ArticlePDF Available

Abstract and Figures

Aim Delimiting recently diverged species is challenging. During speciation, genetic differentiation may be distributed unevenly across the genome, as different genomic regions can be subject to different selective pressures and evolutionary histories. Reliance on limited numbers of genetic markers that may be underpowered can make species delimitation even more challenging, potentially resulting in taxonomic inconsistencies. Rockhopper penguins of the genus Eudyptes comprise three broadly recognized taxa: northern (E. moseleyi), southern (E. chrysocome) and eastern rockhopper (E. filholi). Their taxonomic status has been controversial for decades, with researchers disagreeing about whether E. chrysocome and E. filholi are distinct species or conspecific. Our goal is to evaluate genome‐wide patterns of divergence to evaluate genetic differentiation and species delimitation in rockhopper penguins, and to assess which mechanisms may underlie previous discordance among nuclear versus mitochondrial analyses. Location Sub‐Antarctic and temperate coastal regions of the Southern Hemisphere. Methods We generated reduced‐representation genomic libraries using double digest restriction‐site associated DNA (ddRAD) sequencing to evaluate genetic differentiation, contemporary migration rates and admixture among colonies of rockhopper penguins. Results The extent of genetic differentiation among the three taxa was consistently higher than population‐level genetic differentiation found within these and other penguin species. There was no evidence of admixture among the three taxa, suggesting the absence of ongoing gene flow among them. Species delimitation analyses based on molecular data, along with other lines of evidence, provide strong support for the taxonomic distinction of three species of rockhopper penguins. Main conclusions Our results provide strong support for the existence of three distinct species of rockhopper penguins. The recognition of this taxonomic diversity is crucial for the management and conservation of this widely distributed species group. This study illustrates that widespread dispersive seabird lineages lacking obvious morphological differences may nevertheless have complex evolutionary histories and comprise cryptic species diversity.
This content is subject to copyright. Terms and conditions apply.
Diversity and Distributions. 2021;27:2277–2296.
Received: 26 April 2021 
  Revised: 3 August 2021 
  Accepted: 4 August 2021
DOI: 10.1111/ddi.13399
Taxonomy based on limited genomic markers may
underestimate species diversity of rockhopper penguins and
threaten their conservation
María José Frugone1,2,3 | Theresa L. Cole4,5 | María Eugenia López6|
Gemma Clucas7,8 | Pável Matos- Maraví9| Nicolás A. Lois10,11 | Pierre Pistorius12|
Francesco Bonadonna13| Phil Trathan14| Andrea Polanowski15| Barbara Wienecke15|
Andrea Raya- Rey16,17,18| Klemens Pütz19| Antje Steinfurth20,21| Ke Bi22|
Cynthia Y. Wang- Claypool22| Jonathan M. Waters4| Rauri C. K. Bowie22 |
Elie Poulin1,2 | Juliana A. Vianna23
1Laboratorio de Ecología Molecular, Depar tamento de Ciencias Ecológicas, Facult ad de Ciencias, Universidad de Chile, Santiago, Chile
2Instituto de Ecología y Biodiversidad (IEB), Santiago, Chile
3Instituto de Ciencias Ambientales y Evolutivas, Facultad de Ciencias, Universidad Austral de Chile, Valdivia, Chile
4Department of Zoology, University of Otago, Dunedin, New Zealand
5Department of Biology, Ecology and Evolution, Universit y of Copenhagen, Copenhagen, Denmark
6Department of Aquatic Resources, Swedish University of Agricultural Sciences, Drottningholm, Sweden
7Atkinson Center for a Sustainable Future, Cornell University, Ithaca, NY, USA
8Cornell Lab of Ornithology, Cornell Universit y, Ithaca, NY, USA
9Biology Centre of the Czech Academy of Sciences, Institute of Entomology, České Budějovice, Czech Republic
10Departamento de Ecología, Genética y Evolución, Facultad de Ciencias Exactas y Naturales, Universidad de Buenos Aires, Buenos Aires, Argentina
11Instituto de Ecología Genética y Evolución de Buenos Aires, Consejo Nacional de Investigaciones Científicas y Técnicas, Buenos Aires, Argentina
12DST/NRF Centre of Excellence at the Percy FitzPatrick Institute for African Ornithology, Department of Zoology, Nelson Mandela University, Port Elizabeth,
South Africa
13CEFE, Univ Montpellier, CNRS, EPHE, IRD, Montpellier, France
14British Antarctic Survey, Cambridge, UK
15Australian Antarctic Division, Kingston, Tasmania, Australia
16Centro Austral de Investigaciones Científicas – Consejo Nacional de Investigaciones Científicas y Técnicas (CADIC- CONICET), Ushuaia, Argentina
17Wildlife Conservation Society, Bronx, NY, USA
18Instituto de Ciencias Polares, Ambiente y Recursos Naturales, Universidad Nacional de Tierra del Fuego, Ushuaia, Argentina
19Antarctic Research Trust, Bremervörde, Germany
20FitzPatrick Institute of African Ornithology, University of Cape Town, Rondebosch, South Africa
21RSPB Centre for Conservation Science, Cambridge, UK
22Museum of Ver tebrate Zoology and Department of Integrative Biology, University of California, Berkeley, CA, USA
23Pontificia Universidad Católica de Chile, Center for Genome Regulation, Facultad de Agronomía e Ingeniería Forestal, Departamento de Ecosistemas y Medio
Ambiente, Santiago, Chile
This is an open access article under the terms of the Creat ive Commo ns Attri bution License, which permits use, distribution and reproduction in any medium,
provided the original work is properly cited.
© 2021 The Authors. Diversity and Distributions published by John Wiley & Sons Ltd.
   FRUGONE Et al.
Accurate delimitation of species is essential for the management
and conservation of biodiversity (Agapow et al., 2004; Frankham
et al., 2012). Indeed, conservation action plans and assessments
are often based on biological and demographic attributes focused
primarily at the species level (Agapow et al., 2004). Any failure to
recognize cryptic species diversity could lead to overestimates of
population and range sizes, and thereby misinform the assessment of
threat levels and compromise the prioritization of conservation re-
sources (Bickford et al., 2007). Conversely, over- splitting of lineages
through inaccurate recognition of multiple species might impair
management through the exclusion of potentially valuable source
stocks that could otherwise be used for relocations or genetic res-
cue efforts (Frankham et al., 2012). Furthermore, with limited fund-
ing for biological conservation, over- splitting dilutes and misdirects
available resources (Agapow et al., 2004).
Despite its importance, species delimitation can be a challenging
task (Fujita et al., 2012; Sangster, 2014). The root of this problem
lies in the biological complexity and variability accompanying the
speciation process, and how we define a species (De Queiroz, 2007;
Schilthuizen, 2000). Consequently, there is no universal species
Juliana A . Vianna, Pontificia Universidad
Católica de Chile, Facultad de Agronomía
e Ingeniería Forestal, Departamento de
Ecosistemas y Medio Ambiente, Vicuña
Mackenna 4860, Macul, Santiago, Chile.
Funding information
Instituto Antártico Chileno (INACH),
Grant/Award Number: INACH DT- 11_17
and INACH RT_12– 14; Czech Science
Foundation Junior, Grant/Award Number:
GAČR grant (GJ20- 18566Y); Czech Academy
of Sciences, Grant/Award Number:
PPLZ program (L200961951); Comisión
Nacional de Investigación Científica y
Tecnológica, Grant/Award Number: PIA
ACT172065 GAB; Beca Doctorado Nacional
2016, Grant/Award Number: 2116038;
Fondo Nacional de Desarrollo Científico
y Tecnológico, Grant/Award Number:
1150517 and 1151336
Editor: April Blakeslee
Aim: Delimiting recently diverged species is challenging. During speciation, genetic
differentiation may be distributed unevenly across the genome, as different genomic
regions can be subject to different selective pressures and evolutionary histories.
Reliance on limited numbers of genetic markers that may be underpowered can make
species delimitation even more challenging, potentially resulting in taxonomic in-
consistencies. Rockhopper penguins of the genus Eudyptes comprise three broadly
recognized taxa: northern (E. moseleyi), southern (E. chrysocome) and eastern rock-
hopper (E. filholi). Their taxonomic status has been controversial for decades, with
researchers disagreeing about whether E. chrysocome and E. filholi are distinct spe-
cies or conspecific. Our goal is to evaluate genome- wide patterns of divergence to
evaluate genetic differentiation and species delimitation in rockhopper penguins, and
to assess which mechanisms may underlie previous discordance among nuclear ver-
sus mitochondrial analyses.
Location: Sub- Antarctic and temperate coastal regions of the Southern Hemisphere.
Methods: We generated reduced- representation genomic libraries using double
digest restriction- site associated DNA (ddRAD) sequencing to evaluate genetic dif-
ferentiation, contemporary migration rates and admixture among colonies of rock-
hopper penguins.
Results: The extent of genetic differentiation among the three taxa was consistently
higher than population- level genetic differentiation found within these and other
penguin species. There was no evidence of admixture among the three taxa, sug-
gesting the absence of ongoing gene flow among them. Species delimitation analyses
based on molecular data, along with other lines of evidence, provide strong support
for the taxonomic distinction of three species of rockhopper penguins.
Main conclusions: Our results provide strong support for the existence of three dis-
tinct species of rockhopper penguins. The recognition of this taxonomic diversity
is crucial for the management and conservation of this widely distributed species
group. This study illustrates that widespread dispersive seabird lineages lacking obvi-
ous morphological differences may nevertheless have complex evolutionary histories
and comprise cryptic species diversity.
Eudyptes, genomics, rockhopper penguins, species delimitation
concept that can be applied to all living organisms to confidently
delimit species. Additionally, the most basic assumption of species
delimitation strategies is that “true species” will exhibit variation in
traits of different biological dimensions (e.g. genetics, behaviour,
morphology, ecology). This carries an additional level of complexity,
as different traits along these biological dimensions could diverge
at different times during the speciation process (De Queiroz, 2007).
Furthermore, variation in traits may be the result of local adapta-
tion or phenotypic plasticity and not be indicative of the speciation
process (Frugone et al., 2019; Mason & Taylor, 2015). Thus, inte-
grative taxonomy, in which traits from multiple lines of evidence
(ecology, genomics, phenotype, behaviour, geography) aid the iden-
tification of divergent lineages, has become a widely accepted ap-
proach to delimiting species (Cicero et al., 2021; De Queiroz, 2007;
Sangster, 2014). Using this approach, it is essential to understand the
underlying processes generating trait variation (or the lack of it) that
may be responsible for discordances between independent lines of
Molecular data and the availability of sophisticated analyt-
ical methods (e.g. coalescent- based species delimitation; Fujita
et al., 2012) have provided a useful tool for uncovering cryptic di-
versity and delimiting species. However, discordances may arise
among different sets of nuclear markers or between nuclear and
mitochondrial DNA (mtDNA) when delimiting species, especially
for recently diverged lineages (Pedraza- Marrón et al., 2019; Spinks
et al., 2014). The nuclear genomes of recently diverged species may
contain varying levels of genetic differentiation dispersed through-
out the genome (Nosil et al., 2009; Wu, 2001). Even under complete
reproductive isolation, genomic regions with reduced mutation
rates, or which are subject to stabilizing selection, may remain con-
served for a considerable length of time compared to other regions
with higher mutation rates or those under divergent selection (Nosil
et al., 2009). Furthermore, when speciation occurs in the presence
of gene flow, different regions of the genome may either become
homogenized, or remain differentiated if introgressed alleles carry
negative fitness consequences, and if selection is stronger than
gene flow (Bazykin, 1969; Key, 1968; Wu, 2001). This may lead to
genomic “islands” of differentiation against a highly undifferenti-
ated genomic background (Cutter, 2013; Feder et al., 2012; Nosil
et al., 2009; Wu, 2001). Therefore, to some extent, discrepancies in
species delimitation can be explained by introgressive hybridization
and Incomplete Lineage Sorting (ILS) across segments of the genome
(Cutter, 2013; Maddison, 1997; Pedraza- Marrón et al., 2019).
Discrepancies may also occur between mtDNA and nu-
clear markers due to differences in selective regimes and sex-
biased gene flow, or more commonly, because of the relative
degree of divergence among lineages for each type of marker
(Edwards & Bensch, 2009; Toews & Brelsford, 2012; Zink &
Barrowclough, 2008). The effective population size is four times
smaller for mtDNA than for nuclear DNA, due to intrinsic differ-
ences in ploidy and the maternal mode of inheritance of mtDNA.
Consequently, monophyly after diversification will occur for
mtDNA before nuclear DNA (Zink & Barrowclough, 2008). In
this sense, mtDNA may exhibit greater differentiation and be
better suited for detecting recently diverged taxa than nu-
clear DNA, especially if just a few nuclear genetic markers are
used to assess species- and population- level divergence (Zink &
Barrowclough, 2008). One way to overcome such limitations, and
to estimate the underlying causes of discordance between phy-
logenetic hypotheses generated using mtDNA and nuclear DNA,
is through the use of thousands of unlinked genetic markers. This
multilocus approach encompasses regions subject to a variety of
evolutionary constraints in order to infer genome- wide patterns of
divergence (Esquerre et al., 2019; Guo et al., 2019).
Eudyptes is the most species- rich penguin genus (6– 8 extant spe-
cies), with “rockhopper penguins” comprising three recently diverged
lineages: the subtropical taxon E. moseleyi and the sub- Ant arc tic ta xa
E. chrysocome and E. filholi (Figure 1). Using the fossil record or dates
of geological events, the oldest diversification time across different
studies for the split between E. moseleyi an d the sub- A nta rctic taxa is
3.06 Mya, while the divergence of the sub- Antarctic taxa from each
other occurred from 0.5 to 2.26 Mya (Cole, Dutoit, et al., 2019; Cole,
Ksepka, et al., 2019; de Dinechin et al., 2009; Frugone et al., 2018;
Gavryushkina et al., 2017; Vianna et al., 2020). The taxonomic status
of rockhopper penguins has been repeatedly revised. In the 1990s,
these three lineages were considered to represent a single species
(Martínez, 1992). In 2006, based on differences in their mating calls,
timing of breeding (starting ~2 months earlier in E. moseleyi), ge-
netics (mtDNA; HVRI) and morphological characters, rockhoppers
were classified into two species (García & Boersma, 2013; Jouventin
et al., 2006): the south ern rockhoppe r (E. chr ysocome) and the nort h-
ern rockhopper (E. moseleyi), hereafter the two- species hypothesis.
