University of Groningen
Divergence and adaptive capacity of marine keystone species Fietz, Katharina
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date: 2017
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Fietz, K. (2017). Divergence and adaptive capacity of marine keystone species. Rijksuniversiteit Groningen.
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.
60
Chapter 3
Genetic insights into the extent of gene flow between the two
known humpback whale (Megaptera novaeangliae) breeding
grounds in the North Atlantic, and into the effective size of
the breeding population in Cape Verde
Genetic insights into the extent of gene flow between the two known humpback whale (Megaptera novaeangliae) breeding grounds in the North Atlantic, and into the effective size of the breeding population in Cape Verde
Katharina Fietz1, 2*, Martine Bérubé1, Conor Ryan3,4,5, Simon D. Berrow3,4, Pedro López-Suárez6, Frederick W.Wenzel7, Per J. Palsbøll1*
1
Marine Evolution and Conservation, Faculty of Science and Engineering, University of Groningen, Nijenborgh 7, 9747 AG Groningen, The Netherlands
2
Natural History Museum of Denmark, University of Copenhagen, Section for Evolutionary Genomics, Øster Voldgade 5-7, 1350 Copenhagen Denmark
3
Irish Whale and Dolphin Group, Merchant’s Quay, Kilrush, Co. Clare, Ireland 4
Marine and Freshwater Research Group, Galway-Mayo Institute of Technology, Dublin Road, Galway, Ireland
5
Current address: Hebridean Whale and Dolphin Trust, 28 Main Street, Tobermory, Isle of Mull, PA75 6NU, United Kingdom
6
BIOS.cv, C.P. 100, Sal Rei, Boa Vista, Republic of Cape Verde 7
NOAA, National Marine Fisheries Service, Northeast Fisheries Science Center, 166 Water St., Woods Hole, Ma. 02543 USA
Keywords Genetic connectivity - breeding ground - effective population size - Cape Verde -
humpback whale
Corresponding authors*:
Katharina Fietz, katharina.fietz@snm.ku.dk, Natural History Museum of Denmark, University of Copenhagen, Section for Evolutionary Genomics, Øster Voldgade 5-7, 1350 Copenhagen, Denmark
Per J. Palsbøll: p.j.palsboll@rug.nl, Faculty of Science and Engineering, Groningen Institute for Evolutionary Life Sciences, Nijenborgh 7, 9747 AG Groningen, The Netherlands
Abstract
Knowledge of the genetic connectivity and effective population size is important for conservation management of species whose populations have undergone extensive declines. Humpback whales (Megaptera novaeangliae) in the North Atlantic have undergone a drastic depletion during the 19th and 20th century due to heavy whaling activity and only within the last 60 years have had a chance to recover. Of the two known North Atlantic breeding grounds, the one in the Cape Verde Archipelago has previously been estimated to be home to only ca. 100 individuals and is thought to be reproductively isolated. Using two types of neutral genetic markers (up to 20 microsatellite loci and a fragment of the mitochondrial control region), we inferred the genetic connectivity among Cape Verde and the only other known North Atlantic breeding ground in the West Indies. We further provide an estimate of Cape Verde's contemporary effective population size (Ne). Genetic divergence estimates between Cape Verde
and the West Indies are of the same order of magnitude as between different ocean basins, and the Cape Verde population is currently at or below a minimum viable population size. An Approximate Bayesian Computation (ABC) investigation of Cape Verdes demographic population history supports a population decline in the past predating anthropogenic exploitation. While Cape Verde's humpback whale population is isolated from the West Indies breeding ground, our data suggest that it might be part of a larger, yet undiscovered Eastern Atlantic breeding population.
Introduction
Anthropogenic impacts on the earth's ecosystems are steadily expanding and are the main causes for population declines in many taxa (Morrison et al. 2007; Frankham et al. 2009). Reductions in population sizes typically entail a (temporary) limitation of gene flow and an associated increase in isolation. Small isolated populations often experience a loss of genetic diversity and as a result are particularly prone to the impacts of genetic stochasticity. It is worth noting that there is no straightforward causal link between decreased genetic diversity and inbreeding depression following a population reduction (Bouzat 2010). Populations of some wild species display remarkably low genetic diversity without showing signs of inbreeding depression (Obrien et al. 1985; Merola 1994; Castro-Prieto et al. 2011). Nonetheless, the loss of genetic diversity is a concern as reduced polymorphism increases the likelihood of inbreeding depression and in the long term lowers the evolutionary potential of a population to respond to environmental change (Fisher 1930; Wright 1931; Reed & Frankham 2003). Both the intensity and duration of a bottleneck thereby impact population survival and fitness. The large changes in allele frequencies that are associated with small populations will entail an increase in the frequency of deleterious alleles. The longer a bottleneck lasts, the more frequent these deleterious alleles may become over time, see e.g. Trimble and Keeler (1938).
In order to incorporate information on population size and the level of isolation in a conservation management framework, critical levels of these parameters need to be quantified. We may draw on genetic data to do so, and to thereby delineate appropriate population units for conservation (Palsboll et al. 2007). Regarding population size, an ongoing debate concerns the concept of a minimum viable population size (MVP), the 'population size required to provide some specified probability of persistence for a given time period' (Flather et al. 2011). According to Franklin (1980), as a rule-of-thumb, the effective size of a population should not be smaller than 50 in the short-term (five generations) to avoid inbreeding depression. This rule has been subjected to much controversy (Franklin & Frankham 1998; Frankham 2005; Jamieson & Allendorf 2012; Frankham et al. 2014). A recent perspective recommends a revision of the short-term minimum estimate of 50, and suggests that a short-term minimum viable Ne ≥ 100 may actually be needed to avoid inbreeding depression (Frankham et al. 2014). The authors further suggest setting the approximation to retain evolutionary potential to at least Ne ≥ 1000. Populations whose effective sizes range around these threshold levels should receive particular attention by conservation managers. As part of this MVP debate, the level of isolation between populations need be taken into account as well. Also termed connectivity, its extent impacts population persistency (Frankham et al. 2014). It has been shown that as little as one migrant per generation is sufficient to avoid harmful effects of genetic drift and inbreeding (Wright 1951). This one-migrant-per-generation rule is no threshold above which there will be no harmful effects (Mills & Allendorf 1996). Rather, it is based on the expectation that a significant reduction in the harmful effects of inbreeding can be achieved with one migrant per generation (Lowe & Allendorf 2010). Lowe and Allendorf (2010) termed this 'inbreeding connectivity' and it serves the purpose that many conservation managers will have in mind.
