous and disturbed landscapes prefer habitat characteristics that are intermediate
to the parental forms. These findings can be extended to help predict ecological impacts
of hybridization and introgression. For example, coupling an understanding of hybrid
distribution and niche characteristic to the analysis of hybrid fitness across the hybrid
zone relative to parental groups, will provide a better idea of the long-term consequences
of hybridization. Equally important, fitness consequences of habitat selection should be
investigated relative to environmental gradients across the hybrid zone, to see how this
may vary along key niche axes. This comprehensive approach will promote a better
understanding of the mechanisms underlying hybrid zone dynamics, and allow for the
development of reliable predictive models that potentially lead to effective mitigative
measures, if appropriate (Pearson & Manuwal, 2000).
According to the recommendations put forth by Allendorf et al. (2001), whenever
possible we should seek to maintain pure parental populations when hybridization arises
directly through anthropogenic causes. In this context, it is imperative to assess where and
how many eastern wolves occur in eastern North America and to understand the likely
long-term viability of this species. I demonstrated using niche models that there are few
suitable areas where eastern wolves are likely to be found, and these areas (e.g., northern
Maine, Parc de la Mauricie in Quebec, Adirondack Park in New York) should be
surveyed for canid genotype (e.g., scat surveys; Adams et al., 2003; Rutledge et al.,
2009). Thereafter, habitat models may help to: i) find new areas for eastern wolf (or red
wolf) translocation in North America, and ii) assess potential areas for habitat restoration
to promote eastern wolf population viability. However, it is notable that the candidate
areas for eastern wolf colonization are sufficiently small and disjunct that they are
43
unlikely to suitably connect a meta-population of eastern wolves, or to naturally exclude
other wild canids. Therefore, the potential to re-establish viable and genetically pure
populations of eastern wolves in these areas simply may not be realistic.
Conclusion
Overall, this thesis presents evidence that there is ecological differentiation among
wild canids in eastern North America. Given the high suitability for hybrids and eastern
coyotes across the landscape, and their greater resilience to human disturbance, eastern
wolves may be increasingly challenged to persist. Ultimately, the combination of
continued anthropogenic activities, hybridization, and niche displacement among top
predators in natural and managed landscapes in eastern North America could have
profound impacts on the structure and function of those ecosystems (Whitham et al.,
2003; Fortin et al., 2005).
44
Literature cited
Abbott, R., Albach, D., Ansell, S., Arntzen, J.W., Baird, S.J.E., Bierne, N., Boughman, J.,
et al. (2013) Hybridization and speciation. Journal of evolutionary biology, 26, 229–
46.
Adams, J.R., Kelly, B.T. & Waits, L.P. (2003) Using faecal DNA sampling and GIS to
monitor hybridization between red wolves (Canis rufus) and coyotes (Canis latrans).
Molecular Ecology, 12, 2175–2186.
Allendorf, F.W., Leary, R.F., Spruell, P. & Wenburg, J.K. (2001) The problems with
hybrids: Setting conservation guidelines. Trends in Ecology and Evolution, 16, 613–
622.
Anderson, E. & Stebbins Jr., G.L. (1954) Hybridization as an Evolutionary Stimulus.
Evolution, 8, 378–388.
Anderson, E. (1948). Hybridization of the habitat. Evolution, 2, 1–9.
Arnold, M. L. (1997) Natural hybridization and evolution. Oxford University Press,
Oxford.
Arnold, M.L. (1992) Natural Hybridization as an Evolutionary Process. Annual Review of
Ecology and Systematics, 23, 237–261.
Atwood, T., Weeks, H. & Gehring, T. (2004) Spatial ecology of coyotes along a
suburban-to-rural gradient. Journal of Wildlife Management, 68, 1000–1009.
Barton, N.H. & Hewitt, G.M. (1985) Analysis of Hybrid Zones. Annual Review of
Ecology and Systematics, 16, 113–148.
Barton, N.H. (2001) The role of hybridization in evolution. Molecular Ecology, 10, 551–
568.
Baskett, M.L. & Gomulkiewicz, R. (2011) Introgressive hybridization as a mechanism for
species rescue. Theoretical Ecology, 4, 223–239.
Benson, J.F., Mills, K.J., Loveless, K.M. & Patterson, B.R. (2013) Genetic and
environmental influences on pup mortality risk for wolves and coyotes within a
Canis hybrid zone. Biological Conservation, 166, 133–141.
Benson, J.F., Patterson, B.R. & Wheeldon, T.J. (2012) Spatial genetic and morphologic
structure of wolves and coyotes in relation to environmental heterogeneity in a Canis
hybrid zone. Molecular ecology, 21, 5934–54.
45
Beyer, H. L. (2012). Geospatial Modelling Environment v0. 7.3.0
Boria, R.A., Olson, L.E., Goodman, S.M. & Anderson, R.P. (2014) Spatial filtering to
reduce sampling bias can improve the performance of ecological niche models.
Ecological Modelling, 275, 73–77.
Bozarth, C.A., Hailer, F., Rockwood, L.L., Edwards, C.W. & Maldonado, J.E. (2011)
Coyote colonization of northern Virginia and admixture with Great Lakes wolves.
Journal of Mammalogy, 92, 1070–1080.
Broennimann, O., Fitzpatrick, M.C., Pearman, P.B., Petitpierre, B., Pellissier, L., Yoccoz,
N.G., Thuiller, W., Fortin, M.J., Randin, C., Zimmermann, N.E., Graham, C.H. &
Guisan, A. (2012) Measuring ecological niche overlap from occurrence and spatial
environmental data. Global Ecology and Biogeography, 21, 481–497.
Bullini, L. (1994) Origin and evolution of animal hybrid species. Trends in Ecology and
Evolution, 9, 422–426.
Chambers, S. M., Fain, S. R., Fazio, B., & Amaral, M. (2012) An account of the
taxonomy of North American wolves from morphological and genetic
analyses. North American Fauna, 77, 1–67.