In the same year, based on genetic evidence (mtDNA; 12S, cytb and
COI), Banks et al. (2006) proposed that rockhopper penguins should
instead be classified as three species (hereafter the three- species
hypothesis) comprising E. moseleyi, and E. chrysocome split into two
species: the eastern rockhopper, E. filholi distributed on islands in
the Indian and western Pacific oceans, and the southern rockhopper,
E. chrysocome occupying islands in the eastern Pacific and Atlantic
oceans (Figure 1).
The three- species hypothesis has been supported by recent
studies including more extensive geographic sampling using mtDNA
(Cole, Dutoit, et al., 2019; Cole, Ksepka, et al., 2019; Frugone
et al., 2018). Conflicting with the above, Mays et al. (2019) suggested
that E. filholi and E. chrysocome should be considered conspecific,
in agreement with the two- species hypothesis, as their species de-
limitation analyses did not find sufficient differences to discriminate
among the taxa, and analyses of migration rates between colonies
suggested ongoing gene flow. However, such conclusions were
based on the analysis of only six nuclear introns, and their species
delimitation analysis could not discriminate the even more diver-
gent and widely accepted split (around 5 Mya) between macaroni
(E. chrysolophus) and rockhopper penguins; only one out of their four
Genealogical Diversity Index (GDI) scenarios exceeded the 0.2 GDI
threshold for recognizing potential species, but it did not reach the
0.7 GDI threshold for a strongly divergent species lineage. With a
   FRUGONE Et al.
different set of nuclear introns, Frugone et al. (2018) were also un-
able to discriminate macaroni and rockhopper penguins, nor could
they discriminate among lineages within rockhopper penguins, de-
spite their strong divergence and reciprocal monophyly with mtDNA
If E. chrysocome and E. filholi are conspecific and are not re-
productively isolated, discordances between mtDNA and nuclear
markers might be explained by skewed dispersal (and gene flow)
towards males (i.e. with females being more philopatric). Another
possibility is that the chosen nuclear markers correspond to un-
differentiated regions of the genomes of the taxa under study, a
situation that may arise from different mechanisms, such as intro-
gression or ILS. Unravelling the causes of discordance between
mtDNA and nuclear DNA is important to more accurately infer
species delimitations that could ultimately reinforce biodiversity
conservation. Our reassessment of the evolutionary history of
rockhopper penguins using a reduced genomic representation
library of single nucleotide polymorphisms (SNPs), resulting in
thousands of genetic markers, allow us to evaluate whether the
genomic pattern of divergence is in accordance with that previ-
ously shown by mtDNA or if other processes, unrelated to ge-
nomic coverage, are responsible. In addition to evaluating genetic
differentiation and species delimitation of E. moseleyi, E. filholi and
E. chrysocome, we also evaluate contemporary migration rates
and search for evidence of admixture based on prior suggestions
about ongoing gene flow among rockhopper penguins (Mays
et al., 2019). Finally, we explore regional population genomics
within each taxon to improve our understanding of the extent of
intra- and interspecific genetic differentiation.
Currently, the IUCN recognizes two rockhopper penguin species,
E. moseleyi and E. chrysocome, categorizing them as endangered and
vulnerable, respectively (BirdLife International, 2020a, 2020b). The
conservation actions proposed by the IUCN include the determina-
tion of the taxonomic status of E. filholi and E. chrysocome, repre-
senting a key data source for the development of conservation plans
for these charismatic birds.
2.1 | Blood sampling and ddR AD- seq library
We collected 96 blood samples across an extensive area cover-
ing nearly the entire distribution of the three rockhopper taxa: 24
E. moseleyi (n = 12 from Nightingale Island, Tristan da Cunha; n = 12
from Amsterdam Island); 30 E. chrysocome (n = 13 from Terhalten
Island, Chile; n = 13 from Staten Island, Argentina; n = 4 from the
Falkland/Malvinas Islands); and 42 E. filholi (n = 14 from Marion
Island; n = 10 from Crozet Island; n = 10 from Kerguelen Island; n = 8
FIGURE 1 Sample locations for
rockhopper penguins throughout their
distribution. Eudyptes filholi (eastern
rockhopper), E. chrysocome (southern
rockhopper) and E. moseleyi (northern
from Macquarie Island; Figure 1, Table S1). Access to penguin colo-
nies, permission to collect blood samples and animal ethics approv-
als were granted by the responsible authorities for each sampling
location (Table S2). We used blood samples to prepare double digest
restriction- site associated DNA (ddRAD) libraries, following the pro-
tocol described in Peterson et al. (2012). Laboratory procedures and
the library preparation protocol are described in the Appendix S1
Methods Section II.1.
2.2 | ddRAD- seq data processing, SNP
calling and filtering
We used a custom PERL pipeline encompassing various external
programmes for processing the ddRAD- seq data (https://github.
com/CGRL- QB3- UCBer keley/ RAD). Raw fastq reads were first de-
multiplexed based on the sequences of internal barcodes with a
tolerance of one mismatch. De- multiplexed reads were removed if
the expected cutting sites were not found at the beginning of the
sequences, allowing for one mismatch. The reads were then filtered
using CUTADAPT V. 1.8.1 (Martin, 2011) and TRIMMOMATIC V.
0.36 (Bolger et al., 2014) to trim adapter contamination and low-
quality reads, respectively.
To improve the efficiency and accuracy of short- read mapping,
we used reference genomes for aligning ddRAD- seq data (Shafer
et al., 2017). We used the genomes of E. moseleyi (mean coverage:
24x, 16,344 Scaffolds, N50 length: 5,071,598), E. chrysocome (mean
coverage: 31x, 21,917 Scaffolds, N50 length: 5,071,598) and E. fil-
holi (mean coverage: 26x, 19,210 Scaffolds, N50 length: 5,071,598)
from Vianna et al. (2020). For our population genomic analyses
(i.e. analyses performed within each of the three taxa; Table S3),
we aligned the reads of each taxon to its own reference genome
(Vianna et al., 2020). For analyses involving only E. chrysocome and
E. filholi, we generated a second dataset, aligning the reads to E. mo-
seleyi genome (see Table S3). For analyses involving the three taxa
(see Table S3), reads from all individuals were aligned to an E. chrys-
olophus reference genome (mean coverage: 28x, 18,969 scaffolds,
N50 length: 5,071,598; Frugone et al., 2019; Vianna et al., 2020).
All alignments were performed using the bwa mem algorithms (Li &
Durbin, 2009) sorting and indexing bam files using SAMTOOLS V.
1.9 (Li et al., 2009).
Single nucleotide polymorphisms calling was conducted using
gstacks (Catchen et al., 2013). We used the populations programme
in STACKS V. 2.52 (Catchen et al., 2013) and VCFTOOLS V. 0.1.13
(Danecek et al., 2011) to filter loci. For this step, we used different cri-
teria following the assumptions and recommendation for each analy-
sis as documented in Benestan et al. (2016) and O’Leary et al. (2018).
For all datasets, we filtered out sites with observed heterozygosity
>0.5 and those with a mean depth of <10x or >175x to avoid unre-
liable genotypes or SNPs called from repetitive regions of the ge-
nome (indicated by the hump in distribution of locus depths above
175x, Figure S1). We removed sites that exhibited >20% of missing
data within each population/grouping option, aiming for a similar
distribution of missing data across them. We retained sites that were
present in >70% of individuals in a population/grouping option, as
established on the population map from STACKS (Table S3). Mean
depth and missing data across sites and groups were calculated for
each dataset using VCFTOOLS and then filtered using the blacklist
in STACKS. These filtering criteria were used to generate Manhattan
plots and to perform FINERADSTRUCTURE and OUTFL ANK anal-
yses (see below). We refer to the datasets obtained following these
filtering criteria as the “full datasets.”
To perform STRUCTURE, PCA, pairwise FST and genetic diver-
sity analyses, we further filtered our datasets, as several population
genomic analyses required loci to be in Hardy– Weinberg equilibrium
(HWE) and at linkage equilibrium (e.g. Pritchard et al., 2000, 2010).
We used PLINK V. 1.9 (Purcell et al., 2007) to remove loci in link-
age disequilibrium calculated using a window size of 50 SNPs, two
SNPs to shift the window at each step and a variance inflation factor
of two. Then, using VCFTOOLS, we calculated the deviation from
HWE and removed sites that exhibited significant deviations (after
FDR correction) in over ~50% of the populations. We retained sites
exhibiting a minimum allele frequency (MAF) of 0.05. We refer to
these datasets as “unlinked datasets.” For BAYESASS, SNAPP and
Bayes factor delimitation (BFD*) described below, we did not filter
by MAF and for SNAPP and BFD*, we retained only sites that were
present across all individuals. Detailed descriptions for all filtering
options are summarized in Table S3. Also, we evaluated whether our
dataset included loci under selection, as several analyses of popu-
lation genetics assume neutrality of the data. We ran OUTFLANK
V. 0.2 (Whitlock & Lotterhos, 2015) in R 3.6.3 (R Core Team, 2020)
to detect sites that could be under selection. We conducted these
outlier analyses for all pairwise comparisons among the three taxa
using the unlinked dataset, and also for all intraspecific datasets. In
OUTFLANK, we set a trim factor (left and right) of 0.05, a minimum
heterozygosity of 0.1 and a q threshold of 0.05. We used PGDSPIDER
V. to transform some VCF files into input files for multiple
progra ms (Lisch er & Excoffier, 2012). Fi nally, a preliminar y PCA anal-
ysis revealed the presence of three outliers corresponding to E. filholi
individuals from Kerguelen Island (Figure S2a; Appendix S1 Methods
Section II.2). We further explored the genetic relatedness of these
outliers to all other sampled individuals (including both E. filholi and
E. chrysocome) via an identity- by- state analysis using SNPRELATE
V. 1.20.1 (Zheng et al., 2012). Results showed that the distance be-
tween these individuals was higher than that observed among the
rest of E. filholi individuals, although the analysis assigned them
correctly to the E. filholi group (Figure S2b; Appendix S1 Methods
Section II.2). Such genetic differences in these three E. filholi individ-
uals may be a result of either (a) a hybrid individual from a cross with
other unknown penguin species, (b) migrant individuals from unsam-
pled colonies exhibiting higher levels of genetic differentiation (e.g.
Campbell, Auckland and Antipodes Islands; Cole, Dutoit, et al., 2019;
Cole, Ksepka, et al., 2019) or (c) a laboratory or sequencing artefact.
Given that we do not have additional data needed to test these hy-
potheses, we removed them for subsequent analyses and retained a
total of 93 individuals.
   FRUGONE Et al.
2.3 | Private alleles, FST and Manhattan plots
We used the populations program in STACKS to calculate the num-
ber of private alleles for each population and taxon using unlinked
datasets of 2,302 and 1,211 SNPs, respectively (Table S3). Other
genetic diversity calculations (mean expected heterozygosity and
nucleotide diversity (π) and inbreeding coefficient (FIS) can be found
in Table S6. To estimate genomic differentiation, we calculated pair-
wise FST among all colonies from the three taxa using GENODIVE
V. 3.04 (Meirmans & Van Tienderen, 2004). Significance was deter-
mined with 10,000 permutations of the unlinked dataset. p- values
were corrected using the False Discovery Rate method (Benjamini
et al., 2006).
To evaluate differentiation across genomic sites, we also calcu-
lated FST on a per- site basis using STACKS to produce Manhattan
plots for each pair of taxa (E. moseleyi/E. chrysocome, E. moseley-
i/E. filholi and E. chrysocome/E. filholi) and among all populations
within each taxon. Calculations were made with the full datasets
(Table S3). To generate one intraspecific Manhattan plot for each
taxon, we calculated the mean FST for all sites that were shared
across all populations. Manhattan plots were generated in R V.
3.6.3 (R Core Team, 2020) using the QQMAN V. 0.1.4 package
(Turner, 2014).
2.4 | Clustering analyses
We employed several approaches to evaluate population genetic
structure and genetic differentiation among the three rockhopper
penguin taxa. We used PLINK V. 2.00a3 (Purcell et al., 2007) and
GGPLOT2 V. 3.3.3 (Wickham, 2016) in R 3.6.3 (R Core Team, 2020)
to calculate and visualize a PCA using the unlinked datasets. Missing
data (E. moseleyi- E. chrysocome- E. filholi ~40%; E. chrysocome- E. fil-
holi ~0.46%; E. moseleyi ~0.78%; E. chrysocome ~1.37%; E. filholi
~0.53%) was replaced using the “meanimpute” modifier in PLINK
that performs a mean impute of missing genotype calls.
We also used FINERADSTRUCTURE (Malinsky et al., 2018) to
infer shared ancestry among all individuals, using the full dataset
that included all taxa, and RADPAINTER to generate the input file
from the VCF file and to calculate the co- ancestry matrix. We re-
ordered loci according to linkage disequilibrium with the script pro-
vided with the package and ran FINERADSTRUCTURE using default
In addition, we used the Bayesian clustering method in
STRUCTURE (Falush et al., 2003; Pritchard et al., 2000). All runs
were made using PARALLELSTRUCTURE (Besnier & Glover, 2013)
on the CIPRES V. 3.3 (Miller et al., 2010) platform. We performed this
analysis with (a) all nine populations across the three taxa together
in a single analysis, (b) the seven populations of E. chrysocome and
the E. filholi populations and (c) separately for each taxon. For these
analyses, we used the unlinked datasets (Table S3). For the analyses
involving more than one taxon (a and b), we chose the admixture
ancestry model, uncorrelated allele frequencies (Falush et al., 2003;
Porras- Hurtado et al., 2013) and adjust the alpha parameter to 1/K
due to the differences in sample size for each taxon (Wang, 2017).
For analyses at the taxon level, we chose the admixture ances-
try model and correlated allele frequencies (Falush et al., 2003;
Porras- Hurtado et al., 2013). Given the taxonomic uncertainty of
E. chrysocome and E. filholi, we also used this parameter setting for
the analysis, including all populations of these taxa, to compare re-
sults yielded by the uncorrelated allele frequency model. In all cases,
we did not include the geographical origin of each individual as a
prior, given our interest in detecting strong population structure. We
performed 10 independent runs for each value of K, with 500,000
MCMC permutations and a burn- in of 50,000 permutations. The
number of K value s tested for eac h dat ase t was dete rmi ned based on
the number of geographic populations sampled for each taxon. We
used the online version of STRUCTURE HARVESTER V. 0.6.94 (Earl
& von Holdt, 2012) (http://taylo r0.biolo tureH arves
ter/) to infer the most likely K by the highest mean ln likelihood for
each value of K and compared them with the results from Evanno's
method when testing K > 2 (Evanno et al., 2005). We then used
CLUMPAK (Kopelman et al., 2015), an online server that automati-
cally delivers results from CLUMPP (Jakobsson & Rosenberg, 2007)
and DISTRUCT (Rosenberg, 2003), to visualize and summarize the
results of all previous runs.