In the marine environment, physical barriers to movement are often absent and organisms with strong dispersal abilities may be expected to exhibit high connectivity patterns throughout their range of occurrence. Humpback whales (Megaptera novaeangliae) are a prime example of marine long-distance travelers with annual migration routes of > 8,000km (Darling et al. 1996; Rasmussen et al. 2007; Stevick et al. 2011). They spend the summer months in high-latitude areas for feeding while they migrate to low-latitude areas for breeding and calving during the winter (Dawbin 1966; Stone et al. 1990). In the North Atlantic, humpback whales during the winter mating season aggregate in two known subtropical breeding grounds in the Caribbean West Indies and off the West-African coast in Cape Verde (Jann et al. 2003). During the summer months, they migrate to several distinct feeding grounds in the Gulf of St. Lawrence, the Gulf of Maine, off Western Greenland, around Iceland, and in the Barents Sea (Katona & Beard 1990; Clapham et al. 1993a). Individual humpbacks thereby show a high degree of maternally transmitted site fidelity to their feeding ground, with little exchange between these areas (Katona & Beard 1990). Over many generations, such site fidelity can lead to population subdivision even in highly migratory species due to reproductive isolation; indeed, the complex repertoire of
behaviors in humpback whales has been shown lead to significant population structure even within a single ocean (Baker et al. 1990; Palsbøll et al. 1995). It may therefore be expected that we find small local populations among humpback whales besides their dispersal abilities.
Small populations are no rarity in baleen whales (suborder Mysticeti), and a range of species today range in low numbers. The gray whale (Eschrichtius robustus) population in the Western North Pacific is one example that has been estimated at only 140 individuals (95% confidence interval (CI) 128 - 152) in 2012 (Cooke et al. 2013a). It is listed as Critically Endangered by the IUCN Red List of Threatened Species (Reilly et al. 2008b). Similarly, the blue whale (Balaenoptera musculus) population of the Western North Atlantic is estimated to hold a minimum of 440 individuals (NOAA 2010). The species as a whole has been estimated to have been depleted by 70 – 90% over the last three generations, and today is listed as Endangered (Reilly et al. 2008a). Among humpback whales, the Arabian Sea subpopulation is one of the smallest estimated populations to date. It displays a unique non-migratory behavior and resides off Oman year-round. The current abundance estimate is 82 individuals (95% CI 60 – 111) (Minton et al. 2011).
For humpback whales as for most of the baleen species, the main cause of population decline can be attributed to the extensive whaling that took place throughout the last centuries (Best 1993; Clapham et al. 1999; Baker & Clapham 2004). Humpback whales in the North Atlantic have been subject to whaling activities since at least the 17th century (Stevick et al. 2003). During the 19th century, over 2,000 humpback whales were taken by the American whaling fleet (Smith & Reeves 2003), and between 1885 and 1910, close to 5,000 individuals were caught off Norway and Iceland (Ingebrigtsen 1929; Sigurjónsson 1988). Recent estimates assume the numbers of landings and total removals of humpback whales from the North Atlantic to be 21,476 (95% CI 21,257 – 21,895) and 30,842 (95% CI 29,558 – 32,126), respectively (Smith & Reeves 2010). It has been estimated that the North Atlantic humpback whale population was reduced to < 1,000 individuals before protection measures were put into action (Mitchell & Reeves 1983; Katona & Beard 1990). A study based on coalescent models for mitochondrial DNA even estimated a 95% decline in the North Atlantic humpback whale population pre- versus post-whaling (Roman & Palumbi 2003). Due to this drastic depletion, humpback whales have been protected since 1955 (Best 1993). Within the last few decades, the census size has steadily increased (Stevick et al. 2003). A series of abundance estimates from capture-recapture data for the breeding population in the West Indies has shown an increase of 3.5% individuals per year during the period 1979 - 1993 with the most precise estimate of 10,700 individuals (coefficient of variation = 0.068) in 1992/93 (Stevick et al. 2003). From the only other known, possibly genetically differentiated, North Atlantic breeding population in Cape Verde, only a single census size point estimate for the entire area is available rating the abundance at around 100 individuals (Punt et al. 2006). Due to this low census size, the Cape Verde breeding population may be vulnerable to threats faced
by small populations unless there is gene flow between Cape Verde and another breeding ground. To date, no quantification of gene flow between the two known breeding grounds in the North Atlantic exists, and the degree of genetic differentiation between Cape Verde and West Indies humpback whales remains unknown.
In this study, we set out to assess the degree of gene flow between individuals sampled at the two known North Atlantic breeding grounds. We further determined the effective population size (Ne) of humpback whales frequenting the Cape Verde Archipelago, and estimated their
demographic history. Our results suggest that the amount of genetic exchange between the known breeding grounds in the West Indies (off the Dominican Republic) and Cape Verde is indeed very limited. However, some data point to a potentially heterogeneous use of Caribbean waters, which would entail a yet unknown separation between whales in different parts throughout the Caribbean. Having likely undergone several historic and more recent population declines, the Cape Verde population has been small throughout history, and today may be close to or below a minimum viable size. Some data suggest though that Cape Verde might in fact be part of a larger Eastern Atlantic breeding range.
Materials and Methods
Sample collection and laboratory methods
Skin biopsy (Lambertsen 1987) and sloughed skin samples (Clapham et al. 1993b) were collected from free-ranging humpback whales in the West Indies in 2004-2005. In the Cape Verde Archipelago, individuals were biopsy-sampled in 2012-2013. Samples were conserved in a saturated NaCl solution with 20% dimethylsulphoxide (Amos & Hoelzel 1991) and were stored at either -20°C or -80°C.