Choler, P., Erschbamer, B., Tribsch, A., Gielly, L. & Taberlet, P. (2004) Genetic
introgression as a potential to widen a species’ niche: insights from alpine Carex
curvula. Proceedings of the National Academy of Sciences of the United States of
America, 101, 171–6.
COSEWIC. (2015). Canadian species at risk, May 2015. Committee on the Status of
Endangered Wildlife in Canada, Ottawa. Available from:
http://www.sararegistry.gc.ca/default.asp?lang=En&n=65C48F31-1
COSSARO. (2004). Committee on the Status of Species at Risk in Ontario. Summary of
species at risk in Ontario: eastern wolf. Available from:
http://www.rom.on.ca/ontario/risk. php?doc_type=fact&id=287&lang=en (accessed
January 2015).
Crete, M., Ouellet, J., Tremblay, J. & Arsenault, R. (2001) Suitability of the forest
landscape for coyotes in northeastern North America and its implications for
coexistence with other carnivores. Ecoscience, 8, 311–319.
Crispo, E., Moore, J.S., Lee-Yaw, J.A., Grey, S.M. & Haller, B.C. (2011) Broken
barriers: Human-induced changes to gene flow and introgression in animals.
BioEssays, 33, 508–518.
46
DeFries, R.S., & Belward, A.S. (2000). Global and regional land cover characterization
from satellite data: an introduction to the Special Issue.International Journal of
Remote Sensing, 21, 1083–1092.
Detwiler, K.M., Burrell, A.S. & Jolly, C.J. (2005) Conservation implications of
hybridization in African cercopithecine monkeys. International Journal of
Primatology, 26, 661–684.
Doherty, T.S., Dickman, C.R., Nimmo, D.G. & Ritchie, E.G., (2015) Multiple threats, or
multiplying the threats? Interactions between invasive predators and other ecological
disturbances. Biological Conservation, 190, 60–68.
Elith, J., Kearney, M. & Phillips, S. (2010) The art of modelling range-shifting species.
Methods in Ecology and Evolution, 1, 330–342.
Ellington, E.H., & Murray, D.L. (2015). Influence of hybridization on animal space use: a
case study using coyote range expansion. Oikos, 124, 535-542.
Ellstrand, N.C. & Schierenbeck, K.A. (2000) Hybridization as a stimulus for the
evolution of invasiveness in plants? Proceedings of the National Academy of
Sciences, 97, 7043–7050.
Ellstrand, N.C., Whitkus, R. & Rieseberg, L.H. (1996) Distribution of spontaneous plant
hybrids. Proceedings of the National Academy of Sciences of the United States of
America, 93, 5090–5093.
ESA. (2010). European Space Agency. GlobCover Land Cover Maps. Available from:
http://due.esrin.esa.int/page_globcover.php (accessed October 2014)
ESRI. (2014). Environmental Systems Research Institute. U.S. and Canada Detailed
Streets provides nationwide streets for the United States and Canada. Available
from:https://www.arcgis.com/home/item.html?id=f38b87cc295541fb88513d1ed7ce
c9fd (accessed October 2014).
Falush, D., Stephens, M. & Pritchard, J.K. (2003) Inference of population structure using
multilocus genotype data: linked loci and correlated allele frequencies. Genetics,
164, 1567–87.
Fitzpatrick, B.M. & Shaffer, H.B. (2007) Introduction history and habitat variation
explain the landscape genetics of hybrid tiger salamanders. Ecological Applications,
17, 598–608.
Forbes, G.J. & Theberge, J.B. (1996) Cross-Boundary Management of Algonquin Park
Wolves. Conservation Biology, 10, 1091–1097.
47
Fortin, D., Beyer, H.L., Boyce, M.S., Smith, D.W., Duchesne, T. & Mao, J.S. (2005)
Wolves influence elk movements: Behavior shapes a trophic cascade in Yellowstone
National Park. Ecology, 86, 1320–1330.
Fritts, S.H., Stephenson, R.O., Hayes, R.D., Boitani, L. (2003) Wolves and humans. In:
Mech, L.D., Boitani, L. (Eds.), Wolves: Behavior, Ecology and Conservation.
University of Chicago Press, Chicago.
Galpern, P., Manseau, M., Hettinga, P., Smith, K. & Wilson, P. (2012) Allelematch: an R
package for identifying unique multilocus genotypes where genotyping error and
missing data may be present. Molecular ecology resources, 12, 771–8.
Gese, E.M., Knowlton, F.F., Adams, J.R., Beck, K., Fuller, T.K., Murray, D.L., Steury,
T.D., Stoskopf, M.K., Waddell, W.T. & Waits, L.P. (2015) Managing hybridization
of a recovering endangered species : The red wolf Canis rufus as a case study.
Current Zoology, 61, 191–205.
Gill, F.B. (2004) Blue-winged warblers (Vermivora pinus) versus golden-winged
warblers (V. chrysoptera). The Auk, 121, 1014–1018.
Gompper, M. (2002) Top carnivores in the suburbs? Ecological and conservation issues
raised by colonization of north eastern North America by coyotes. Bioscience, 52,
185–190.
Grant, P.R. & Grant, B.R. (1992) Hybridization of bird species. Science, 256, 193–7.
Grant, P.R. & Grant, B.R. (2002) Unpredictable Evolution in a 30-Year Study of
Darwin’s Finches. Science, 296, 707–711.
Grewal, S.K., Wilson, P., Kung, T.K., Shami, K., Theberge, M.T., Theberge, J.B. &
White, B.N. (2004) A genetic Assessment of the Eastern Wolf (Canis Lycaon) in
Algonquin Provincial Park. Journal of Mammalogy, 85, 625–632.
Grubbs, S. & Krausman, P. (2009) Use of urban landscape by coyotes. The Southwestern
Naturalist, 54, 1–12.
Harrison, R.G. (1993) Hybrids and hybrid zones: historical perspective. In: Harrison R.G.