Finally, we conducted an analysis of molecular variance (AMOVA)
to assess how much of the molecular variation could be explained
by differences among populations in contrast with that which could
be explained by separately grouping populations of E. filholi and
E. chrysocome. The AMOVA was performed using the unlinked data-
set of 922 SNPs (Table S3) using POPPR V. 2.8.5 (Kamvar et al., 2014,
2015) in R V. 3.6.3 (R Core Team, 2020). The statistical significance
was assessed through 10,000 permutations.
2.5 | Species delimitation and phylogeographic
For species delimitation analyses and phylogenetic reconstruction,
we included five macaroni penguin individuals (E. chrysolophus)
from Frugone et al. (2019) as an outgroup. This was done to test
the one- species hypothesis for rockhopper penguins. We used the
SNAPP package V. 1.5.1 (Bryant et al., 2012) from BEAST V. 2.6.1
(Bouckaert et al., 2014) to delimit species using SNP data. We
evaluated five competing species models using the BFD* method
(Leaché et al., 2014). We included models representing the one- to
three- species hypotheses of rockhopper penguins (Models a, b, c
in Table 1) (Banks et al., 2006; de Dinechin et al., 2009; Frugone
et al., 2018; Jouventin et al., 2006; Mays et al., 2019) and two extra
models that further split divergent populations of E. moseleyi (Model
d, Table 1), and divergent populations of both E. moseleyi and E. chry-
socome (Model e, Table 1). These last two models were based on
previous studies showing high levels of population divergence
within E. moseleyi (Frugone et al., 2018) and E. chrysocome (Frugone
et al., 2018; Lois et al., 2020).
Because BFD* is computationally demanding, we ran two in-
dependent analyses to check for consistency by twice randomly
selecting five macaroni penguin individuals as an outgroup and 18
rockhopper penguin individuals as an ingroup, which included six
samples from each of the three rockhopper penguin taxa, repre-
senting 1– 3 individuals from nine populations. The selected individ-
uals for both runs and their populations of origin are summarized
in Table S4a and S4b. We used unlinked loci datasets, and we re-
tained only the sites that were present in all individuals (Table S3).
The total numbers of SNPs for each dataset were 1,798 and 1,806
We used the Python script (Ortiz, 2019), available
at domor tiz/vcf2p hylip, to convert the
VCF files to binary nexus format. After fixing the mutation rates (u
and v) to 1 and assigning a gamma distribution G (1, 250) to the prior
for the expected genetic divergence (parameter θ) and a gamma dis-
tribution G (2, 200) for the speciation rate parameter λ, we carried
out stepping- stone path sampling analyses for every competing spe-
cies model. Marginal likelihood estimates were obtained using alpha
=0.3 and 36 MCMC steps, each consisting of 100,000 cycles with
a pre- burn- in run of 10,000 cycles. All models were compared by
calculating Bayes factors. We followed Kass and Raftery (1995) and
considered that, for two competing models, no evidence favours any
model when BF <2, and that evidence for the most- likely model is
positive for 2 <BF <6, strong for 6 < BF <10 and decisive for BF >10 .
To evaluate the phylogenetic relationships among rockhopper
penguins, we performed the Bayesian coalescent method imple-
mented in SNAPP (Bryant et al., 2012) and performed a maximum
likelihood estimate of tree topology using IQTREE V. 2.1.2 (Nguyen
et al., 2015). The dataset for SNAPP included the five macaroni
penguin individuals as an outgroup and 34 rockhopper penguin in-
dividuals (12 E. moseleyi, 10 E. chrysocome and 12 E. filholi, aiming
to include at least three individuals from each study population) as
the ingroup. We followed filtering criterion similar to what we used
for the BFD* analyses (Table S3). The MCMC was run for 600,000
generations and sampled every 500 cycles, and a 20% burn- in rate
was applied. We performed three independent runs and joined the
estimated parameters and trees to evaluate consistency and conver-
gence using TRACER V. 1.7.1 (Rambaut et al., 2018). All estimated
posterior distributions of parameters, except for a few θ values of in-
ternal branches, had ESS values >200. We used DENSITREE V. 2.2.7
(Bouckaert, 2010) to plot the sampled posterior tree topologies for
all combined runs. We summarized the posterior distribution into
a Maximum Clade Credibility (MCC) tree using TREEANNOTATOR
(part of the BEAST V. 2 2.6.1 package).
For the maximum likelihood phylogeny, we selected the
same individuals analysed using SNAPP, but given that this is a
concatenation- based analysis, we did not filter SNPs by linkage dis-
equilibrium. The dataset consisted of 2,270 variable sites with no
missing information. To estimate the best substitution model, we
used MODELFINDER (Kalyaanamoorthy et al., 2017) implemented
in IQTREE V. 2.1.2, and applied the Ascertainment Bias Correction
(+ASC) as this is strongly recommended when analysing SNP data
(Lewis, 2001). The best substitution model estimated using the
Bayesian Information Criterion (BIC) was K3Pu+F+ASC+R2. To
evaluate branch support values, we used 10,000 ultrafast boot-
straps (Hoang et al., 2017) optimized by NNI, and we used 10,000
replicates for the Shimodaira- Hasegawa- like approximate likelihood
test (SH- aLRT) (Guindon et al., 2010).
2.6 | Gene flow among rockhopper taxa
In the light of previous studies suggesting the existence of contem-
porary gene flow among rockhopper penguins (Mays et al., 2019),
we used BAYESASS V. 3.0.4 (Wilson & Rannala, 2003) to test for
recent migration between the three taxa using an unlinked dataset
(Table S3). BAYESASS uses a Bayesian approach with MCMC to es-
timate migrant ancestries of each individual based on an assignment
method and infers migration rates based on the frequency of indi-
viduals exhibiting migrant ancestry. We set the mixing parameters
for allele frequencies and inbreeding coefficients to 0.25 and 0.05,
Bayes Factor
(subset 1)
Bayes Factor
(subset 2)
a) 2 species = E. chrysolophus/E. moseleyi + E. chrysocome
+ E. filholi
- - - -
b) 3 species = E. chrysolophus/E. moseleyi/E. chrysocome +
E. filholi
−3187.32 −3407.18
c) 4 species =
E. chrysolophus/E. moseleyi/E. chrysocome/E. filholi
−384 4.5 −4020.74
d) 5 species = E. chrysolophus/Nightingalea /
Amsterdama /E. chrysocome/E. filholi
−3832.32 −4 013.34
e) 6 species = E. chrysolophus/Nightingalea /Amsterdama /
Falklandsb /Statenb+ Terhaltenb /E. filholi
−3828.72 −4000.64
Note: Bayes factor was calculated against the two- species hypothesis (a). Negative Bayes factor
indicates that the alternative model (b to e) is favoured.
aE. moseleyi populations.
bE. chrysocome populations.
TABLE 1 Bayes factor delimitation
results for all models tested
   FRUGONE Et al.
respectively, after achieving acceptance rates of 20%– 60% on the
test run. The MCMC was run for 50,000,000 iterations, discarding
the first 4,000,000 as burn- in, with a sampling interval of 4,000. We
performed five independent runs with different random seeds and
compared the results from each run. We checked for convergence
using TRACER V. 1.7.1 (Rambaut et al., 2018). We plotted the re-
sults from BAYESASS in R V. 3.6.3 (R Core Team, 2020) using TIDYR
V. 1.0.2 (Wickham & Henry, 2018) and DPLYR V. 0.8.5 (Wickham
et al., 2018) to prepare the data and used CIRCLIZE V. 0.4.13 (Gu
et al., 2014) to plot the results.
3.1 | SNP calling
For all ddRAD datasets, the mean effective per- sample depth of cov-
erage ranged from 73.0× to 98.2×, with a minimum per- sample cov-
erage of >18.2× (Table S5). We did not find any evidence for a site
being under selection or out of Hardy– Weinberg equilibrium in more
than ~50% of the populations or under different grouping criteria
(described on Table S3). The total number of SNPs retained after
each filtering step and the final number of SNPs from each dataset
(full datasets ranging from 2,975 to 14,226 SNPs; unlinked datasets
ranging from 898 to 4,674 SNPs) are presented in Table S3.
3.2 | Genetic diversity and differentiation
All populations of E. moseleyi, E. chrysocome and E. filholi exhibited
private alleles, ranging from 63 (2.7%) for E. chrysocome on the
Falkland/Malvinas Islands to 11 (0.47%) for E. chrysocome and E. fil-
holi on Staten and Crozet islands respectively (Table S6). When we
calculated genetic diversity within each taxon, we determined that
around 10% of the alleles were private to each taxon (E. moseleyi
=121, E. chrysocome =99 and E. filholi =106).
Among populations of each taxon, pairwise FST values were
low, ranging from 0.002 to 0.053, and most were not significantly
differentiated (Figure 2a). The highest intraspecific differentiation
was found among E. moseleyi (0.053) populations; FST values within
E. chrysocome (FST =0.002– 0.032) and E. filholi (FST =0.004– 0.013)
were consistently low. In contrast, interspecific FST values were all
significant, ranging from 0.140 to 0.497. The highest FST was recov-
ered for pairwise comparisons between E. moseleyi and populations
of E. chrysocome/E. filholi (Figure 2a; Table S7).
Manhattan plots revealed clear contrasts in the extent of genomic
differentiation across sites within versus among taxa. Specifically,
intraspecific Manhattan plots exhibited a narrower distribution of
FST values than those involving different taxa. The maximum FST for
the intraspecific calculations was ~0.4 between E. moseleyi popula-
tions. Considering 0.4 as a cut- off, Manhattan plots between E. mo-
seleyi and E. chrysocome/E. filholi revealed high variation in FST values
across the genome, with around 800 sites exhibiting FST values >0.4
(Figure 3a,b). E. chrysocome and E. filholi exhibited lower FST val-
ues, although 64 sites with FST >0.4 were detected (Figure 3c). In
contrast, intraspecific differentiation across genomic regions was
characterized by far lower FST values (Figure 3d,e,f); the only site ex-
hibiting FST >0.4 corresponded to a comparison between two E. mo-
seleyi populations.
3.3 | Clustering analyses
A PCA including all three taxa revealed three separate, tightly
clustered groups according to species boundaries (Figure 2b).
Specifically, the first principal component separated E. moseleyi from
E. chrysocome and E. filholi, explaining 45.45% of the variance, and
the second principal component separated E. filholi from the E. chry-
socome populations, explaining 22.19% of the variance. When a PCA
was conducted with just E. filholi and E. chrysocome, the first princi-
pal component separated these taxa, explaining 43.34% of the total
variance, and the second principal component separated individu-
als from populations of E. chrysocome from the Falkland/Malvinas
Islands from those on Terhalten and Staten islands, explaining 7.8%
of the variance (Figure 2c). At the population level, PC1 (16.42%)
separated E. moseleyi populations at Nightingale and Amsterdam
islands (Figure 2d) and the E. chrysocome (12.01%) population on
the Falklands/Malvinas Islands from the neighbouring mainland
Patagonian populations (Figure 2e). In contrast, E. filholi did not
exhibit clear geographic structure among the sampled populations
(Figure 2f).
Similar results were observed when the SNPs were analysed
using FINERADSTRUCTURE (Figure 4a): three groups were iden-
tified by shared ancestry, each corresponding to one of the three
rockhopper taxa. These analyses revealed no evidence of hybrid
individuals; that is, no individuals exhibited similar levels of co-
ancestry between groups (Malinsky et al., 2018).
For STRUCTURE analyses of E. moseleyi, E. chrysocome and
E. filholi performed in the same run (Figure 4b), the mean LnP(K)
identified the best K was four; however, no additional groups were
formed when K > 3 (Figure S3). On the other hand, the Evanno
method suggested two clusters (Table S8), reflecting the relatively
deep genetic differentiation between E. moseleyi and the remaining
rockhopper taxa. Across all successive K, we did not find evidence
of shared ancestry among the three rockhopper penguin taxa
(Figure S3). We obtained similar results from testing E. chrysocome
and E. filholi, without including E. moseleyi, despite using different
allele frequency models. Both methods indicated that K = 2 was
the best K ( Table S8), an d no additional structure was identified for
K > 2 (Figure S4a,b).
In the intraspecific analyses, K = 2 exhibited a higher proba-
bility than K = 1 in E. moseleyi, corresponding to separation of the
populations on Nightingale and Amsterdam islands (Table S8). We
found low levels of adm ixt ure between indi viduals from both pop-
ulations (Figure 4c). For E. chrysocome, the optimal K suggested
by both methods was K = 2, where one group corresponded
to the Falkland/Malvinas population and the other to those of
Staten/Terhalten islands. We found low levels of admixture be-
tween these groups (Figure 4d) but did not detect any popula-
tion structure across the distributional range of E. filholi. The
Evanno method suggested K = 3, but closer inspection revealed
that ancestries for each cluster were evenly distributed among
all individuals, indicative of a panmictic population (i.e. K = 1;
Figure S5). The highest value of the mean LnP(K) also supported
K = 1 (Figure 4e; Table S8). Finally, AMOVA revealed that 22.93%
(p =.026) of the molecular variance was explained by differenti-
ation between E. chrysocome and E. filholi, whereas only 1.63%
(p <.01) was explained by differentiation among samples within
each taxon (Table S9).
3.4 | Species delimitation and phylogeographic
The BFD* analysis for both independent runs using a different sub-
set of individuals yielded similar results; the most favoured model
was the one with four species (i.e. the three rockhopper species hy-
pothesis +macaroni penguins; Table 1).
The phylogenetic tree inferred using the coalescent method im-
plemented in SNAPP recovered E. filholi sister to E. chrysocome, and
this clade sister to E. moseleyi, all with high posterior probabilities
(PP = 1). In contrast, posterior probabilities were much lower for the
nodes within species (i.e. populations). This tree topology was highly
consiste nt am ong diffe rent tree estim ati ons (Figu re 5a). The ma ximu m
FIGURE 2 Genetic differentiation within and among rockhopper penguin taxa. (a) Pairwise FST among all sampled colonies. NS, not
significant. (b) PCA results of individuals from the three taxa (E. moseleyi, E. chrysocome and E. filholi). (c) PCA between E. chrysocome and
E. filholi. (d) Intraspecific PCA of E. moseleyi individuals. (e) Intraspecific PCA of E. chrysocome individuals. (f) Intraspecific PCA of E. filholi
individuals. Population codes (colour) are indicated on the side of the plot. Eudyptes moseleyi populations: Amsterdam and Nightingale,
E. chrysocome populations: Terhalten, Staten and Falkland/Malvinas, E. filholi populations: Marion, Crozet, Kerguelen and Macquarie. Replicas
of these figures with an alternative colour scheme and additional visual patterns are available in Appendix S1 (section IV: Alternative figures)
   FRUGONE Et al.