Total cell-DNA was extracted from all samples following the protocol in Palsbøll et al. (1997). We PCR-amplified and genotyped 10 microsatellite loci (EV096, GATA028, GATA053, GATA098, GATA417, GT015, GT211, GT271, GT575, TAA031) in all samples, and an additional 10 microsatellite loci (AC087, EV001, EV037, EV094, GGAA520, GT011, GT023, GT101, GT195, GT307) in all Cape Verde samples. The microsatellites had di-, tri- or tetramer repeat motifs, respectively. For amplification and genotyping of six loci (GATA028, GATA053, GATA098, GATA417, GGAA 520 and TAA031), we followed the description of Palsbøll et al. (1997). EV001, EV037, EV094, and EV096 were amplified and genotyped as described by Valsecchi and Amos (1996). AC087, GT211, GT271, and GT575 were amplified and genotyped as described by Bérubé et al. (2005). GT023, GT101, GT195, and GT307 were amplified and genotyped as described by Bérubé et al. (2000). For amplification and genotyping of GT011 we followed the protocol of Bérubé et al. (1998). GT015 was amplified and genotyped as described
by Andersen et al. (2001). Further, part of the mitochondrial control region (mtDNA) was amplified and sequenced following Palsbøll et al. (1995), and a 288 bp fragment was used for analyses. Finally, the genetic sex of each sample was determined following the protocol as described by Bérubé and Palsbøll (1996).
Data analyses
Multiple samples from the same individual were identified by aligning and visually comparing data with no mismatches allowed in GENEIOUS ver. 6.0.6 (Drummond et al. 2009) (mtDNA) and in Microsoft EXCEL (Microsoft 2007) (microsatellite data). The number of samples from different individuals with identical genotypes across 10 loci arising by chance was estimated from the probability of identity (I) (Paetkau & Strobeck 1994) using the software GENECAP ver. 1.4 (Wilberg & Dreher 2004). If all microsatellite and mtDNA markers were identical, the samples were then identified as coming from the same individual. Mother-calf-dyads were visually identified during sample collection, and all duplicate samples and calves were removed from the dataset before further analysis. All analyses of microsatellite data were conducted with a dataset of 10 microsatellite loci unless stated otherwise.
Genetic diversity, Hardy-Weinberg proportions, linkage disequilibrium and neutrality
For mtDNA, we determined the number of unique control region haplotypes, haplotype diversity, nucleotide diversity (π) (Nei & Li 1979), and the neutrality test Tajima's D (Tajima 1989) for the West Indies and Cape Verde separately and combined using ARLEQUIN ver. 3.5 (Excoffier & Lischer 2010). For microsatellite data, we used Fstat 2.9.3 (Goudet 1995) to estimate allelic richness (AR), a measure of allelic diversity corrected for sample size. The package adegenet (Jombart 2008) implemented in R Software (R Development Core Team 2015) was used to determine the observed (Hobs) and expected heterozygosity (Hexp) (Nei 1978) per
locality and overall. We tested differences in genetic diversity between breeding grounds using a Wilcoxon signed-rank (WSR) test for paired samples (Wilcoxon 1945) also implemented in R Software. Linkage disequilibrium (LD) and FIS, a measure of deviation from Hardy-Weinberg equilibrium (HWE), estimated following Weir and Cockerham (1984), were assessed using GENEPOP ver. 4.2.1 (Rousset 2008). We tested for LD among loci for each population and for deviations from HWE using a Monte Carlo Markov Chain approach. For each assessment, the number of dememorizations was set to 10,000, the number of batches to 1,000, and the number of iterations to 10,000. Sequential Bonferroni corrections (Holm 1979; Rice 1989) were conducted for both LD and HWE calculations, and statistical significances are reported after sequential Bonferroni corrections.
Genetic differentiation between North Atlantic breeding grounds
We assessed the genetic differentiation between Cape Verde and the West Indies with hierarchical analyses of molecular variance (AMOVA) conducted in ARLEQUIN 3.5 (Excoffier & Lischer 2010). We estimated pairwise differentiation using FST (Weir & Cockerham 1984) for mtDNA and microsatellite data and 95% CIs were computed over 20,000 bootstraps. In order to determine whether samples from different years and between genders may be pooled, we estimated pairwise differentiation using FST (Weir & Cockerham 1984) between years and gender. For each breeding ground, we therefore used randomly drawn equal numbers of samples per year and per gender, respectively. We tested whether FST estimates were significantly larger than the range expected under panmixia using 10,000 permutations.
Recent effective population size estimates
We estimated the recent Ne’s of humpback whales at both known North Atlantic breeding
grounds one generation ago from microsatellite data with the LD method (Hill 1981; Waples 2006) implemented in NEESTIMATOR ver.2 (Do et al. 2014). We used the random mating system setting, and estimated Ne bounded by 95% CI. Critical values of allele frequencies to screen out
rare alleles were set to pcrit > 1/(2S) (S = the number of individuals with data at both pairs of
loci), so as to ensure that alleles which occur in only a single copy in the sample were excluded (Do et al. 2014). We excluded all samples with incomplete genotypes from the West Indies dataset, and all loci with incomplete genotypes from the Cape Verde dataset.
Detection and quantification of population size change
We inferred the demographic history of humpback whales in Cape Verde using a coalescent-based Approximate Bayesian Computation (ABC) approach (Beaumont et al. 2002; Csillery et
al. 2010) implemented in the software DIYABC ver. 2.0.4 (Cornuet et al. 2014). We compared seven demographic models (Fig. 1). S1 consists of a null hypothesis assuming a constant Ne
through time; S2 – S4 are models that assume a recent bottleneck event. S2 assumes a historic effective size (N1) that declined to its current Ne t1 generations ago; S3 assumes a historic
bottleneck t2 generations ago with a change from population size N3 to N2, followed by a second, more recent bottleneck t1 generations ago to current Ne; S4 assumes a historic expansion
t3 generations ago from effective population size N5 to N4, and a recent bottleneck t1 generations ago to current Ne. S5 also assumes a bottleneck, but more historically (t3 generations
ago) from a historic effective population size N1 to current Ne. The final class of models (S6 and
S7) assumes population expansions: S6 assumes a historic effective size (N6) that increased to its current Ne t1 generations ago; and S7 assumes a historic bottleneck event t3 generations ago
with a change from population size N7 to N6, followed by a recent expansion t1 generations ago to current Ne.