(Ed.), Hybrid Zones and the Evolutionary Process. Oxford University Press, Oxford.
Heaney, L.R. & Timm, R.M. (1985) Morphology, genetics, and ecology of pocket
gophers (genus Geomys) in a narrow hybrid zone. Biological Journal of the Linnean
Society, 25, 301–317.
Hebblewhite, M. & Smith, D.W. (2010) Wolf community ecology: Ecosystem effects of
recovering wolves in Banff and Yellowstone national parks. In: Musiani, M.,
48
Boitani, L. & Paquet, P. (Eds), The world of wolves: new perspectives on ecology,
behavior and policy. University of Calgary Press, Calgary.
Hedrick, P.W., Lee, R.N. & Garrigan, D. (2002) Major histocompatibility complex
variation in red wolves: Evidence for common ancestry with coyotes and balancing
selection. Molecular Ecology, 11, 1905–1913.
Hijmans, R.J., Cameron, S.E., Parra, J.L., Jones, P.G. & Jarvis, A. (2005) Very high
resolution interpolated climate surfaces for global land areas. International Journal
of Climatology, 25, 1965–1978.
Howard, D.J., Waring, G.L., Tibbets, C. a & Gregory, P.G. (1993) Survival of Hybrids in
a Mosaic Hybrid Zone. Evolution, 47, 789–800.
Hubbs, C. L.. (1955). Hybridization between Fish Species in Nature. Systematic
Zoology, 4, 1–20.
Hudson, R.R. & Turelli, M. (2003) Stochasticity overrules the “three-times rule”: genetic
drift, genetic draft, and coalescence times for nuclear loci versus mitochondrial
DNA. Evolution; international journal of organic evolution, 57, 182–190.
Huxel, G. (1999) Rapid displacement of native species by invasive species: effects of
hybridization. Biological Conservation, 89, 143–152.
Jombart T (2008) adegenet: a R package for the multivariate analysis of genetic markers.
Bioinformatics, 24, 1403–5.
Kays, R., Curtis, A. & Kirchman, J.J. (2010) Rapid adaptive evolution of northeastern
coyotes via hybridization with wolves. Biology letters, 6, 89–93.
Kays, R.W., Gompper, M.E. & Ray, J.C. (2008) Landscape ecology of eastern coyotes
based on large-scale estimates of abundance. Ecological Applications, 18, 1014–
1027.
Kolenosky, G. B., & Standfield, R. O. (1975) Morphological and ecological variation
among grey wolves (Canis lupus) of Ontario, Canada. In: Fox, M.W. (Ed.), The wild
canids: their systematics, behavioral ecology and evolution. Van Nostrand Reinhold,
New York.
Kremen, C., Cameron, A., Moilanen, A., Phillips, S.J., Thomas, C.D., Beentje, H.,
Dransfield, J., Fisher, B.L., Glaw, F., Good, T.C., Harper, G.J., Hijmans, R.J., Lees,
D.C., Louis, E., Nussbaum, R.A., Raxworthy, C.J., Razafimpahanana, A., Schatz,
G.E., Vences, M., Vieites, D.R., Wright, P.C. & Zjhra, M.L. (2008) Aligning
Conservation Priorities Across Taxa in Madagascar with High-Resolution Planning
Tools. Science, 320, 222–226.
49
Kyle, C.J., Johnson, A.R., Patterson, B.R., Wilson, P.J., Shami, K., Grewal, S.K. &
White, B.N. (2006) Genetic nature of eastern wolves: Past, present and future.
Conservation Genetics, 7, 273–287.
Landry, C.R., Hartl, D.L. & Ranz, J.M. (2007) Genome clashes in hybrids: insights from
gene expression. Heredity, 99, 483–493.
Larivière, S. & Crête, M. (1993) The size of eastern coyotes (Canis latrans): a comment.
Journal of Mammalogy, 74, 1072–1074.
Lehman, N., Eisenhawer, A., Hansen, K., Mech, L.D., Peterson, R.O., Gogan, P.J.P. &
Wayne, R.K. (1991) Introgression of Coyote Mitochondrial-DNA into Sympatric
North-American Grey Wolf Populations. Evolution, 45, 104–119.
Lewontin, R.C. & Birch, L.C. (1966) Hybridization as a Source of Variation for
Adaptation to New Environments. Evolution, 20, 315–336.
Mallet, J. (2005) Hybridization as an invasion of the genome. Trends in Ecology and
Evolution, 20, 229–237.
Mank, J.E., Carlson, J.E. & Brittingham, M.C. (2004) A century of hybridization:
Decreasing genetic distance between American black ducks and mallards.
Conservation Genetics, 5, 395–403.
McDevitt, A.D., Mariani, S., Hebblewhite, M., Decesare, N.J., Morgantini, L., Seip, D.,
Weckworth, B. V. & Musiani, M. (2009) Survival in the Rockies of an endangered
hybrid swarm from diverged caribou ( Rangifer tarandus ) lineages. Molecular
Ecology, 18, 665–679.
Mesinger, F., DiMego, G., Kalnay, E., Mitchell, K., Shafran, P.C., Ebisuzaki, W., Jovic,
D., Woollen, J., Rogers, E., Berbery, E.H. & Ek, M.B. (2006) North American
regional reanalysis. Bulletin of the American Meteorological Society, 87, 343–360.
Milanovich, J.R., Peterman, W.E., Nibbelink, N.P. & Maerz, J.C. (2010) Projected Loss
of a Salamander Diversity Hotspot as a Consequence of Projected Global Climate
Change. PLoS ONE, 5, e12189.
Murray, D.L. & Waits, L.P. (2007) Taxonomic status and conservation strategy of the
endangered red wolf: a response to Kyle et al. (2006). Conservation Genetics, 8,
1483–1485.