FIGURE 3 Manhattan plots comparing interspecific versus intraspecific differentiation across genomic regions. Interspecific locus- by-
locus pairwise FST values are shown for (a) E. moseleyi– E. chrysocome, (b) E. moseleyi– E. filholi and (c) E. chrysocomeE. filholi. Intraspecific
locus- by- locus FST (mean) is shown for (d) E. moseleyi, (e) E. chrysocome and (f) E. filholi
likelihood reconstruction yielded similar results: a similar topology
with reciprocal monophyly for each taxon with high support values
(SH- aLRT and ultrafast bootstrap support values ~100%; Figure 5b).
3.5 | Gene flow among rockhopper taxa
Migration rates, estimated as the fraction of migrant individuals per
generation and calculated using BAYESASS, yielded almost identical
results across different runs, starting with different random seeds
(Table S10). The proportion of migrant individuals between each
taxon (all results expressed as mean, 95% credibility interval) are:
E. moseleyi to E. chrysocome, 0.010 (−0.009– 0.027); E. chrysocome
to E. moseleyi, 0.012 (−0.012– 0.036); E. moseleyi to E. filholi, 0.008
(−0.008– 0.024); E. filholi to E. moseleyi, 0.012 (−0.012– 0.036);
E. chrysocome to E. filholi, 0.008 (−0.008– 0.024); and E. filholi to
E. chrysocome, 0.010 (−0.009– 0.027). These results are summarized
in Figure 4f.
FIGURE 4 Genetic clustering analyses and estimated migration rates among rockhopper penguins. (a) Co- ancestry matrix derived by
FINERADSTRUCTURE showing the estimated co- ancestry among all individuals; relationships between populations are shown in the top
of the plot, with values corresponding to posterior assignment probabilities. Population codes for each individual are denoted by the colour
of each bar. (b) Structure results (K = 3) involving individuals from the three rockhopper penguins taxa. Intraspecific runs of structure for
(c) E. moseleyi (K = 2), (d) E. chrysocome (K = 2) and (e) E. filholi (K = 1). (f) Circular migration plot exhibiting, for each taxon, the fraction of
migrant individuals (size of the band and numbers on each band) per generation and its origin. Replicas of this figure with an alternative
colour scheme and additional visual patterns are available on Appendix S1 (section IV: Alternative figures)
   FRUGONE Et al.
4.1 | Molecular evidence for species delimitation
One of the challenges in delimiting recently divergent species is
the ab sence of a cle ar geno mic dif ferent iation threshold fo r distin-
guishing intraspecific versus interspecific lineages. Pragmatically,
comparing patterns observed among related well- delimited spe-
cies sharing similar ecological, behavioural and life- history traits
with those observed in the taxa under consideration is one way
to help guide taxonomic decisions, especially within an integrative
taxonomic framework. We therefore investigated genomic differ-
entiation among E. moseleyi, E. chrysocome and E. filholi, a species
complex of penguins whose taxonomy has been the subject of
considerable debate for several decades. Our results consistently
provided support for the recognition of three species of rockhop-
per penguins. The global PCA clearly distinguishe d the subtropical
E. moseleyi from su b- A nta rcti c E. chrysocome and E. filholi (fi r st axis;
45% of genetic variance; Figure 2b) and also separated E. chryso-
come from E. filholi (second axis 22%; Figure 2b). The genetic vari-
ance explaining the separation of E. chrysocome from E. filholi was
even higher when performing the PCA not including E. moseleyi
(First axis 43%; Figure 2c). In contrast, intraspecific comparisons
explained relatively small proportions of genetic variance (11%–
16%; Figure 2d,e,f). This contrast was also evident in the range
of inter- taxon FST (0.140– 0.497) relative to intraspecific values
(0.002– 0.053). Similarly, when testing FST across sites throughout
the penguin genomes, several sites exhibited moderate to high ge-
netic differentiation (FST >0.4) when comparing taxa, in contrast
to intraspecific comparisons where a much narrower distribution
of FST values occurred (most sites exhibited FST <0.2 and almost all
were <0.4; Figure 3).
The range of intraspecific variation observed in our study is
similar to that reported previously for rockhopper penguins (Cole,
Dutoit, et al., 2019; Cole, Ksepka, et al., 2019) and within other
penguin species (broader discussion on Appendix S1 section III).
For example, a recent genomic study found that the intraspecific
variance explained (PCA1) among individuals from different popu-
lations of king (Aptenodytes patagonicus), emperor (A. forsteri), chin-
strap (Pygoscelis antarctica) and Adélie (P. adeliae) penguins ranged
from 1.28% to 2.02%, and FST ranged from 0.002 to 0.003 (Clucas
et al., 20 18) . Gento o pengu ins (P. papua), on the other hand, exhibited
greater genomic structure among colonies geographically isolated
on different islands (15.81% and FST =0.217). However, the anal-
ysis of gentoo penguins involved divergent lineages that were re-
cently proposed to represent at least four distinct species, based on
their genomic divergence and fine- scale morphometric differences
(Clucas et al., 2018; de Dinechin et al., 2012; Pertierra et al., 2020;
Tyler et al., 2020; Vianna et al., 2017). Another recent study using
mitochondrial genomes from all penguin species found that the
range of sequence divergence among well- defined species is 0.8%–
5.2% (Cole, Dutoit, et al., 2019; Cole, Ksepka, et al., 2019). These
authors found that E. chrysocome and E. filholi are 0.7% divergent,
while the range of differentiation for macaroni (E. chrysolophus) and
royal (E. schlegeli) penguins, which probably comprise a single spe-
cies, is 0.2% (Cole, Dutoit, et al., 2019; Cole, Ksepka, et al., 2019;
Frugone et al., 2019). Thus, compared to several other penguin spe-
cies, the range of variation we found among the three rockhopper
taxa exceeded the expected values for intraspecific variation across
genomic regions. Furthermore, our AMOVA results showed that the
variance explained by grouping populations to reflect each taxon
was ~23%, whereas differentiation among populations within each
taxon was only around 2%.
In further support of recognizing three rockhopper penguin spe-
cies, we did not find evidence of admixture among the three taxa
sults suggest that only ~0.01 individuals per generation are migrants
from another taxon. However, because all credibility intervals over-
lapped with 0, we are unable to reject a scenario of zero gene flow and,
hence, complete genetic isolation among the three taxa. We found
hi gh pro p ortion s of pri v ate allele s (~10%) within E. moseleyi, E. chryso-
come and E. filholi, consistent with little or no gene flow among the
three rockhopper taxa. The apparent absence of admixture, the high
proportions of private alleles and the 1317 mutational steps sepa-
rating their mtDNA haplogroups (Frugone et al., 2018) together pro-
vide strong support for the historical isolation of these taxa. While
it is possible that hybrid individuals have been unsampled (Crofts
& Robson, 2015), hybridization among these three taxa appears to
be negligible. Low frequencies of hybridization have been reported
between other penguin sister lineages that are accepted as distinct
species, suc h as the Humbol dt (Spheniscus humboldti) and Magellanic
(S. magellanicus) penguins (Simeone et al., 2009), and among several
species of Eudyptes penguins (Hull & Wilthshire, 1999; Napier, 1968;
White & Clausen, 2002), including hybridization between E. filholi
and E. chrysolophus (Woehler & Gilbert, 1990) and E. chrysocome
and E. chrysolophus (White & Clausen, 2002). Further, high levels
(up to 25%) of ancestral introgression were detected among several
Eudyptes species (Vianna et al., 2020). In this sense, the strong ge-
nomic differentiation found among the three rockhopper penguins
may be the result of their geographic isolation and not necessarily
implying reproductive isolation, and thus, it is not possible to discard
the possibility of hybridization if they come into geographic contact.
Finally, the best model found by our species delimitation anal-
yses using SNP data supported the existence of three species of
rockhopper penguins. Despite earlier reports that the multispecies
coalescent model as implemented in SNAPP and other programs
such as BPP (Yang & Rannala , 2010) might delimit popu lat ion struc-
ture and not species (Sukumaran & Knowles, 2017), our compre-
hensive population- level sampling of all three putative rockhopper
penguin taxa enabled us to clearly delineate taxonomic boundar-
ies, separating intraspecific from interspecific variation. The rec-
ognition of three rockhopper penguin species is also supported
using different species delimitation methods including the tree-
based ABGD (Puillandre et al., 2012), the threshold- based GMYC
(Pons et al., 2006)— both of which are presented in Frugone et.
al. (2018)a nd the BPP (Yang & Rannala, 2010) anal yses presented
by Mays et al. (2019). Mays et al. (2019) rejected the findings of
their BPP analyses because their 6- locus dataset yielded ambigu-
ous results and were unable to reliably distinguish any taxa based
on the GDI (Jackson et al., 2017; Leache et al., 2019). This result
was surprising because it also failed to recognize even the highly
divergent and widely recognized species E. chrysolophus, suggest-
ing that their six- locus dataset was underpowered for evaluating
species boundaries in these recently diverged penguin species.
FIGURE 5 Phylogenetic relationships among rockhopper penguins. (a) Posterior distribution of different topologies yielded by SNAPP
analyses. (b) Maximum likelihood reconstruction from IQ- tree. Posterior probabilities are indicated on the nodes of each phylogeny
   FRUGONE Et al.
In summary, at the molecular level, several lines of evidence
support recognition of three species of rockhopper penguins: 1) ge-
nomic differentiation among E. moseleyi, E. chrysocome and E. filholi
substantially exceeds differentiation observed within each taxon;
2) there is no genomic evidence of admixture or hybrid individuals
among these taxa; 3) our analyses suggest no contemporary gene
flow among the three taxa; and 4) species delimitation analyses
strongly support a three- species designation to the exclusion of a
model designating E. chrysocome/E. filholi as conspecific. Our results,
thus, confirm the findings of previous studies revealing reciprocal
monophyly with mtDNA markers, supporting this three- species des-
ignation (Banks et al., 2006; Cole, Dutoit, et al., 2019; Cole, Ksepka,
et al., 2019; de Dinechin et al., 2009; Frugone et al., 2018; Mays
et al., 2019).
Our analyses provide insights into the causes of discordance
among nuclear/mitochondrial markers leading to different conclu-
sions regarding species delimitation in this group. Previous studies
(e.g. Frugone et al., 2018; Mays et al., 2019) included sampling from
the entire range of each taxon, and the high mtDNA divergence
found among taxa is therefore not a consequence of sparse sam-
pling, which may result in isolation- by- distance being undetected
(Chambers & Hillis, 2020). In this study, we used thousands of un-
linked genetic markers and found that results are in accordance with
the pattern exhibited across all studies using mtDNA. This finding
suggests that the previous discordance among studies did not result
from sex- biased dispersal of rockhopper penguins, but rather from a
lack of power owing to low levels of divergence in the nuclear mark-
ers previously utilized. The value of using a few nuclear markers for
species delimitation is likely to be further eroded under introgres-
sion scenarios, and/or when lineage sorting among closely related
lineages is incomplete.
4.2 | Integrative taxonomy
Over- splitting of lineages potentially presents a valid concern in
taxonomy given the power of modern genomic methods to detect
rapidly evolved, fine- scale differentiation among populations. One
could argue that should enough SNPs be used in an analysis, any
two populations could be delimited through either different allele
frequencies or by a few fixed mutational differences. As such, under
an integrative taxonomic framework, it is important that additional
lines of evidence (e.g. ecology, phenotypes, behaviour, geography,
parasites) are investigated and contrasted with molecular species
delimitations (Matos- Maraví et al., 2019; Solís- Lemus et al., 2015). In
the case of rockhopper penguins, additional factors in support of the
molecular results that reinforce the recognition of three rockhopper
penguin species include: (a) the non- overlapping geographical distri-
butions of each lineage— E. moseleyi (subtropical area in the Atlantic
and Indian Oceans), E. chrysocome (sub- Antarctic area of the eastern
Pacific and Atlantic Oceans) and E. filholi (sub- Antarctic area of the
western Pacific and Indian oceans; Figure 1); (b) different dispersal
capabilities in winter, with E. filholi traversing greater distances than
E. chrysocome (Raya Rey et al., 2007; Thiebot et al., 2012); (c) the dis-
tinct species of parasitic chewing lice found on each taxon, which is
consistent with little to no gene flow among taxa (Banks et al., 2006);
(d) differences in the timing of reproduction, starting about 2 months
earlier for E. moseleyi populations, with E. chrysocome and E. filholi
colonies exhibiting variation in the timing of the onset of repro-
duction (Clausen & Pütz, 2002; Hiscock, 2013; Hull et al., 2004;
Morrison, 2016; Raya Rey et al., 2009); and (e) the presence of
morphological and vocal differences among taxa. Eudyptes moseleyi
can be distinguished from the sub- Antarctic E. chrysocome/E. filholi
by its larger size and longer crests (Banks et al., 2006), whereas
the main morphological differentiation between the sub- Antarctic
taxa corresponds to a more pronounced pink to white gape (bare
skin around the bill) on E. filholi which is black on E. chrysocome
(Tennyson & Miskelly, 1989). Other morphological features differ-
entiating E. filholi from E. chrysocome include a narrower bill and the
shape of the black mark on the undersurface of the apex of the wing
(Hutton, 1879). Finally, vocalizations are an important behavioural
trait in penguins, since they may rely more strongly on auditory cues
for mate selection and individual recognition than on morphological
characters (Aubin & Jouventin, 2002; Jouventin et al., 2006; Searby
& Jouventin, 2005). Differences have been found in the mating calls
of E. moseleyi in comparison with those of E. chrysocome (when con-
sidered a single species to E. filholi), providing an additional line of
evidence for delimiting those taxa (Jouventin et al., 2006). However,
as far as we know, vocalizations of E. chrysocome and E. filholi have
not been directly compared and should be a priority for future data
collection and study.