Approximate Bayesian computations
For each model, we simulated 1 million datasets based on the demographic history that describes the respective model. The parameters defining each model were considered as random variables that follow a predefined prior distribution (Table 5). We determined these priors based on initial DIYABC analyses to ensure that the distributions encompassed the parameter distribution. We jointly and separately analyzed a set of 20 microsatellite loci and a 288 bp fragment of the mtDNA control region. The microsatellite data were assumed to follow a generalized stepwise mutation model (Estoup et al. 2002) with two parameters: the mean mutation rate (µ) and the mean parameter of the geometrical distribution assumed for the length in repeat numbers of mutation events (P) drawn from uniform distributions (µ: 10-5 – 10-3; P: 0.1 – 0.3). All but two loci had a possible range of 40 contiguous allelic states; GATA417 and GGAA520 had a larger possible range of 60 contiguous allelic states. Each locus was characterized by individual µ and
P values drawn from gamma distributions with respective means µ and P, and shape parameters
of 2. For mtDNA, we selected the Hasegawa-Kishino-Yano (HKY) DNA substitution model (Hasegawa et al. 1985) as empirically determined in JMODELTEST ver. 2.1.5 (Darriba et al. 2012) and assumed a per-site and generation µ ranging uniformly between 10-7 and 10-5. We further used the proportion of invariant sites of 63% and the shape of the gamma distribution of mutations among sites equal to 0.23 likewise determined by JMODELTEST.
Summary statistics
We calculated a range of summary statistics for each competing demographic model: For mtDNA data, the number of haplotypes, the mean of pairwise differences, Tajima’ D (Tajima 1989), and the mean number of the rarest nucleotide at segregating sites were estimated. For microsatellite data, the mean number of alleles and the mean Garza-Williamson M-ratio (Garza & Williamson 2001) were estimated.
Model choice procedure and performance analyses
The posterior probability of each demographic model was estimated using a polychotomous logistic regression (Cornuet et al. 2008; Cornuet et al. 2010) on the 1% of simulated datasets closest to the observed dataset. In order to evaluate the confidence in model choice, type I error rates were estimated following the method described in (Cornuet et al. 2010).
Parameter estimation and model checking
Based on the best demographic model, we estimated posterior parameter distributions using a logit transformation of parameter values and subsequently conducted a local linear regression on the 1% of closest simulated parameters. Finally, following Gelman et al. (1995), we evaluated
the ability to reproduce the observed data as described in Cornuet et al. (2010). The test quantities chosen to do so for microsatellite loci were the mean genic diversity (= mean Hexp) and
the mean allelic size variance. For mtDNA data, we calculated the number of segregating sites, the variance of pairwise differences, the number of private sites, and the variance of numbers of the rarest nucleotide at segregating sites for mtDNA data.
Results
After removal of duplicate samples and calves, our microsatellite dataset consisted of 1,745 individual humpback whales sampled in the West Indies and 36 individuals sampled in Cape Verde; our mtDNA dataset consisted of 1,739 individuals from the West Indies and 36 individuals from Cape Verde. The probability of different individuals having identical genotypes across 10 microsatellite loci by chance was estimated under the assumption of HWE at 1.62*10 -11
for the West Indies, and at 7.29*10-9 for Cape Verde. Allele frequencies between humpback whales of different gender and sampled in different years were not significantly different in either sampling locality (Table 1); we therefore per locality pooled samples of different years and gender in subsequent analyses.
Genetic diversity, Hardy-Weinberg proportions, linkage disequilibrium and neutrality
We identified 27 mtDNA haplotypes among the West Indies samples and seven among the Cape Verde samples. All Cape Verde haplotypes were shared with the West Indies. Haplotype diversity was significantly higher in the West Indies than in Cape Verde and π was almost twice as high in the West Indies (Table 2). Our observed estimates of Tajima's D did not differ significantly from the expectation under neutrality for either breeding ground (Table 2).
The microsatellite dataset contained very few missing data (Cape Verde = 0.27%, West Indies = 0.05%, Combined = 0.05%). Genetic diversity estimates, Hardy-Weinberg proportions, and neutrality test estimates per locus for each breeding ground and for both breeding grounds combined are presented in Table 3. The nuclear genetic diversity was comparable between Cape Verde and the West Indies (WSP tests, P-values > 0.05 for all pairwise comparisons). AR over all loci was also comparable between the two breeding grounds, as were Hobs and Hexp. We failed
to reject HWE expectations in all but two loci in the West Indies after sequential Bonferroni corrections (GATA028 and GT417). No evidence was found for LD among loci after sequential Bonferroni corrections.
Genetic differentiation between North Atlantic breeding grounds
Genetic differentiation between Cape Verde and the West Indies was highly significant based on mtDNA and microsatellite data (mtDNA: FST = 0.116 (95% CI = 0.077 - 0.172); microsatellites:
FST = 0.02 (95% CI = 0.012 - 0.029)).
Recent effective population size estimates
After exclusion of all samples with incomplete genotypes from the West Indies dataset and all loci with incomplete genotypes from the Cape Verde dataset, 1738 West Indies samples genotyped at 10 loci, and 36 Cape Verde samples genotyped at 17 loci remained for analysis. To evaluate the influence of rare alleles on the Ne results, we excluded rare frequency alleles at
frequencies (pcrit) smaller than 0.02 through 0.05 for Cape Verde, and smaller than 0.01 through
0.05 for the West Indies. Estimates among pcrit values within the same breeding ground
overlapped (Table 4). For Ne evaluation we selected pcrit = 0.02 with estimates bound by 95%
CIs. Estimates of Ne using the LD method differed widely between the two breeding grounds and
were 20 times higher for the West Indies than for Cape Verde (Table 4). We also estimated Ne
from LD for the West Indies for a dataset with five microsatellite loci including 1,594 additional samples from 1992/93 to investigate potential differences due to sample size and variation between time periods (data and results not shown). As no differences were detected (all CIs overlapped), we reported only results based on 1,738 samples.
Detection and quantification of population size change
To compute the timing and the nature of demographic changes in the Cape Verde breeding population, all ABC analysis steps were conducted with three datasets: one combining microsatellite and mtDNA data, and two using only one of these, respectively. We first evaluated the posterior probability of each competing demographic model. This model checking pointed to the group modeling a population bottleneck (S2 - S4) in the microsatellite and combined dataset (Table 6). Using mtDNA data alone, none of the models significantly differed from the observed data. When including microsatellites, models S2 through S4 received significantly greater support than the other models, but had very similar posterior probabilities among each other (Table 7). We next estimated the type I error probability by evaluating the power of the model choice procedure. The model with the highest posterior probability (S3) was selected in slightly more than 50% of cases (microsatellite dataset: 52.0%; microsatellite & mtDNA dataset: 51.6%), leaving a high type I error rate. When taking the group of bottleneck models together (S2 - S4), the proportion of times that those were not selected dropped to 19.8% or 18.6% for the microsatellite and combined dataset, respectively (Table 8). The posterior parameter estimates and their 95% CIs were then computed under the best-supported model (S3) (Table 9). CIs
between the three datasets largely overlapped. According to the estimation using both nuclear and mtDNA data, a historic population with an Ne of 175,000 (95% CI 65,200 – 488,000)
declined to 11,300 breeding whales (95% CI 6,630 – 145,000) approximately 20,200 generations ago (95 % CI 6,560 – 24,600) and subsequently declined further to a current Ne of 2,550 (95 %
CI 1,780 – 5,700) approximately 4,560 generations ago (95 % CI 1,620 – 14,700). The model checking procedure provided support that the observed data were plausible under the selected posterior predictive distribution of S3 as the probabilities Prob (Ssimul < Sobs) for all summary
statistics were not significantly different from 0.5 (Table 10).