Murray, D.L., Bastille-Rousseau, G., Adams, J.R. & Waits, L.P. (2015) The Challenges
of Red Wolf Conservation and the Fate of an Endangered Species Recovery
Program. Conservation Letters
50
Nakazato, T., Warren, D.L. & Moyle, L.C. (2010) Ecological and geographic modes of
species divergence in wild tomatoes. American Journal of Botany, 97, 680–693.
Neira, C., Levin, L.A. & Grosholz, E.D. (2005) Benthic macrofaunal communities of
three sites in San Francisco Bay invaded by hybrid Spartina, with comparison to
uninvaded habitats. Marine Ecology Progress Series, 292, 111–126.
Newbold, T. (2010) Applications and limitations of museum data for conservation and
ecology, with particular attention to species distribution models. Progress in
Physical Geography, 34, 3–22.
Nowak, R. M. (1995). Another look at wolf taxonomy. Ecology and conservation of
wolves in a changing world. Canadian Circumpolar Institute, Edmonton, 375.
Olden, J.D. (2006) Biotic homogenization: A new research agenda for conservation
biogeography. Journal of Biogeography, 33, 2027–2039.
Page, B., Goldsworthy, S.D. & Hindell, M.A. (2001) Vocal traits of hybrid fur seals:
intermediate to their parental species. Animal Behaviour, 61, 959–967.
Pamilo, P. & Nei, M. (1988) Relationships between gene trees and species trees.
Molecular biology and evolution, 5, 568–583.
Parkes, K.C. (1951) The genetics of the Golden-winged x Blue-winged Warbler complex.
The Wilson Bulletin, 61, 5–15.
Pearson, S. & Manuwal, D. (2000) Influence of niche overlap and territoriality on
hybridization between Hermit Warblers and Townsend’s Warblers. The Auk, 117,
175–183.
Peers, M.J.L., Thornton, D.H. & Murray, D.L. (2012) Reconsidering the specialistgeneralist paradigm in niche breadth dynamics: resource gradient selection by
Canada lynx and bobcat. PloS one, 7, e51488.
Peterson, a. T. (2011) Ecological niche conservatism: a time-structured review of
evidence. Journal of Biogeography, 38, 817–827.
Phillips, S.J., Anderson, R.P. & Schapire, R.E. (2006) Maximum entropy modeling of
species geographic distributions. Ecological Modelling, 190, 231–259.
Pritchard, J.K., Stephens, M. & Donnelly, P. (2000) Inference of population structure
using multilocus genotype data. Genetics, 155, 945–959.
Pulliam, H.R. (2000) On the relationship between niche and distribution. Ecology Letters,
3, 349–361.
51
Rhymer, J.M. & Simberloff, D. (1996) Extinction by Hybridization and Introgression.
Annual Review of Ecology and Systematics, 27, 83–109.
Richens, V. & Hugie, R. (1974) Distribution, taxonomic status, and characteristics of
coyotes in Maine. The Journal of Wildlife Management, 38, 447–454.
Rieseberg, L.H. & Wendel, J.F. (1993) In: Harrison R.G. (Ed.), Hybrid Zones and the
Evolutionary Process. Oxford University Press, Oxford.
Rieseberg, L.H., Archer, M.A. & Wayne, R.K. (1999) Transgressive segregation,
adaptation and speciation. Heredity, 83, 363–372.
Rieseberg, L.H., Raymond, O., Rosenthal, D.M., Lai, Z., Nakazato, T., Durphy, J.L.,
Schwarzbach, A.E., Donovan, L.A., Lexer, C. & Livingstone, K. (2003) Major
Ecological Transitions in Wild Sunflowers Facilitated by Hybridization. Science,
301, 1211–1216.
Ripple, W.J., Estes, J.A., Beschta, R.L., Wilmers, C.C., Ritchie, E.G.,, Hebblewhite, M.,
Berger, J., Elmhagen, B., Letnic, M., Nelson, M.P., Schmitz, O.J., Smith, D.W.,
Wallach, A.D. & Wirsing, A.J. (2014) Status and ecological effects of the world’s
largest carnivores. Science, 343, 1241484.
Rutledge, L.Y., Devillard, S., Boone, J.Q., Hohenlohe, P.A., White, B.N., Drive, E.B.,
Canada, K.J. & Biome, L. De (2015) RAD sequencing and genomic simulations
resolve hybrid origins within North American Canis. Biology Letters, 11, 20150303.
Rutledge, L.Y., Garroway, C.J., Loveless, K.M. & Patterson, B.R. (2010a) Genetic
differentiation of eastern wolves in Algonquin Park despite bridging gene flow
between coyotes and grey wolves. Heredity, 105, 520–31.
Rutledge, L.Y., Holloway, J.J., Patterson, B.R. & White, B.N. (2009) An Improved Field
Method to Obtain DNA for Individual Identification From Wolf Scat. Journal of
Wildlife Management, 73, 1430–1435.
Rutledge, L.Y., Patterson, B.R., Mills, K.J., Loveless, K.M., Murray, D.L. & White, B.N.
(2010b) Protection from harvesting restores the natural social structure of eastern
wolf packs. Biological Conservation, 143, 332–339.
Rutledge, L.Y., White, B.N., Row, J.R. & Patterson, B.R. (2012b) Intense harvesting of
eastern wolves facilitated hybridization with coyotes. Ecology and evolution, 2, 19–
33.
Rutledge, L.Y., Wilson, P.J., Klütsch, C.F.C., Patterson, B.R. & White, B.N. (2012a)
Conservation genomics in perspective: A holistic approach to understanding Canis
evolution in North America. Biological Conservation, 155, 186–192.
52
Ryan, M.E., Johnson, J.R. & Fitzpatrick, B.M. (2009) Invasive hybrid tiger salamander
genotypes impact native amphibians. Proceedings of the National Academy of
Sciences of the United States of America, 106, 11166–11171.
Sacks, B.N., Bannasch, D.L., Chomel, B.B. & Ernest, H.B. (2008) Coyotes demonstrate
how habitat specialization by individuals of a generalist species can diversify
populations in a heterogeneous ecoregion. Molecular biology and evolution, 25,
1384–94.