There are several well- documented historic cases of over-
lumping of species due to the absence of obvious morphological
variation among taxa (Adams et al., 2014; Bickford et al., 2007;
Grosser et al., 2015). The absence of apparent diagnostic morpho-
logical variation in such cryptic or sibling species may be explained
by recent diversification, morphological constraints imposed by se-
lective pressures, or by convergence (Fiser et al., 2018). Conversely,
high variability in morphological characters within a species may
lead to over- splitting taxa. For example, in birds, variation in meristic
or plumage characters may result from local (population) adaptation
across an environmental gradient (Mason & Bowie, 2020) or could
be due to phenotypic plasticity (e.g. Mason & Taylor, 2015). Penguin
species exhibit very low variation in their diet, and have similar ecol-
ogy, behaviour and morphology when compared to other groups of
birds, such as ducks (Anseriformes), shorebirds (Charadriiformes),
pheasants (Galliformes) or songbirds (Passeriformes) (Winkler
et al., 2015). Despite not overlapping in their distributions, both
E. chrysocome and E. filholi breed in similar marine environments, al-
though in different ocean basins, prey on similar items, and differ
very slightly in their timing of reproduction (Clausen & Pütz, 2002;
Hull, 1999). These aspects are also shared with other penguin species
distributed in the sub- Antarctic area. For example, similarities in diet
and reproductive timing are also present when comparing E. chryso-
come and E. filholi with other Eudyptes species like the macaroni pen-
guin (Thiebot et al., 2011; Williams & Croxall, 1991). This implies that
because E. chrysocome, E. filholi and several other penguin species
overlap in ecological niche space (e.g. oceanographic conditions over
their distributional range, diet and breeding period), it may constrain
differentiation in some phenotypic traits (Fiser et al., 2018) if they
have moved towards the same adaptive peak. The extent of niche
overlap might also explain the limited morphological differentiation
among Eudyptes penguins, not only for the rockhopper penguin, but
also among other penguin species complexes (Tyler et al., 2020). In
this sense, the use of genomic approaches with extensive sampling
across the species complex range is critically important if inter- ver-
sus intraspecific taxonomic boundaries are to be delineated.
Most bird species were initially delimited under the
Morphological Species Concept (MSC) (Cronquist, 1978) and/
or Biological Species Concept (BSC) (Cracraft, 1992; Mayr, 1942).
However, under the MSC, cryptic species or subtle morphologi-
cal variation can lead to underestimation of species diversity, and
hence, a poor under sta nding of the underly ing evolutionar y histor y,
which can have significant impacts on the management of threat-
ened taxa. On the other hand, in birds, hybridization might be more
common than in other groups and has been documented in ~16%
of their species (Ottenburghs et al., 2015). Furthermore, multispe-
cies hybridization (i.e. hybridization between more than two spe-
cies) might also be common in this group (Ottenburghs, 2019). In
this sense, divergent bird species that occasionally hybridize might
be considered conspecific under the BSC, given that it assumes
reproductive isolation. Moreover, to test for the BSC criterion of
reproductive isolation in allopatric taxa may be dif ficult without the
use of molecular markers to estimate gene flow. The phylogenetic
species concept (PSC) (Cracraft, 1983; Nelson & Platnick, 1981)
clarified the taxonomy of several bird species (e.g. Cracraft, 1992)
including penguins, facilitated by the advances in molecular biol-
ogy and genomics. For example, macaroni and royal penguins were
considered by several authors as different species mainly based
on their allopatric distribution and morphological differentiation.
However, genomic data revealed absence of reciprocal monophyly
(Frugone et al., 2018, 2019), suggesting that they should be con-
sidered a single species. Here, we recognized the three rockhop-
per species under the unified species concept proposed by De
Queiroz (2007) within an integrative taxonomy framework (Cicero
et al., 2021; Sangster, 2014), which reconciles elements from sev-
eral species concepts including the MSC, BSC and the PSC. We
suggest the distinction of three rockhopper species based on the
criteria described above: concordance and genealogical agreement
across multiple loci (e.g. mtDNA and genomic data) with sampling
encompassing most of the species distributional ranges, concor-
dance with biogeographical patterns, morphological characters and
no evidence of recent gene flow.
Population structure could be viewed as a first step in the spe-
ciation process. In general, seabirds, especially those inhabiting the
Southern Ocean, exhibit very low levels of genetic structure (e.g.
Burg & Croxall, 2001; Milot et al., 2008; Quillfeldt et al., 2017).
Our results, however, reveal an opposite pattern and highlight that
even widespread seabirds with great dispersal potential can house
cryptic diversity previously undetected due to limited morpholog-
ical differentiation. We also suggest that it is especially important
to understand the role of introgression, as seabird colonies may
have been connected multiple times in the past, and account for
this by sampling across the genome to have confidence in species
4.3 | Conservation implications
Declining numbers recently observed for populations of all three
rockhopper penguin species are of major concern for contempo-
rary conservation. Rising oceanic temperatures and associated
oceanographic changes, the introduction of predators and over-
fishing are among the threats thought to have influenced these
demographic declines (Crawford et al., 2003; Cuthbert et al., 2009;
Trathan et al., 2015; Wilson et al., 2010). Currently, E. moseleyi
is considered endangered (BirdLife International, 2020b) and
E. chrysocome/E. filholi, as a single taxon, is considered vulnerable
(BirdLife International, 2020a). Delimiting E. chrysocome and E. filholi
as distinct species may heighten the severity of their conservation
status due to more restricted ranges and smaller population sizes
and acknowledge their independent evolutionary histories. Eudyptes
filholi has a wider geographical range than E. chrysocome, and census
data suggest smaller population sizes for E. filholi (422,000 breed-
ing pairs) than E. chrysocome (850,000 breeding pairs). Overall,
while populations of both species are decreasing, E. filholi has ex-
perienced particularly severe population contractions over recent
decades (BirdLife International, 2020a). Whereas some authors sug-
gest that the largest E. chrysocome breeding colony in the Falkland/
Malvinas Islands has increased (Baylis et al., 2013), E. filholi popula-
tions have significantly declined across most of the species distri-
butional range, including Marion (Crawford et al., 2003), Campbell
(Morrison et al., 2015) and Crozet islands (Barbraud et al., 2020).
Following the recommendations of the most recent IUCN report
(BirdLife International, 2020a), we also suggest the need for con-
tinuing monitoring of population trends for these two species
separately, and especially for E. filholi. Additionally, it is of utmost
importance to prioritize research efforts aiming to understand the
underlying causes of these severe population declines. In this sense,
the accurate delimitation of these birds is of paramount importance
to the conservation of rockhopper penguins (Bickford et al., 2007;
Cuthbert et al., 2009) and will be necessary for effective manage-
ment decisions.
We thank Daly Noll and Alison Cleary for help in the labora-
tory. We thank Milan Malinsky for useful comments and help
with FINERADSTRUCTURE analysis. This study was funded by
INACH DT- 11_17 and INACH RT_12– 14 grants, FONDECYT pro-
jects 1150517, Centro de Regulación del Genoma (CRG) ANID/
FONDAP/15200002 and 1151336 CONICYT PIA ACT172065 GAB.
PMM was supported by a Czech Science Foundation Junior GAČR
   FRUGONE Et al.
grant (GJ20- 18566Y) and the Czech Academy of Sciences PPLZ
programme (L200961951), and acknowledges the computational
resources provided by the CESNET LM2015042 and the CERIT
Scientific Cloud LM2015085. FB was supported by the French Polar
Institute Paul- Emile Victor (IPEV; Prog. 354). We also would like to
thank Damien Esquerré and anonymous referees whose comments
and suggestions helped to improve this manuscript.
The authors declare no conflict of interest.
The peer review history for this article is available at https://publo n/10.1111/ddi.13399.
DNA Sequences: BIOPROJECT: PRJNA745960. Methodological
specifications and input files used in analyses: https://doi.
org/10.5061/dryad.mw6m9 05v6.
María José Frugone
Theresa L. Cole
María Eugenia López
Gemma Clucas
Pável Matos- Maraví
Nicolás A. Lois
Jonathan M. Waters
Rauri C. K . Bowie
Elie Poulin
Juliana A. Vianna
Adams, M., Raadik, T. A., Burridge, C. P., & Georges, A. (2014). Global
biodiversity assessment and hyper- cryptic species complexes: More
than one species of elephant in the room? Systematic Biology, 63(4),
518– 5 33 . sb i o/syu 017
Agapow, P.- M., Bininda- Emonds, O. R., Crandall, K. A., Gittleman, J. L .,
Mace, G. M., Marshall, J. C., & Pur vis, A. (2004). The impact of spe-
cies concept on biodiversity studies. The Quarterly Review of Biology,
79(2), 161– 179.
Aubin, T., & Jouventin, P. (2002). How to vocally identify kin in a crowd:
The penguin model. Advances in the Study of Behavior, 31(31), 243–
2 7 7 . h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / S 0 0 6 5 - 3 4 5 4 ( 0 2 ) 8 0 0 1 0 - 9
Banks, J., Van Buren, A., Cherel, Y., & Whitfield, J. B. (2006). Genetic
evidence for three species of rockhopper penguins, Eudyptes chryso-
come. Polar Biology, 30 (1), 61 67.
0- 006- 0160- 3
Barbraud, C., Delord, K., Bost, C. A., Chaigne, A., Marteau, C., &
Weimerskirch, H. (2020). Population trends of penguins in the
French Southern Territories. Polar Biology, 43(7), 835– 850. https://
d o i . o r g / 1 0 . 1 0 0 7 /s 0 0 3 0 0 - 0 2 0 - 0 2 6 9 1 - 6
Baylis, A. M. M., Wolfaardt, A. C., Crof ts, S., Pistorius, P. A., & Ratcliffe,
N. (2013). Increasing trend in the number of southern rockhopper
penguins (Eudyptes c. chrysocome) breeding at the Falkland Islands.
Polar Biology, 36(7), 1007– 1018. 0
0 - 0 1 3 - 1 3 2 4 - 6
Bazykin, A. D. (1969). Hypothetical mechanism of speciation. Evolution,
23(4), 685– 687. 5646.1969.tb035
Benestan, L. M., Ferchaud, A.- L., Hohenlohe, P. A., Garner, B. A.,
Naylor, G. J. P., Baums, I. B., Schwartz, M. K., Kelley, J. L., &
Luikart, G. (2016). Conservation genomics of natural and man-
aged populations: Building a conceptual and practical framework.
Molecular Ecology, 25(13), 2967– 2977. ht tps://doi.o rg /10.1111/
mec .136 47
Benjamini, Y., Krieger, A. M., & Yekutieli, D. (20 06). Adaptive linear
step- up procedures that control the false discovery rate. Biometrika,
93(3), 491– 507. t/93.3.491
Besnier, F., & Glover, K. A. (2013). ParallelStructure: A R package to dis-
tribute parallel runs of the population genetics program STRUCTURE
on multi- core computers. PLoS O ne, 8(7), e70651. https://doi.
org/10.1371/journ al.pone.0070651
Bickford, D., Lohman, D. J., Sodhi, N. S., Ng, P. K. L., Meier, R., Winker, K.,
Ingram, K. K., & Das, I. (2007). Cryptic species as a window on diver-
sity and conservation. Trends in Ecology & Evolution, 22(3), 148– 155.
BirdLife International (2020a). Eudyptes chrysocome. The IUCN Red List of
Threatened Species 2020. e.T22735250A182762377.
BirdLife International (2020b). Eudyptes moseleyi. The IUCN Red List of
Threatened Species 2020. e.T22734408A18 4698049.
Bolger, A. M., Lohse, M., & Usadel, B. (2014). Trimmomatic: A flexible
trimmer for Illumina sequence data. Bioinformatics, 30(15), 2114–
2120. forma tics/btu170
Bouckaert, R. R. (2010). DensiTree: Making sense of sets of phylogenetic
trees. Bioinformatics, 26(10), 1372– 1373.
bioin forma tics/btq110
Bouckaert, R., Heled, J., Kühnert, D., Vaughan, T., Wu, C.- H., Xie, D.,
Suchard, M. A., Rambaut, A., & Drummond, A. J. (2014). BEAST
2: A software platform for Bayesian evolutionary analysis. PLoS
Computational Biology, 10 (4), e1003537.
journ al.pcbi.1003537
Bryant, D., Bouckaert, R., Felsenstein, J., Rosenberg, N. A., &
RoyChoudhury, A. (2012). Inferring species trees directly from bial-
lelic genetic markers: Bypassing gene trees in a full coalescent anal-
ysis. Molecular Biology and Evolution, 29(8), 19171932. https://doi.
org /10.1093/molbe v/mss086
Burg, T. M., & Croxall, J. P. (2001). Global relationships amongst
black- browed and grey- headed albatrosses: Analysis of pop-
ulation structure using mitochondrial DNA and microsat-
ellites. Molecular Ecology, 10 (11), 2647– 2660. https://doi.
org/10.1046/j.0962- 1083.2001.01392.x
Catchen, J., Hohenlohe, P. A., Bassham, S., Amores, A., & Cresko, W.
A. (2013). Stacks: An analysis tool set for population genomics.
Molecular Ecology, 22(11), 3124314 0. htt ps://doi.or g/10.1111/
Chambers, E. A., & Hillis, D. M. (2020). The multispecies coalescent
over- splits species in the case of geographically widespread taxa.
Systematic Biology, 69(1), 184– 193. o/
Cicero, C., Mason, N. A., Jiménez, R. A., Wait, D. R., Wang- Claypool, C. Y.,
& Bowie, R. C. K. (2021). Integrative taxonomy and geographic sam-
pling underlie successful species delimitation. Ornithology, 138(2), 1–
15. holog y/ukab009
Clausen, A. P., & Pütz, K. (2002). Recent trends in diet composition
and productivity of gentoo, magellanic and rockhopper penguins
in the Falkland Islands. Aquatic Conservation- Marine and Freshwater
Ecosystems, 12(1), 51– 61.
Clucas, G. V., Younger, J. L., Kao, D., Emmerson, L., Southwell, C.,
Wienecke, B., Rogers, A. D., Bost, C. A., Miller, G. D., Polito, M. J.,
Lelliott, P., Handley, J., Crofts, S., Phillips, R. A., Dunn, M. J., Miller,
K. J., & Hart, T. (2018). Comparative population genomics reveals key
barriers to dispersal in Southern Ocean penguins. Molecular Ecology,
27(23), 4 68 0– 4697. ht tps:// /10.1111/mec.14896
Cole, T. L., Dutoit, L., Dussex, N., Hart, T., Alexander, A., Younger, J. L.,
Clucas, G. V., Frugone, M. J., Cherel, Y., Cuthbert, R., Ellenberg, U.,
Fiddaman, S. R., Hiscock, J., Houston, D., Jouventin, P., Mattern, T.,
Miller, G., Miskelly, C., Nolan, P., … Waters, J. M. (2019). Receding ice
drove parallel expansions in Southern Ocean penguins. Proceedings
of the National Academy of Sciences, 116(52), 2669026696. https:// 48116
Cole, T. L., Ksepka, D. T., Mitchell, K. J., Tennyson, A. J. D., Thomas, D. B.,
Pan, H., Zhang, G., Rawlence, N. J., Wood, J. R., Bover, P., Bouzat, J.
L., Cooper, A., Fiddaman, S., Hart, T., Miller, G., Ryan, P. G., Shepherd,
L. D., Wilmshurst, J. M., & Waters, J. M. (2019). Mitogenomes un-
cover extinct penguin taxa and reveal island formation as a key driver
of speciation. Molecular Biology and Evolution, 36(4), 784– 797.