Discussion
While humpback whales are among the widest-traveling mammals on earth with annual migration routes of > 8,000km, see e.g. (Rasmussen et al. 2007; Stevick et al. 2011), they have developed a complex repertoire of behaviors that can lead to significant population structure even on relatively small spatial scales (Baker et al. 1990; Palsbøll et al. 1995). To date, the extent of genetic connectivity between the two known humpback whale breeding grounds in the North Atlantic was not known and it was suspected that the whales in Cape Verde might form a small, genetically isolated population (Wenzel et al. 2009).
How much gene flow exists between Cape Verde and the West Indies breeding grounds?
Our study revealed that Cape Verdean humpback whales are considerably differentiated from the West Indies on an evolutionary time scale. Genetic divergence estimates between the two breeding grounds suggest very low long-term average gene flow; the matrilineal differentiation is of the same order of magnitude as genetic differentiation between ocean basins that was estimated from a global humpback whale dataset from breeding and feeding grounds (Jackson et
al. 2014). This degree of isolation even led Jackson et al. (2014) to propose that different
subspecies inhabit different ocean basins. Nuclear differentiation estimates in our study are less pronounced than in Jackson et al.'s study, and less pronounced than mitochondrial ones. Nonetheless do they demonstrate the considerable limitation of gene flow. Nuclear divergence estimates are approximately 10 times higher between the two breeding grounds than between the North Atlantic feeding areas (unpublished data). The higher divergence at the maternally inherited mtDNA marker in comparison to divergence at the Mendelian-inherited nuclear markers suggests male-mediated gene flow (Karl et al. 1992). This is in accordance with suggestions in previous humpback whale studies (Palumbi & Baker 1994; Palsbøll et al. 1995; Palsboll et al. 2004), and is supported by the fact that genetic diversity estimates in the two breeding grounds are comparable in the nuclear dataset while they differ in the maternally inherited mtDNA.
Data encompassing a more recent time scale suggest a similar extent of divergence. A comparison of 88 individual fluke photographs from the Cape Verde region collected between 1990 and 2009 to over 6,500 individual fluke photographs maintained in the North Atlantic
Humpback Whale Catalogue yielded three matches with whales photographed in the feeding
grounds of the Eastern North Atlantic, but no matches were found with the Western North Atlantic (Jann et al. 2003; Wenzel et al. 2009). While the low match-rate with the Eastern North Atlantic might be comprehensible given the limited photo-ID effort throughout the last decades in this area (Wenzel et al. 2009), the lack of any matches with the extensive data base from the Western North Atlantic is surprising and designative. It suggests very little migration between this part of the Atlantic and Cape Verde. Photo identification of some few individuals (four sightings) that have been sighted in both Cape Verde and Guadeloupe in the Eastern Caribbean provides the only direct proof of current migration between Cape Verde and the Western North Atlantic to date (Stevick pers. com.). Along with a high interannual resighting rate suggesting strong site fidelity (Wenzel et al. 2009), this low match rate points to a high contemporary degree of isolation between these two known North Atlantic breeding grounds; It hence supports our findings from genetic data and the hypothesis that Cape Verde might constitute a distinct breeding population.
A logical next step is to test the degree of genetic differentiation between the humpback whales within the Caribbean - between the West Indies and Guadeloupe in the East. Given the high maternally induced population structure between Cape Verde and the West Indies and the absence of photographic matches between these areas while some whales seem to be migrating between Guadeloupe and Cape Verde, we might expect to find whales in the West Indies and Guadeloupe to be distinct from each other at a molecular level.
How many whales belong to the Cape Verde population?
Recent research has suggested that the population size of the potentially isolated Cape Verde humpback whales might be very small (Jann et al. 2003; Wenzel et al. 2009; Ryan et al. 2014). On an evolutionary time scale, the small size of Cape Verdes humpback whale population is supported by our genetic diversity estimates for mitochondrial data. Low genetic diversity can be caused by a small Ne, see e.g. Fontaine et al. (2014). The small number of haplotypes, as well as
the significantly smaller haplotype and nucleotide diversities in Cape Verde in comparison to the West Indies suggest that many more whales were part of the West Indies population than of the Cape Verde one.
It is also possible to infer a direct estimate of the historic Cape Verde Ne from the data we have
available. This can be achieved by assuming, for the purpose of comparability, that the LD Ne
factor between the LD Ne and ABC estimates of current Ne, and then multiplying the historic
ABC Ne estimate by that same factor provides a historic Ne estimate of 3,836 individuals for the
Cape Verde population. Reason for the initial difference between ABC and LD Ne estimates
likely are the different time scales at which the respective methods are informative. Whereas the estimate based on LD provides a population size estimate of the parent generation of the sampled individuals, the ABC estimate represents an average Ne estimate since the last change in
population size (t1), hence applying to a larger temporal scale. This difference in population size and the time scales which the two estimates apply to supports the depletion of the Cape Verde humpback whales through heavy whaling in the last 150 years (Mitchell & Reeves 1983; Smith & Reeves 2003, 2010).