Schwartz, M.K., Pilgrim, K.L., McKelvey, K.S., Lindquist, E.L., Claar, J.J., Loch, S. &
Ruggiero, L.F. (2004) Hybridization between Canada lynx and bobcats: Genetic
results and management implications. Conservation Genetics, 5, 349–355.
Seehausen, O. (2004) Hybridization and adaptive radiation. Trends in ecology &
evolution, 19, 198–207.
Seehausen, O., Takimoto, G., Roy, D. & Jokela, J. (2007) Speciation reversal and
biodiversity dynamics with hybridization in changing environments. Molecular
Ecology, 17, 30–44.
Stockwell, C.A., Hendry, A.P. & Kinnison, M.T. (2003) Contemporary evolution meets
conservation biology. Trends in Ecology and Evolution, 18, 94–101.
Stone, G. (2000) Phylogeography, hybridization and speciation. Trends in Ecology and
Evolution, 15, 354–355.
Stronen, A. V., Tessier, N., Jolicoeur, H., Paquet, P.C., Nault, M.H., Villemure, M.,
Patterson, B.R., Sallows, T., Goulet, G. & Lapointe, F.J. (2012) Canid hybridization:
Contemporary evolution in human-modified landscapes. Ecology and Evolution, 2,
2128–2140.
Tauleigne-Gomes, C. & Lefèbvre, C. (2008) Natural hybridisation between two coastal
endemic species of Armeria (Plumbaginaceae) from Portugal. 2. Ecological
investigations on a hybrid zone. Plant Systematics and Evolution, 273, 225–236.
Thornton, D.H. & Murray, D.L. (2014) Influence of hybridization on niche shifts in
expanding coyote populations. Diversity and Distributions, 20, 1355–1364.
Vähä, J.P. & Primmer, C.R. (2006) Efficiency of model-based Bayesian methods for
detecting hybrid individuals under different hybridization scenarios and with
different numbers of loci. Molecular Ecology, 15, 63–72.
van Dongen, W.F.D., Lazzoni, I., Winkler, H., Vásquez, R.A. & Estades, C.F. (2013)
Behavioural and genetic interactions between an endangered and a recently-arrived
hummingbird. Biological Invasions, 15, 1155–1168.
53
Verardi, A., Lucchini, V. & Randi, E. (2006) Detecting introgressive hybridization
between free-ranging domestic dogs and wild wolves (Canis lupus) by admixture
linkage disequilibrium analysis. Molecular Ecology, 15, 2845–2855.
Warren, D.L., Glor, R.E. & Turelli, M. (2008) Environmental niche equivalency versus
conservatism: Quantitative approaches to niche evolution. Evolution, 62, 2868–2883.
Warren, D.L., Glor, R.E. & Turelli, M. (2010) ENMTools: a toolbox for comparative
studies of environmental niche models. Ecography, 1, 607–611.
Way, J. (2007) A comparison of body mass of Canis latrans (coyotes) between eastern
and western North America. Northeastern Naturalist, 14, 111–124.
Way, J.G., Rutledge, L., Wheeldon, T. & White, B.N. (2010) Genetic Characterization of
Eastern “Coyotes” in Eastern Massachusetts. Northeastern Naturalist, 17, 189–204.
Wheeldon, T. & Patterson, B. (2012) Genetic and morphological differentiation of wolves
(Canis lupus) and coyotes (Canis latrans) in northeastern Ontario. Canadian Journal
of Zoology, 1230, 1221–1230.
Wheeldon, T., Patterson, B. & White, B. (2010) Colonization history and ancestry of
northeastern coyotes. Biology letters, 6, 246–7; author reply 248–9.
Whitham, T.G., Young, W.P., Martinsen, G.D., Catherine, A., Schweitzer, J.A., Shuster,
S.M., Wimp, G.M., Fischer, D.G., Bailey, J.K., Lindroth, R.L., Woolbright, S. &
Kuske, C.R. (2003) Community and Ecosystem Genetics : A Consequence of the
Extended Phenotype. Ecology, 84, 559–573.
Wilson, P., Grewal, S., Mallory, F. & White, B. (2009) Genetic Characterization of
Hybrid Wolves across Ontario. Journal of Heredity, 100, S80–S89.
Wilson, P.J., Grewal, S., Lawford, I.D., Heal, J.N., Granacki, A.G., Pennock, D.,
Theberge, J.B., Theberge, M.T., Voigt, D.R., Waddell, W., Chambers, R.E., Paquet,
P.C., Goulet, G., Cluff, D. & White, B.N. (2000) DNA profiles of the eastern
Canadian wolf and the red wolf provide evidence for a common evolutionary history
independent of the grey wolf. Canadian Journal of Zoology, 78, 2156–2166.
Wilson, R.R., Blankenship, T.L., Hooten, M.B. & Shivik, J.A. (2010) Prey-mediated
avoidance of an intraguild predator by its intraguild prey. Oecologia, 164, 921–929.
54
Supporting information 1
Genetic data sources
Published data
Benson, J.F., Patterson, B.R. & Wheeldon, T.J. (2012) Spatial genetic and morphologic
structure of wolves and coyotes in relation to environmental heterogeneity in a Canis
hybrid zone. Molecular ecology, 21, 5934–54.
Rutledge, L.Y., Garroway, C.J., Loveless, K.M. & Patterson, B.R. (2010) Genetic
differentiation of eastern wolves in Algonquin Park despite bridging gene flow
between coyotes and grey wolves. Heredity, 105, 520–31.
Way, J.G., Rutledge, L., Wheeldon, T. & White, B.N. (2010) Genetic Characterization of
Eastern “Coyotes” in Eastern Massachusetts. Northeastern Naturalist, 17, 189–204.
Wheeldon, T. & Patterson, B. (2012) Genetic and morphological differentiation of wolves
(Canis lupus) and coyotes (Canis latrans) in northeastern Ontario. Canadian Journal
of Zoology, 1230, 1221–1230.