Cracraft, J. (1983). Species concepts and speciation analysis. In R. F.
Johnston (Ed.), Current Ornithology (pp. 159– 187). Springer.
Cracraft, J. (1992). The species of the birds- of- paradise (Paradisaeidae):
Applying the phylogenetic species concept to a complex pattern of
diversification. Cladistics: The International Journal of the Willi Hennig
Society, 8(1), 1– 43. ht tps:// /10.1111/j.1096- 0 031.1992.
tb000 49.x
Crawford, R . J. M., Cooper, J., Dyer, B. M., Greyling, M. D., Klages, N. T.
W., Nel, D. C., Nel, J. L., Petersen, S. L., & Wolfaardt, A. C. (2003).
Decrease in numbers of the eastern rockhopper penguin Eudyptes
chrysocome filholi at Marion Island, 1994/95– 2002/03. Africa n Journal
of Marine Science, 25(1), 4 87– 498.
32030 9504037
Crofts, S., & Robson, B. (2015). First record of hybridisation between
northern Eudyptes moseleyi and southern rockhopper penguins
E.c.chrysocome. Seabird, 28, 37– 42.
Cronquist, A. (1978). Once again, what is a species? In J. A. Romberger
(Ed.), Biosystematics in Agriculture (pp. 3– 20). Montclair, NJ: John
Wiley & Sons.
Cuthbert, R., Cooper, J., Burle, M. H., Glass, C. J., Glass, J. P., Glass, S.,
Glass, T., Hilton, G. M., Sommer, E. S., Wanless, R. M., & Ryan, P.
G. (2009). Population trends and conservation status of the north-
ern rockhopper penguin Eudyptes moseleyi at Tristan da Cunha and
Gough Island. Bird Conservation International, 19(1), 109– 1 20 . https:// 27090 8007545
Cutter, A. D. (2013). Integrating phylogenetics, phylogeography and
population genetics through genomes and evolutionary theory.
Molecular Phylogenetics and Evolution, 69(3), 1172– 1185. https://doi.
Danecek, P., Auton, A., Abecasis, G., Albers, C. A., Banks, E., DePristo, M.
A., Handsaker, R. E., Lunter, G., Marth, G. T., Sherry, S. T., McVean,
G., Durbin, R., & Genomes Project Analysis Group (2011). The variant
call format and VCFtools. Bioinformatics, 27(15), 2156– 2158. https:// forma tics/btr330
de Dinechin, M., Dobson, F. S., Zehtindjiev, P., Metcheva, R., Couchoux,
C., Martin, A., Quillfeldt, P., & Jouventin, P. (2012). The biogeog-
raphy of gentoo penguins (Pygoscelis papua). Canadian Journal of
Zoology- Revue Canadienne De Zoologie, 90(3), 352– 360. https://doi.
o r g / 1 0 . 1 1 3 9 / z 2 0 1 2 - 0 1 6
de Dinechin, M., Ottvall, R., Quillfeldt, P., & Jouventin, P. (2009).
Speciation chronology of rockhopper penguins inferred from molec-
ular, geological and palaeoceanographic data. Journal of Biogeog raphy,
36(4), 693– 702. 2699.2008.02014.x
De Queiroz, K. (20 07). Species concepts and species delimitation.
Systematic Biology, 56(6), 879– 886.
15070 1701083
Earl, D. A., & vonHoldt, B. M. (2012). STRUCTURE HARVESTER: A web-
site and program for visualizing STRUCTURE output and implement-
ing the Evanno method. Conservation Genetics Resources, 4(2), 359–
3 6 1 . h t t p s : / / d o i . o r g / 1 0 . 1 0 0 7 / s 1 2 6 8 6 - 0 1 1 - 9 5 4 8 - 7
Edwards, S., & Bensch, S. (2009). Looking forwards or looking backwards
in avian phylogeography? A comment on Zink and Barrowclough
2008 . Molecular Ecology, 18(14), 2930– 2933. https://doi.
org /10.1111/j.1365- 294X .2009.0 4270.x
Esquerre, D., Ramirez- Alvarez, D., Pavon- Vazquez, C. J., Troncoso-
Palacios, J., Garin, C. F., Keogh, J. S., & Leache, A. D. (2019).
Speciation across mountains: Phylogenomics, species delimita-
tion and taxonomy of the Liolaemus leopardinus Glade (Squamata,
Liolaemidae). Molecular Phylogenetic s and Evolution, 139, 17. https://
Evanno, G., Regnaut, S., & Goudet, J. (2005). Detecting the number
of clusters of individuals using the software STRUCTURE: A sim-
ulation study. Molecular Ecology, 14 (8), 26112620. https://doi.
org/10.1111/j.1365- 294X.2005.02553.x
Falush, D., Stephens, M., & Pritchard, J. K . (2003). Inference of popula-
tion structure using multilocus genot ype data: Linked loci and cor-
related allele frequencies. Genetics, 164(4), 1567– 1587. https://doi.
org/10.1093/genet ics/164.4.1567
Feder, J. L., Egan, S. P., & Nosil, P. (2012). The genomics of speciation
with gene flow. Trends in Genetics: TIG, 28(7), 342– 350. https://doi.
Fiser, C., Robin son, C. T., & Malar d, F. (2018). Cryp tic spe cies as a win dow
into the paradigm shift of the species concept. Molecular Ecology,
27(3), 613– 635.
Frankham, R., Ballou, J. D., Dudash, M. R ., Eldridge, M. D. B., Fenster,
C. B., Lacy, R. C., Mendelson, J. R., Porton, I. J., Ralls, K., & Ryder,
O. A. (2012). Implications of different species concepts for conser v-
ing biodiversity. Biological Conservation, 153, 25– 31. https://doi.
Frugone, M. J., López, M. E., Segovia, N. I., Cole, T. L., Lowther, A.,
Pistorius, P., Dantas, G. P. M., Petry, M. V., Bonadonna, F., Trathan, P.,
Polanowski, A., Wienecke, B., Bi, K., Wang- Claypool, C. Y., Waters,
J. M., Bowie, R. C. K., Poulin, E., & Vianna, J. A. (2019). More than
the eye can see: Genomic insights into the drivers of genetic differ-
entiation in Royal/Macaroni penguins across the Southern Ocean.
Molecular Phylogenetics and Evolution, 139, 106563. https://doi.
Frugone, M. J., Lowther, A., Noll, D., Ramos, B., Pistorius, P., Dantas, G. P.
M., Petry, M. V., Bonadonna, F., Steinfurth, A., Polanowski, A., Raya
Rey, A., Lois, N. A., Pütz, K., Trathan, P., Wienecke, B., Poulin, E., &
Vianna, J. A. (2018). Contrasting phylogeographic pattern among
Eudypte s penguins around the Southern Ocean. Scientific Reports,
8( 1 ) , 1 7 4 8 1 . h t t p s : / / d o i . o r g / 1 0 . 1 0 3 8 / s 4 1 5 9 8 - 0 1 8 - 3 5 9 7 5 - 3
Fujita, M. K., Leaché, A . D., Burbrink, F. T., McGuire, J. A., & Moritz, C.
(2012). Coalescent- based species delimitation in an integrative tax-
onomy. Trends in Ecology & Evolution, 27(9), 480– 488. https://doi.
García, P., & Boersma, D. (2013). Penguins: Natural History and
Conservation (pp. 360). University of Washington Press.
Gavryushkina, A., Heath, T. A., Ksepka, D. T., Stadler, T., Welch, D., &
Drummond, A. J. (2017). Bayesian total- evidence dating reveals the
recent crown radiation of penguins. Systematic Biology, 66(1), 57– 73. o/syw060
Grosser, S., Burridge, C. P., Peucker, A . J., & Waters, J. M. (2015).
Coalescent modelling suggests recent secondary- contact of cryp-
tic penguin species. PLoS One, 10 (12), e0144966. https://doi.
org/10.1371/journ al.pone.0144966
Gu, Z., Gu, L., Eils, R., Schlesner, M., & Brors, B. (2014). Circlize imple-
ments and enhances circular visualization in R. Bioinformatics, 30(19),
28112812. forma tics/btu393
Guindon, S., Dufayard, J.- F., Lefort, V., Anisimova, M., Hordijk, W.,
& Gascuel, O. (2010). New algorithms and methods to estimate
maximum- likelihood phylogenies: Assessing the performance
of PhyML 3.0. Systematic Biology, 59(3), 307– 321. https://doi.
org/10.1093/sysbi o/syq010
   FRUGONE Et al.
Guo, B. C., Fang, B. H., Shikano, T., Momigliano, P., Wang, C., Kravchenko,
A., & Merila, J. (2019). A phylogenomic perspective on diversity,
hybridization and evolutionary affinities in the stickleback genus
Pungitius. Molecular Ecology, 28(17), 4 046– 406 4.
Hiscock, J. A. (2013). Monitoring penguins in the Antipodes Island Group:
Methods and baseline data (Department of Conservation Technical
Series 37). Department of Conservation.
Hoang, D. T., Chernomor, O., von Haeseler, A., Minh, B. Q., & Vinh, L.
S. (2017). UFBoot2: Improving the ultrafast bootstrap approxima-
tion. Molecular Biolog y and Evolution, 35(2), 518– 522. https://doi.
org /10.1093/molbe v/msx281
Hull , C. L. (199 9). The for agin g zone s of bre edi ng roya l (Eudyptes schlegeli)
and rockhopper (E. chrysocome) penguins: An assessment of tech-
niques and species comparison. Wildlife Research, 26(6), 789– 803.
Hull, C. L., Hindell, M., Le Mar, K., Scofield, P., Wilson, J., & Lea, M. A.
(2004). The breeding biology and factors affecting reproductive
success in rockhopper penguins Eudyptes chrysocome at Macquarie
Island. Polar Biology, 27 (11), 711– 720.
0- 004- 0643- z
Hull, C., & Wilthshire, A. (1999). An apparent hybrid royal x rockhopper
penguin at Macquarie Island. Australian Bird Watcher, 183, 95– 100.
Hutton, F. W. (1878). On an apparently new species of penguin, from
Campbell Island. Proceedings of the Linnean Society of New South
Wales, 3(12), 334– 335.
Jackson, N. D., Carstens, B. C., Morales, A. E., & O’Meara, B. C. (2017).
Species delimitation with gene flow. Systematic Biology, 66(5), 799–
812. o/syw117
Jakobsson, M., & Rosenberg, N. A. (2007). CLUMPP: A cluster matching
and permutation program for dealing with label switching and mul-
timodality in analysis of population structure. Bioinformatics, 23(14),
1801– 1806.
Jo uvent in, P., Cuth ber t , R. J., & Ott vall, R. (2006) . Genetic isolat ion and di -
vergence in sexual traits: Evidence for the northern rockhopper pen-
guin Eudyptes moseleyi b ei n g a s ib li n g s p ec i e s. Molecular Ecology, 15(11),
3413– 3423. 294X.2006.03028.x
Kalyaanamoorthy, S., Minh, B. Q., Wong, T. K. F., von Haeseler, A., &
Jermiin, L. S. (2017). ModelFinder: Fast model selection for accurate
phylogenetic estimates. Nature Methods, 14(6), 587– 589. https://doi.
Kamvar, Z. N., Brooks, J. C., & Grünwald, N. J. (2015). Novel R tools for
analysis of genome- wide population genetic data with emphasis
on clonality. Frontiers in Genetics, 6, 208.
Kamvar, Z. N., Tabima, J. F., & Grünwald, N. J. (2014). Poppr: An R pack-
age for genetic analysis of populations with clonal, partially clonal,
and/or sexual reproduction. Pee rJ, 2, e281. ht tps://
Kass, R. E., & Raftery, A. E. (1995). Bayes factors. Journal of the
American Statistical Association, 90 (430), 773– 795. https://doi.
org/10.1080/01621 459.1995.10476572
Key, K. H. L. (1968). The concept of stasipatric speciation. Systematic
Biology, 17(1), 14– 22. o/17.1.14
Kopelman, N. M., Mayzel, J., Jakobsson, M., Rosenberg, N. A., &
Mayrose, I. (2015). Clumpak: A program for identifying cluster-
ing modes and packaging population structure inferences across
K. Molecular Ecology Resources, 15(5), 1179– 1191. https://doi.
org /10.1111/1755- 0998.12387
Leache, A. D., Fujita, M. K., Minin, V. N., & Bouckaer t, R. R. (2014). Species
delimitation using genome- wide SNP data. Systematic Biology, 63(4),
534– 542. o/syu018
Leache, A. D., Zhu, T. Q., Rannala, B., & Yang, Z. H. (2019). The spectre
of too many species. Systematic Biology, 68(1), 168– 181. https://doi.
org/10.10 93/sysbi o/sy y051
Lewis, P. O. (2001). A likelihood approach to estimating phylogeny from
discrete morphological character data. Systematic Biology, 50(6),
913– 925. /10.1080/10635 15017 53462876
Li, H., & Durbin, R. (2009). Fast and accurate short read alignment with
Burrows- Wheeler transform. Bioinformatics, 25(14), 17541760. forma tics/btp324
Li, H., Handsaker, B., Wysoker, A., Fennell, T., Ruan, J., Homer, N., Mar th,
G., Abecasis, G., & Durbin, R. (2009). The sequence alignment/Map
format and SAMtools. Bioinformatics, 25(16), 2078– 2079. https://doi.
org/10.1093/bioin forma tics/btp352
Lischer, H. E. L., & Excoffier, L. (2012). PGDSpider: An automated data
conversion tool for connecting population genetics and genomics
programs. Bioinformatics, 28(2), 298– 299.
bioin forma tics/btr642
Lois, N. A., Campagna, L., Balza, U., Polito, M. J., Pütz, K., Vianna, J. A.,
Morgenthaler, A., Frere, E., Sáenz- Samaniego, R., Raya Rey, A., &
Mahler, B. (2020). Metapopulation dynamics and foraging plasticity
in a highly vagile seabird, the southern rockhopper penguin. Ecology
and Evolution, 10, 3346– 3355.
Maddison, W. P. (1997). Gene trees in species trees. Systematic Biology,
46(3), 523– 536. o/46.3.523
Malinsky, M., Trucchi, E., Lawson, D. J., & Falush, D. (2018). RADpainter
and fineRADstructure: Population inference from RADseq data.
Molecular Biolog y and Evolution, 35(5), 1284– 1290. https://doi.
org /10.1093/molbe v/msy023
Martin, M. (2011). Cutadapt removes adapter sequences from high-
throughput sequencing reads. EMBnet.journal, 17, 10– 12. ht tps://doi.
org/10.148 06/ ej.17.1.200
Martínez, I. (1992). Order sphenisciformes. In J. del Hoyo, A. Elliott, & J.