On a more recent time scale, estimates of Ne based on the LD method have been shown to be reliable for small Ne’s when at least 25-50 individuals are sampled at 10-20 polymorphic loci (Waples & Do 2010). For larger populations, reliable estimates are more difficult to obtain as the genetic signal becomes weak in relation to sampling noise (Ansmann et al. 2013). In our study, we included a similar number of samples and microsatellite loci as recommended by (Waples & Do 2010). However, the calculation of our CIs is based on the assumption of a homogeneous population with non-overlapping generations. As our focal population does not meet this assumption, the true CIs may be larger than our reported ones. The only abundance estimate of the entire Cape Verde Archipelago to date rates the total abundance at 99 individuals (Punt et al. 2006), while a yet more recent study incorporating recapture data only from around the island of Boa Vista concluded that the census population size around Boa Vista ranged between 171 and 260 individuals (Ryan et al. 2014). Taking into account that spatial and temporal research efforts were not completely comparable between studies, our Ne estimates overall support that Cape Verde hosts few humpback whales. A small and potentially isolated population is further affirmed by high interannual re-sighting rates of whales in the archipelago (>22%) (Wenzel et al. 2009).
Our estimate of 43 to 76 breeding individuals one generation ago places the Cape Verde humpback whale breeding population at a precarious state, the small Ne making it vulnerable to
stochastic effects and anthropogenic impacts. It is by a multitude smaller than the West Indies breeding population, and is within the range of or below a minimum viable short-term population size that is required to avoid inbreeding depression according to Franklin’s rule-of-thumb. As such, any management should be conservative and treat the Cape Verde population as one that is currently potentially at risk.
What do genetics reveal about the demographic history of the Cape Verde humpback whale population?
Our ABC analyses support the hypothesis of a population decline among Cape Verde humpback whales. Among the three bottleneck models that we included in our analysis, each had an approximate posterior probability of 1/3. While the demographic model modeling a two-stage decline was consistently best-supported, the high type I error rate (55%) suggested that our data did not allow distinguishing between those three most likely models without doubt. We therefore caution to put too much emphasis on the precise estimates, but to rather take notion of the presence of a substantial population size reduction throughout the last millennia. This decline does not however reflect the intense recent decline caused by whaling as we would have expected, but instead denotes more historic events. Of the best-supported model, the more recent of the two bottleneck events dates back ca. 66,000 years (95% CI 23,500 - 213,000) when assuming a generation time of 14.5 years (Taylor et al. 2007). This time period - let alone the more ancient bottleneck estimate of ca. 290,000 years (95% CI 95,000 - 357,000) - by far pre-dates any noteworthy anthropogenic impacts; humpback whales in the North Atlantic were hunted most intensively within the last 150 years (Mitchell & Reeves 1983; Smith & Reeves 2003, 2010), and even early whaling may date back only as far as 6,000 BC (Lee & Robineau 2004). Instead, the detected bottleneck signal possibly relates to a decreased ocean productivity during the Pleistocene which may have limited the food resources available to humpback whales at the time and hence may have caused population decline(s) (Thomas et al. 1995).
However, the interpretation of our results need be viewed also in the light of the limitations that the applied analytical framework poses. Our ABC approach does not take any migration between Cape Verde and a breeding ground other than the West Indies into account. New information suggests that indeed this may be inaccurate and that whales from the Northern and Southern hemispheres might come into contact in the Cape Verde area. In September 2014 two humpback whale biopsy samples were collected in Cape Verde (unpublished data, not included in this study) whose mtDNA sequences were identical to whales from the Southern hemisphere (Rosenbaum et al. 2009). If present in our dataset, whales from the Southern hemisphere likely add genetic diversity to our local population (Palsbøll et al. 2013). Any derived abundance estimate will hence reflect the total of these populations rather than of the Cape Verde area alone. Besides, while we selected an appropriate prior range of mutation rates for mtDNA and nuclear data, we caution to put too much emphasis in the precise bottleneck timing estimates given above, as they depend on these selected mutation rates and may hence vary considerably.
Conclusions
We investigated the degree of gene flow between the only two known North Atlantic humpback whale breeding populations and estimated the size of the allegedly very small population in the Cape Verde Archipelago. While the amount of genetic exchange between the known breeding grounds in the West Indies (off the Dominican Republic) and Cape Verde is very limited, confirmed migrations between Cape Verde and the East Indies (Guadeloupe) point to existing gene flow between this Eastern part of the Caribbean and Cape Verde. If this indeed is the case, a yet unknown separation might exist between whales in different parts throughout the Caribbean. A heterogeneous use of Caribbean waters for breeding would have important implications for conservation management. Consequently, a next step in humpback whale research should focus on testing the degree of molecular distinction between Guadeloupe and the main breeding ground around the Dominican Republic in the West Indies.
Having likely undergone several historic and more recent population declines, the Cape Verde population has been small throughout history, and today may be close to or below a minimum viable size. If indeed the Cape Verde humpback whales have as little genetic exchange with any other, yet unknown breeding ground as they have with the West Indies, this population might be very sensitive to stochastic and anthropogenic effects and should be handled with great care in a conservation context. This being said, some data suggest that Cape Verde might in fact be part of a larger Eastern Atlantic breeding range. Genetic distances of mtDNA data collected in the North Atlantic feeding areas in 1992/93 suggest that only 40 to 60% of Icelandic humpback whales winter in the West Indies, and that as many as 90% of humpback whales feeding in the Barents Sea may winter elsewhere than the West Indies (unpublished data). To date this may only be speculation as sampling efforts in the Eastern Tropical Atlantic have been very limited. Future research would therefore greatly benefit from increasing sampling efforts in the Eastern Tropical Atlantic to discover any potential additional breeding grounds and to shed light on the nescience of how isolated Cape Verdes humpback whales indeed are.
Table 1: Estimates of genetic differentiation between samples of humpback whales of different gender, and between samples of different years, for the Cape Verde and the West Indies breeding grounds.
Breeding
Ground Molecular Marker FST
P-value
Gender Cape Verde mtDNA 0.000 0.863
West Indies mtDNA 0.002 0.058
Sampling
Year Cape Verde mtDNA 0.000 0.430
Microsatellites 0.015 0.510
West Indies mtDNA 0.001 0.184
Microsatellites 0.000 1.000
Table 2: Genetic diversity estimates and neutrality estimates for mtDNA per humpback whale North Atlantic breeding ground and for both breeding grounds combined. Number of individuals (N); number of haplotypes (HT); HT diversity with 95% confidence intervals (CI); nucleotide diversity (π) averaged over loci with 95% CI; Tajima’s D; and significance (P-value) of Tajima’s D.