Wheeldon, T., Patterson, B. & White, B. (2010) Colonization history and ancestry of
northeastern coyotes. Biology letters, 6, 246–7; author reply 248–9.
Wheeldon, T.J., Rutledge, L.Y., Patterson, B.R., White, B.N. & Wilson, P.J. (2013) Ychromosome evidence supports asymmetric dog introgression into eastern coyotes.
Ecology and evolution, 3, 3005–3020.
Unpublished data
Holloway, J. (2009) Size dependent resource use of a hybrid wolf (C. lycaon x C. lupus)
population in northeast Ontario.
Wheeldon, T. (2009) Genetic Characterization of Canis Populations in the Western Great
Lakes Region.
Wolf and Coyote DNA Bank at Trent University (http://wolf.nrdpfc.ca/)
Organizations/Institutions who contributed tissue samples
-Manitoba Conservation and Water Stewardship
(http://www.gov.mb.ca/conservation/index.html)
-New York State Department of Environmental Conservation (http://www.dec.ny.gov/)
-New York State Museum (https://www.nysm.nysed.gov/)
-North American Fur Auctions (http://www.nafa.ca/)
55
Supporting information 2
Additional details on genetic methods employed in this manuscript
Genetic analysis
All samples with more than 4 missing alleles were excluded from the analysis. The
dataset was screened with Allelematch 2.03 (Galpern et al., 2012), a package for R (R
Core Team, 2011), which allowed us to identify unique multilocus genotypes by grouping
similarities according to the “alleleMismatch-value” defined by quality of the complete
data set. The alleleMismatch-value evaluated was four alleles, meaning that every sample
that differed by fewer than 5 alleles or involved duplicate genotypes was removed from
the final dataset.
Removing grey wolf ancestry; assignment analysis
We removed animals with strong grey wolf ancestry by assessing the probability of
ancestry, i.e. Q-value, for all individuals using program Structure (Pritchard et al., 2000;
Falush et al., 2003). To sort individuals by species, we used standard groups provided by
the Natural Resources DNA Profiling and Forensic Centre (NRDPFC) at Trent University
and which are available on DRYAD (Rutledge et al., 2010a,
http://dx.doi.org/10.5061/dryad.q9d6s). These groups included as reference 43 grey
wolves from Northwest Territories and 50 coyotes from Saskatchewan, originating from
regions where resident canids have not hybridized with eastern wolves (Mech, 2011). As
a standard group for eastern wolf designation, we used the top 50, i.e. highest Q-value,
individuals assigned as eastern wolves, from the study of Rutledge et al. (2010). At first,
we ran the dataset including standard groups without prior population information (i.e. no
Popflag, which predefines genotype references for the different groups to be assigned), to
56
determine whether 3 groups was the appropriate choice of canid groups. We used the
algorithm developed by Evanno et al. (2005) to identify canid group clusters (K), as per
Structure harvester (http://taylor0.biology.ucla.edu/structureHarvester/). The statistics
derived from this procedure included the most appropriate K, as defined by the highest
Delta K value relative to the plateau value of MeanLnP(K) (see Earl & VonHoldt 2012).
The highest Delta K was at K=2, reflecting that eastern wolves and coyotes have new
world origin and are clearly distinct from old world (grey wolf) genotypes. Accordingly,
we adopted K=3 to reflect the comparable Mean LnP(K) statistics and Delta K (Fig.
S2.1). In addition, we used further isolated 50 individuals that were most highly assigned
to each of the 3 clusters as standard groups, by using the PopFlag designation, to validate
our K selection. The assignment results were the same with and without the predefined
reference groups, so here we only present the predefined, standard group method.
57
Figure S2.1. Plots of K determination criteria values, Delta K and Mean LnP(D) from 5
runs, for Structure analysis of eastern North America Canis microsatellite genotype data
based on 12 loci, including grey wolf, eastern wolf and eastern coyote reference groups.
We used the F-model (Admixture Ancestry model) with a PopFlag (i.e. USEPOPINFO)
in the program Structure to sort unknown individuals into one of three pre-assigned
reference groups. The burn-in period was set at 500 000 iterations followed by 106
Markov chain Monte Carlo (MCMC), for the three populations (K=3). We performed 10
runs and the output of the different runs for K=3 were sorted and averaged with CLUMPP
1.1.2 (Jakobsson & Rosenberg, 2007). Individuals were classified in a specific cluster at
Q ≥ 0.8 and considered as part of a parental group or as an admixed individual (i.e.
hybrid) if Q ≤ 0.8 (Fig. S2.2) (Vähä & Primmer, 2006, Rutledge et al., 2010a). All
individuals assigned in the grey wolf cluster, and all admixed individuals with Q ≥ 0.2
(i.e. Q = 1 (-) Q ≤ 0.8) for grey wolf were removed (Fig. S2.2). For the next analysis,
58
individuals assigned as coyote were kept inside our data set, but the 50 coyotes used as
standard group were removed because they were out of our study area, i.e. Saskatchewan.
Eastern wolf
Coyote
Grey wolf
1
Q-value
0.8
0.6
0.4
0.2
0
Figure S2.2. Plot of 1325 individuals proportional membership to K=3 genetic clusters
inferred by Structure. Verticals lines represent individuals with their proportional ancestry
(Q-value) for each parental group represented by different colors and the shaded box
captured individuals having Q ≥ 0.2 for grey wolf that were removed from the analysis.
Clustering eastern wolves and eastern coyotes
The 648 remaining individuals were reassigned with the F-model, but without PopFlag
to allow the software to find its own population division. The burn-in period was set at
500 000 iterations followed by 500 000 iterations for K=1-10 for 5 repetitions. The more
representative K was then K=2 (Evanno et al., 2005; Earl & VonHoldt, 2012; Fig. S2.3)
so we ran Structure at K=2 with a burn-in period of 500 000 iteration followed by 106
iterations for 10 runs. Again, following sorting of the different runs in CLUMMP
(Jakobsson & Rosenberg, 2007) individuals were classified in a parental group at Q ≥ 0.8
and as an admixed individual if Q < 0.8 (Fig. S2.4).