Sargatal (Eds.), Handbook of the birds of the world, Ostrich to Ducks (1,
pp. 140– 160). Lynxs Edicions.
Mason, N. A., & Bowie, R. C. K. (2020). Plumage patterns: Ecological
functions, evolutionary origins, and advances in quantification. The
Auk, 137(4), 1– 29.
Mason, N. A., & Taylor, S. A. (2015). Differentially expressed genes match
bill morphology and plumage despite largely undifferentiated ge-
nomes in a Holarctic songbird. Molecular Ecology, 24(12), 3009– 3025.
https://doi.or g/10.1111/m ec .1314 0
Matos- Maraví, P., Wahlberg, N., Antonelli, A ., & Penz, C. M. (2019).
Species limits in butterflies (Lepidoptera: Nymphalidae): Reconciling
classical taxonomy with the multispecies coalescent. Systematic
Entomology, 44(4), 745– 756.
Mayr, E. (1942). Systema tics and the Origi n of Species, from th e Viewpoint of
a Zoologist. Columbia University Press.
Mays, H. L., Oehler, D. A., Morrison, K. W., Morales, A. E., Lycans,
A., Perdue, J., Battley, P. F., Cherel, Y., Chilvers, B. L., Crofts, S.,
Demongin, L., Fry, W. R., Hiscock, J., Kusch, A., Marin, M., Poisbleau,
M., Quillfeldt, P., Raya Rey, A., Steinfurth, A., … Weakley, L. A. (2019).
Phylogeography, population structure, and species delimitation in
rockhopper penguins (Eudyptes chrysocome and Eudyptes moseleyi).
The Journal of Heredity, 110(7), 801– 817.
jhere d/esz051
Meirmans, P. G., & Van Tienderen, P. H. (2004). Genotype and gen-
odive: Two programs for the analysis of genetic diversity of asex-
ual organisms. Molecular Ecology Notes, 4(4), 792– 794. https://doi.
org /10.1111/j.1471- 8286.20 04.0 0770.x
Miller, M. A., Pfeiffer, W., & Schwartz, T. (2010). Creating the CIPRES
Science Gateway for inference of large phylogenetic trees (pp. 1– 8).
Proceedings of the Gateway Computing Environments Workshop
(GCE), 14 Nov. 2010.
Milot, E., Weimerskirch, H., & Bernatchez, L . (2008). The seabird para-
dox: Dispersal, genetic structure and population dynamics in a highly
mobile, but philopatric albatross species. Molecular Ecology, 17(7),
1658– 1673. 294X.2008.03700.x
Morrison, K. W. (2016). Individual repeatability in laying behaviour does
not support the migratory carry- over effect hypothesis of egg- size
dimorphism in Eudyptes penguins. Journal of Avian Biolog y, 47(4),
466– 475.
Morrison, K. W., Battley, P. F., Sagar, P. M., & Thompson, D. R. (2015).
Population dynamics of Eastern Rockhopper Penguins on Campbell
Island in relation to sea surface temperature 19422012: Current
warming hiatus pauses a long- term decline. Polar Biology, 38(2), 163–
1 7 7 . h t t p s : / / d o i . o r g / 1 0 . 1 0 0 7 / s 0 0 3 0 0 - 0 1 4 - 1 5 7 5 - x
Napier, R. B. (1968). Erect- crested and rockhopper penguins inter-
breeding in the Falkland Islands. British Antarctic Survey Bulletin, 16,
71– 7 2 .
Nelson, G. J., & Platnick, N. I. (1981). Systematic and Biogeography:
Cladistics and Vicariance. Columbia University Press.
Nguyen, L.- T., Schmidt, H. A ., von Haeseler, A., & Minh, B. Q. (2015).
IQ- TREE: A fast and effective stochastic algorithm for estimating
maximum- likelihood phylogenies. Molecular Biology and Evolution,
32(1), 268– 274. v/msu300
Nosil, P., Funk, D. J., & Ortiz- Barrientos, D. (2009). Divergent selection
and heterogeneous genomic divergence. Molecular Ecology, 18(3),
375– 402. 294X.200 8.03946.x
O’Lear y, S. J., Puritz, J. B., Willis, S. C., Hollenbeck, C . M., & Portnoy, D.
S. (2018). These aren’t the loci you're looking for: Principles of effec-
tive SNP filtering for molecular ecologists. Molecular Ecology, 27(16),
3193– 320 6. https://doi .org/10.1111/mec.14792
Ortiz, E. M. (2019). vcf2phylip v2.0: Conver t a VCF matrix into several ma-
trix formats for phylogenetic analysis.
Ottenburghs, J. (2019). Multispecies hybridization in birds. Avian
Research, 10( 1 ) , 2 0 . h t t p s : / /d o i . o r g / 1 0 . 1 1 8 6 / s 4 0 6 5 7 - 0 1 9 - 0 1 5 9 - 4
Ot tenbu rgh s, J., Ydenber g, R. C., Van Hoof t, P., Van Wieren, S. E., & Prin s,
H. H. T. (2015). The avian hybrids project: Gathering the scientific
literature on avian hybridization. Ibis, 157(4), 892– 894. https://doi.
org /10.1111/ibi.12285
Pedraza- Marrón, C. D. R., Silva, R., Deeds, J., Van Belleghem, S. M.,
Mastretta- Yanes, A., Domínguez- Domínguez, O., Rivero- Vega,
R. A., Lutackas, L., Murie, D., Parkyn, D., Bullock, L. H., Foss, K.,
Ortiz- Zuazaga, H., Narváez- Barandica, J., Acero, A., Gomes, G., &
Betancur- R, R. (2019). Genomics overrules mitochondrial DNA,
siding with morphology on a controversial case of species delimita-
tion. Proceedings of the Royal Society - Biological Sciences, 286(190 0),
Pertierra, L. R., Segovia, N. I., Noll, D., Martinez, P. A., Pliscoff, P., Barbosa,
A., Aragón, P., Raya Rey, A., Pistorius, P., Trathan, P., Polanowski, A.,
Bonadonna, F., Le Bohec, C., Bi, K., Wang- Claypool, C. Y., González-
Acuña, D., Dantas, G. P. M., Bowie, R. C. K., Poulin, E., & Vianna, J. A.
(2020). Cryptic speciation in gentoo penguins is driven by geographic
isolation and regional marine conditions: Unforeseen vulnerabilities
to global change. Diversity and Distributions, 26(8), 958– 975. https://
Peterson, B. K., Weber, J. N., Kay, E. H., Fisher, H. S., & Hoekstra, H. E.
(2012). Double digest radseq: An inexpensive method for de novo
snp discover y and genotyping in model and non- model species. PLoS
One, 7(5), e37135. al.pone.0037135
Pons, J., Barraclough, T. G., Gomez- Zurita, J., Cardoso, A., Duran,
D. P., Hazell, S., Kamoun, S., Sumlin, W. D., & Vogler, A. P. (20 06).
Sequence- based species delimitation for the DNA taxonomy of un-
described insects. Systematic Biology, 55(4), 595– 609. https://doi.
org/10.1080/10635 15060 0852011
Porras- Hurtado, L., Ruiz, Y., Santos, C., Phillips, C., Carracedo, Á., &
Lareu, M. (2013). An overview of STRUCTURE: Applications, param-
eter settings, and supporting software. Frontier s in Genetics, 4, 98.
Pritchard, J. K., Stephens, M., & Donnelly, P. (2000). Inference of popula-
tion structure using multilocus genotype data. Genetics, 155(2), 945–
959. ics/155.2.945
Pritchard, J., Wen, X., & Falush, D. (2010). Documentation for structure
software: Version 2.3. Retrieved from https://web.stanf
g r o u p / p r i t c h a r d l a b / s t r u c t u r e _ s o f t w a r e / r e l e a s e _ v e r s i o n s / v 2 . 3 . 4 /
struc ture_doc.pdf
Puillandre, N., Lamber t, A., Brouillet, S., & Achaz, G. (2012). ABGD,
automatic barcode gap discovery for primary species de-
limitation. Molecular Ecology, 21(8), 1864– 1877. https://doi.
org /10.1111/j.1365- 294X .2011.05239.x
Purcell, S., Neale, B., Todd- Brown, K., Thomas, L., Ferreira, M. A., Bender,
D., Maller, J., Sklar, P., de Bakker, P. I., Dal y, M. J., & Sha m, P. C. (2007).
PLINK: A tool set for whole- genome association and population-
based linkage analyses. American Journal of Human Genetics, 81(3),
559– 575.
Quillfeldt, P., Moodley, Y., Weimerskirch, H., Cherel, Y., Delord, K.,
Phillips, R. A., Navarro, J., Calderón, L., & Masello, J. F. (2017). Does
genetic structure reflect differences in non- breeding movements? A
case study in small, highly mobile seabirds. BMC Evolutionary Biology,
17( 1 ) , 1 6 0 . h t t p s : / / d o i . o r g / 1 0 . 1 1 8 6 / s 1 2 8 6 2 - 0 1 7 - 1 0 0 8 - x
Rambaut, A., Drummond, A. J., Xie, D., Baele, G., & Suchard, M. A. (2018).
Posterior summarization in Bayesian phylogenetics using Tracer 1.7.
Systematic Biology, 67(5), 901– 904. o/
Rey, A. R., Pütz, K ., Luna- Jorquera, G., Lüthi, B., & Schiavini, A. (2009).
Diving patterns of breeding female rockhopper penguins (Eudyptes
chrysocome): Noir Island, Chile. Polar Biology, 32(4), 561– 568. ht tps://
d o i . o r g / 1 0 . 1 0 0 7 /s 0 0 3 0 0 - 0 0 8 - 0 5 5 0 - 9
Rey, A. R ., Trathan, P., Pütz, K., & Schiavini, A. (2007). Effect of ocean-
ographic conditions on the winter movements of rockhopper
penguins Eudyptes chrysocome chrysocome from Staten Island,
Argentina. Marine Ecology Progress Series, 330, 285– 295. https://doi.
org/10.3354/meps3 30285
Rosenberg, N. A. (2003). Distruct: A program for the graphical display of
population structure. Molecular Ecology Notes, 4(1), 137– 138. https:// 8286.2003.00566.x
Sangster, G. (2014). The application of species criteria in avian taxonomy
and its implications for the debate over species concepts. Biological
Reviews of the Cambridge Philosophical Society, 89(1), 199– 214.
https://doi.or g/10.1111/ br v.12051
Schilthuizen, M. (2000). Dualism and conflicts in understanding spe-
ciation. BioEssays: News and Reviews in Molecular, Cellular and
Developmental Biology, 22(12), 1134– 1141.
Searby, A., & Jouventin, P. (2005). The double vocal signature of crested
penguins: Is the identity coding system of rockhopper penguins
Eudyptes chrysocome due to phylogeny or ecology? Journal of Avian
Biology, 36(5), 449– 460.
Shafer, A . B. A., Peart, C. R., Tusso, S., Maayan, I., Brelsford, A., Wheat,
C. W., & Wolf, J. B. W. (2017). Bioinformatic processing of RAD- seq
data dramatically impacts downstream population genetic infer-
ence. Methods in Ecology and Evolution, 8(8), 907– 917. https://doi.
org /10.1111/2041- 210X.12700
Simeone, A., Hiriart- Bertrand, L., Reyes- Arriagada, R., Halpern, M.,
Dubach, J., Wallace, R., Pütz, K., & Luthi, B. (2009). Heterospecific
pairing and hybridization between wild Humboldt and Magellanic
penguins in Southern Chile. The Condor, 111(3), 544– 550. https://doi.
Solís- Lemus, C., Knowles, L. L., & Ané, C. (2015). Bayesian species delim-
itation combining multiple genes and traits in a unified framework.
Evolution, 69(2), 492507.
Spinks, P. Q., Thomson, R. C., & Shaffer, H. B. (2014). The advantages of
going large: Genome- wide SNPs clarify the complex population his-
tory and systematics of the threatened western pond turtle. Molecular
Ecology, 23(9), 2228– 2241.
Sukumaran, J., & Knowles, L. L. (2017). Multispecies coalescent delim-
its structure, not species. Proceedings of the National Academy of
Sciences, 114(7), 1607. 21114
   FRUGONE Et al.
Team, R Core (2020). R: A language and environment for statistical com-
puting. R Foundation for Statistical Computing. Retrieved from https://
w w w . R - p r o j e c t . o r g /
Tennyson, A. J. D., & Miskelly, C. (1989). “Dark- faced” rockhopper pen-
guins at the Snares Islands. Notorn is, 36(3), 183– 189.
Thiebot, J. B., Cherel, Y., Trathan, P. N., & Bost, C. A. (2011). Inter-
population segregation in the wintering areas of macaroni pen-
guins. Marine Ecology Progress Series, 421, 279– 290. https://doi.
org/10.3354/meps0 8907
Thiebot, J. B., Cherel, Y., Trathan, P. N., & Bost, C. A . (2012). Coexistence
of oceanic predators on wintering areas explained by population-
scale foraging segregation in space or time. Ecology, 93(1), 122– 130. 0385.1
Toews, D. P. L ., & Brelsford, A. (2012). The biogeography of mitochon-
drial and nuclear discordance in animals. Molecular Ecology, 21(16),
39073930. 294X.2012.05664.x
Trathan, P. N., García- Borboroglu, P., Boersma, D., Bost, C.- A., Crawford,
R. J. M., Crossin, G. T., Cuthbert, R. J., Dann, P., Davis, L. S., De La
Puente, S., Ellenberg, U., Lynch, H. J., Mattern, T., Pütz, K., Seddon, P.
J., Trivelpiec e, W., & Wienecke, B. (2015). Pol lu tion, habit at los s, fish-
ing, and climate change as critical threats to penguins. Conservation
Biology, 29(1), 31– 41.
Turner, S. D. (2014). qqman: An R package for visualizing GWAS re-
sults using Q- Q and Manhattan plots. bioR xiv, 5165. https://doi.
Tyler, J., Bonfitto, M. T., Clucas, G. V., Reddy, S., & Younger, J. L. (2020).
Morphometric and genetic evidence for four species of gentoo pen-
guin. Ecology and Evolution, 10 (24), 13836– 13846.