Parameters Cape Verde West Indies Combined
N 36 1,739 1,775 HT 7 27 27 HT diversity (95% CI) 0.6810 (0.664-0.698) 0.8264 (0.826-0.827) 0.8267 (0.826-0.827) π (95% CI) 0.0131 (0.011-0.016) 0.0229 (0.022-0.023) 0.0229 (0.022-0.023) Tajima's D -0.638 3.001 2.982 Tajima's D P-value < 0.303 < 0.995 < 1
77
Table 3: Genetic diversity estimates, Hardy-Weinberg proportions, and neutrality test estimates for 10 microsatellite loci among humpback whales in North Atlantic breeding grounds (Combi = Cape Verde and West Indies combined). Number of samples (N); Number of alleles per locus (NA); Allelic richness (AR) (based on a minimum sample size of 35); Observed heterozygosity (Hobs); Expected heterozygosity (Hexp); P-values with 95% CIs as a measure of deviation from Hardy-Weinberg equilibrium (HWE); and FIS, a measure of deviation from HWE estimated following Weir and Cockerham (1984). * indicate significant P
-values at the 0.05 level.
EV096 GATA028 GATA053 GATA098 GATA417 GT015 GT211 GT271 GT575 TAA031
Cape N 36 36 36 36 35 36 36 36 36 36 Verde NA 9 6 7 10 14 9 7 7 8 10 AR 8.972 5.999 6.972 7.972 14 8.943 6.972 6.916 7.944 9.971 Hobs 0.861 0.556 0.778 0.944 0.886 0.639 0.778 0.722 0.833 0.778 Hexp 0.832 0.564 0.779 0.860 0.849 0.743 0.789 0.574 0.758 0.775 HWE (95% CI) 0.883 (0.8827-0.8833) 0.901 (0.9007-0.9013) 0.715 (0.7147-0.7153) 0.243 (0.2423-0.2437) 0.054 (0.0533-0.0547) 0.026 (0.0257-0.0263) 0.103 (0.1027-0.1033) 0.489 (0.4883-0.4897) 0.769 (0.7683-0.7697) 0.053 (0.0527-0.0533) FIS 0.154 0.010 -0.029 0.028 0.095 -0.245 0.016 -0.085 -0.021 0.030 West N 1,745 1,743 1,745 1,744 1,743 1,743 1,745 1,745 1,745 1,744 Indies NA 10 10 10 10 16 15 7 15 12 16 AR 8.591 6.881 8.146 7.016 11.262 10.711 6.79 7.225 9.065 9.612 Hobs 0.811 0.464 0.831 0.954 0.837 0.801 0.818 0.567 0.708 0.816 Hexp 0.821 0.467 0.832 0.851 0.874 0.796 0.812 0.581 0.703 0.817 HWE (95% CI) 0.397 (0.3968-0.3972) 0* (0.000) 0.975 (0.9750-0.9750) 0.665 (0.6649-0.6651) 0.004* (0.0040-0.0040) 0.630 (0.6298-0.6302) 0.773 (0.7729-0.7731) 0.390 (0.3897-0.3903) 0.950 (0.9500-0.9500) 0.944 (0.9439-0.9441) FIS 0.007 0.042 0.025 0.012 -0.005 -0.008 -0.006 0.002 -0.007 0.001 Combi N 1,781 1,779 1,781 1,780 1,778 1,779 1,781 1,781 1,781 1,780 NA 10 11 10 10 19 15 7 15 12 16 AR 8.626 7.051 8.128 7.041 11.496 10.697 6.81 7.217 9.053 9.659 Hobs 0.812 0.465 0.830 0.954 0.838 0.798 0.817 0.570 0.710 0.815 Hexp 0.822 0.470 0.831 0.851 0.874 0.796 0.814 0.581 0.705 0.817
78
Table 4: Estimates of contemporary effective population sizes (Ne) with 95% CIs of
humpback whales in Cape Verde (based on 17 loci) and in the West Indies (based on 10 loci) for different critical values (pcrit) using the LD method implemented in NEESTIMATOR
ver.2 (Do et al. 2014).
Cape Verde West Indies
Pcrit Ne Lower 95% CI Upper 95% CI Ne Lower 95% CI Upper 95% CI 0.05 41.6 32.6 55.4 913.7 756.8 1,120.9 0.04 49 37.9 66.7 957.3 796.9 1,168.2 0.03 49 37.9 66.7 956.6 811.8 1,141.4 0.02 55.9 43.3 76.2 1,039 896.1 1,217.6 0.01 - - - 1,028.6 901.5 1,183.5
Table 5: Prior distributions for demographic parameters and for locus-specific mutation models, and simulation conditions for demographic parameter estimation implemented in DIYABC ver. 2.0.4 (Cornuet et al. 2014). UN = uniform distribution, GA = gamma
distribution.
Priors for demographic parameters
Ne UN ̴[10; 6,000] N1 UN ̴[100; 100,000] N2 UN ̴[10; 150,000] N3 UN ̴[100; 500,000] N4 UN ̴[100; 500,000] N5 UN ̴[10; 250,000] N6 UN ̴[10; 3,000] N7 UN ̴[100; 500,000] t1 UN ̴[1; 20,000] t2 UN ̴[10; 25,000] t3 UN ̴[10; 200,000]
Priors for mutation model
microsatellites
Mean - µmic UN ̴[10-5 - 10-3] Gam - µmic GA ̴[10-5 - 10-3, 2]
Mean - P UN ̴[0.1 - 0.3]
Gam - P GA ̴[0.01 - 0.9, 2]
mtDNA control region
µsec UN ̴[10-7 - 10-5] K1 UN ̴[0.05 - 20] % invar. sites 63 shape 0.23 Conditions N1>Ne t2>t1 N2>Ne t3>t2 N3>N2 N5<N4 N4>Ne N6<Ne N7>N6
80
Table 6: Model checking for each summary statistic for each demographic model (S1 – S7, see Fig. 1) for one dataset including both mtDNA and microsatellite data, as well as datasets that include only mtDNA and only microsatellite data, respectively. The probability Prob (Ssimul < Sobs) was computed from 500 virtual datasets simulated from the posterior distributions of parameters obtained under a given demographic model. Values indicate the proportion of simulated datasets which have a value below the observed one (Ssimul < Sobs). NAL = number of alleles, MGW = Garza-Williamson’s M, NHA = number of haplotypes, MPD = mean of pairwise differences, DTA = Tajima’s D, MNS = mean number of rarest nucleotide at segregating sites. **, *** = tail-area probability < 0.01, < 0.001, respectively.