59
Figure S2.3. Plots of K determination criteria values, Delta K and Mean LnP(K) for
Structure analysis of eastern North America Canis microsatellite genotype data based on
12 loci, including eastern wolf and eastern coyote ancestry only.
Eastern coyote
Eastern wolf
1
Q-value
0.8
0.6
0.4
0.2
0
Figure S2.4. Plot of 648 individuals proportional membership to K=2 genetic cluster
inferred by Structure. Verticals lines represent individuals with their proportional ancestry
(Q-value) for each parental group represented by different colors and the shaded box
captured hybrid individuals; Q ≤ 0.8 for both parental groups.
60
Confirming the Structure assignment
Structure is a Bayesian approach that assumes that populations are in Hardy-Weinberg
equilibrium, which may not always be the case for natural populations (Jombart et al.,
2009). Accordingly, we used a multivariate ordination method that relaxes the HardyWeinberg assumption using principal component analysis (PCA) implemented in the
adegenet package (Jombart, 2008) for R (R Core Team, 2013). The PCA was centered,
but not scaled and only the first two axes were kept to plot the data (Fig. S2.5). Results
from the PCA were qualitatively consistent with the Structure results, confirming our
choice of K=2 parental canid groups with a third hybrid group. We acknowledge the
possible variability in individual assignment depending on the threshold used, but based
on our PCA, we would not expect to have any individual from a parental group to move
into the other parental group, but maybe we could have different number of individuals
from parental considered as hybrids under a different threshold.
61
Figure S2.5. Individual eastern wolves, eastern coyotes and their hybrids in eastern North
America arranged along axes 1 and 2 of principal components analysis (PCA), which
explain 8.3% and 4.8% of the total variation, respectively. PCA was used to corroborate
the individual assignments to genetic clusters and admixed category made at K=2 with
results from Structure. Eastern wolves (EW) are represented in orange and eastern coyote
(EC) in blue with their hybrids (HYB) in red, with 95% confidence ellipses. The PCA
shows that highly assigned EW and EC at K=2 do not overlap, but their hybrids have
more variability.
62
Literature Cited
Earl, D.A. & vonHoldt, B.M. (2012) STRUCTURE HARVESTER: A website and
program for visualizing STRUCTURE output and implementing the Evanno method.
Conservation Genetics Resources, 4, 359–361.
Evanno, G., Regnaut & S., Goudet, J. (2005) Detecting the number of clusters of
individuals using the software STRUCTURE: a simulation study. Molecular
ecology, 14, 2611–20.
Falush, D., Stephens, M. & Pritchard, J.K. (2003) Inference of population structure using
multilocus genotype data: linked loci and correlated allele frequencies. Genetics,
164, 1567–87.
Galpern, P., Manseau, M., Hettinga, P., Smith, K. & Wilson, P. (2012) Allelematch: an R
package for identifying unique multilocus genotypes where genotyping error and
missing data may be present. Molecular ecology resources, 12, 771–8.
Jakobsson, M. & Rosenberg, N.A. (2007) CLUMPP: A cluster matching and permutation
program for dealing with label switching and multimodality in analysis of population
structure. Bioinformatics, 23, 1801–1806.
Jombart, T., Devillard, S., Dufour, A.-B. & Pontier, D. (2008) Revealing cryptic spatial
patterns in genetic variability by a new multivariate method. Heredity, 101, 92–103.
Jombart, T., Pontier, D. & Dufour, A.-B. (2009) Genetic markers in the playground of
multivariate analysis. Heredity, 102, 330–341.
Mech, L.D. (2011) Non-Genetic Data Supporting Genetic Evidence for the Eastern Wolf.
Northeastern Naturalist, 18, 521–526.
Pritchard, J.K., Stephens, M. & Donnelly, P. (2000) Inference of population structure
using multilocus genotype data. Genetics, 155, 945–59.
R Core Team. (2011) R: A language and environment for statistical computing. R
Foundation for Statistical Computing, Vienna, Austria.
Rutledge, L.Y., Garroway, C.J., Loveless, K.M. & Patterson, B.R. (2010) Genetic
differentiation of eastern wolves in Algonquin Park despite bridging gene flow
between coyotes and grey wolves. Heredity, 105, 520–31.
Vähä, J.P. & Primmer, C.R. (2006) Efficiency of model-based Bayesian methods for
detecting hybrid individuals under different hybridization scenarios and with
different numbers of loci. Molecular Ecology, 15, 63–72.
63
Supporting information 3
Niche model development details
Our broad objective included modeling the environmental niche of each canid group
across the entire study area. Hybrids and eastern wolves do not occupy the entire study
region, and efforts would model beyond the spatial extent of our observations and thereby
potentially lead to model mis-specification (see Lonergan, 2014). However, our initial
approach involved clamping data in MaxEnt to restrict the extrapolation of values and
thereby minimize speculation outside the calibration range of the species (Phillips et al.
2006; Anderson & Raza, 2010). Then, we evaluated spatial extent and importance of
clamping by assessing the multivariate environmental similarity surface (MESS; Elith et
al., 2010) calculated for each model projection. The MESS calculation shows how similar
a given point or location is to a reference set of points, given a certain set of predictor
variables (Elith et al. 2010). The MESS calculates, with respect to the predictor climatic
variables used in this analysis, how similar a point is to a reference set of points (in this
case, the reference set corresponds to the buffered minimum convex polygon specific at
each species). For the three Canis, the MESS maps reveals that there was limited
extrapolation to novel environments, which improves confidence in the models (Fig.
S3.1).