Vianna, J. A., Fernandes, F. A . N., Frugone, M. J., Figueiró, H. V., Pertierra,
L. R., Noll, D., Bi, K. E., Wang- Claypool, C. Y., Lowther, A., Parker, P.,
Le Bohec, C., Bonadonna, F., Wienecke, B., Pistorius, P., Steinfurth,
A., Burridge, C. P., Dantas, G. P. M., Poulin, E., Simison, W. B., …
Bowie, R. C. K. (2020). Genome- wide analyses reveal drivers of pen-
guin diversification. Proceedings of the National Academy of Sciences
of the United States of America, 117(36), 22303– 22310. https://doi.
org/10.1073/pnas.20066 59117
Vianna, J. A ., Noll, D., Dantas, G. P. M., Petry, M. V., Barbosa, A.,
Gonzalez- Acuna, D., Le Bohec, C., Bonadonna, F., & Poulin, E. (2017).
Marked phylogeographic structure of gentoo penguin reveals an on-
going diversification process along the Southern Ocean. Molecular
Phylogenetics and Evolution, 107, 486– 498.
Wang, J. (2017). The computer program structure for assigning individu-
als to populations: Easy to use but easier to misuse. Molecular Ecology
Resources, 17(5), 981– 990. 0998.12650
White, R., & Clausen, A. (2002). Rockhopper Eudyp tes chrysocome ch ryso-
come × macaroni E. chrysolophus penguin hybrids apparently breed-
ing in the Falkland islands. Marine Ornithology, 30, 40– 42.
Whitlock, M. C., & Lotterhos, K. E. (2015). Reliable detection of loci
responsible for local adaptation: Inference of a null model through
trimming the distribution of F(ST). The Americ an Naturalist, 186(Suppl
1), S24– S36.
Wickham, H. (2016). ggplot2: Elegant graphics for data analysis. Springer-
Verlag. ISBN 978- 3- 319- 24277- 4.
Wickham, H., François, R., Henr y, L., & Müller, K . (2018). dplyr: A gram-
mar of data manipulation. R package version 0.7 (R package version
0. 7. 6 . Retrieved from https://CRAN.R- proje ge=dplyr
[Computer software]
Wickham, H., & Henr y, L. (2018). tidyr: Easily tidy data with “spread”
and “gather” functions. R package version 0.8 (R package version
0.8.2. Retrieved from https://CRAN.R- proje ge=tidyr
[Computer software]
Williams, T. D., & Croxall, J. P. (1991). Annual variation in breeding biology
of macaroni penguins, Eudyptes chrysolophus, at Bird island, South
Georgia. Journal of Zoology, 223, 189– 2 02. ht tps:// /10.1111/
j . 1 4 6 9 - 7 9 9 8 . 1 9 9 1 . t b 0 4 7 5 9 . x
Wilson, G. A., & Rannala, B. (20 03). Bayesian inference of recent migra-
tion rates using multilocus genotypes. Genetics, 163(3), 1177– 1191. ics/163.3.1177
Wilson, J. W., Burle, M. H., Cuthber t, R ., Stirnemann, R. L., & Ryan, P. G.
(2010). Breeding success of northern rockhop per pengu ins (Eudyptes
moseleyi) at Gough Island, South Atlantic Ocean. The Emu, 110 (2),
Winkler, D. W., Billerman, S. M., & Lovette, I. J. (2015). Bird families of the
world: An invitation to the spectacular diversity of birds. Lynx Edicions.
Woehler, E. J., & Gilbert, C. A. (1990). Hybrid rockhopper- macaroni pen-
guins, interbreeding and mixed species pairs at Heard and Marion
islands. The Emu, 90, 198– 201. htt ps:// U990 0198
Wu, C. I. (2001). The genic view of the process of speciation.
Journal of Evolutionary Biology, 14(6), 851865. https://doi.
org/10.1046/j.1420- 9101.2001.0 0335.x
Yang, Z., & Rannala, B. (2010). Bayesian species delimitation using multi-
locus sequence data. Proceedings of the National Academy of Sciences,
107(20), 9264. 22107
Zheng, X., Levine, D., Shen, J., Gogarten, S. M., Laurie, C., & Weir, B. S.
(2012). A high- performance computing toolset for relatedness and
principal component analysis of SNP data. Bioinformatics, 28(24),
3326– 3328. forma tics/bts606
Zink, R. M., & Barrowclough, G. F. (2008). Mitochondrial DNA under
siege in avian phylogeography. Molecular Ecology, 17(9), 2107– 2121.
https://doi.or g/10.1111/ j.1365- 294X. 20 08.03737.x
María José Frugone is a postdoctoral researcher at Universidad
Austral de Chile interested in the evolution and diversification
of the Southern Ocean marine biota. This study was part of her
PhD thesis.
Author contributions: M.J.F., R.C.K.B., J.A.V. and E.P. wrote the
paper. M.J.F., M.E.L., G.C., P.M.M., N.A.L. and K.B. analysed the
data. J.M.W. and T.L.C. discussed results extensively and helped
prepare the manuscript. P.P., F.B., P.T., A.P., B.W., A.R.R., K.P., A.S
and C.Y.W.C performed field or laboratory work and contributed
to the manuscript.
Additional supporting information may be found online in the
Supporting Information section.
How to cite this article: Frugone, M. J., Cole, T. L., López, M.
E., Clucas, G., Matos- Maraví, P., Lois, N. A., Pistorius, P.,
Bonadonna, F., Trathan, P., Polanowski, A., Wienecke, B.,
Raya- Rey, A., Pütz, K., Steinfurth, A., Bi, K., Wang- Claypool,
C. Y., Waters, J. M., Bowie, R. C. K., Poulin, E., & Vianna, J. A.
(2021). Taxonomy based on limited genomic markers may
underestimate species diversity of rockhopper penguins and
threaten their conservation. Diversity and Distributions, 27,
2277– 2296.
[OPEN ACCESS] The Fairy Tern Sternula nereis is an Australasian tern that breeds in Australia, New Caledonia and New Zealand, with the latter having the smallest breeding population and is listed as ‘Threatened – Nationally Critical’ by the New Zealand Department of Conservation. Here, we investigate the genetic relatedness and level of endemism (gene flow) of the New Zealand Fairy Tern S. n. davisae population compared to the larger breeding populations in Australia S. n. nereis and New Caledonia S. n. exsul using the NADH subunit 2 (ND2) region of the mitochondrial DNA. We found that the three main populations (n = 86) were genetically distinct with a different fixed haplotype restricted to New Zealand (n = 15) and New Caledonia (n = 16), and that the estimated gene flow was low to zero, indicating no interbreeding between the populations. The current genetic evidence is consistent with observations of morphological and behavioural differences among the populations, and we suggest continued independent management of the population in New Zealand and further surveys and independent management of the New Caledonia population.
Full-text available
Species delimitation requires a broad assessment of population-level variation using multiple lines of evidence, a process known as integrative taxonomy. More specifically, studies of species limits must address underlying questions of what limits the distribution of populations, how traits vary in association with different environments, and whether the observed trait differences may lead to speciation through reproductive isolation. While genomic data have revolutionized the process of delimiting species, such data should be analyzed along with phenotypic, behavioral, and ecological traits that shape individuals across geographic and environmental space. The integration of multiple traits promotes taxonomic stability and should be a major guiding principle for species delimitation. Equally important, however, is thorough geographic sampling to adequately represent population-level variation—both in allopatry and across putative contact zones. We discuss the importance of both of these factors in the context of species concepts and traits and present different examples from birds that illustrate criteria for species delimitation. In addition, we review a decade of proposals for species-level taxonomic revisions considered by the American Ornithological Society’s North American Classification Committee, and summarize the basis for decisions on whether to split or lump species. Finally, we present recommendations and discuss challenges (specifically permits, time, and funding) for species delimitation studies. This is an exciting time to be studying species delimitation in birds: many species-level questions remain, and methodological advances along with increased access to data enable new approaches to studying age-old problems in avian taxonomy.
Full-text available
Gentoo penguins (Pygoscelis papua) are found across the Southern Ocean with a circumpolar distribution and notable genetic and morphological variation across their geographic range. Whether this geographic variation represents species‐level diversity has yet to be investigated in an integrative taxonomic framework. Here, we show that four distinct populations of gentoo penguins (Iles Kerguelen, Falkland Islands, South Georgia, and South Shetlands/Western Antarctic Peninsula) are genetically and morphologically distinct from one another. We present here a revised taxonomic treatment including formal nomenclatural changes. We suggest the designation of four species of gentoo penguin: P. papua in the Falkland Islands, P. ellsworthi in the South Shetland Islands/Western Antarctic Peninsula, P. taeniata in Iles Kerguelen, and a new gentoo species P. poncetii, described herein, in South Georgia. These findings of cryptic diversity add to many other such findings across the avian tree of life in recent years. Our results further highlight the importance of reassessing species boundaries as methodological advances are made, particularly for taxa of conservation concern. We recommend reassessment by the IUCN of each species, particularly P. taeniata and P. poncetii, which both show evidence of decline.
Full-text available
Birds exhibit remarkable variation in plumage patterns, both within individual feathers and among plumage patches. Differences in the size, shape, and location of pigments and structural colors comprise important visual signals involved in mate choice, social signaling, camouflage, and many other functions. While ornithologists have studied plumage patterns for centuries, recent technological advances in digital image acquisition and processing have transformed pattern quantification methods, enabling comprehensive, detailed datasets of pattern phenotypes that were heretofore inaccessible. In this review, we synthesize recent and classic studies of plumage patterns at different evolutionary and organismal scales and discuss the various roles that plumage patterns play in avian biology. We dissect the role of plumage patches as signals within and among species. We also consider the evolutionary history of plumage patterns, including phylogenetic comparative studies and evolutionary developmental research of the genetic architecture underlying plumage patterns. We also survey an expanding toolbox of new methods that characterize and quantify the size, shape, and distribution of plumage patches. Finally, we provide a worked example to illustrate a potential workflow with dorsal plumage patterns among subspecies of the Horned Lark (Eremophila alpestris) in western North America. Studies of plumage patterning and coloration have played a prominent role in ornithology thus far, and recent methodological and conceptual advances have opened new avenues of research on the ecological functions and evolutionary origins of plumage patterns in birds.
Full-text available
Genome-wide association studies (GWAS) have identified thousands of human trait-associated single nucleotide polymorphisms. Here, I describe a freely available R package for visualizing GWAS results using Q-Q and manhattan plots. The qqman package enables the flexible creation of manhattan plots, both genome-wide and for single chromosomes, with optional highlighting of SNPs of interest. Availability: qqman is released under the GNU General Public License, and is freely available on the Comprehensive R Archive Network ( The source code is available on GitHub (
Full-text available
Penguins are the only extant family of flightless diving birds. They currently comprise at least 18 species, distributed from polar to tropical environments in the Southern Hemisphere. The history of their diversification and adaptation to these diverse environments remains controversial. We used 22 new genomes from 18 penguin species to reconstruct the order, timing, and location of their diversification, to track changes in their thermal niches through time, and to test for associated adaptation across the genome. Our results indicate that the penguin crown-group originated during the Miocene in New Zealand and Australia, not in Antarctica as previously thought, and that Aptenodytes is the sister group to all other extant penguin species. We show that lineage diversification in penguins was largely driven by changing climatic conditions and by the opening of the Drake Passage and associated intensification of the Antarctic Circumpolar Current (ACC). Penguin species have introgressed throughout much of their evolutionary history, following the direction of the ACC, which might have promoted dispersal and admixture. Changes in thermal niches were accompanied by adaptations in genes that govern thermoregulation and oxygen metabolism. Estimates of ancestral effective population sizes (N e ) confirm that penguins are sensitive to climate shifts, as represented by three different demographic trajectories in deeper time, the most common (in 11 of 18 penguin species) being an increased N e between 40 and 70 kya, followed by a precipitous decline during the Last Glacial Maximum. The latter effect is most likely a consequence of the overall decline in marine productivity following the last glaciation.
Full-text available
Penguins are important top consumers in marine food webs and are one of the most threatened bird families, especially by climate change and food web alterations by marine fisheries. Yet, long-term population trends are lacking or are uncertain for many populations. Seven species of penguins breeding at the French Southern Territories in the southern Indian Ocean on the Crozet, Kerguelen, Saint-Paul–Amsterdam archipelagos and in Terre Adélie/Adelie Land, Antarctica are monitored regularly. This monitoring started in the early 1950s and most populations have been surveyed during the past four years, allowing assessments of population trends. King penguins increased at nearly all breeding sites within the Crozet and Kerguelen archipelagos. Emperor penguins have decreased at Terre Adélie/Adelie Land, with a partial recovery of the colony during the 2010s. Gentoo penguin populations at Crozet and Kerguelen are highly variable but stable. Adélie penguins have been increasing in Terre Adélie/Adelie Land. The trends in eastern rockhopper penguins vary between colonies and archipelagos. Northern rockhopper penguins have continuously decreased in numbers at Amsterdam Island, but appear to have increased at the nearby Saint-Paul Island. Macaroni penguins have first increased and then stabilized since the 2000s at Kerguelen and are stable at the Crozet Islands. Overall, most penguin populations breeding in the French Southern Territories increased or were stable over the past 30–60 years, with the exception of the northern rockhopper penguin, king and gentoo penguins on Crozet and the emperor penguin. The ecological reasons for these trends are poorly understood and require further investigation.
Full-text available
Aim The conservation of biodiversity is hampered by data deficiencies, with many new species and subspecies awaiting description or reclassification. Population genomics and ecological niche modelling offer complementary new tools for uncovering functional units of phylogenetic diversity. We hypothesize that phylogenetically delineated lineages of gentoo penguins (Pygoscelis papua ) distributed across Antarctica and sub‐Antarctic Islands are subject to spatially explicit ecological conditions that have limited gene flow, facilitating genetic differentiation, and thereby speciation processes. Location Antarctica and sub‐Antarctic area. Methods We identify divergent lineages for gentoo penguins using ddRAD‐seq and mtDNA, and generated species distribution models (SDMs) based on terrestrial and marine parameters. Results Analyses of our genomic data supports the existence of four major lineages of gentoo penguin: (i) spanning the sub‐Antarctic archipelagos north of the Antarctic Polar Front (APF); (ii) Kerguelen Island; (iii) South America; and (iv) across maritime Antarctic and the Scotia Arc archipelagos. The APF, a major current system around Antarctica, acts as the most important barrier separating regional sister lineages. Our ecological analyses spanning both the terrestrial (breeding sites) and marine (feeding sites) realms recover limited niche overlap among the major lineages of gentoo penguin. We observe this pattern to correspond more closely with regional differentiation of marine conditions than to terrestrial macroenvironmental features. Main conclusions Recognition of regional genetic lineages as discrete evolutionary entities that occupy distinct ecological niches and also differ morphologically should be considered a priority for conservation. Gentoo penguins provide a good example of how conservation policy can be directly impacted by new insights obtained through the integration of larger genomic datasets with novel approaches to ecological modelling. This is particularly pertinent to polar environments that are among the most rapidly changing environments on earth.