Microsatellite data
Summary Statisics Observed value S1 S2 S3 S4 S5 S6 S7
NAL 7.45 0.7072 0.6139 0.5498 0.5749 0.7013 0.6797 0.672
MGW 0.5017 0.0002(***) 0.0746 0.3621 0.2327 0.0076(**) 0.0001(***) 0.0035(**)
MtDNA data
Summary Statisics Observed value S1 S2 S3 S4 S5 S6 S7
NHA 7 0.4169 0.4155 0.4103 0.4078 0.4224 0.3044 0.3143
MPD 3.7825 0.6344 0.5967 0.5627 0.5739 0.6353 0.5605 0.5679
DTA -0.4279 0.2538 0.2426 0.2361 0.2386 0.2527 0.2584 0.2614
MNS 5.1765 0.4794 0.4627 0.454 0.4604 0.4781 0.4682 0.4705
Microsatellite and mtDNA data
Summary Statisics Observed value S1 S2 S3 S4 S5 S6 S7
NAL 7.45 0.7071 0.6133 0.551 0.5745 0.7015 0.6802 0.6721 MGW 0.5017 0.0001(***) 0.0747 0.3618 0.2331 0.0074(**) 0.0001(***) 0.0036(**) NHA 7 0.42 0.4184 0.413 0.4118 0.4242 0.3086 0.3079 MPD 3.7825 0.6547 0.6162 0.5798 0.5929 0.6547 0.5857 0.5813 DTA -0.4279 0.2316 0.2232 0.2165 0.2188 0.2311 0.232 0.2316 MNS 5.1765 0.4782 0.4623 0.451 0.4586 0.4773 0.4675 0.4657
Table 7: Relative posterior probability for each competing demographic model with 95% CIs based on summary statistics for the model choice analysis using a logistic approach. Results are shown for one dataset including both mtDNA and microsatellite data, as well as for datasets that include only mtDNA and only microsatellite data, respectively. Individual values calculated for each summary statistic per scenario are provided in table 6.
Microsatellite data MtDNA data Microsatellite and mtDNA data
Model Post. prob. (95% CI) Model Post. prob. (95% CI) Model Post. prob. (95% CI)
S1 0 % (0 - 0) S1 13.6 % (12.9 - 14.3) S1 0.05 % (0 - 0.9) S2 21.8 % (21.5 - 22.1) S2 16.3 % (15.5 - 17.1) S2 26.3 % (25.2 - 27.4) S3 40.0 % (39.6 - 40.1) S3 12.6 % (11.9 - 13.4) S3 37.9 % (36.8 - 39.1) S4 35.2 % (34.8 - 35.6) S4 12.0 % (11.3 - 12.7) S4 32.4 % (31.3 - 33.5) S5 2.1 % (1.9 - 2.4) S5 15.3 % (14.5 - 16.1) S5 2.6 % (1.7 - 3.5) S6 0 % (0 - 0) S6 15.7 % (14.9 - 16.5) S6 0.03 % (0 - 0.9) S7 0.9 % (0.6 - 1.1) S7 14.5 % (13.8 - 15.3) S7 0.7 % (0 - 1.6)
Table 8: Instances (in %) of the selected demographic model exhibiting the highest posterior probability compared with all competing models for 500 simulated datasets that are generated under the best-supported model (S3) using a logistic regression. Results are shown for one dataset including both mtDNA and microsatellite data, as well as for datasets that include only mtDNA and only microsatellite data, respectively.
Model Posterior probability (%) - log reg
S1 S2 S3 S4 S5 S6 S7
Microsatellites 11.2 19 52 9.2 7 1.6 0
MtDNA 13 5.2 13.2 1.2 27 31.2 9.2
Microsatellites
Table 9: Demographic parameters (mode and 95% CI) estimated under the best-supported demographic model (S3) for one dataset including both mtDNA and microsatellite data, as well as for datasets that include only mtDNA and only microsatellite data, respectively.
Microsatellite data
Parameter Mode 95% CI
Current effective population size (Ne) 2,760 (1,800 - 5,820)
Population size before most recent bottleneck (N2) 16,000 (7,780 - 145,000)
Population size before historic bottleneck (N3) 218,000 (72,300 - 489,000)
Time since recent bottleneck (t1) 6,120 (1,280 - 15,700)
Time since historic bottleneck (t2) 19,600 (6,040 - 24,600)
MtDNA data
Parameter Mode 95% CI
Current effective population size (Ne) 2,250 (1,090 - 5,810)
Population size before most recent bottleneck (N2) 15,200 (5,430 - 145,000)
Population size before historic bottleneck (N3) 235,000 (52,700 - 487,000)
Time since recent bottleneck (t1) 3,530 (963 - 18,500)
Time since historic bottleneck (t2) 22,000 (4,830 - 24,600)
Microsatellite and mtDNA data
Parameter Mode 95% CI
Current effective population size (Ne) 2,550 (1,780 - 5,700)
Population size before most recent bottleneck (N2) 11,300 (6,630 - 145,000)
Population size before historic bottleneck (N3) 175,000 (65,200 - 488,000)
Time since recent bottleneck (t1) 4,560 (1,620 - 14,700)
Table 10: Cumulative distribution function values of each test quantity (= Prob (Ssimul < Sobs) to evaluate the fit of the observed data with respect to simulated data from 1000 simulated datasets under the best-supported model (S3). Results are shown for one dataset including both mtDNA and microsatellite data, as well as for datasets that include only mtDNA and only microsatellite data, respectively. HET = mean expected heterozygosity, VAR = mean allelic size variance, NSS = Number of segregating sites, VPD = variance of pairwise differences, PSS = private segregating sites, VNS = variance of numbers of the rarest nucleotide at segregating sites.
Dataset Summary Statisics Observed value Proportion (Ssimul < Sobs) Microsatellite HET 0.716 0.472 data VAR 22.5436 0.282 MtDNA data NSS 18 0.748 VPD 15.4932 0.763 PSS 18 0.748 VNS 36.9689 0.886 Microsatellite HET 0.716 0.507 and mtDNA data VAR 22.5436 0.249 NSS 18 0.783 VPD 15.4932 0.789 PSS 18 0.783 VNS 36.9689 0.905
Figure 1: Alternative demographic models of Cape Verde humpback whale demography tested by implementing the ABC approach.