64
Eastern wolves
Eastern coyotes
Hybrids
Figure S3.1. MESS maps for eastern wolves, eastern coyotes and hybrids. Negative
values (shown in red) indicate novel climate (where values of one or more predictors fall
outside the range of those layers in the buffered minimum convex polygon of the
species). Positive values (shown in blue) indicate values more similar to the median
values in the buffered minimum convex polygon of the species.
65
In addition, we also restricted the background to a buffered area (200 km) around the presence
data of the focal canid group to further calibrate models to known environmental variable values
(Anderson & Raza, 2010). Finally, models were projected across the study area, defined as a
buffered minimum convex polygon containing all canid locations. Also, to balance goodness-offit and model complexity we used R package ENMeval (Muscarella et al., 2014) to define the
optimal feature class (linear (L), quadratic (Q), product (P), threshold (T) and hinge (H))
combinations and to find the best regularization coefficients (default = 1) of MaxEnt models.
This approach executes MaxEnt models across a range of feature-class combinations and
regularization multipliers to calculate different evaluation criteria based on spatially-independent
test data; this approach promotes model generality while also facilitating projections across time
and space (Radosavljevic & Anderson, 2013; Muscarella et al., 2014). First, we partitioned
occurrence localities into testing and training bins (folds) for 10-fold cross-validation using the
“block” method, which use four spatially independent bins. We tested possible combinations of
models using regularization coefficients ranging from 0.5 to 10 (increment of 0.5), and five
feature class combinations (L, LQ, H, LQH, LQHP) for a total of 100 possible combinations.
resulting in 4000 individual model runs. However, it is notable that because we worked on a
relatively small geographic extent with relatively few observations (50 to 300 depending the
canid group), we reduced the number of background points to 1000 (default 10 000) in order to
be properly weighted against observations (Barbet-Massin et al., 2012).
Finally, with the lack of consensus according to the most appropriate metric to evaluate
performance of MaxEnt models (Lobo et al., 2008; Peterson et al., 2011; Warren & Seifert,
2010), and following the recommendation of Muscarella et al. (2014), we evaluated model
66
performance using three additional metrics to the default threshold-independent value of AUC
(area under the curve, Philips et al., 2006). To quantify over-fitting, we used the difference
between training and testing AUC (AUC.DIFF, Warren & Seifert, 2010) and omission rate using
the lowest presence threshold (ORmin, Peterson et al., 2011; Radosavljevic & Anderson, 2013).
In addition, we used change in Akaike information criterion corrected for small sample sizes
(∆AICc) to evaluate goodness-of-fit as well as model complexity (Burnham & Anderson, 2004;
Warren & Seifert, 2010). The best model would be the one that optimizes model performance and
avoids over-fitting which will be in order of importance a model having the 1) lowest
AUC.DIFF, 2) lowest ORmin, 3) highest Mean AUC and, 4) lowest ∆AICc. Therefore, the
feature class combination and the regularization coefficient resulting with the best model for each
species were used as the final model in analysis (see Fig. S3.2). The final models were run for 10
replicates used during the cross-validation process, and projected over the range of the study area
to produce our suitability score maps to characterize niche ecology of canids in northeastern
North America.
67
Eastern wolves
Eastern coyotes
Hybrids
Figure S3.2) Four different evaluation metrics for Maxent output models; omission rate
using the lowest presence threshold (Mean.ORmin), difference between training and
testing area under the curve (Mean.AUC.DIFF), evaluation AUC (Mean.AUC) and delta
of Akaike information criterion corrected for small sample sizes (delta.AICc). Arrows
indicate the final setting for each species; eastern wolves (LQ 0.5), eastern coyotes (LQ
1.5) and hybrids (LQ 1.0).
68
Literature Cited
Anderson, R.P. & Raza, A. (2010) The effect of the extent of the study region on GIS
models of species geographic distributions and estimates of niche evolution:
Preliminary tests with montane rodents (genus Nephelomys) in Venezuela. Journal
of Biogeography, 37, 1378–1393.
Barbet-Massin, M., Jiguet, F., Albert, C.H. & Thuiller, W. (2012) Selecting pseudoabsences for species distribution models: How, where and how many? Methods in
Ecology and Evolution, 3, 327–338.
Burnham, K.P. & Anderson, D.R. (2004) Multimodel Inference: Understanding AIC and
BIC in Model Selection. Sociological Methods & Research, 33: 261–304,
Elith, J., Kearney, M. & Phillips, S. (2010) The art of modelling range-shifting species.
Methods in Ecology and Evolution, 1, 330–342.
Lobo, J.M., Jiménez-valverde, A. & Real, R. (2008) AUC: A misleading measure of the
performance of predictive distribution models. Global Ecology and Biogeography,
17, 145–151.
Lonergan, M. (2014) Modelling beyond data is uninformative: A comment on “Statespace modelling reveals proximate causes of harbour seal population declines” by
Matthiopoulos et al. Oecologia, 175, 1063–1067.
Muscarella, R., Galante, P.J., Soley-Guardia, M., Boria, R.A., Kass, J.M., Uriarte, M. &
Anderson, R.P. (2014) ENMeval: An R package for conducting spatially
independent evaluations and estimating optimal model complexity for MaxEnt
ecological niche models. Methods in Ecology and Evolution, 5, 1198–1205.
Peterson, A.T., Soberon, J., Pearson, R.G., Anderson, R.P., Martinez-Meyer, E.,
Nakamura, M. & Araujo, M.B. (2011) Ecological niches and geographic
distributions. Princeton University Press, Princeton.
Phillips, S.J., Anderson, R.P. & Schapire, R.E. (2006) Maximum entropy modeling of
species geographic distributions. Ecological Modelling, 190, 231–259.
Radosavljevic, A. & Anderson, R.P. (2013) Making better MaxEnt models of species
distributions: Complexity, overfitting and evaluation. Journal of Biogeography, 41,
629–643.
Warren, D.L. & Seifert, S.N. (2010) Ecological niche modeling in MaxEnt: the
importance of model complexity and the performance of model selection criteria.
Ecological Applications, 21, 335–342.
69